8.

EUROPEAN CONFERENCE
ON
RENEWABLE ENERGY SYSTEMS

ECRES 2020
24-25 August 2020
Istanbul / TURKEY
www.ecres.net

PROCEEDINGS
Edited by
Prof. Dr. Erol Kurt
ISBN: 978-605-86911-8-6

1
 Organizing Institutions

Published by
Erol Kurt
on 30th August 2020
(Ankara/Turkey)

All rights of the abstract book of European Conference on Renewable Energy Systems
(ECRES) are preserved. No part of this publication may be produced, stored in retrieval system,
or transmitted in any form of electronic, mechanical and photocopying or reproduction
technique without the prior permission of the publisher.
The responsibility for the ingredients covering information and opinion rests exclusively with the
authors and independent from the organizers and publisher.

Book typesetted and designed by

Assoc. Prof. Dr.Yunus Uzun and Bekir Dursun

2
Cooperative Institutions

Supporting Institutions

3
 Committees

Honorary Chairman
Prof. Dr. Musa Yıldız, Rector of Gazi University, Turkey

Local Organizing Committee

Prof. Dr. Erol Kurt, Gazi University, Turkey
Assoc. Prof. Dr. Yunus Uzun, Aksaray University, Turkey

International Organizing Committee

Prof. Dr. Adnan Sozen, Gazi University, Turkey

Prof. Dr. Alfredo Vaccaro, Sannio University, Italy
Prof. Dr. Carlos Rubio-Maya, Univ. Michoacana de San Nicolás de Hidalgo, Mexico
Assoc. Prof. Dr. Fontina Petrakopoulou-Robinson, Universidad Carlos III de Madrid, Spain
Prof. Dr. Ian Hunter, University of Leeds, UK
Prof. Dr. Jose Manuel Lopez Guede, University of Basque Country, Spain
Prof. Dr. Mehmet Tekerek, Kahramanmaras Sutcu Imam University, Turkey
Prof. Dr. Mordjaoui Mourad, Skikda University, Algeria
Prof. Dr. Murat Kunalbayev, Inst. of Information and Computational Technologies, Kazakhstan
Prof. Dr. Nicu Bizon, Pitesti University, Romania
Prof. Dr. Peter Childs, Imperial College London, UK
Prof. Dr. Poul Alberg Østergaard, Aalborg University, Denmark
Prof. Dr. Raoul Rashid Nigmatullin, Kazan National Research Tech. Univ., Tatarystan, Russia
Prof. Dr. Saad Mekhilef, University of Malaya, Malaysia
Prof. Dr. Saeed Badshah, Int. Islam. Uni. Islamabad, Pakistan
Prof. Dr. Sagdulla L. Lutpullaev, Uzbekistan Academy of Sciences, Uzbekistan
Prof. Dr. Serguei Martemianov, University of Poitiers, France
Prof. Dr. Shadi Shahedipour-Sandvik, University at Albany, USA
Prof. Dr. Shadia J. Ikhmayies, Al Isra University, Jordan
Prof. Dr. Waqar Mahmood, University of Engineering and Technology, Pakistan
Assoc. Prof. Dr. Yussupova Gulbakhar, Turan University, Kazakhstan

International Scientific Committee

Prof. Dr. Alessandro Zanarini, University of Bologna, Italy
Prof. Dr. Antonio Soria Verdugo, Universidad Carlos III de Madrid, Spain
Dr. Athar Waseem, International Islamic University, Pakistan
Assoc. Prof. Dr. Aybaba Hançerlioğulları, Kastamonu University, Turkey
Prof. Dr. Ayman El-Hag, American University of Sharjah, UAE
Prof. Dr. Bernabé Marí Soucase, Polytechnical University of Valencia, Spain
Assist. Prof. Dr. Burak Akin, Yildiz Technical University, Turkey
Assoc. Prof. Dr. Bünyamin Tamyürek, Eskisehir Osmangazi University, Turkey
Prof. Dr. Carolina Marugan Cruz, Universidad Carlos III de Madrid, Spain
Assoc. Prof. Dr. Christiane Hennig, German Biomass Research Center, Germany
Prof. Dr. Corneliu Marinescu, Transilvania University of Brasov, Romania
Assoc. Prof. Coşku Kasnakoğlu, TOBB Ekonomy and Technology University, Turkey
Assoc. Prof. Dr. Çigdem Yangin Gömeç, Istanbul Technical University, Turkey
Dr. Danny Müeller, Technische Universität Wien, Austria
Assoc. Prof. Dr. Diana Zalostiba, Riga Technical University, Latvia
Dr. Eduar Eduardo Zarza, CIEMAT Solar Platform of Almeria, Spain
Prof. Dr. Eleonora Guseinoviene, Klaipeda University, Lithuania

4
Prof. Dr. Farqad Al-Hadeethi, Royal Scientific Society of Jordan, Jordan
Assoc. Prof. Dr. Fontina Petrakopoulou-Robinson, Universidad Carlos III de Madrid, Spain
Prof. Dr. Francesco Calise, University of Naples Federico II, Italy
Assoc. Prof. Dr. Francesco Cottone, University of Perugia, Italy
Prof. Dr. Guang-Bin Huang, Nanyang Technological University, Singapore
Prof. Dr. Guido Van Oost, University of Gent, Belgium
Prof. Dr. Güngör Bal, Gazi University, Turkey
Prof. Dr. H. Hilal Kurt, Gazi University, Turkey
Prof. Dr. H. Mehmet Şahin, Karabük University, Turkey
Prof. Dr. Haitham Abu-Rub, Texas A&M University at Qatar, Qatar
Prof. Dr. Herman Vermaak, Central University of Technology, Free State, South Africa
Prof. Dr. Ibrahim Dincer, University of Ontario, Canada
Prof. Dr. Ibrahim Sefa, Gazi University, Turkey
Prof. Dr. Ilya Galkin, Riga Technical University, Latvia
Assoc. Prof. Dr. Ilona Sárvári Horváth, University of Borås, Sweden
Prof. Dr. Jongho Yoon, Hanbat National University, S. Korea
Prof. Dr. Jongsoon Song, Chosun University, S. Korea
Prof. Dr. Jorge R. Frade, University of Aveiro, Portugal
Prof. Dr. Jose A. Aguado, University of Malaga, Spain
Prof. Dr. Jose A. Ramos Hernanz, Universidad del Pais Vasco, Spain
Prof. Dr. Jose M. Lopez Guede, Universidad del Pais Vasco, Spain
Prof. Dr. Josep Guerrero, Aalborg University, Denmark
Assoc. Prof. Dr. K.Premkumar, Rajalakshmi Engineering College, Chennai, India
Prof. Dr. Kozo Taguchi, Ritsumeikan University, Japan
Prof. Dr. Leijun Xu, Jiangsu University, China
Prof. Dr. Jun Yang, Huazhong University of Science and Technology, China
Dr. Loreto V. Gutierrez, CIEMAT Solar Platform of Almeria, Spain
Prof. Dr. Mahmood Ghoranneviss, Islamic Azad University, Iran
Prof. Dr. Maria Venegas, Universidad Carlos III de Madrid, Spain
Assoc. Prof. Dr. Mario E. Magana, Oregon state University, USA
Prof. Dr. Maris Klavins, University of Latvia, Latvia
Prof. Dr. Mehmet Önder Efe, Hacettepe University, Turkey
Prof. Dr. Mehmet Tekerek, Kahramanmaras Sutcu Imam University, Turkey
Assoc. Prof. Dr. Merih Palandöken, İzmir Katip Çelebi University, Turkey
Prof. Dr. Metin Gürü, Gazi University, Turkey
Prof. Dr. Milan Stork, University of West Bohemia, Czech Republic
Prof. Dr. Mohammad N. A. Hawlader, International Islamic University, Malaysia
Prof. Dr. Munir Nayfeh, University of Illinois at Urbana-Champaign, USA
Prof. Dr. Muris Torlak, Sarajevo University, Bosnia and Herzegovina
Prof. Dr. Mustafa Ilbas, Gazi University, Turkey
Prof. Dr. N. Nasimuddin, Institute for Infocomm Research, Singapore
Dr. Nam Choon Baek, Korea Institute of Energy Research, S. Korea
Prof. Dr. Namazov Subhan Nadiroglu, Azerbaijan Technical University, Azerbaijan
Prof. Dr. Narasimha G. Reddy, Lamar University, USA
Prof. Dr. Nicolae Paraschiv, Petroleum - Gas University of Ploiesti, Romania
Prof. Dr. Nicu Bizon, Pitesti University, Romania
Prof. Dr. Nikolay Djagarov, Nikola Vaptsarov Naval Academy, Bulgaria
Dr. Nilufar R. Avezova, Uzbekistan Academy of Sciences, Uzbekistan
Prof. Dr. Peter Lund, Aaalto University, Finland
Prof. Dr. Poul Alberg Østergaard, Aalborg University, Denmark
Prof. Dr. Rachid Chenni, University Mentouri of Constantine, Algeria
Prof. Dr. Rafael K. Jardan, Budapest Univ. of Technology and Economics, Hungary
Prof. Dr. Rafaela Hillerbrand, RWTH Aachen University, Germany
Prof. Dr. Ramazan Bayindir, Gazi University, Turkey
Prof. Dr. Raoul Rashid Nigmatullin, Kazan National Research Technical University,Russia
Dr. Rosaria Villari, Italian National Agency for New Technologies, Italy
Prof. Dr. Saffa Riffat, Nottingham University, UK

5
Assist. Prof. Dr. Sertac Bayhan, Texas A&M University at Qatar, Qatar
Prof. Dr. Shadia J. Ikhmayies, Al Isra University, Jordan
Prof. Dr. Sing Lee, Institute for Plasma Focus Studies, Australia
Prof. Dr. Sor Saw Heoh, Nilai University, Malaysia
Prof. Dr. Souad A.M. Albathi, Int. Islamic Uni. Malaysia, Malaysia
Assoc. Prof. Dr. T.Thamizhselvan, Rajalakshmi Engineering College, Chennai, India
Prof. Dr. Tae Hee Lee, Hanyang University, S. Korea
Prof. Dr. V. Jagannathan, Bhabha Atomic Research Center, India
Prof. Dr. Wail N. Al-Rifaie, Philadelphia University, Jordan
Prof. Dr. Wang Ru-Zhu, Shanghai Jiao Tong University, China
Assoc. Prof. Dr. Yong Song, Institute of Nuclear Energy Safety Technology, China
Prof. Dr. Zhiqiang Zhu, Institute of Nuclear Energy Safety Technology, China

6
 FOREWORD
Dear Colleagues,
We are glad to see you and your contribution for the Eight European Conference on Renewable Energy Systems
(ECRES 2020). The event has been organized in Istanbul/Turkey on 24-25 August 2020 by the organizers Gazi
University and Projenia. Besides, many institutions world-widely take a part as the cooperating institutions.
Turkish Science-Research Foundation (TÜBAV) and Journal of Energy Systems (JES) support the event.
Historically, the first, second, third, fourth, fifth, sixth and seventh conferences were completed very successfully
in Alanya/Antalya (2012), Antalya (2013), Kemer/Antalya (2015), Istanbul (2016), Sarajevo / Bosnia and
Herzegovina (2017), Istanbul (2018) and Madrid/Spain (2019), respectively. In averaged, 170 papers/abstracts
were presented in each event from all over the world, and published the highly amount of selected ones in
Thompson SCI and SCOPUS indexed reputable journals. This year, 304 papers are received. Among them, 128
papers from 35 countries are accepted and 164 papers are included into the final event programme after
withdrawals. We gratitute all authors for their respects to the organizing committee on the mutual communication
in order to make a good event from preperation stage to the post-event one. Besides, we acknowledge the
reviewers for their efforts to select the high scientific level studies.
The purpose of the ECRES is to bring together researchers, engineers and natural scientists from all over the
world, interested in the advances of all branches of renewable energy systems such as wind, solar, hydrogen,
hydro-, geothermal, solar concentrating, fuel-cell. It aims to present and disseminate the cutting-edge results to
the international community of energy in the form of research, development, applications, design and
technology. It is thereby expected that it can assist researchers, scientists, manufacturers, companies,
communities, agencies, associations and societies to keep abreast of new developments in their specialist fields
and to find innovative solutions in their problems.
All accepted papers will be published in a special Conference Proceeding
after conference and will be delivered online from www.ecres.net to only participants. In addition, high amount
of good papers, which are going to be presented in the Conference will be published in Science Citation Index
(SCI-indexed), SCOPUS-indexed and EBSCO-indexed journals, after the peer-review.
We would like to send our greetings to all of you and looking forward to having your future contribution to the
future events for a much green and peaceful word.

Prof. Dr. Erol KURT

Chairman of ECRES Series
Gazi University, Technology Faculty
Department of Electrical and Electronics Engineering
06500 Besevler ANKARA TURKEY
E-mail: [email protected]

7
8
 CONTENTS
KEYNOTES
Title Presenter Page
ELECTRICAL ENERGY STORAGE – THE FUTURE ROLES AND CHALLENGES Ahmed Faheem Zobaa 17
WIDE BAND-GAP SEMICONDUCTOR POWER ELECTRONICS Shadi Shahedipour-Sandvik 18
GEOTHERMAL INVESTIGATION FOR THE SOURCE OF RENEWABLE ENERGY Nur Islami 19
THE NEED FOR INCREASED FLEXIBILITY IN MODERN POWER SYSTEM
Ozan Erdinç 20
OPERATION
RECENT ADVANCES IN WAVE ENERGY CONVERSION Ossama Abdelkhalik 21
HETEROGENEOUS CATALYST ASSISTED TRIGLYCERIDE AND GLYCEROL
Amjad Ali 22
CONVERSION INTO VALUE ADDED PRODUCTS
SINGLE COMPONENT SOLUTIONS OF CONDUCTING POLYMERS AS INTERFACIAL
Ferda Hacıvelioğlu 23
LAYERS FOR ORGANIC ELECTRONICS

REGULAR ABSTRACTS
Paper ID Title Authors Page
Yagmur Tugba Unal, Fevzi Buyuksolak,
INVESTIGATION OF A FRANCIS TURBINE PERFORMANCE USING CFD
82 Burak Altintas, Kutay Celebioglu, Ece Ayli, 26
AND SITE EFFICIENCY MEASUREMENTS
Oguzhan Ulucak, Selin Aradag
Renzo Miguel Caballero Rosas, Raul Fidel
90 STOCHASTIC OPTIMAL CONTROL OF RENEWABLE ENERGY 27
Tempone
APPLICATION OF ELECTROCHEMICALLY DEPOSITED MOS2 FILMS
Ramunas Levinas, Natalia Tsyntsaru,
214 FOR HETEROGENOUS HYDROGEN EVOLUTION REACTION 28
Henrikas Cesiulis
CATALYSIS
THE EFFECTS OF BOUNDARY LAYER TRANSITION ON THE OVERALL
220 Sercan Acarer 29
EFFICIENCIES OF WIND TURBINES
NUMERICAL INVESTIGATION OF HIGH TURNING BLADES WITH
222 Hasan Çelik, Sercan Acarer 30
BOUNDARY LAYER BLOWING TOWARDS WIND TURBINES
Aliona Nicolenco, Xiang-Zhong Chen,
NANOPOROUS FE-GA ALLOYS WITH MAGNETOELECTRIC
223 Natalia Tsyntsaru, Henrikas Cesiulis, 31
CAPABILITIES
Salvador Pané, Eva Pellicer, Jordi Sort
IRON GROUP-BASED ELECTRODEPOSITS FOR EFFECTIVE Natalia Tsyntsaru, Edita Vernickaite,
231 32
ELECTROCATALYSIS Henrikas Cesiulis
Aliaksei Bakavets, Yauhen Aniskevich,
ELECTROCHEMICAL FABRICATION OF (BI2)M(BI2TE3)N THIN FILMS
233 Genady Ragoisha, Natalia Tsyntsaru, 33
WITH CONTROLLABLE BISMUTH CONTENT
Henrikas Cesiulis, Eugene Streltsov
ULTRASONIC (CONTACT METHOD) WELD PENETRATION DEPTH Bulent Kılıçaslan Serkan, Koten Hasan,
235 34
MEASURING STUDIES Yucel Necla
EXERGY ANALYSES AND INTEGRATED GAS COMBINED CYCLE FOR A
243 Rauf Terzi, Erol Kurt 35
HIGH EFFICIENCY OF A NUCLEAR POWER PLANT
ELECTRODEPOSITION OF CO AND CO-PT METAL FOAMS FOR WATER Modestas Vainoris, Natalia Tsyntsaru,
248 36
ELECTROCATALYSIS Henrikas Cesiulis
ELECTROCHEMICAL IMPEDANCE SPECTROSCOPY AS AN EFFECTIVE
252 Henrikas Cesiulis, Natalia Tsyntsaru 37
TOOL TO CHARACTERIZE MATERIALS FOR ENERGY APPLICATIONS
HYBRIDIZATION OF A GEOTHERMAL POWER PLANT WITH BIOMASS Balkan Mutlu, Derek Baker, Feyza Kazanç
267 38
DRIVEN SCO2 CYCLE Özerinç
Bilgehan Barış Öner, İrem Öner Alp, Esra
OPTIMIZATION OF ZNO-BASED NEAR UV SOLAR CELL AND LIGHT
297 Eroğlu, Saime Şebnem Aydin, Yasemin 39
EMITTING DEVICES
Öztekin Çiftci

REGULAR FULL PAPERS

Paper ID Title Authors Page
A CLASSIFICATION TOOL FOR THE MOTION OF A PIEZOELECTRIC Erol Kurt, Mehmet Tekerek, Kayhan Çelik,
12 42
ENERGY HARVESTER VIA NEURAL NETWORK Mehmet Gök
NUMERICAL INVESTIGATION OF HEAT TRANSFER AND PRESSURE Aslan Erman, Yildiz Omer, Kartun Hagop,
20 49
DROP ON FIN-TUBE HEAT EXCHANGERS Gungor Gokhan
TRANSESTERIFICATION OF WASTE COOKING OIL USING BONE AND
42 Ali A. Jazie, Amar J. Albaaji 55
EGG SHELL MIXTURE AS A CATALYST
INFLUENCE OF ELECTROLESS COATING OF NANO REINFORCEMENTS
47 Amar J. Albaaji Ali A. Jazie 68
ON MAGNETIC PROPERTIES OF FECO ALLOY COMPOSITES
EFFECT OF OPTIMIZED INTAKE MANIFOLD GEOMETRY ON Rafaa Saaidia, Raid Oyouni, Haider
50 74
BEHAVIORS AND EMISSION LEVEL OF H2CNG FUELED ENGINE Almuslam
AN FDTD-BASED SIMULATION STUDY ON DISTINGUISHING
Guomin Ji, Sudantha Perera, Yihru Huang,
66 CHARACTERISTICS OF INVERTER SWITCHING AND PULSE FAULT 87
Yan Zhang, Di Wu
DURING SUB-CYCLE TRANSIENT PERIOD
Varun Perumalla, Gopakumar Gopinathan,
LIMITATION OF FUNDAMENTAL ANALYTICS FOR FAULT
Milad Javadi, Almir Ekic, Di Wu, Rama
67 TRANSIENTS ON PLANNING OF TRANSMISSION SYSTEMS WITH HIGH 95
Ramakumar, Dhruv Sharma, Guomin Ji,
PENETRATION OF SOLAR PV RESOURCES
John Jiang

9
 USING FLOATING-INTERLEAVING BOOST CONVERTERS FOR SMALL
69 Moncef Justin Lalou, Capone Michele 103
WIND ENERGY CONVERSION
COMPARISON AND ANALYSIS OF PERFORMANCE VALUES OF FIXED
72 ANGLED SYSTEMS AND RADIATION TRACKING SYSTEMS IN SOLAR Kaplan Yusuf Alper, Saldamli Mehmet Ali 110
POWER PLANTS
ON RENEWABLES AND THE RESILIENCE OF SOCIO-TECHNOLOGICAL
76 Amadi-Echendu, Joe, Thopil, George Alex 116
ENERGY SYSTEMS
CFD INVESTIGATION ON PISTON HEAD GEOMETRY EFFECT ON IN-
80 Raid Oyouni, Rafaa Saaidia 123
CYLINDER FLOW FOR HYDROGEN FUELED ENGINE
81 RENEWABLE ENERGY UTILIZATION AND POLICIES OF TURKEY Ünaldı, Gülizar Gizem, Kaplan, Yusuf Alper 135
CFD ANALYSIS OF THE EFFECT OF H2CNG BLEND NATURE ON IN-
83
CYLINDER FLOW Haider Almusallm, Rafaa Saaidia 141
EFFECT OF MOLAR FRACTION OF THE INTRINSIC LAYER ON THE PIN
Mousli Latifa, Benmoussa Dennai, Azeddine
86 STRUCTURE FOR SOLAR CELL BASED AN INDIUM GALLIUM NITRIDE 153
Bouziane
USING AMPS-1D
UPGRADING OF PERFORMANCE OF AIR TO AIR HEAT PIPE HEAT Kerim Martin, İpek Aytaç, Çağdaş Filiz,
93 160
EXCHANGER BY USING CUO+ZNO HYBRID NANO FLUID Adnan Sözen, Ümit İskender
EXPERIMENTAL INVESTIGATION OF THE USE OF MGO+ZNO MIXTURE
Kerim Martin, İpek Aytaç, Çağdaş Filiz,
94 IN THERMOSIPHON TYPE HEAT PIPES WITHIN THE SCOPE OF AIR-TO- 165
Adnan Sözen, Cuma Kılınç
AIR HEAT EXCHANGER DESIGN
Mykola Radchenko, Eugeniy Trushliakov,
METHOD OF DETERMINING A RATIONAL DESIGN REFRIGERATION
95 Andrii Radchenko, Dmytro Konovalov, 171
CAPACITY OF OUTDOOR AIR CONDITIONING SYSTEMS
Veniamin Tkachenko, Oleksii Zielikov
THE IMPACT OF EXTREME WEATHER ON ELECTRICITY DEMAND Chesser Michael, Lyons Pádraig, O’Reilly
98 178
FROM HOMES HEATED BY AIR SOURCE HEAT PUMPS Padraic, Carroll Paula
Dmytro Konovalov, Mykola Radchenko,
RESEARCH OF CHARACTERISTICS OF THE FLOW PART OF AN
107 Halina Kobalava, Victoria Kornienko, 184
AEROTHERMOPRESSOR FOR GAS TURBINE INTERCOOLING AIR
Roman Radchenko, Vitaliy Maksymov
DYNAMIC ANALYSIS OF TRI-FLOATER WITH VERTICAL AXIS WIND Asadulla Thoppil, M. Abdul Akbar, V.
114 190
TURBINE SUPPORTED AT ITS CENTROID Mustafa, DadiRamBabu
AN APPLICATION OF A WIRELESS POWER TRANSFER AT LOW Erol Kurt, Fatemeh Zafarmand, Hikmet
116 196
FREQUENCY RANGE Fidanboy, Şevki Demirbaş
IMPLEMENTATION OF 30KHZ SINUSOIDAL CURRENT GENERATOR BY
119 Liang-Rui Chen, Zi-Yu Lin, Ying-Lin Chen 203
USING GAN-FET
123 WIND POWER FARMS MODELED BASED ON THE WAKE EFFECT Victorita Radulescu 209
124 FREQUENCY AND VOLTAGE CONTROL OF SEIG-WT SYSTEM Nasir Bilal Abdullah 215
ALGEBRAIC MODELING OF A NEW ELECTROMAGNETIC Erol Kurt, Aigerim Issimova, Bekbolat
127 221
MECHANICAL VIBRATION ENERGY HARVESTING Medetov
EXPERIMENTAL VALIDATION OF AN ARTIFICIAL NEURAL
Amadou BA, Sengahne Mbodji,
128 NETWORKS (ANN) MPPT CONTROLLER OF AN INSTALLED SOLAR 227
Alphousseyni Ndiaye
POWER PLANT AT POLYTECHNIC HIGHER SCHOOL
Andrii Radchenko, Mykola Radchenko,
133 A NOVEL DEGREE-HOUR METHOD FOR RATIONAL DESIGN LOADING Dariusz Mikielevicz, Roman Radchenko, 238
Andrii Andreev, Igor Esin
THE RECOVERY OF HISTORIC CENTRES: A MULTIPLE STRATEGY FOR Losco Giuseppe, Pierleoni Andrea,
134 245
THE ENERGY SAVINGS Roncaccia Elisa
FEATURES OF MOBILE PLATFORMS' EQUIPMENT SECOND ORDER Hanna Ukhina, Ivan Afanasyev,
137 DIGITAL FREQUENCY-DEPENDENT COMPONENTS PHASE- ValeriiSytnikov, Oleg Streltsov, Pavel 256
FREQUENCY RESPONSE AT TUNING Stupen
HOW CAN ROMANIA IMPROVE THE POTENTIAL OF SUSTAINABLE
138 Laurențiu Ciornei, Paula Munteanu 264
ENERGY DEVELOPMENT USING FOREST BIOMASS
ARTIFICIAL NEURAL NETWORK DESIGN FOR AN AXIAL FLUX
140 Adem Tekerek, Erol Kurt 273
PERMANENT MAGNET GENERATOR
NUMERICAL APPROXIMATION OF THE HYDROLOGICAL TIME OF
144 Barrón-Fernández, J.R., Calvo-Jurado C. 285
CONCENTRATION
INFLUENCE OF TRAFFIC AND ROAD SURFACE PROPERTIES ON
145 Roldán-Oliden P., Calvo-Jurado C. 291
ELASTIC BEHAVIOR OF LAYERED PAVEMENTS
SUPER TWISTING ALGORITHM- BASED DISCRETE-TIME CONTROL OF Farhi Salah Eddine, Sakri Djamel, Goléa
147 297
INDUCTION MOTOR Noureddine
Mykola Radchenko, Andrii Radchenko,
RATIONAL LOADS OF TURBINE INLET AIR ABSORPTION-EJECTOR
152 Serhiy Kantor, Bohdan Portnoi, Anatolian 308
COOLING SYSTEMS
Zubarev, Ivan Kalinichenko
INTERRUPTION-LESS CHARGING OF PLUG-IN ELECTRIC VEHICLES Abdul Rauf Bhatti, Arslan Dawood Butt,
153 315
USING THE GRID-CONNECTED PHOTOVOLTAIC SYSTEM Yawar Ali Sheikh, Kashif Nisar Paracha
HOW DO STUDENTS’ COMPETITIONS LIKE SOLAR DECATHLON
154 INCREASE RENEWABLE ENERGY AWARENESS AND WHO DOES Liudmila Cazacova, Eka Sediadi 322
BENEFIT?
MAPPING THE SOLAR ROOF POTENTIAL OF THE NORTH CAMPUS Joseph Nofech, Nima Narjabadifam, Hadia
158 330
BUILDINGS AT THE UNIVERSITY OF ALBERTA Awad, Mustafa Gül
Аndrey Lebedev, Anna Naumova, Boris
INXGA1-XP AND INXAL1-XP HETEROSTRUCTURES FOR CASCADE Zhalnin, Nina Vagapovа, Еvgeny lyschenko,
159 338
SOLAR CELLS INVESTIGATION Sergey Sharov, Ekaterina Obrucheva, Sergey
Didenko, Anatoliy Izotov

10
 BOOSTING THE ELECTRIC OUTPUT OF A CANTILEVER
Claudia Irina Borzea, Constantin Daniel
162 PIEZOELECTRIC HARVESTER BY TIP CURVATURE BLOCKING 344
Comeagă, Adrian Săvescu
ELEMENTS
PERFORMANCE OF THE FIRE-FIGHTING SYSTEM: ALGERIAN CASE
166 Marref Souad, Chettouh Samia 351
STUDY
THE STUDY AND OPTIMIZATION OF PRODUCTION/FERMENTATION
172 PROCESSES IN BIOFUEL PRODUCTION FOR MAXIMIZING ENERGY Amardeep Singh 360
OUTPUTS
A MULTI-AGENT APPROACH FOR DECENTRALIZED VOLTAGE
175 REGULATION IN ELECTRICAL DISTRIBUTION SYSTEM WITH Ömer Gül, Emre Köksal 366
RENEWABLE GENERATION
NUMERICAL MODELING OF A NEW WIND FARM BASED ON LOCAL
177 Victorita Radulescu 375
MEASURED ENVIRONMENTAL PARAMETERS
ENERGY AND EXERGY ANALYSIS OF SOLAR BASED HYDROGEN
181 PRODUCTION BY HELIOSTAT FIELD AND PARABOLIC TROUGH Alper Karakoç, Yalım Gültekin, Birol Kılkış 382
COLLECTOR
PRELIMINARY EXPERIMENTAL CALIBRATION AND VALIDATION OF Antonio Rosato, Francesco Guarino,
185 A DYNAMIC SIMULATION MODEL FOR FAULT DETECTION AND Vincenzo Filomena, Sergio Sibilio, Luigi 386
DIAGNOSIS OF AIR-HANDLING UNITS Maffei
PRELIMINARY ASSESSMENT OF IMPACT OF TYPICAL FAULTS ON Antonio Rosato, Francesco Guarino,
186 ENERGY PERFORMANCE OF AIR-HANDLING UNITS BY MEANS OF A Vincenzo Filomena, Sergio Sibilio, Luigi 393
DYNAMIC SIMULATION MODEL: A CASE STUDY Maffei
188 A SIMPLE METHOD FOR PARAMETER EXTRACTION OF A PV MODULE Nader Anani 400
ADAPTIVE OVERCURRENT PROTECTION FOR ELIMINATE TO
192 Ömer Gül, Anıl Çağlar Doğancı 409
MISOPERATION IN ELECTRICAL DISTRIBUTION NETWORKS
195 THD ANALYSIS FOR A PMSG WITH DIFFERENT MAGNET STRUCTURE Adem Dalcalı, Erol Kurt 418
1D MODELING OF OLIVE OIL MILL WASTEWATER EVAPORATION
199 Özel Binboğa, M. Zeki Yılmazoğlu 425
WITH A PARABOLIC TROUGH SOLAR COLLECTOR
Ahmet Yıldız, Onder Özgener, Leyla
202 HIGH VOLTAGE SOLAR INVERTER STRUCTURES 437
Özgener
NUMERICAL SIMULATION OF AN ON-GRID NATURAL GAS PEMFC -
Renato de Oliveira Gabriel, Sergio Leal
SOLAR PHOTOVOLTAIC MICRO CHP UNIT: AN ENERGY AND
203 Braga, Florian Pradelle, Eduardo Torres 444
FINANCIAL ANALYSIS FOR RESIDENTIAL AND INDUSTRIAL
Serra, Cesar Luiz Coutinho Sobral Vieira
APPLICATIONS
Aybaba Hançerliogulları, Erol Kurt, Abbas
NAVIER–STOKES-MAXWELL EQUATIONS FOR MHD EFFECTS WITH
212 Ghasemizad, Rezvan Rezaeizadeh, Yosef G. 454
TURBULENT FLOW IN ADVANCED ARIES-RS FUSION TECHNOLOGY
Ali Madee, Ettahir El Hadi Ali Omar Swese
Ashraf Mohammed Abusida, Aybaba
THE LOAD SHEDDING PREDICTION IN GECOL USING ARTIFICIAL Hançerlioğullari, Javad Rahebi, Erol Kurt,
213 469
NEURAL NETWORK Seçil Karatay, Yasemin Gültepe, Rezvan
Rezaeizadeh
ECO-TOXIC METAL POLLUTION IN SURFACE WATER AND ITS
Şeref Turhan, Aybaba Hançerliogulları, Aslı
216 SUITABILITY FOR IRRIGATION AROUND A LIGNITE-FIRED THERMAL 479
Kurnaz, Cellattin Duran, Onur Metin
POWER PLANT IN TURKEY
METAMODELING THE INFLUENCE OF FORM AND SHADING ON THE
Aristotelis Vartholomaios, Angeliki
219 HEATING AND COOLING LOADS OF A RESIDENTIAL ZONE IN THE 487
Chatzidimitriou, Konstantinos Ioannidis
MEDITERRANEAN
A FEASIBILITY ASSESSMENT OF OFFSHORE WIND POWER POTENTIAL
Fatih Karipoğlu, Mustafa Serdar Genç,
225 BASED ON ENVIRONMENTAL, SOCIAL AND ECONOMIC IMPACTS: A 493
Kemal Koca
CASE STUDY ON THE KIRKLARELI COASTLINE IN TURKEY
DETAILED FEASIBILITY STUDY ON ENERGY AND ECONOMIC
Fatih KAripoğlu, Mustafa Serdar Genç,
228 APPLICABILITY OF OFFSHORE HYBRID ENERGY SYSTEMS: A CASE 501
Kemal koca, Mert Gezici
STUDY SAROS BAY
THERMAL-FSI OF MINICHANNEL HEAT SINK FOR CALCULATING THE Shoukat Ahmad Adnan, Anwar Muhammad,
229 509
OPTIMUM FLOW RATE: A NUMERICAL APPROACH Koten Hasan
THERMODYNAMIC ANALYSIS OF A REFRIGERATION CYCLE USING
230 Touaibi Rabah, Koten Hasan 515
NATURAL REFRIGERANTS WITH LOW GLOBAL WARMING POTENTIAL
EXPERIMENTAL STUDIES OF NOVEL HYDROGEN PRODUCTION S.Senthilraja, R.Gangadevi, Hasan Köten,
232 519
SYSTEM INTEGRATED WITH NANOFLUID BASED PVT SYSTEM P.Ravichandran
A WIND EMULATOR ASSOCIATED WITH DOUBLY-FED INDUCTION Hallouz Mohamed, Kabeche Nadir,
239 525
GENERATOR Moulahoum Samir
STUDY OF THE EFFECT OF PREPARATION PARAMETERS ON THERMAL
240 CONDUCTIVITY OF METAL OXIDE NANOFLUIDS USING TAGUCHI Nikhil S. Mane, Vadiraj Hemadri 531
METHOD
CHB CONVERTER WITH BBESS SUPPORT FOR LARGE SCALE Álvaro Pérez Mayo, Ainhoa Galarza,
241 537
PHOTOVOLTAIC POWER PLANTS Fernando Martin, Javier Vadillo
PHOTODEPOSITION OF PD ON MOS2/TIO2 NANOCOMPOSITES FOR
242 ENHANCED PHOTOCATALYTIC HYDROGEN PRODUCTION UNDER Güy Nuray 545
VISIBLE LIGHT
DESIGN OF A VISUAL TESTING SYSTEM IN A ROBOTISED WELDING
Adolfo Lamas Rodríguez, Javier Pernas
244 WORKSHOP OF JACKET NODES BY MEANS OF DES AND ERGONOMIC 549
Álvarez, Inés Taracido López
SOFTWARE

11
 ACCELEROMETER MASS LOADING STUDY BASED ON A DAMAGE
246 IDENTIFICATION METHOD USING FUNDAMENTAL LAWS IN CLOSED Tarık Tufan, Hasan Köten 556
SYSTEMS
FORECASTING THE POWER PRODUCTION FROM A MICRO GRID PV
247 Penka V. Georgieva 563
SYSTEM
PMSM PARAMETER ESTIMATION USING A KIEFER-WOLFOWITZ
250 Artun Sel, Cosku Kasnakoglu 570
BASED SEARCH ALGORITHM
MULTICRITERIA OUTRANKING METHODS AND THE HYDROPOWER
251 Raupp Igor, Clímaco João, Costa Fernanda 578
EXPANSION PLANNING IN BRAZIL
SATELLITE CARRIER STRUCTURE ANALYSIS AND AERODYNAMIC
253 Koten Hasan, Tufan Tarık, Demir Enes 585
CHARACTERISTICS FOR LOW ALTITUDE DELIVERY SYSTEM
MAXIMUM POWER POINT TRACKING UNDER PARTIAL SHADING
254 Denis Karabetsky, Victor Sineglazov 590
CONDITIONS USING NOVEL MINE BLAST ALOGRITHM
255 HYBRID ENERGY STORAGE SYSTEMS Victor Sineglazov, Denis Karabetsky 596
ASSESSING S-CO2 MIXTURES FOR POTENTIALLY IMPROVING Robert Valencia-Chapi, Luis Coco-Enríquez,
256 601
SUPERCRITICAL POWER PLANTS PERFORMANCE Javier Muñoz-Antón
NUMERICAL SIMULATION ON MATHEMATICAL MODELING OF
257 Shimna Shafeek, Mohammad Asim Riaz 607
VERTICAL CONTACT MODE TRIBOELECTRIC GENERATORS
MICROCONTROLLER-BASED, LOW COST IMAGE-GUIDED SOLAR Rughooputh Hridayesh Deepak,
258 614
TRACKING Gooroochurn Mahendra
NUMERICAL INVESTIGATION OF NACA 2412 AIRFOIL USING CFD
259 Köten Hasan, Akkaya Kerim 621
TECHNIQUE
FIRE MODELLING AND SMOKE EXTRACTION IN THE ENCLOSED CAR
260 Koten Hasan, Akkaya Kerim 625
PARK ZONE USING CFD TECHNIQUE
DEVELOPMENT OF HEMP SHIVE WALLBOARD WITH INTEGRATED Edgars Kirilovs, Gusovius Hans-Jörg, Inga
261 631
PHASE CHANGE MATERIAL Zotova, Silvija Kukle
PERFORMANCE EVALUATION OF POWER CONVERTERS & ARTIFICIAL Vidhya K Viswambaran, Akram Bati,
262 637
INTELLIGENCE BASED MPPT TECHNIQUES IN GRID-TIED PV SYSTEMS Erping Zhou
SUITABLE SITE SELECTION TO WIND POWER PLANTS FOR
Fatih Karipoğlu, Mustafa Serdar Genç,
265 ENVIRONMENTAL AND SOCIAL IMPACTS BY USING GIS AND MCDM: 645
Kemal Koca
A CASE STUDY DEVELI KAYSERI
RESEARCH OF ENERGY AND EXERGY MODEL OF THERMOSIPHON Yedilkhan Amirgaliyev, Murat Kunelbayev,
266 CIRCULATION SOLAR COLLECTORS WITH THERMAL PUMPS FOR Aliya Kalizhanova, Omirlan Auelbekov, 653
HEATING AND HOT WATER SUPPLY Nazbek Katayev, Ainur Kozbakova
DRY COOLING OPTION FOR CONCENTRATING (CSP SOLAR POWER)
269 Satya Sekhar Bhogilla, Rama Sreekanth P.S 659
PLANT FOR THE DESERT REGION, INDIA
Nabeel Shaway Shyaa Al–Atwan, Mircea
271 POWER MANAGEMENT SYSTEMS FOR AN INTELLIGENT HOME 665
Nitulescu
METAL ORGANIC FRAMEWORK MATERIAL (MOF) FOR ATMOSPHERIC Ibrahim Al-Hashemi, Raya Al-Dadah, Saad
272 672
WATER HARVESTING IN ARID REGIONS Mahmoud
THE EFFECT OF CATHODE CAVITIES TO THE MODES OF A CAVITY
273 Yunus Yıldız, Erol Kurt 679
MAGNETRON
OVERVIEW OF PHOTO-BIOREACTORS USING COMPUTATIONAL FLUID Yucel Necla, Koten Hasan, Kılıcaslan
274 688
DYNAMICS TECHNIQUE Bulent Serkan
DESIGN OF ARTIFICIAL NEURAL NETWORKS FOR FUNCTION
Vidhya K Viswambaran, Akram Bati,
275 APPROXIMATION OF MAXIMUM POWER POINT TRACKING IN PV 695
Erping Zhou
SYSTEMS
MACHINE LEARNING MODELS FOR PREDICTING PV GENERATED
276 Penka V. Georgieva, Alexander Ivanov 703
POWER
NEW ANALYTICAL METHOD DEDICATED TO CALCULATING THE
277 ENERGY GAP VALUE OF NIO/FTO THIN-FILM MATERIAL OF Slimani Hamza, Bessous Noureddine 710
PHOTOVOLTAIC SYSTEMS
SIMULATIVE EVALUATION OF TRIBOELECTRIC POLYMER PAIRS IN
278 Shimna Shafeek, Sibgatulla Sharieef 717
VERTICAL CONTACT MODE
THEORETICAL STUDY ON PRESSURE EFFECTS ON PHYSICAL
279 Yasemin Öztekin Çiftci 725
PROPERTIES OF SEMICONDUCTOR COMPOUND LICDP
COMPOSITION EFFECTS ON ELASTIC AND ELECTRONIC PROPERTIES
280 Yasemin Öztekin Çiftci 730
OF RU DOPED OSHF (RUXOS1-XHF) ALLOYS: AB-INITIO STUDY
FINE STRUCTURED RED-BAND TAIL PHOTOLUMINESCENCE (PL)
281 Shadia J. Ikhmayies 735
SPECTRA OF NANOCRYSTALLINE CDS:IN THIN FILMS
ARITHMETIC STUDY ON ENERGY SAVING FOR SOME COMMON DATA Pedro Juan Roig, Salvador Alcaraz, Katja
282 744
CENTRE TOPOLOGIES Gilly
PROTON EXCHANGE MEMBRANE FUEL CELL DEGRADATION: AN
284 Sujit Sopan Barhate, Rohini Mudhalwadkar 752
EXPERIMENTAL ANALYSIS
Penka V. Georgieva, Milen Baltov, Maria
KINATION- STUDENTS’ EXPLORATION OF A FUTURE RENEWABLE
285 Daskalova, Nikolai Tsonev, Kalina Hristova, 759
ENERGY SOURCE
Georgi Karavasilev
3D FLOW SIMULATION OF A HORIZONTAL AXIS WIND TURBINE WITH Mustafa Serdar Genç, Khalida Sekhoune
287 764
A FLAT PLATE ON PRESSURE SIDE Özden, Tuna Murat Bodur
FLOW-INDUCED MECHANICAL ANALYSIS OF A HORIZONTAL AXIS Mustafa Serdar Genç, Tuna Murat Bodur,
289 770
WIND TURBINE WITH A FLAT PLATE ON PRESSURE SIDE Khalida Sekhoune Özden

12
 Fleury Gabriela Rocha de Oliveira,
Nascimento Douglas Vieira do, Galvão Filho
IMAGE-BASED WATER LEVEL ESTIMATION FOR REDUNDANCY
290 Arlindo Rodrigues, Carvalho Rafael Viana, 776
INFORMATION USING CONVOLUTIONAL NEURAL NETWORK
Ribeiro Filipe de Souza Lima, Coelho
Clarimar José
A PRELIMINARY IN-SITU MEASUREMENT AND SIMULATION OF THE
Gooroochurn Mahendra, Seegobin Budhiraj,
291 PERFORMANCE OF GROUND-COUPLED HEAT EXCHANGER SYSTEM 781
Jankee Lowkhaushalsing
IN THE TROPICAL CONTEXT OF MAURITIUS
THERMODYNAMIC MODELLING TO OPTIMIZE THE PERFORMANCE OF Ibrahim Albaik, Gavin Tozer, Raya Al-
292 789
SMALL SCALE ORGANIC RANKINE CYCLE (ORC) SYSTEMS Dadah, Saad Mahmoud
COMPARATIVE STUDY OF DISCHARGE EMISSION AMPLIFICATION IN
293 AR-DRIVEN GAS DISCHARGE SYSTEM WITH MICROPOROUS ZEOLITE B.G. Salamov, H. Hilal Kurt 796
CATHODE
294 THE PLASMA SYSTEM WITH INP PHOTODETECTOR H. Hilal Kurt, F. Karip 802
THE ELECTRON CURRENT DENSITY PATTERNS IN SEMICONDUCTOR-
295 H. Hilal Kurt, B.G. Salamov, E.Tanrıverdi 805
GAS DISCHARGE SYSTEM
THE CHARACTERISTIC FEATURES OF THE COLD PLASMA WITH ZNSE
296 H. Hilal Kurt, S.Kızışar 808
ELECTRODE
STEP BY STEP MODELLING, ANALYSIS AND SIMULATION OF A HIGH-
298 EFFICIENCY BATTERY CONTROL SYSTEM BY USING MATLAB Mohammed Ayad Alkhafaji, Yunus Uzun 811
ENVIRONMENT PROGRAM
Alejandro Ballesteros-Coll, Unai Fernandez-
PERFORMANCE STUDY OF THE CELL-SET MODEL FOR A GURNEY
299 Gamiz, Iñigo Aramendia, Ekaitz Zulueta, 820
FLAP IMPLEMENTATION
Jose Manuel Lopez-Guede
Iosu Ibarra-Udaeta, Koldo Portal-Porras,
LARGE EDDY SIMULATION OF A RECTANGULAR VORTEX Unai Fernandez-Gamiz, Iñigo Aramendia,
300 826
GENERATOR ON A FLAT PLATE WITH AN INCIDENT ANGLE OF 25º Ekaitz Zulueta, Javier Sancho, Jose Manuel
Lopez-Guede
ATOM SEARCH OPTIMIZATION ALGORITHM FOR OPTIMIZATION OF Wafeeqa Abdulrazak Hasan, Issa Ahmed
301 831
POWER GENERATION IN MICROGRID Abed, Diyah Kammel Shary
Ekaitz Zulueta Guerrero, Ander Sánchez
ARTIFICIAL NEURAL NETWORK-BASED STOP CRITERIA FOR A Chica, Daniel Teso-Fz-Betoño, Unai
302 841
GENETIC ALGORITHM Fernández-Gámiz, Jose Antonio Ramos-
Hernanz, Jose Manuel Lopez-Guede
HIGH TEMPERATURE LEAD ACID BATTERY SOC AND SOH Javier Olarte, Ekaitz Zulueta, Raquel Ferret,
303 CHARACTERIZATION BASED ON ELECTROCHEMICAL IMPEDANCE Erol Kurt, Jaione Martínez de Ilarduya, Jose 847
SPECTROSCOPY DATA Manuel Lopez-Guede
EFFECTS OF SUBSTRATES ON THE GROWTH OF BETA VULGARIS Tran Trong Nhan, Ngo Van Tuan, Nguyen
304 853
SUBSP. VULGARIS IN HYDROPONIC SYSTEMS Phuc Thien

13
14
8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

KEYNOTES

15
8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

16
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

ELECTRICAL ENERGY STORAGE – THE FUTURE ROLES AND

CHALLENGES

Ahmed Faheem Zobaa

Institute of Energy Futures at Brunel University London, UK

Zobaa, AF. Electrical energy storage – the future roles and challenges. 8th Eur. Conf. Ren. Energy
Cite this paper as:
Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The recent IEC white paper on Electrical Energy Storage presented that energy storage has played three main
roles. First, it reduces cost of electricity costs by storing electricity during off-peak times for use at peak
times. Secondly, it improves the reliability of the power supply by supporting the users during power
interruptions. Thirdly, it improves power quality, frequency and voltage. Energy storage is expected to solve
many problems including excessive power fluctuation and undependable power supply due to the use of
large penetration levels of renewable energy. Electric vehicles with batteries are the most promising
technology to replace fossil fuels by electricity from mostly renewable energy sources. This tutorial presents
the future roles and challenges of electrical energy storage.
Contents
The roles of energy storage technologies in electricity use
Types and features of energy storage systems
Markets for energy storage
Forecast of energy storage market potential by 2030.
Keywords:
© 2020 Published by ECRES

17
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

WIDE BAND-GAP SEMICONDUCTOR POWER ELECTRONICS

Shadi Shahedipour-Sandvik
Nanoscale Engineering, SUNY Polytechnic Institute, USA

Shahedipour-Sandvik, S. Wide band-gap semiconductor power electronics. 8th Eur. Conf. Ren. Energy
Cite this paper as:
Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: According to International Energy Agency (IEA), Global electricity demand in 2018 increased considerably
in 2018 to nearly 900 TWh with about 70% of that growth by the US and China. Renewables were responsible
for a major part of the increase in power demand. In the US, nearly 40% of the total energy consumption is
due to electricity. Given the fast growth in world population and the move to a more plugged-in world that
works toward cutting carbon emissions and shift to renewables, power electronics are becoming ever more
important as an enabling technology toward greater electrification. The current incumbent power
semiconductor is silicon switches. Relative to wide bandgap (WBG) Si devices face multiple challenges
including high losses due to relatively low bandgap of Si, low switching frequency and limited performance
at elevated temperatures. Wide band-gap semiconductors have already shown great potential in offering
higher efficiency power semiconductor devices by overcoming physical limitations of Si. In this talk I will
offer an overview of the state of the technology in WBG, discuss potential new opportunities in application
of ultra wide band-gap, and my research progress in developing WBG power devices..
Keywords:
© 2020 Published by ECRES

18
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

GEOTHERMAL INVESTIGATION FOR THE SOURCE OF

RENEWABLE ENERGY

Nur Islami
University of Riau, Indonesia

Islami, N. Geothermal investigation for the source of renewable energy. 8th Eur. Conf. Ren. Energy
Cite this paper as:
Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: One of the main problems of the earth is the global warming. This is caused by the level of air pollution, and
one of them due to the use of fossil fuels. The search for renewable energy sources is something that must
be done. Geothermal energy is a renewable energy source that is not making the negative impact to the
environment. This study is showing an investigation of the potential of geothermal energy sources in Rokan
Hulu, Indonesia, using integrated geophysical methods. The results show that there were sufficient heat and
water sources in the study area, which were shown from the geomagnetic, VLF and resistivity data. The head
source and water is connected by the fractured, and therefore the study area is potential to be developed as
the new source of the renewable energy..
Keywords: tidal, energy, CFD, turbine
© 2020 Published by ECRES

19
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

THE NEED FOR INCREASED FLEXIBILITY IN MODERN POWER

SYSTEM OPERATION

Ozan Erdinç
Yıldız Technical University, Turkey

Erdinç, O.. The need for increased flexibility in modern power system operation. 8th Eur. Conf. Ren.
Cite this paper as:
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The power system operation is currently facing a vital level of complexity together with the increase in the
integration of nondispatchable renewable production units, new loads with different characteristics like
electric vehicles, etc. Therefore, the flexibility of operation has become a prior issue to reconsider from the
system operators’ point-of-view. In this invited speech, the need for increased flexibility in modern power
system operation and the relevant possible flexibility resources will be discussed from different perspectives.
Keywords:
© 2020 Published by ECRES

20
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

RECENT ADVANCES IN WAVE ENERGY CONVERSION

Ossama Abdelkhalik
Iowa State University, USA

Abdelkhalik, O. Recent Advances in Wave Energy Conversion. 8th Eur. Conf. Ren. Energy Sys. 24-25
Cite this paper as:
August 2020, Istanbul, Turkey

Abstract: In 2012 the world wave installed capacity was less than 5 MW, mostly produced on a pre-commercial basis.
In 2016, the global ocean energy capacity jumped up to 536MW; this is still less than 0.05% of the global
capacity of hydro power, and is less than 0,03% of the total renewable energy, in the same year. The sea,
however, is a colossal reservoir of energy of particularly high density. The total theoretical ocean energy
potential is estimated to be 29.5 PWh/yr. Despite its potential, wave energy conversion technology is not yet
mature due to few challenges. This talk highlights some of these challenges, and presents some recent
advances in the area of modelling and control of wave energy converters.
One of the main challenges is the need for bidirectional power flow power take-off (PTO) units. Many of the
existing Wave Energy Converters (WECs) use buoys to harvest power from the ocean, kinetically, before it
is converted to electric power. In order to increase the harvested kinetic energy, a spring-like control force
should be applied on the buoy, in addition to the damping control force that is usually obtained from the PTO
unit. This spring-like control force implies that the PTO unit should have a bidirectional power flow
capability. This bidirectional power flow capability increases the complexity and cost of the PTO unit.
Keywords:
© 2020 Published by ECRES

21
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

HETEROGENEOUS CATALYST ASSISTED TRIGLYCERIDE AND

GLYCEROL CONVERSION INTO VALUE ADDED PRODUCTS

Amjad Ali
Thapar Institute of Engineering and Technology, India

Ali, A. Heterogeneous catalyst assisted triglyceride and glycerol conversion into value added products.
Cite this paper as:
8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In order to address the issues related to the bio-waste management, global warming and renewable fuel
production biomass valorization has become extremely important. The main components of biomass could
be categorized as carbohydrates, lignin, protein and lipids. These constituents could be utilized as a
renewable and eco-friendly feedstock to obtain fine-chemicals, pharma products, eco-friendly and renewable
alternatives to the fossil based fuel and fuel additives. Lipids could be used as a renewable feedstock for the
production of other value added products such as fatty acid alkyl esters, commonly known as biodiesel (BD),
fatty acid amides, epoxidized fatty acid esters, etc. Glycerol, which is created as a by-product during these
chemical transformations, could also be utilized as a substrate for the synthesis of another valuable chemicals
such as glycerol carbonate, glycerol monoacetate, glycerol diacetate and glycerol triacetate, etc. Application
of the appropriate catalysts during the above mentioned cemical reactions could not only improve the
economy of the processes but also make the reactions less energy intensive. The presentation would highlight
the application of the mixed metal oxides based heterogeneous catalysts for the chemical transformation of
triglyceride and glycerol into value added products.
Keywords:
© 2020 Published by ECRES

22
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

SINGLE COMPONENT SOLUTIONS OF CONDUCTING POLYMERS

AS INTERFACIAL LAYERS FOR ORGANIC ELECTRONICS

Ferda Hacıvelioğlu
Gebze Technical University, Department of Chemistry, Turkey, [email protected]

Hacıvelioğlu, F. Single Component Solutions of Conducting Polymers as Interfacial Layers for Organic
Cite this paper as:
Electronics. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Optoelectronic devices (OEDs) are electrical-to-optical or optical-to-electrical converter systems which must
contain at least one transparent electrode to the target wavelength of the light. These devices are usually built
by deposition of different layers, which have different electrical and/or optical properties. Electrically
conductive polymers (CPs) are one of the key materials form optoelectronic devices such as touch panels,
electrochromic devices, light emitting diodes, light sensors and solar cells associated with their unique
properties such as transparency, tuneable electrical conductivity, high mechanical strength and flexibility.
Poly(3,4-ethylenedioxythiophene)- Polystyrene sulfonic acid, i.e. PEDOT-PSS is the most famous one of
these type polymeric materials that is used for optoelectronic device applications. Different types or grades
of PEDOT-PSS have been used as different layer of OEDs such as transparent conducting electrode, active
or interfacial layers. The interactions between the PEDOT and PSS does not consist of covalent bonding,
meaning that the PEDOT-PSS water dispersions have at least two or more components. However, a robust
highly efficient and reliable OED requires each component or layer with very well-known properties. It is
therefore very important to develop single component solutions of conjugated materials for OED
applications.
In this lecture we are going to focus our attention to new developments on the single component solution and
processable conducting conjugated materials as interfacial layers for organicelectronics applications..
Keywords: Solution Processability, Conducting Polymer, PEDOT:PSS, Optoelectronic Device
© 2020 Published by ECRES

Acknowledgement:
We are greatly acknowledged to the Royal Society to support our work under the grant number of NAF-R2-180625.

23
8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

REGULAR
ABSTRACTS

24
8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

25
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

INVESTIGATION OF A FRANCIS TURBINE PERFORMANCE USING

CFD AND SITE EFFICIENCY MEASUREMENTS

Yagmur Tugba Unal

TOBB ETU, Ankara, Turkey, [email protected], ORCID: 0000-0001-7272-4591

Fevzi Buyuksolak
TOBB ETU, Ankara, Turkey, [email protected], ORCID: 0000-0002-3105-6328

Burak Altintas
TOBB ETU, Ankara, Turkey, [email protected], ORCID: 0000-0001-9297-2138

Kutay Celebioglu
TOBB ETU Hydro, Ankara, Turkey, [email protected], ORCID: 0000-0001-8845-4928

Ece Ayli
Cankaya University, Ankara, Turkey, [email protected], ORCID: 0000-0002-6209-161X

Oguzhan Ulucak
TED University, Ankara, Turkey, [email protected], ORCID: 0000-0002-2063-2553

Selin Aradag
TED University, Ankara,Turkey, [email protected], ORCID: 0000-0002-2034-0008

Unal, Y.T, Buyuksolak, F., Altintas, B., Celebioglu, K., Ayli, E., Ulucak, O., Aradag, S,. Investigation
Cite this paper as: of a Francis Turbine Performance Using CFD and Site Efficiency Measurements, 8th Eur. Conf. Ren.
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In the last decades, the refurbishment and upgrading of hydroturbines have been increasingly demanded
according to the performance loss, reliability reduction, and often maintenance need. Before, starting to the
rehabilitation process, a key role is presenting the present state of the power plant. Without understanding
the current problems, which cause a reduction in the efficiency, generated power and reliability, starting to
the rehabilitation process can cause unnecessary expenses and modification with no increment in the
performance. The objective of this research is to undertake Computational Fluid Dynamics analyses of
selected hydroelectric power plants for the purpose of evaluating upgrade potential as well as identification
of operational characteristics. For this purpose, firstly turbine parts were scanned in the field and hybrid
reverse engineering methodology were developed which combines the traditional reverse engineering steps
with flow dynamics of hydraulic turbines. In the second step, steady and unsteady analyses are performed to
determine the efficiency and power output of the turbine. A hill chart is generated to investigate the behavior
of the turbine at different head and flow rate values. Obtained CFD results are compared with the site
measurements. The CFD results showed that at the full-load condition the turbine has 92.61% efficiency with
174 MW power output which are consistent with the guaranteed values from manufacturer. In the third part
of the study, the problems that decrease the turbine efficiency are detected with performing transient analysis.
According to the CFD results, cavitation zones are observed on the runner blades. Transient analysis and
visualization techniques are used for diagnostics of turbine and enables identification of different cavitation
mechanisms. Results depicted that blades suffers from leading edge cavitation which is the most hazardous
cavitation type. It is recommended that; regarding the economic aspects modifying the conventional runner
blade to the X-blade will increase the performance and prevent the leading-edge cavitation.
Keywords: Rehabilitation, CFD, Reverse Engineering, Francis Turbine, Renewable Energy, Transient Analysis
© 2020 Published by ECRES

26
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

STOCHASTIC OPTIMAL CONTROL OF RENEWABLE ENERGY

Renzo Miguel Caballero Rosas

Kaust, Thwal, KSA, [email protected], ORCID: 0000-0003-3220-0923

Raul Fidel Tempone

Rwth Aachen University, Aachen, Germany, [email protected], ORCID: 0000-0003-1967-4446

Caballero, R., Tempone, R.. Stochastic Optimal Control of Renewable Energy. 8th Eur. Conf. Ren.
Cite this paper as:
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Renewable energy is becoming more relevant as time advances. Unlike other sources of energy, they can
be uncontrollable and hard to predict, such as solar and wind power. Uruguay is a pioneer in the use of
renewable energies. Nowadays, it can satisfy its total demand from renewable sources, but almost half
of the installed power is uncontrollable. Aside from fuel power stations, Uruguay has hydropower, solar,
and wind power.
Overproduction of energy gives Uruguay the possibility of exporting. However, to achieve the optimal
use of energy, it is necessary to overcome the uncertainty and high variability in other countries' demand
and uncontrollable generators. We are looking for continuous-time optimal controls in a problem with
uncertainty. Thus, we propose to solve the Hamilton-Jacobi-Bellman equation associated with the
electrical system. We explore the addition of a battery to achieve less final cost and a more reliable
supply.
The uncertainties in the uncontrollable energy production are modeled using stochastic differential
equations, and the deterministic dynamics using partial differential equations.
Keywords: Stochastic Optimal Control, Hamilton-Jacobi-Bellman, Renewable Energy, Optimal Control, Continuous-
Time Control
© 2020 Published by ECRES

27
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

APPLICATION OF ELECTROCHEMICALLY DEPOSITED MOS2

FILMS FOR HETEROGENOUS HYDROGEN EVOLUTION REACTION
CATALYSIS

Ramunas Levinas
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania,
[email protected], ORCID: 0000-0003-3324-6472

Natalia Tsyntsaru
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania/Institute of
Applied Physics, Chisinau, Moldova, [email protected], ORCID: 0000-0002-9813-2460

Henrikas Cesiulis
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania,
[email protected], ORCID: 0000-0002-5077-7884

Levinas, R., Tsyntsaru, N., Cesiulis, H., Application of electrochemically deposited MoS 2 films for
Cite this paper as: heterogenous hydrogen evolution reaction catalysis. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
2020, Istanbul, Turkey

Abstract: Hydrogen is often considered, as an alternative to fossil fuels as a liquid form energy carrier, but generation
of large amounts of clean hydrogen fuel remains an issue. Electrochemical water splitting is a process
whereby an electric current splits water into its two constituents – hydrogen and oxygen, yielding an
unpolluted product. However, the kinetics of the hydrogen evolution reaction (HER) are sluggish, and
extensive research is being devoted to developing catalysts that would increase the rate of water splitting.
In this study, we use electrochemical deposition to cover a large surface area of 3D copper foam electrode
with molybdenum sulfide (MoS2), which is known HER electrocatalyst in acidic media. We demonstrated
that tuning electrodeposition conditions, as well as using certain additives, has a significant impact on the
electrocatalytic properties of the electrodes. Electrochemical impedance spectroscopy (EIS) was used to
approximate the electrocatalytically active surface area, and important kinetic parameters of the catalytic
material. The best-performing electrodes have been successfully engineered to have both a large amount of
catalytically active sites, and improved per-site activity.
Authors acknowledge funding from H2020 project MSCA-RISE-2017-778357-SMARTELECTRODES.
Keywords: Hydrogen evolution reaction, electrocatalysis, electrodeposition, molybdenum sulfide, electrochemical
impedance spectroscopy
© 2020 Published by ECRES

28
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

THE EFFECTS OF BOUNDARY LAYER TRANSITION ON THE

OVERALL EFFICIENCIES OF WIND TURBINES

Sercan Acarer
İzmir Katip Çelebi University, İzmir, Turkey, [email protected], ORCID: 0000-0002-5891-7458

Acarer, S. The Effects of Boundary Layer Transition on the Overall Efficiencies of Wind Turbines. 8th
Cite this paper as:
Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: The boundary layer transition from laminar-to-turbulent flow generally reduces lift and increases drag of
airfoils. Many studies related to medium-size HAWTs and VAWTs do not provide information about such
effects. This reflects into uncertainties in the reported power coefficients or incomplete success criteria for
CFD models. Dust and insect contamination and any other roughness elements may induce fully turbulent
flow over airfoils. This study investigates the efficiency degradation of medium-sized wind turbines and
requirements for CFD simulations. Published experimental results for the modern 27% thick DU12W262
airfoil both for clean and dirty surface conditions are taken as a baseline. The rough (dirty) surface is
simulated by forced transition near the leading edge by zigzag tapes. CFD is employed to reproduce airfoil
characteristics both for clean and dirty surfaces. Then, both experimental and CFD-predicted airfoil
characteristics are provided as an input to Blade Element Momentum (BEM) Models to simulate their
corresponding performances in HAWTs and VAWTs, where 2D URANS CFD is also employed for the
latter. This study aims to reveal the generally neglected differences between wind turbine performances in
laboratory conditions and likely operational conclusions and presents the effect of employing CFD to predict
airfoil characteristics.
Keywords: Boundary Layer Transition, Wind Turbine, Rough Surface, Roughness Contamination.
© 2020 Published by ECRES

29
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

NUMERICAL INVESTIGATION OF HIGH TURNING BLADES WITH

BOUNDARY LAYER BLOWING TOWARDS WIND TURBINES
Hasan Çelik
İzmir University of Economics, İzmir, Turkey, [email protected], ORCID: 0000-0002-2512-8196

Sercan Acarer
İzmir Katip Çelebi University, İzmir, Turkey, [email protected], ORCID: 0000-0002-5891-7458

Çelik, H. and Acarer, S. Numerical Investigation of High Turning Blades with Boundary Layer
Cite this paper as:
Blowing. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: Modern wind turbines require high performance, i.e. glide ratio, airfoils. This necessitates laminar flow over
the airfoil without boundary layer separation. Nevertheless, laminar flow is prone to separation, which may
eliminate the gains due to avoiding turbulent flow. Flow control is a promising method to prevent boundary
layer separation, which may be essential to achieve higher glide ratios which is difficult to achieve by airfoils
without flow control. Passive flow control methods have been applied with success previously. Active flow
controls are generally not commonly used due to their complexity. Nevertheless they may promise even
higher performance gains. In this study, published experimental results for NACA0015 airfoil at Re=1.5x105
with trailing edge flaps located at 75% of the chord is employed as a baseline. A two dimensional slot exists
over the hinge of the flap, at the suction side. The flap angle considered is 20 degrees downwards. Transition
(-Re) k-omega Shear Stress Transport (SST), baseline k-omega SST and Spalart-Allmaras (SA) models
are compared against experimental data of with and without blowing. It is shown that all the three models
accurately capture majority of the flow physics in such a high turning airfoil. Therefore, the boundary layer
transition plays a minor role in such high turning blades and the flow is mainly governed by the turbulent
separation towards the aft of the airfoil.
Keywords: Airfoil, Active, Flow Control, Flap, NACA, CFD, Blowing.
© 2020 Published by ECRES

30
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

NANOPOROUS FE-GA ALLOYS WITH MAGNETOELECTRIC

CAPABILITIES
Aliona Nicolenco
Departament de Física, Facultat de Ciències, Universitat Autònoma de Barcelona, Bellaterra, Spain/ Institute of Applied Physics,
Chisinau, Moldova, [email protected], ORCID: 0000-0003-4624-2163

Xiang-Zhong Chen
Multi-Scale Robotics Lab, Institute of Robotics & Intelligent Systems, ETH Zurich, Switzerland, [email protected], ORCID:
0000-0002-2294-7487

Natalia Tsyntsaru
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania/Institute of
Applied Physics, Chisinau, Moldova, [email protected], ORCID: 0000-0002-9813-2460

Henrikas Cesiulis
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania,
[email protected], ORCID: 0000-0002-5077-7884

Salvador Pané
Multi-Scale Robotics Lab, Institute of Robotics & Intelligent Systems, ETH Zurich, Switzerland, [email protected] , ORCID: 0000-
0003-0147-8287

Eva Pellicer
Departament de Física, Facultat de Ciències, Universitat Autònoma de Barcelona, Bellaterra, Spain, [email protected], ORCID:
0000-0002-8901-0998

Jordi Sort
Departament de Física, Facultat de Ciències, Universitat Autònoma de Barcelona, Bellaterra/Institució Catalana de Recerca i Estudis
Avançats (ICREA), Barcelona, Spain, [email protected], ORCID: 0000-0003-1213-3639

Nicolenco, A., Chen, X.-Z., Tsyntsaru, N., Cesiulis, H., Pané, S., Pellicer, E., Sort, J., Nanoporous Fe-Ga alloys
Cite this paper as:
with magnetoelectric capabilities. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Magnetoelectric (ME) coupling in multiferroic heterostructures is the keystone in energy-efficient magneto-
electronic technologies, as it allows for (i) wireless modulation of electric polarization by external magnetic
fields, or (ii) electric-field control of magnetism. Nevertheless, in conventional multiferroic bilayers, ME
coupling is commonly limited by substrate clamping which hinders effective strain propagation. Nanoporous
materials can circumvent this problem producing strain-gradients instead of homogeneous strains. In this
work, we investigate the effects of nanoporosity on the magnetostrictive properties of electrodeposited Fe-
Ga alloys. First, continuous Fe-Ga layers were electrodeposited onto rigid substrates. The magnetostrictive
response of the deposits was investigated by X-ray diffraction under the application of external magnetic
fields in-situ. The experimental results suggest that while the bottom of the Fe-Ga layer is clamped, its air
side exhibits a pronounced tetragonal deformation due to the residual nanoporosity existing between the
columnar grains, i.e., a strain gradient develops across the thickness of the Fe-Ga film. Subsequently, a high
degree of nanoporosity was induced in Fe-Ga films through electrodeposition onto colloidal templates. The
pores of magnetostrictive Fe-Ga alloys can accommodate various ferroelectric materials, such as P(VDF-
TrFE), thus rendering composites with enhanced ME response integrated on rigid substrates. First results on
strain-gradient mediated magnetoelectric effects in Fe-Ga/P(VDF-TrFE) composites will be presented.
Keywords: Magnetoelectric coupling, electrodeposition, Fe-Ga, nanoporous materials
© 2020 Published by ECRES
Acknowledgment:
Authors acknowledge funding from H2020 project MSCA-RISE-2017-778357-SMARTELECTRODES, the Spanish Government (MAT2017-
86357-C3-1-R and associated FEDER), the Generalitat de Catalunya (2017-SGR-292), and European Research Council (SPIN-PORICS 2014-
Consolidator Grant, Agreement n° 648454).

31
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

IRON GROUP-BASED ELECTRODEPOSITS FOR EFFECTIVE

ELECTROCATALYSIS
Natalia Tsyntsaru
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania/Institute of
Applied Physics, Chisinau, Moldova, [email protected], ORCID: 0000-0002-9813-2460

Edita Vernickaite
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania,
[email protected]

Henrikas Cesiulis
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania,
[email protected], ORCID: 0000-0002-5077-7884

Cite this paper Tsyntsaru, N., Vernickaite E., Cesiulis, H., Iron group-based electrodeposits for effective electrocatalysis. 8th Eur. Conf. Ren.
as: Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Our previous works demonstrate that electrodeposited iron group metal (Ni, Co, Fe)-based deposits
containing W/Mo exhibit an enhanced catalytic performance towards hydrogen evolution reaction (HER) in
30 wt.% NaOH solution. The activity reduces in the order (in at.%): Co–52Mo>Ni–54Mo>Fe–54Mo>Ni-
29W>Co-33W>Fe-30W [1,2]. Besides, it was shown that ultra-nanocrystalline Co-W coating can be used
for anodic oxidation of methanol in H2SO4 solution [3]. Furthermore, electrodeposited nanowire (NW) arrays
exhibit an enhanced catalytic performance in comparison with their thin films due to the greatly enlarged
specific surface areas and number of exposed active sites. Thus, NWs of Ni can be used for urea electro-
oxidation and subsequently for hydrogen production and remediation of urea-rich wastewater, demonstrating
higher current efficiency than those of bulk Ni film [2]. The electrocatalytic activity of the NWs can be
enhanced additionally by the incorporation of the second component in the system. For example,
codeposition P with Ni raised HER performance in 0.5 M H2SO4 electrolyte of NWs [3]. Meanwhile,
electrodeposited Co-Ni nanowires have greater electrocatalytic performance for HER in 0.025 M H2SO4
solution than that of either the Co or the Ni nanoarrays under the same electrodeposition conditions [4].
Furthermore, electrodeposited Co-Ni alloy nanowires showed higher electrocatalytic activity in terms of
lower onset potential and higher peak current density compared to the sample synthesized by conventional
deposition route for urea electro-oxidation [5]. The rational design of electrode architecture is also an
effective way to improve ultimate catalytic activity. By combining conductive nanostructured scaffold and
highly active outer layers, the as-prepared Ni-Mo/Cu nanowires deliver superior hydrogen and oxygen
evolution performance in 1 M KOH electrolyte [6]. Meanwhile a hierarchical Ni96W4/Cu NW catalyst for
the acidic HER offered a performance that was ∼80% of that of a Pt-cathode-embedded electrolyzer [7].
Keywords: Electrodeposition, nanowires, iron group metals, electrocatalysis
© 2020 Published by ECRES

Acknowledgment:
Authors acknowledge funding from H2020 project MSCA-RISE-2017-778357-SMARTELECTRODES.

References:
[1] E. Vernickaite, N. Tsyntsaru, K. Sobczak, H. Cesiulis. https://doi.org/10.1016/j.electacta.2019.06.087 .
[2] E. Vernickaite, O. Bersirova, H. Cesiulis, N. Tsyntsaru. https://doi.org/10.3390/coatings9020085.
[3] E. Vernickaite, N. Tsyntsaru, H. Cesiulis. https://doi.org/10.1016/j.surfcoat.2016.07.049 .
[2] W. Yan, D. Wang, L.A. Diaz, G. G. Botte. https://doi.org/10.1016/j.electacta.2014.03.134.
[3] H.B. Lee, J.-C. Tsau, C.-Y. Lee. https://doi.org/10.1155/2013/191728.
[4] M. Nie, H. Sun, Z.D. Gao, Q. Li, Z.H. Xue, J. Luo, J.M. Liao. https://doi.org/10.1016/j.elecom.2020.106719.
[5] X. Gao Y. Wang, W.Li, F.Li, H.Arandiyan, H.Sun, Y.Chen. https://doi.org/10.1016/j.electacta.2018.07.033.
[6] S. Zhao et al. https://doi.org/10.1039/C6TA10749D.
[7] H. Kim et al. https://doi.org/10.1021/acssuschemeng.8b06643.

32
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

ELECTROCHEMICAL FABRICATION OF (BI2)M(BI2TE3)N THIN FILMS

WITH CONTROLLABLE BISMUTH CONTENT
Aliaksei Bakavets
Research Institute for Physical Chemical Problems, Belarusian State University, Minsk, Belarus, [email protected],
ORCID: 0000-0002-3348-8650

Yauhen Aniskevich
Chemistry Department, Belarusian State University, Minsk, Belarus, [email protected], ORCID: 0000-0003-2619-7019

Genady Ragoisha
Research Institute for Physical Chemical Problems, Belarusian State University, Minsk, Belarus, [email protected]

Natalia Tsyntsaru
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania/Institute of
Applied Physics, Chisinau, Moldova, [email protected], ORCID: 0000-0002-9813-2460

Henrikas Cesiulis
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania,
[email protected], ORCID: 0000-0002-5077-7884

Eugene Streltsov
Chemistry Department, Belarusian State University, Minsk, Belarus, [email protected], ORCID: 0000-0003-2939-8502

Bakavets, A., Aniskevich, Y., Ragoisha, G., Tsyntsaru, N., Cesiulis, H., Streltsov, E., Electrochemical fabrication
Cite this paper
of (Bi2)m(Bi2Te3)n thin films with controllable bismuth content. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
as:
Istanbul, Turkey

Abstract: Bismuth telluride is one of the best materials available for thermoelectric applications, though thermoelectric
performance of bare bismuth telluride is not sufficient for the waste heat conversion on a large scale. To increase
the overall performance of a thermoelectric material, its thermal conductivity should be reduced. Although it is
hard to suppress electronic heat transport while maintaining a high power factor, lattice thermal conductivity
can be reduced without any fallout. Formation of Bi 2Te3-based superlattices has proven to be useful as a way
to reduce the lattice thermal conductivity. We have developed the electrochemical pulse potential controlled
deposition of (Bi2)m(Bi2Te3)n superlattices with controllable Bi-content. The superlattice structure of the
deposits was proved via XRD analysis, XPS analysis, and TEM. The composition is controlled by the ratio of
Bi3+ and TeO2 precursor concentrations in the electrolyte. Moreover, we have shown that bismuth interlayers
can be removed from the superlattice structure by selective anodic oxidation, with the oxidation resulting in the
product corresponding to Bi2Te3 by stoichiometry but having an expanded crystal structure. These results clear
the way for new synthetic strategies of high performance Bi 2Te3-based thermoelectric materials preparation.
We also consider effect of pulse deposition parameters, electrolyte composition, pretreatment of substrate on
the nucleation and growth of the superlattice structure, morphology and uniformity of the deposit.
Acknowledgments: Authors acknowledge funding from H2020 project MSCA-RISE-2017-778357-
SMARTELECTRODES.
Keywords: Bismuth telluride, bismuth, superlattice, electrodeposition, thermoelectric material
© 2020 Published by ECRES

33
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

ULTRASONIC (CONTACT METHOD) WELD PENETRATION DEPTH

MEASURING STUDIES

Bulent Kılıçaslan Serkan

Department of Nanoscience and Nanoengineering, İstanbul Medeniyet University, Istanbul, 34700, Turkey, ORCID: 0000-0002-
0842-3636

Koten Hasan
Istanbul Medeniyet University, Mechanical Engineering Department, Istanbul, 34700, Turkey, ORCID: 0000-0002-1907-9420

Yucel Necla
Department of Nanoscience and Nanoengineering, İstanbul Medeniyet University, Istanbul, 34700, Turkey, ORCID: 0000-0003-
0093-2521

Kılıçarslan, S., Koten, H., Yucel, N. Ultrasonic (Contact Method) Weld Penetration Depth Measuring
Cite this paper as:
Studies. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In industrial production facilities (field or factory) the most used quality assurance tool is still Non-
Destructive tests. Due to the portability, safety and the cost efficiency, Ultrasonic method has a big share in
the inspection industry. One of the industries which ultrasonic inspection is nearly essential for production
is automotive manufacturing. For spot weld inspections ultrasonic is the only NDT method. For years,
automotive industry has been looking for a solution for TIG arc welds to inspection the penetration depth of
the weld. Many studies were made by contact ultrasonic, EMAT, Laser Ultrasonic (Non-Contact) techniques.
Most of the studies used LUT technique, due to its novelty and fresh research areas. Contact ultrasonic
technique become less studied, and most studies were mainly on theoretical parts. In our planned study, we
would like to see the results of the contact ultrasonic method for measuring the weld penetration depth with
different contact technique and our study will differ from previous studies that contact technique will be used
in Welding process. For this aim, we reviewed four studies of ten researchers which were conducted with
different techniques to measure the weld penetration depth, between the years of 1984 and 2012.
Keywords: Ultrasonic, Non Destructive Test, Weld penetration Depth
© 2020 Published by ECRES

34
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

EXERGY ANALYSES AND INTEGRATED GAS COMBINED CYCLE

FOR A HIGH EFFICIENCY OF A NUCLEAR POWER PLANT

Rauf Terzi
Nuclear Regulatory Authority, Ankara, Turkey, [email protected], ORCID: 0000-0001-6491-5005

Erol Kurt
Gazi University, Faculty of Technology, Department of Electrical and Electronics Engineering, Ankara, Turkey,
[email protected], ORCID: 0000-0002-3615-6926

Terzi, R.,Kurt, E. Exergy analyses and integrated gas combined cycle for a high efficiency of a nuclear
Cite this paper as:
power plant. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,

Abstract: The most practical and effective ways for increasing the thermal efficiency of power generation system is
based on a gas turbine combined conversion cycle. In this study, an integrated Gas Combined Conversion
Cycle (GCCC) is proposed as a means of integrating a gas turbine cycle with a 1000 MW water-cooled
water-moderated power reactor (VVER) type nuclear power plant that is thermodynamically analysed. In
this cycle, saturated steam produced in the steam generators of the nuclear power plant is superheated by use
of the gas turbine. An exergy analysis is carried out actual operating parameters based on the second law of
thermodynamics and the effects of gas turbine inlet temperature are investigated on the net power output, the
first and second law efficiencies. The results of analysis show that it is possible to obtain for analysed cycles
estimated thermal efficiency of an 50%. The development of the proposed nuclear gas turbine combined
conversion cycle (NGTCCC) concept could benefit from present technology bases and this could lead to
commercial operation of multi-module or hybrid plants in the near future
Keywords: Exergy analysis, nuclear power plant, gas turbine.
© 2020 Published by ECRES

35
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

ELECTRODEPOSITION OF CO AND CO-PT METAL FOAMS FOR

WATER ELECTROCATALYSIS

Modestas Vainoris
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania,
[email protected], ORCID: 0000-0001-8488-9545

Natalia Tsyntsaru
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania/Institute of
Applied Physics, Chisinau, Moldova, [email protected], ORCID: 0000-0002-9813-2460

Henrikas Cesiulis
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania,
[email protected], ORCID: 0000-0002-5077-7884

Vainoris M., Tsyntsaru, N., Cesiulis, H., Electrodeposition of Co and Co-Pt metal foams for water
Cite this paper as:
electrocatalysis. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Cobalt and its alloys have various applications, e.g. anti-corrosion and wear-protecting coatings, magnetic
recording and medical devices, catalysts, sensors, etc. With the increase in the need for eco-friendly and
efficient energy sources, such as O2 and H2 fuel cells, alternatives to catalysts like Pt, Re, Au, and other
precious metals are being investigated. Co or Co alloys metal foams, with its high surface area, could be used
as an efficient and cheap catalyst for oxygen evolution reaction (OER) and hydrogen evolution reaction
(HER).
This work aimed to find an easy and cost-efficient way to manufacture Co/Co-Pt metal foams via
electrodeposition, using hydrogen bubbles as a dynamic template. Metal foams were deposited under
galvanostatic conditions onto a copper substrate using high current densities. Obtained Co/Co-Pt foams
morphology was investigated using scanning electron microscope and energy dispersive spectroscopy.
Cylindrical pores were visible on the coatings, however, the density of the pores was rather low and neither
pores diameter nor their density was dependant on the current density applied. In order to increase Co
coatings porosity, the solution surface tension was reduced with the addition of 2 M isopropanol. The
diameter of pores, overall porosity as well as the surface area were increased, however, the efficiency of
deposition was reduced. Chemical composition of the Co metal foams shows the presence of oxygen up to
50 at.% - meaning obtained high surface area foams, are easily oxidized in the air. To even further increase
Co foams’ surface area, the effect of angle between the working electrode and counter electrode during
electrodeposition was also investigated. Two angles of 45° and 90° were chosen. Samples obtained at 45°
had more pores on both sides, in turn, foams obtained under 90° angle had bigger porosity, but only on the
side facing the bottom of the electrochemical cell. The catalytic activity of Co/Co-Pt metal foams for
HER/OER was tested in 0.1 M KOH solution, performing polarization from open circuit potential to -1.2 V.
HER started at the potential region of -0.75-0.85 V. Pure Co foams showed low overpotential needed to start
the HER, but 10mA/cm² current density was reached only using 150mV overpotential, however, using Co-
Pt foams (Pt amount ~20-30 at. %) the overpotential was significantly decreased.
Keywords: Cobalt, electrodeposition, foams, electrocatalytic activity
© 2020 Published by ECRES

Acknowledgment:
Authors acknowledge funding from H2020 project MSCA-RISE-2017-778357-SMARTELECTRODES

36
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

ELECTROCHEMICAL IMPEDANCE SPECTROSCOPY AS AN

EFFECTIVE TOOL TO CHARACTERIZE MATERIALS FOR ENERGY
APPLICATIONS
Henrikas Cesiulis
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania,
[email protected], ORCID: 0000-0002-5077-7884

Natalia Tsyntsaru
Vilnius University, Faculty of Chemistry and Geosciences, Department of Physical Chemistry, Vilnius, Lithuania/Institute of
Applied Physics, Chisinau, Moldova, [email protected] , ORCID: 0000-0002-9813-2460

Cesiulis, H., Tsyntsaru N. Electrochemical impedance spectroscopy as an effective tool to characterize

Cite this paper as:
materials for energy applications. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The electrochemical impedance spectroscopy (EIS) being a non-destructive technique can provide time
dependent quantitative information about the electrode processes and complex interfaces; therefore it is
widely used for characterization of electrochemical systems and surfaces. The specific area or
electrochemically active area is an important factor influencing on the energetic efficiency of materials. The
mentioned areas are essentially different from geometric areas. The voltammetric methods to determine
active area are described in [1]. For this purposes the EIS is a convenient method. The interpretation of EIS
data is built upon reaction model (or equivalent circuit) consisting on passive elements connected in some
order. The roughness factor, and sequentially the specific surface area can be estimated by normalizing the
double layer capacitance (DLC) of the electrode with the double layer capacitance of a flat surface. DLC is
easily extracted from EIS data recorded in the range of relatively high frequencies, i.e. 300-5000 Hz. In this
way, the active area of metallic foams was determined [2] and is 100-300 times larger than that of flat surface.
However, the complications are raised when the electrochemical reaction on metallic foams undergoes with
some diffusion limitations. In this case, the active area is only 3-10 times higher than that of the flat surface.
The semiconducting materials are also widely used for energy conversion. Currently, research on the rarer
metal oxides (such as tungsten or molybdenum) remains important, especially in the increasingly vital solar
energy conversion field. Anodization is a convenient and effective way to form an oxide film, because the
anodic dissolution of these metals is followed by immediate formation of semiconducting oxide films. By
means of EIS it is possible to investigate kinetics of an oxide layer formation, also by applying Mott-Schottky
analysis to determine the type of semiconductor and donor site density, active site density for hydrogen
adsorption [3,4].
Keywords: EIS, surface area, semiconductors, metal foams, active sites
© 2020 Published by ECRES

Acknowledgment:
Authors acknowledge funding from H2020 project MSCA-RISE-2017-778357-SMARTELECTRODES.

References:
[1] D. Voiry, M.Chhlowalla, Y. Gotosi, N.A. Kotov, Y. Li, R.M. Penner, E. Schaak, P.S. Weiss., DOI: 10.1021/acsnano.8b07700
[2] M. Vainoris, N. Tsyntsaru, H. Cesiulis. https://doi.org/10.3390/coatings9050306
[3] H. Cesiulis, N. Tsyntsaru, A. Ramanavicius, G. Ragoisha. DOI 10.1007/978-3-319-30198-3_1
[4] R. Levinas, N.Tsyntsaru, H. Cesiulis. 10.1016/j.electacta.2019.06.002

37
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

HYBRIDIZATION OF A GEOTHERMAL POWER PLANT WITH

BIOMASS DRIVEN SCO2 CYCLE

Balkan Mutlu
Mechanical Engineering Department, Middle East Technical University, Ankara, Turkey, [email protected], ORCID:
0000-0002-7526-6538

Derek Baker
Mechanical Engineering Department, Middle East Technical University, Ankara, Turkey, [email protected], ORCID: 0000-
0003-4163-1821

Feyza Kazanç Özerinç

Mechanical Engineering Department, Middle East Technical University, Ankara, Turkey, [email protected], ORCID: 0000-
0002-3850-3071

Mutlu, Balkan, Baker, Derek, Kazanç Özerinç, Feyza. An Investigation of Opportunities to Increase the
Cite this paper as: Flexibility of Flash Geothermal Power Plant Through Hybridization with Biomass driven sCO2 Cycle:
Case Study of Kizildere-1. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Due to the increasing penetration of variable renewable energy technologies, the need for flexible power
plants is becoming increasingly important. This study investigates the hybridization scenario of Kizildere-1
single flash geothermal power plant with a biomass driven recompression sCO2 topping cycle where a solid
local biomass source, olive residue, is used as a fuel. The combustion characteristics, kinetic parameters and
slagging-fouling potential of the fuel are determined using TGA, Coats-Redfern Method and XRF analysis,
respectively. The hybrid power plant is modeled using the simulation software EBSILON Professional.
Supercritical CO2 (sCO2) cycle is specifically chosen for evaluation of future scenarios due to having a more
compact footprint compared to steam cycles and the potential to be ramped up and down more rapidly, thus
enabling flexible operation. A synergy between the topping sCO2 and bottoming geothermal cycle is
achieved by a good temperature match between two cycles at the coupling heat exchanger where the waste
heat from the topping cycle is utilized in the bottoming cycle. The intrinsic limitation of sCO2 cycles, i.e.
the external heat addition to the cycle takes place between a certain temperature range, is also eliminated
utilizing the whole heat in the flue gas in the geothermal cycle.
Keywords: hybridization, geothermal, biomass, sCO2 cycle
© 2020 Published by ECRES

38
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

OPTIMIZATION OF ZNO-BASED NEAR UV SOLAR CELL AND LIGHT

EMITTING DEVICES

Bilgehan Barış Öner

Gazi University-Department of Physics, 06500, Ankara, Turkey, [email protected], ORCID: 0000-0001-9440-2235

İrem Öner Alp

Gazi University-Department of Physics, 06500, Ankara, Turkey, [email protected], ORCID: 0000-0002-6937-7864

Esra Eroğlu
Gazi University-Department of Physics, 06500, Ankara, Turkey, [email protected], ORCID: 0000-0002-6848-5142

Saime Şebnem Aydin

Gazi University-Department of Physics, 06500, Ankara, Turkey, [email protected], ORCID: 0000-0002-1617-1643

Yasemin Öztekin Çiftci

Gazi University-Department of Physics, 06500, Ankara, Turkey, [email protected], ORCID: 0000-0003-1796-0270

Öner, BB., Alp, İÖ., Eroğlu, E., Aydin, SŞ, Çiftçi YÖ. Optimization of ZnO-based Near UV Solar Cell
Cite this paper as:
and Light Emitting Devices. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In this study, ZnO based ultraviolet optoelectronic devices (a solar cell and a light emitting diode) have been
designed by nextnano optoelectronic device software package. These types of LED devices suffer from
possible maximum p-doping concentrations. However, it can be shown that it is possible to optimize p-i-n
junction layer thickness, doping and alloy concentration values to have relatively increased solar and internal
quantum efficiencies (IQEs) of each devices. Each material corresponding to the optoelectronic device layers
are analyzed via CASTEP to investigate structural, electronic and optical properties of the related materials.
Refractive index and extinction / absorption coefficient values of the LED layers are obtained to compute
also light transmission and external quantum efficiency values. In conclusion, efficient solar cells (solar
efficiency over 3%) and single quantum well light emitting devices (IQE over 30%) operating within near
ultraviolet region have been achieved computationally. The ZnO material is also utilized within hybrid solar
cells to increase the energy efficiency of a device by an additional absorption of the UV region of the solar
spectrum.
Keywords: Solar Cells, UV-LEDs, ZnO Based Devices, Ultraviolet Optoelectronic Devices, Optical Properties
© 2020 Published by ECRES

Figure 1. Schematic design of ZnO based near UV (a) LED and (b) solar cell devices.

39
8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

40
8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

REGULAR FULL
PAPERS

41
 A CLASSIFICATION TOOL FOR THE MOTION OF A
PIEZOELECTRIC ENERGY HARVESTER VIA NEURAL
NETWORK
Erol Kurt
Gazi University, Technology Faculty, Department of Electrical and Electronics Engineering, Ankara, Turkey, E-mail:
[email protected], ORCID: 0000-0002-3615-6926

Mehmet Tekerek
Kahramanmaraş Sütçüimam University, Computer Education and Instructional Technology Department,
Kahramanmaraş, Turkey, E-mail: [email protected], ORCID: 0000-0001-6112-3651

Kayhan Çelik
Gazi University, Department of Electrical and Electronics Engineering, Faculty of Technology, Ankara, Turkey,
[email protected], ORCID: 0000-0003-0371-0473

Mehmet Gök
Ministry of National Education, Kahramanmaraş, Turkey, [email protected], ORCID: 0000-0003-1656-5770

Kurt, E, Tekerek, M, Çelik, K, Gök, M.. A Classification Tool for The Motion of A Piezoelectric
Cite this paper as: Energy Harvester Via Neural Network. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: A new classification tool has been design and implemented for the sake of the dynamic responses of
a new built wind energy harvester. The tool indeed uses an artificial neural network algorithm in
order to determine displacement and velocity of the tip of the piezoelectric layer. It trains time series
data of the displacement and velocity and processes them into a two-layer network for a given set of
excitation frequency and amplitude. According to one of our previous work in Ref. [1], it was proven
that such a system creates different types of dynamic responses from periodic to hyper chaotic. All
these dynamic responses yield to different power generation for this harvester; thereby an estimation
of the dynamic character (i.e. motion) of a piezoelectric beam is vital to predict the output power of
the harvester. Following the network training procedure, it has been proven that the classification
can be performed with a high accuracy. This tool can then be recommended for the engineering
applications on the vibration-based systems such as transducers and sensors.

Keywords: Wind energy, piezoelectric, motion classification, dynamics, neural network

© 2020 Published by ECRES

1. INTRODUCTION

Since the existing energy is limited and the increasing of the energy demand in the world makes it imperative to
develop new and effective energy harvesting systems and techniques [1]. These systems mainly focus to meet
the power requirements of the low power electronic devices such as wireless sensor networks and actuators.
Because the harvester makes it possible to extend the lifecycle of these kind of systems via the by acting as a
source of battery [2].

In the literature, there are many types of energy harvesters such as vibration [3], solar [4] and wind [5] based.
Vibration based piezoelectric harvesters (PZT) have the highest power density and they exhibit the much longer
life span due to contactless operation mechanism with the help of the magnetic field [6] [7]. In addition to this,
PZT has the high efficiency at its own operating frequency. For this reason, it is the most important issue to
convert the frequencies of natural vibrations in the environment to the resonance frequency of PZT in order to
obtain maximum power. The main reason for that is the random vibrations at different frequencies, which cause
loss of power and decrease the efficiency. The wide band spectrum of the ambient vibrations gives us chance

42
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

for the exploring the characteristics of the pendulum-like PZT layers by the changing of the magnetic force
volume. So that, the lifetime of the PZT can be extended a little bit more.

2. THE WIND ENERGY HARVESTER AND THEORETICAL BACKGROUND

The wind harvester has a four unis which are the rotating propeller with wind power, shaft which transmits the
rotational motion to the magnet, piezoelectric unit which produces the alternating voltage and lastly the output
port regulates the harvested power as shown in Fig. 1. The other physical properties and important parameters
of the harvester can be learned from the previous work [1].

Figure 1. The piezoelectric wind harvester.

To summarize briefly, the harvester operates as follows: The shaft, which is connected to the propeller, is rotated
by the effect of the wind. Two magnets connected to the shaft and PZT approach each other and move away
with the rotation of the shaft. This movement repeats in a certain period depending on the wind speed. Since the
magnets with the same poles will push each other when they approach to each other; with this magnetic force,
the piezoelectric layer will bend and restore periodically and produce an electric current as shown in Figure 2.

Figure 2. The operation mechanism and experiment setup of the wind harvester.

2.1. The Vibration Tests of the Piezoelectric Beam

The experimental setup for the vibration characteristic of the PZT consist of the ferromagnetic beam with a PZT,
the magnetic excitation unit, rectifier and storage circuit, signal generator, laser displacement sensor (LDS) and
data acquisition and monitoring units. As seen in Figure 2, the electromagnet is made active and passive with
the periodic signal produced from the signal generator. When it is activated, the magnet which is positioned to
the bottom tip of the PZT layer as a pendulum is pushed and PZT layer is bent. When it is done passive, PZT
layer becomes straight again periodically with the help of magnetic force. The electrical power is obtained as a
result of this periodic movement. LSD measures the displacement of the PZT layer and data acquisition and
monitoring units saves the data on the computer such as displacement, applied magnetic force as a voltage,
harvested output power. This kind of system can be used to analyses the dynamical vibration characteristics of
the pendulum like PZT under the different magnetic excitation frequencies (ωc) and the magnetic force strength
(applied voltage-Uc). The displacement occurs in the system with the effect of the magnetic force generated by

43
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the applied periodic signal, which creates voltage and current on the load. The graph of displacement of the PZT
according to the different excitation frequencies constitutes the different chaotic attractors such as regular, quasi-
periodic, saw-teeth chaotic, the double, triple and many scroll chaotic as seen from the Figure 3. It is time-
consuming to classify these attractors visually. Because of this, in this paper, the classification of the attractor
are made via the Neural Network. In the next chapter details of this process is explained briefly.

Figure 3. The attractors (i.e. phase space representations) from displacement versus velocity data from simulations.

3. ARTIFICIAL NEURAL NETWORKS

An artificial nerve cell can be defined as an algorithm that reveals a mathematical model developed based on
the basic behaviors of a biological nerve cell or a physical tool [8]. As in the definition of the biological nerve
cell, it can be said that an artificial nerve cell collects the signals from other nerve cells and as soon as the total
signal accumulation exceeds a certain threshold, this artificial nerve cell transmit its signal to another nerve cell.
As biological neural networks have nerve cells, artificial neural networks also have artificial neural cells.
Artificial neural cells in engineering science is also called as a process element. As shown in Figure 4, basic
elements of each process element are the inputs, weights, transfer function, activation function and outputs [9].

Figure 4. The general structure of an artificial nerve cell.

Artificial neural networks are databased systems created by connecting artificial neural cells in layers. It aims
to use the abilities of the human brain, such as learning and ability to make decisions very quickly under different
conditions to solve complex problems with the help of simplified models. The task of the artificial neural

44
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

network is largely determined by the connections between process elements. An artificial neural network can be
trained by adjusting the weights between the process elements to perform a particular task. Artificial neural
networks can approach any function with a high accuracy rate. As it is a non-linear model, it is flexible enough
to be used to reveal complex relationships in practically any problem. Since it has a data-driven and self-adapting
structure, it works without the need for additional functions or distributed definitions for the targeted model [10].

Generally, an ANN consists of three types of layers, the input layer where external information is applied to the
network, hidden layers that are capable of processing information and an output layer that exports the neural
network's decisions as shown in Figure 5 [11]. In the input layer, the process elements transmit the information
to the hidden layer without making any changes. The information mentioned here is expressed by weights on
the connecting lines between the processing elements of the neural network. This information is obtained during
the training of ANN and it is used to interpret the situations after training of the ANN. It is possible to use all
ANN models as a pattern classifier. However, the most widely used ANN is multi-layer feed forward back
propagation network structure [12]. Figure 5 shows a multi-layered feed forward ANN model.

Figure 5. Multi-layer feed forward ANN model.

The number of the layers of the ANN and the number of the neurons in the layers, the type of the activation
function are usually application dependent [13] [14]. Artificial neural networks are an auxiliary tool for
classification problems and have been successfully applied in many classification problems in industrial and
scientific fields. In this study, "Back Propagation Network" is used as artificial neural network [15] [16] [17].

3.1. Estimation of the chaotic attractor type of Piezoelectric Layer under Magnetic Stimulation with
Artificial Neural Network

As mentioned before, the graph of displacement of the PZT according to the different excitation frequencies
constitutes the different chaotic attractors. In this part of the paper, the process of the classification of the
attractors via the back propagation neural network are presented rather than visually classifying. The excitation
voltages and frequencies are used for the training of the ANN, however, it is observed that the values of
excitation voltages and frequencies are insufficient to train the artificial neural network. In addition to these
values, RMS values of the displacement and THD values of the velocity corresponding to each voltage and
frequency values are extracted for using in the training of the artificial neural network. In addition to the
excitation voltage and frequency, one of the RMS or THD values will be used in ANN training. Which one of
the RMS and THD inputs will get better results will continue to be used. The multi-layered ANN structure which
is trained with back propagation algorithm is used which has 3 input variables. The structure has 2 hidden layers;
at the 1st layer there is 14 neurons and at the 2nd layer there is 20 neuron. The vector length of the output
classification is 7. The learning algorithm is selected as Gradient Descent. MSE target value and the learning
coefficient are taken as 0.00001 and 0.07 respectively.

45
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1. Sample assessments of classification

Frequency value of Excitation RMS value of displacement THD values of velocity
Class
excitation voltage (F) voltage (U) values (RMS-D) values (THD-V)
190 30 0,1542 0,7313 multiple scroll chaotic
190 90 0,0971 0,2528 multiple scroll chaotic
190 2 0,0731 0,1011 periodic regular
190 4 0,0736 0,1041 periodic regular
190 8 0,0747 1,6042 periodic regular
190 12 0,0761 0,1093 periodic regular
190 20 0,0804 0,1079 periodic regular
200 25 0,1493 0,0734 multiple scroll chaotic
200 90 0,1681 0,2025 multiple scroll chaotic
200 16 0,0891 31,7236 quasi periodic
200 50 0,0877 0,2151 double scroll chaotic
200 60 0,0928 0,1008 double scroll chaotic

Table 2. Numbering values of the class tags

Class Tags Value
multiple scroll chaotic 0
quasi periodic 1
chaotic 2
double scroll chaotic 3
periodic regular 4
saw-teeth chaotic 5
triple scroll chaotic 6

4. EXPERIMENTAL RESULTS

The dataset used for the ANN training is reapplied for the verification process of the system at the experimental
study. RMS-D and THD-V values are used as a third input variable respectively. When RMS-D is used as the
third feature, the achieved success rate of the ANN system is 81.05 % for the classification of the chaotic
attractors. In this process, the 124 of the total 153 cases are classified correctly. The cross-validation matrix for
this classification is given below. Diagonal elements in the cross-validation matrix show the number of states
correctly classified. The other elements show the number of the wrong classified status.

Table 3. Cross verification chart of the 1st process.

0 1 2 3 4 5 6
0 17 2 0 8 0 0 0
1 0 33 0 0 0 0 0
2 0 1 8 0 3 0 0
3 1 0 0 43 0 0 0
4 0 0 2 0 22 0 0
5 0 6 3 0 0 1 0
6 0 0 0 3 0 0 0

Considering this table, it is seen that the correct classification could not be made in any way for case 6 (triple
scroll chaotic) since the number of the data in the training set is low. However, it is clearly seen that the
classification rate of the 3 rd third case (double scroll chaotic) is higher than the 6th case since due to the large
amount of data. When THD-V is used as the third feature, the achieved success rate of the ANN system is 81.7%
for the classification of the chaotic attractors. In this process, the 125 of the total 153 cases are classified
correctly. The cross-validation matrix for this classification is given below. Diagonal elements in the cross-
validation matrix show the number of states correctly classified. The other elements show the number of the
wrong classified status. When this chart is examined, the almost the same result as the previous one is obtained.

46
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 4. Cross verification chart of the 2nd process.

0 1 2 3 4 5 6
0 18 2 0 5 0 0 2
1 0 33 0 0 0 0 0
2 0 1 8 0 3 0 0
3 2 0 0 42 0 0 0
4 0 0 2 0 22 0 0
5 0 5 3 0 1 1 0
6 0 0 0 2 0 0 1

After both training procedures, it is seen that more data are needed especially for cases 5 and 6. The highest
performance is achieved when using Gradient Descent as a learning algorithm. The training process is completed
very quickly with Levenberg-Marquadt algorithm; however, it is observed that the performance is quite low in
this test procedure.

Figure 6. The graph of the classified attractor types via ANN.

Figure 7. The graph of the visually classified attractor types.

The result of the classification process is given graphically in Figure 6 and the classification made visually in
the previous work is also given in Figure 7. According to the graphs, it can be seen that ANN based classification
method is successful and it has a high performance.

5. CONCLUSIONS

In this study we have performed a new ANN model to classify the dynamics of a piezoelectric energy harvester.
The tool indeed uses an artificial neural network algorithm in order to determine displacement and velocity of
the tip of the piezoelectric layer. It trains time series data of the displacement and velocity and processes them

47
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

into a two-layer network for a given set of excitation frequency and amplitude. It has been proven that the new
model can predict and classify the dynamic responses of the new harvester within a good accuracy.

REFERENCES
[1] Çelik, Kayhan, Erol Kurt, and Yunus Uzun. "Experimental and theoretical explorations on the buckling piezoelectric
layer under magnetic excitation." Journal of Electronic Materials 46.7 (2017): 4003-4016.
[2] Dondi, Denis, et al. "Modeling and optimization of a solar energy harvester system for self-powered wireless sensor
networks." IEEE Transactions on industrial electronics 55.7 (2008): 2759-2766.
[3] Sun, Shilong, and W. Tse Peter. "Modeling of a horizontal asymmetric U-shaped vibration-based piezoelectric energy
harvester (U-VPEH)." Mechanical Systems and Signal Processing 114 (2019): 467-485.
[4] Saha, C. R., et al. "Photovoltaic (PV) and thermo-electric energy harvesters for charging applications." Microelectronics
Journal 96 (2020): 104685.
[5] Hu, Gang, et al. "Experimental investigation on the efficiency of circular cylinder-based wind energy harvester with
different rod-shaped attachments." Applied energy 226 (2018): 682-689.
[6] Uzun, Yunus, S. Demirbas, and E. Kurt. "Implementation of a new contactless piezoelectric wind energy harvester to a
wireless weather station." Elektronika ir Elektrotechnika 20.10 (2014): 35-39.
[7] Karami, M. Amin, Justin R. Farmer, and Daniel J. Inman. "Parametrically excited nonlinear piezoelectric compact wind
turbine." Renewable energy 50 (2013): 977-987.
[8] Shanmuganathan, Subana. "Artificial neural network modelling: An introduction." Artificial neural network modelling.
Springer, Cham, 2016. 1-14.
[9] Dongare, A. D., R. R. Kharde, and Amit D. Kachare. "Introduction to artificial neural network." International Journal
of Engineering and Innovative Technology (IJEIT) 2.1 (2012): 189-194.
[10] 1) Zhang, G.P., "Neural Networks for Classification: A Survey", IEEE Trans. on Systems, Man, and Cybernetics—
part c: applications and reviews, 30 (4): 451-462 (2000).
[11] Samarasinghe, Sandhya. Neural networks for applied sciences and engineering: from fundamentals to complex pattern
recognition. Crc Press, 2016.
[12] 2) Chen, C., “A two-stage solution procedure for digital power metering according to IEEE Standart 1459-2010 in
single-phase system”, IEEE Transactions on Industrial Electronics, 60 (12): 5550-5552 (2013).
[13] Huang, Guang-Bin, and Haroon A. Babri. "Upper bounds on the number of hidden neurons in feedforward networks
with arbitrary bounded nonlinear activation functions." IEEE transactions on neural networks 9.1 (1998): 224-229.
[14] Karlik, Bekir, and A. Vehbi Olgac. "Performance analysis of various activation functions in generalized MLP
architectures of neural networks." International Journal of Artificial Intelligence and Expert Systems 1.4 (2011): 111-
122.
[15] LeCun, Yann, et al. "Handwritten digit recognition with a back-propagation network." Advances in neural information
processing systems. 1990.
[16] Hecht-Nielsen, Robert. "Theory of the backpropagation neural network." Neural networks for perception. Academic
Press, 1992. 65-93.
[17] Nekovei, Reza, and Ying Sun. "Back-propagation network and its configuration for blood vessel detection in
angiograms." IEEE Transactions on Neural Networks 6.1 (1995): 64-72.

48
 NUMERICAL INVESTIGATION OF HEAT TRANSFER AND
PRESSURE DROP ON FIN-TUBE HEAT EXCHANGERS

Aslan Erman
Istanbul University-Cerrahpasa, Department of Mechanical Engineering, Istanbul, Turkey,
[email protected], ORCID: 0000-0001-8595-6092

Yildiz Omer
Istanbul University-Cerrahpasa, Department of Mechanical Engineering, Istanbul, Turkey,
[email protected], ORCID: 0000-0001-6311-1670

Kartun Hagop
Karyer Heat Exchangers, R&D Department, Istanbul, Turkey, [email protected],
ORCID: 0000-0003-0816-3755

Gungor Gokhan
Karyer Heat Exchangers, R&D Department, Istanbul, Turkey, [email protected],
ORCID: 0000-0002-4143-224X

Aslan, E, Yildiz, O, Kartun, H, Gungor, G. Numerical Investigation of Heat Transfer and

Cite this paper as: Pressure Drop on Fin-Tube Heat Exchangers. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
2020, Istanbul, Turkey

Abstract: The characteristics of heat transfer and pressure drop for fin-tube heat exchangers are investigated
numerically. In the numerical study, finite volume based commercial code Ansys-Fluent 19.2 is
used. Three different fin types are considered, namely, plain, sharp wave, and sine wave. As a
working fluid, water and air are used for inner and outer flow, respectively. Hot water passes
through the tubes and cold air progress across the fins and tubes. Material of tubes is copper and
material of fins is aluminum. Three-dimensional numerical analysis is performed. In order to avoid
large mesh numbers, whole heat exchangers are not modelled. Three fins and twenty-four tubes
are modelled in partial modelling. Tubes are arranged in staggered form. While six tubes are
located in flow direction as tube rows, four tubes are settled perpendicular the flow direction as
tubes in height. Eight different tube configurations which are 50×25, 35×30, 30×26, 40×34.64,
40×24.66, 40×20, 32×28 and 44×38.1 are determined. Fin spacing is 2.5mm. Nusselt number,
pressure drop and fin efficiency are discussed for three different fins and eight different tube
configurations.
Keywords: Heat transfer, pressure drop, fin efficiency, fin-tube heat exchanger
© 2020 Published by ECRES

Nomenclature
A fin Fin area [m2]
Cp Specific heat [j/kgK]
Di Inner tube diameter [m]
Do Outer fin diameter [m]
P Pressure drop [Pa]
 Fin efficiency
f Friction factor
h Convective heat transfer coefficient of air [W/m2K]
k Thermal conductivity of air [W/mK]
L Length of domain at axial direction [m]
mair Mass flow rate of air [kg/s]
 Thermal performance factor
Nu Nusselt number
Nu wo  fin Nusselt number in case without fin
ST Transverse distance of tubes [m]
SL Longitudinal distance of tubes [m]

49
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Ti , a Temperature of inlet air [ºC]

Ti , w Temperature of inlet water [ºC]
To , a Temperature of outlet air [ºC]
T fin Fin temperature [ºC]
 Density of air [kg/m3]
q Heat, transfer rate [W]
qwo  fin Heat transfer rate in case without fin [W]
Vi , a Velocity of inlet air [ºC]
Vi , w Velocity of inlet water [ºC]

1. INTRODUCTION

The fin-tube heat exchangers have been widely used in many industrial applications, such as refrigeration
system, air conditioning system, food processing system and petrochemical systems. In these applications, heat
transfer is normally limited by the thermal resistance on the air side of the heat exchanger. One solution to
increase heat transfer on the air side of the heat exchanger is to modify the fin geometry [1]. As the sharp wave
and sine wave can extend the surface area and provide better air flow mixing, they are currently used. For
demonstrating the air side performance of fin and tube heat exchangers, numerical simulations can be used,
Also, the inner flow is solved via numerical simulations too.
The air side of performance of different fin geometries and tube heat exchangers have been studied
experimentally and numerically, by a number of researches. Wen and Ho made experimental study about fin-
tube heat exchanger [2]. In this experimental study, three different fins are used, namely, plate fin, wavy fin and
compounded fin. The heat transfer and pressure drop performances are discussed against Reynolds number.
According to results, usage of compounded fins is strongly suggested by them [2]. Thermal-hydraulic
performance of fin and tube heat exchangers was experimentally investigated by Okbaz et al [3]. Effect on
thermal-hydraulic performance of three most dominant geometry parameters, which are fin type (louvered and
wavy fins), fin pitch and number of tube-rows, are discussed. Jang and Chen are performed numerical simulation
to predict heat transfer and fluid flow characteristics in a three dimensional wavy fin and tube heat exchangers
[4]. The effects of different geometrical parameters, containing tube row numbers (1-4 rows), wavy angles
(8.95º, 17.05º and 32.21º) and wavy heights (0.751, 1.500 and 3.003) are investigated in detail for Reynolds
number varying from 400 to 1200, which is based on fin spacing and frontal velocity. In numerical simulations,
control volume based finite difference formulation is used. They can be observed that, row effect is not crucial
in wavy fin as compared to plain fin counterpart. Airside performance of sinusoidal wavy fin-tube heat
exchangers wavy fin-tube heat exchangers with circular tube configurations is studied experimentally and
numerically by Chu et. al [5]. The finite volume based commercial code Ansys-Fluent is used to analyze local
heat transfer performance. The existing experimental correlations for wavy fin with circular tube fail to present
experimental data due to different range wave of fin pitch, tube pitch and tube diameter. On the other hand,
numerical predictions are suitable with present experimental results.
Alternative methods can be used for increasing heat transfer in fin-tube heat exchangers. One of method through
these alternatives is using delta winglets [6-8] and slits [6] on the fins. Also, some amendments are can be
applied in the tubes, for example using elliptical tubes [5, 9, 10] and elliptical tubes with protrusion [11], hereby,
thermal performance of inner flow is increased, therefore, total heat transfer of fin-tube heat transfer is enhanced.
In this paper, heat transfer, pressure drop characteristics are investigated numerically for fin-tube heat
exchangers. For solving governing equation numerically, finite volume based commercial code Ansys-Fluent
19.2 [12] is used. Three different fin types, which are plain, sharp wave and sine wave, are applied. Circular
tubes are used for inner domain. Three dimensional numerical analysis is performed. Instead of modelling of
whole domain, partial modelling is performed, for avoiding large mesh numbers. Eight different tube
configurations (transverse pitch × longitudinal pitch) are determined, which are 50×25, 35×30, 30×26,
40×34.64, 40×24.66, 40×20, 32×28 and 44×38.1. Fin spacing is kept constant as 2.5mm. Nusselt number, fin
efficiency, friction factor and thermal performance factor are discussed for three different fins and eight different
tube configurations.

2. PROBLEM DEFINITION

Figure 1 shows the schematic figure fin-tube heat exchangers and the used fin types. While air flow (outer flow)
passes across the fin and tubes, water flow (inner flow) is proceed the through the tubes. Inlet velocity of air is

50
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(Vi,a=3.36 m/s) for outer flow, and inlet velocity of water is (Vi,w=1.79 m/s) for inner flow. Cold air (Ti,a=15 ºC)
enters to the outer domain, at the same time, hot water (Ti,w =90 ºC) gets in the inner domain. Therefore, heat
transfer occurs between two flow phenomenon via fin and tubes. Tube and fin are made of, copper and
aluminum, respectively. Tube inner diameter is taken as 12.7mm, the thickness of tube is 0.40mm. Three
different fin types are used, which names are (a) plain, (b) sharp wave and (c) sine wave. The fin thickness and
spacing are taken as 0.10mm and 2.5mm, respectively.

Figure 1. The schematic figure of fin-tube heat exchangers with boundary conditions and the used fin types (a)plain
(b)sharp wave and (c)sine wave

Figure 2 shows the staggered tube arrangements over the sharp wave fin as an example. Staggered arrangements
has been used for both plain, sharp and sine wave fin. The considered variable for staggered arrangement are
inner tube diameter (Di), outer fin diameter (Do), transverse distance (ST) and longitudinal distance (SL). Eight
different transverse and longitudinal distance combinations are determined. The determined combinations are
listed in Figure 2, again.
Reynolds number is based on tube inner diameter and outer fin diameter, for inner flow and outer flow,
respectively. For inner flow, Reynolds number is taken as 2333 and for outer flow, Reynolds number is kept as
3091.
Thirty-two cases are determined to solve numerically. Cases of three different fin types and eight different
spacings are solved, additionally, cases with without fin are also solved with eight spacing. For better classify,
fins are named (a) plain, (b) sharp wave and (c) sine wave like in Figure 1, Similar in Figure 2, spacing are listed
from 1 to 8. Therefore, combine these, new naming is occurred. For example, If we mentions, sharp wave with
30×36, 3-b is new name.

3. MATHEMATICAL MODELLING AND ASSUMPTIONS

The time-dependent incompressible Navier Stokes equations coupled with the continuity equation ad energy
equation are solved using the finite volume based commercial computational fluid dynamics code Ansys-Fluent
19.2 [12]. According to Reynolds number of inner and outer flow, laminar flow assumptions is used for both
flow.
As mentioned before in Figure 1, four domains, which are inner flow domain (water flow), tube walls (copper),
fin walls (aluminum), and outer flow domain (air flow) are modelled. Between these domain, bonded contacts
have to be arranged for transferring heat, accurately. Generally, three different bonded contact are determined,
and the contacts are inner flow domain and tube walls coupling, tube walls and fin walls coupling, fin walls and
outer domain coupling.

51
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. The staggered tube bundle arrangements with transverse and longitudinal distances

For outer flow, velocity inlet is applied in left yz plane with specific velocity and temperature. Pressure outlet is
used in right yz plane with zero-gauge pressure. Symmetry boundary conditions are used for four planes (left
xy, right xy, bottom xz and top xz planes) For inner flow, velocity inlets are considered at inner flow areas on
left xy plane with specific velocity and temperatures. Pressure outlets are used at inner flow areas on right xy
plane with zero-gauge pressure. For inner and outer flow, at pressure outlet boundary conditions, outlet
temperatures are not given. Outlet temperatures are predicted by Ansys-Fluent via other boundary conditions.
For discretizing convective terms, a second-order upwinding scheme is used [12]. For pressure-velocity coupling
SIMPLE algorithm is used. As an under-relaxation factor, pressure is 0.3, momentum is 0.5 and energy is 0.5.
As the converge criteria, a residual value 10-6 for all equations except energy equation. For energy equation,
residual value 10-8 is used. An unstructured mesh strategy is applied. Triangular prisms are used for meshes.
The convective heat transfer coefficient of air (h) is calculated as;

q  mair C p  To , a  Ti , a   hAfin T fin  Tair  (1)

Nusselt number (Nu) is based on outer diameter of fin, and determined as;

hDo
Nu 
k (2)

Fin efficiency (ε) is ratio of heat transfer rate with fin to heat transfer rate without fin if the entire fin is at the
base temperature.

q
 (3)
qwo  fin

Pressure drop (ΔP) is calculated via pressure differences via inlet of outer domain and outlet of outer domain.
And the friction factor (f) is;

P
f  Do 0.5Vi ,2a (4)
L

For a combined assessment of the heat transfer enhancement and the associated pressure drop, namely, thermo-
hydraulic performance factor (η) suggested by Chang et al. [13] is used.

Nu Nuwo  fin

f f wo  fin 
13 (5)

4. RESULTS

Figure 3 shows the prediction of the (a) x-velocity (b) temperature distribution between two fins for 50×25
spacing for sine wave fins. The other fin types and spacing’s predictions are not shown here. The flow patterns

52
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

do not show a remarkable change. The maximum x-velocities are observed the between tube rows and negative
velocity zones occurs behind the tubes. One can see that flow patterns attain quite a repeating pattern behind the
first row. It is known that in flows around cylinders an unsteady vortex shedding occurs behind a cylinder, which
can attain different modes, depending on the Reynolds number. In the present case, due to presence of fins with
spacing 2.5mm, this unsteady motion is dampened. Beside, some unsteadiness can be observed near the top and
bottom symmetry planes.
Cold air gains the temperature via hot inner flow, similar in fin temperatures. In other words, air temperature
increases gradually because of hot inner flow. Again, there are no significant differences between air
temperatures of different fin types.

(a) (b)
Figure 3. Distribution of x-velocity between two fins of 50×25 for sine wave

Nusselt number for all cases are presented in Figure 4(a). One can see results, regardless the fin spacing, sharp
wave fin produce highest Nusselt number, and plain fin produce the lowest Nusselt number. Also, the maximum
Nusselt number occurs for 30×26 and 40×20 tube spacings. As mentioned before, tubes arranged as staggered,
and maximum Nusselt number is produced cases which small spacings. With increasing tube spacings, Nusselt
number decreases. Figure 4(b) represents the fin efficiency for all cases. Generally, regardless of tube spacing,
plain fins produce small fin efficiencies. However, two exceptions can be observed for 50×25 and 44×38.1.
These results are unexpected for us. Sharp wave produce high fin efficiencies in first four cases (50×25, 35×30,
30×26 and 40×34.64), however, in the other cases (40×24.66, 40×20, 32×28 and 44×38.1), sine wave fin produce
high fin efficiencies And the maximum fin efficiency observed at spacing of 44×38.1, after that 50×25 and
40×24.66 produce high efficiencies. The minimum fin efficiency is observed at 30×26 tube spacings. Friction
factor values for all cases are shown in Figure 4(c). Generally, regardless of the tube spacings, the minimum and
maximum friction factor is produced in plain fin and sharp-wave fin types, respectively. Because of sharp wave,
flow mixing better and it produces highest friction factor. While, tube spacing of 40×20 generate the maximum
friction factor, tube spacing of 44×38.1 minimum friction factor. Beside 40×34.66 tube spacing bring out the
less friction factors. Figure 4(d) shows the

(a) (b)

(c) (d)
Figure 4. (a) Nusselt number, (b) fin efficiency, (c) friction factor and (d) thermal performance factor for all cases

53
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

thermal performance factor for all cases. Generally, if we do not consider the tube spacings, plain fins have
greater thermal performance factor than the other fin (sharp wave and sine wave fin). This situation is not an
unexpected, because plain fin produces lowest friction factor ratio ( f f wo  fin ). However, there is one exception
for only 30×26 tube spacing. One can observe that; sine wave fin produce a little bit higher thermal performance
factor. All thermal performance factor is greater than one, therefore, using fin is suitable method for enhancing
convective heat transfer and decreasing pressure drop. While, tube spacings of 40×24.66 and 30×26 produce
high thermal performance factor, 50×25 tube spacing generate the minimum thermal performance factor.

5. CONCLUSIONS

This paper presents a three dimensional numerical simulation of fin-tube heat exchanger for three different fin
types, which are plain, sharp wave and sine wave fins, and eight different tube spacing combinations The effects
of fin types and tube spacing combinations on flow and heat transfer characteristics are investigated.
Due to presence of fins, unsteady flow is dampened. Irrespective of tube spacings, sharp wave fin and plain fin
generate the highest and lowest Nusselt number, respectively. And small spacings triggers to increase Nusselt
number. Plain fins generally produce small fin efficiencies regardless of tube spacings. While 30×26 tube
spacing produce minimum fin efficiency, the 44×38.1 tube spacing generate maximum fin efficiency. If we do
not consider the tube spacing, plain fin has a minimum friction factor, and sharp-wave fin has a maximum
friction factor. In sharp wave fin, flow is mixing better, so it produces maximum friction factors. Because of
miminum friction factor ratio, generally plain fin has maximum thermal performance factor, regardless of tube
spacings. Thermal performance factor of sine wave fin is a little bit greater than sharp wave fin. As thermal
performance factors are greater than unity, usage of fins proper for augmenting convection heat transfer and
diminishing pressure drops.

REFERENCES

[1] Wongwises, S, Chokeman, Y., Effect of fin pitch and number of tube rows on the air side performance of herringbone wavy
fin and tube heat exchangers. Energy Conversion and Management 2005; 46: 2216-2231.
https://doi.org/10.1016/j.enconman.2004.09.011
[2]Wen, MY, Ho CY., Heat-transfer enhancement in fin-and-tube heat exchanger with improved fin design. Applied Thermal
Engineering 2009; 29: 1050-1057. https://doi.org/10.1016/j.applthermaleng.2008.05.019.
[3] Okbaz, A, Pınarbaşı, A, Olcay, AB., Experimental investigation of effect of different tube row-numbers, fin pitches and
operating conditions on thermal and hydraulic performances of louvered and wavy finned heat exchangers. International
Journal of Thermal Sciences 2020; 151: 106256. https://doi.org/10.1016/j.ijthermalsci.2019.106256.
[4] Jang, JH, Chen LK., Numerical analysis of heat transfer and fluid flow in a three-dimensional wavy-fin and tube heat
exchanger. International Journal of Heat and Mass Transfer 1997; 40(16): 3981-3990. https://doi.org/10.1016/S0017-
9310(97)00047-1.
[5] Chu, WX, Sheu, WJ, Hsu, CC, Wang, CC., Airside performance of sinusoidal wavy fin-and-tube heat exchangers subject tı
large-diameter tubes with round or oval configuration. Applied Thermal Engineering 2020; 164: 114469.
https://doi.org/10.1016/j.applthermaleng.2019.114469.
[6] Sadeghianjahromi, A, Kheradmand S, Nemati, H, Wang CC., Heat transfer enhancement of wavy fin-and-tube heat exchangers
via innovative compound designs. International Journal of Thermal Sciences 2020; 149: 106221.
https://doi.org/10.1016/j.ijthermalsci.2019.106211.
[7] Tain, L, He, Y, Tao, Y, Tao, W., A comparative study on the air-side performance of wavy fin-and-tube heat exchanger with
punched delta winglets in staggered and in-line arrangements. International Journal of Thermal Sciences 2009; 48: 1765-
1776. https://doi.org/10.1016/j.ijthermalsci.2009.02.007.
[8] Hwang, SW, Kim, DH, Min, JK, Jeong, JH., CFD analysis of fin tube heat exchanger with a pair of delta winglet vortex
generator. Journal of Mechanical Science and Technology 2012; 26(9): 2949-2958.
https://link.springer.com/article/10.1007/s12206-012-0702-2.
[9] Sun, L, Zhang, CL., Evaluation of elliptical finned-tube-heat exchanger performance using CFD and response surface
methodology. International Journal of Thermal Sciences 2014; 75: 45-53. https://doi.org/10.1016/j.ijthermalsci.2013.07.021
[10] Zhao, L, Wang, B, Wang, R, Yang, Z., Aero-thermal behavior and performance optimization of rectangular finned elliptical
heat exchangers with different tube arrangements. International Journal of Heat and Mass Transfer 2019; 133: 1196-1218.
https://doi.org/10.1016/j.ijheatmasstransfer.2018.12.138.
[11] Huang, S, Wan, Z, Tang, Y., Manufacturing and single-phase thermal performance of an arc-shaped inner finned tube for
heat exchanger. Applied Thermal Engineering 2019; 159: 113817.
[12] Ansys Fluent 19.2 Theory Guide, Southpointe 2600 Ansys Drive, PA 15317: Ansys Inc, 2018
[13] Chang, SW, Su, LM, Yang, TL, Chiou, SF., Enhanced heat transfer of shaker bored piston cooling channel with twisted tape
insert. Heat Transfer Engineering 2007; 28(4): 321-334. https://doi.org/10.1080/01457630601122740

54
 TRANSESTERIFICATION OF WASTE COOKING OIL USING
BONE AND EGG SHELL MIXTURE AS A CATALYST

Ali A. Jazie*
Chemical Engineering Dep., Engineering College, University of Al-Qadisiyah, Al-Qadisiyah, Iraq ,
[email protected], ORCID: 0000-0002-0409-7001

Amar J. Albaaji
Materials Engineering Dep., Engineering College, University of Al-Qadisiyah, Al-Qadisiyah, Iraq ,
[email protected], ORCID: 0000-0002-3551-5974

Jazie Ali A., Albaaji Amar J.. Transesterification of Waste Cooking Oil Using Bone and Egg
Cite this paper as: Shell Mixture as a Catalyst. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,
Turkey

Abstract: Reuse of waste such as eggshells and animal bones as a catalyst to convert waste cooking oil into
biodiesel fuel were investigated. A mixture of bone and eggshell waste (50:50) was used as a
catalyst for the biodiesel production from waste cooking oil. Different calcination temperatures
had been investigated for the catalyst and the highest production of biodiesel was found at 900 ˚C.
The experimental results showed that a methanol /oil ratio of 15:1, a mixture of bone and eggshell
catalyst (900 ˚C, 2 h) of 12 wt%, a reaction temperature of 60 ˚C gave a biodiesel yield of 94% at
4 h. An effect of calcination temperature and catalyst leaching were examined. The catalysts were
characterized by X-ray diffraction (XRD), N2 adsorption-desorption, X-Ray Photoelectron
Spectroscopy (XPS) and Fourier transform infrared spectrometry (FTIR).
Keywords: Biodiesel, Waste cooking oil, Eggshell waste, Animal bone waste, Catalyst characterization.
© 2020 Published by ECRES

1. INTRODUCTION

As a consequence of increasing global demand for energy, many of the works and research focused on the
production of biodiesel using waste materials as an effective low-cost heterogeneous catalyst. Presently, the
yield of biodiesel using sodium or potassium hydroxide as a homogeneous catalyst is used efficiently
industrially. Although there is a need for further operations to remove these catalysts that are generating large
amounts of sewage and time consuming, making the production process is expensive [1]. In addition, the
presence of high impurities in the crude glycerol by-product, making it of low value. Several different processes
should be used to purify glycerin like filtration, vacuum distillation, deodorizing, and many others which they
are costly processes. In fact, there is a need for an integrated process for low-cost biodiesel production, high-
quality glycerol by-product using moderate conditions and environmentally friendly catalysts. Heterogeneous
catalysts are the appropriate technology to overcome these problems [2]. Many researchers attempted variety of
high-cost analytical reagent grade solid chemicals as a catalyst for biodiesel production [3] Also, many other
researchers reported recycling of wastes such as shell of chicken egg [4], oyster [5], mud crab [6], mollusk [7],
snail, and Venus [8], in addition to animal bone [9] as cost effective resources for the providing of a suitable
heterogeneous catalyst for the transesterification process. The benefits of using wastes in the biodiesel
production eliminate the wastes and lowering the cost of the process. Eggshells consist nearly of 10% of the
chicken egg mass, with approximate analysis of dry eggshell consist of (CaCO3: 85-95%, MgCO3: 1.4%,
phosphate: 0.76%, organic matter: 4%, in addition to trace amounts of copper, manganese, zinc, potassium, iron,
and sodium) [10]. Hydroxyapatite can be extracted from bones and many researchers used it as catalyst for
different chemical reactions because of it is catalytic activity like adsorption efficiency, the ability of the
functional group exchange, acid and base properties [11]. Hence, waste of animal bone also can be used as a
low cost source for the preparation of catalysts. So far no literature is available on the utilization of a mixture of
bone and egg shell wastes as a catalyst in the transesterification reaction. Therefore, detailed study of the
characteristics of the animal bone and egg shell and the catalytic activity together with the optimization of

55
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

process parameters for the transesterification of waste cooking oil is required to establish economic feasibility
and sustainability of the process. In our previous work, the activity of bone as a catalyst for the transesterification
of vegetable oil was investigated. In the present work, the two sources of bone and eggshell are characterized in
details by using X-Ray Photoelectron Spectroscopy (XPS), X-ray diffraction (XRD), N2 adsorption-desorption
for BET surface area and pore volume, and Fourier transform infrared spectrometry (FTIR). The suitability of
mixture of bone and eggshell as a catalyst in the transesterification of waste cooking oil is examined.

2. EXPERIMENTAL

Materials:
Waste cooking oil and egg shells were obtained from a local restaurant. Goat bones were obtained from a local
butcher. Initially, bones were cut into small pieces. For removing impurities and adherent tissue, small pieces of
bone were boiled at a temperature of 105 ˚C under normal atmospheric pressure and washed several times with
hot water in order to reach a complete removal of both fat and tissues. Moreover, small pieces of bone were
dried at 105 ˚C degrees for 24 hours. Finally, the small pieces of bones were ground to get a fine powder and
then the calculation process was done for the powder. As for the bone, egg shells have been washed several
times using a hot water in order to remove the organic material, then it follows the drying process for the shells
at 105 ˚C for 24 hours using a hot air oven. Finally, grinding process for the shells into a fine powder was done
before the calculation process. Both sources of catalysts were calcined at different temperatures in the range
(200˚C–1000˚C) for 2 hours in a muffle furnace. Methanol and Phosphoric acid were procured from Fisher
Scientific, India. Standards of 13-Octadecenoic acid methyl ester, Hexadecanoic acid methyl ester, Octadecanoic
acid methyl ester, 9,12-Octadecadienoic acid methyl ester, 9,12,15-Octadecatrie acid methyl ester were obtained
from Sigma-Aldrich.
Catalyst Examinations:
The elemental composition of the surface of catalyst, oxidation state and electronic environment of each
component was obtained using X-ray photoelectron spectroscopy technology (Shemadzu XPS Instrument,
Kratos Amicus model, Japan). A monochromated Al Kα source as the electron spectrometer was used. The
binding energy (BE) measure was referenced to the C1s line of aliphatic carbon, set at 285 eV. The precision of
BE determination is 0.1 eV. X-ray diffraction technology using (Philips 1710 X-ray diffractometer) was used
for the analysis of the crystalline phases of calcined samples and for the calculation of catalyst crystallite size.
The BET surface area, pore volume and pore size of the different samples were characterized using (N2
adsorption-desorption technology, Micromertics, ASAP 2020). Fourier transform infrared spectroscopy (FTIR)
technology using (Thermo-Nicolet 5700 models) was used for functional group's characterization in the 500–
4000 cm−1 region, with an average of 32 scans and a resolution of 4 cm−1.
Batch Reactor Arrangement:
Experiments were conducted for the production of bio-diesel using a glass three-neck reactor (500 ml) tied to a
reflux water-condenser and thermometer. The reaction mixture is heated in the reactor using a hot plate heater
and a water bath with thermostat for temperature stabilization. Experiments were conducted within the range of
30-70˚C. Mixing the reaction mixture with catalyst particles was done using a magnetic stirrer at a mixing rate
of 700 RPM for all tests.
Reaction:
Initially the waste cooking oil was filtered and heated to 105 ˚C for 1 h in a nitrogen-purge to remove water and
impurities and then cooled down to room temperature. Subsequently, the prepared mixture of catalyst in
methanol at a specified quantity was added to the oil. A number of tests were conducted in different conditions.
For each test run, after cooling the reaction mixture, leaving the mixture to separate into two distinguished layers.
The biodiesel product and traces of unconverted triglycerides were found in the upper layer. While, the lower
layer consisted of glycerol, all the catalyst amounts, excess methanol, traces of some methyl esters and soap.
The separation of the two layers was conducted by sedimentation and the biodiesel (upper layer) was washed
several times with water and then dried at 105 ˚C for 4 h in order to remove the water content from the product.
The catalyst was recovered by centrifugation and filtration from the lower layer in order to test the reusability
of catalyst.
Analysis and yield of FAME:
The gas chromatograph–mass spectrometry (GC–MS) was used for the analysis of the composition of FAME.
Mass analyzer (Shimadzu GC–MS system model QP 2010 PLUS with a Quadra pole) used in the analysis
process. The specification of the GC column where (the carrier gas: helium, length of the column: 29.4 m,

56
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

thickness: 0.25 mm, with film thickness: 0.25 μm). The upper layer (mainly FAME) in the separating funnel,
after being washed by water are dried with anhydrous sodium sulfate. Then, before being examined by GC-MS,
the samples (1mL aliquot) were taken from the reaction mixture and neutralized by adding 1 ml (5% diluted
phosphoric acid). Finally, the samples and the chemical standards were diluted in n-hexane then analyzed using
GC-MS.

3. FINDINGS AND DISCUSSION

XRD Findings:
After comparison with the diffraction standards (JCPDS), the XRD results for the catalysts derived from goat
bone and egg shell showed a high similarity with the diffraction standards for HA file and CaO file. Figure1
showed the XRD spectra for bone catalyst calcined at (800-1000˚C). For the calcined bone catalyst at the range
of calcination temperature (800-1000 ˚C), it was found that the catalyst calcined at 900˚C showed a sharper peak
which is a meaning of better crystallinity. The spectra for the calcined catalyst at 900˚C showed a good
agreement with the HA JCPDS file No. (09-0432) with a lattice parameter (a=b=0. 942 NM, c=0. 688 NM) and
no peaks found for impurities in the spectra for 900˚C. The structure of the catalyst calcined at 900˚C showed a
HA with a hexagonal structure. All the peaks in the spectra of calcined catalyst at 900˚C belongs to the standard
hydroxyapatite and showed that all the collagen and organic compounds have been eliminated from the goat
barn. The planer spacing and intensity for the spectra of calcined catalyst at 900˚C for 2 h were shown the
characteristics of HA and it is different from that for catalysts calcined at other calcination temperatures (800˚C,
1000˚C).
Also, the same results were concluded for the egg shell catalyst calcined at 900˚C and 1000 ˚C (Figure2). The
spectra for the catalyst calcined at 900˚C and 1000˚C showed a similar pattern to that of the standard CaO JCPDS
file No. (4-777) with a peak of pure CaO. While, the other catalysts calcined at other temperature showed a
different pattern from that one for the standard CaO. Many beaks were founded in the spectra of 800˚C (Figure2).
The peaks found in the spectra of 800˚C catalyst belongs to the CaCO3 (peak at 2θ=29.478˚), Ca(OH)2 (peaks
at 2θ = 14.772˚ and 17.8690˚) and CaO (peaks at 2θ=32.318˚, 37.455˚, and 53.930˚). The XRD spectra for the
uncalcined catalyst showed the characteristics of CaCO3 compound while for the catalyst calcined at 900 ˚C for
2 h showed the characteristics of pure Cao as shown in the present work and elsewhere in the literature [4]. The
structure of the catalyst calcined at 900˚C showed a cubic structure and can be concluded from the sharp peaks
in the XRD spectra. It is clear that the calcination process at 900 ˚C removed all the CaCO3 from the catalyst
and convert it to CaO and evolved CO2 [12]. Table. 1 showed the crystallite sizes for the different catalysts
calcined at different temperatures. The crystallite sizes were calculated using Scherrer's equation as explained
elseware [13]. The results of the crystallite sizes for the catalysts calcined at 900˚C showed reduction to 41 and
25 nm for bone and egg shell calcined catalysts respectively as shown in the table. 1.
FTIR Findings:
The FTIR Spectroscopy results for the calcined catalyst showed an agreement with the results of the XRD
results. For the bone catalyst calcined at a 900˚C, it is clear from the spectra that the peaks for OH and CaCO3
which belongs to the hydroxyapatite were found and support the XRD results as explained in our previous paper
[9]. For the egg shell catalyst, the spectra of catalyst calcined at 900˚C showed that CO3-2 peak was absent and
that means it is converted to CaO as explained in our previous paper [14]
Surface Area Findings:
The results of the surface area also support the XRD and FTIR results. Table.2 showed that the BET surface
area for the catalyst calcined at 900˚C were the highest (90 m2g-1 for bone catalyst, 60 m2g-1 for egg shell
catalyst). It is concluded from the BET surface area result that the decrease in the crystallite size increases the
surface area and increase the activity of the catalyst. Also the porosity of the catalyst at 900 ˚C increased due to
the carbonate decomposition by calcination process. The reduction in BET surface area at 1000˚C may be as a
result of high decrease of the unit cell [15,16]. The sintering process was clear in changing CaCO3 to CaO as
provided in the literature [17]. The result of XRD and BET surface area show that the crystallite size of the
catalysts affect the adsorption type. Where, the catalyst with large crystallite size reveals the type II adsorption
with a marginal hesterisis at p/p0> 0.9 and there is no noticeable rise at p/p0< 0.1. Moreover, the dominant pores
are the macropores (>50 nm) which belongs to inter-particle pores as can be seen in the case of fresh bone and
egg shell catalysts. While, the catalysts calcined at temperature range (800˚C-1000˚C) show a downward shift
of the hysteresis ending point as a result for the small inter-particle pore size (Tables 2 and 3).

57
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

■ CaO
▲ Ca(OH)2
□ □ HA 1000oC
1200
● CaCO3

□ □□
Intensity (a.u.)

□ □ □
■ ▲▲ ● ■ □ □
900□oC
600

800oC

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85
2 theta
Figure 1. XRD of the calcined bone catalyst at 800˚C, 900˚C and 1000˚C.

■ CaO
▲ CaCO3
■ □ Ca(OH)
1000
■ 1000oC
■ ■
■
□□ ▲ ■■
Intensity (a.u.)

500 900oC

800oC
0

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85
2 theta
Figure 1. XRD of the calcined egg shell catalyst at 800˚C, 900˚C and 1000˚C.

XPS Findings:
The survey and detailed XPS results for the bone catalyst fresh and calcined showed clear observation of
removing CaCO3 upon calcination at 900˚C as shown in figure3a-e. The different atomic ratio for Ca/P, Ca/O,

58
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

P/O and CInorg./Ca was calculated and showed in table 3. The peaks in the XPS spectra as expected were C,
Ca, P and O. Two types of C in the samples one is the inorganic carbon as impurity with atomic percentages
varying from 0.71-0.03 for the different samples as shown in table 3. Some of the inorganic carbon comes from
the atmosphere and adsorbed on the surface of the samples and called the adventitious carbon. The adsorbed
carbon can be used for binding energy calibration and is set by 285 eV in order to correct the sample charging
[18]. The intensities of spectral components is related to hydrocarbon chains at 285.0 eV, and to carbon singly
bound to oxygen at 286.4 eV. The presence of CO2 in the atmosphere and solutions is the main source for the
impurity as carbonate in the various calcium phosphates [19]. Also, the carbonate is a major part of the biological
minerals [20]. C 1s spectra (Fig.3b) display a noticeable decrease in the relative intensity of spectra, indicating
structural re-arrangements and also removing the carbon causing the reduction in Cinorg/Ca, Ca/O, Ca/P, and
P/O ratios, as given in Table 3. It can be concluded from the XPS results that the carbonate was removed by
calcination process at 900˚C and 1000˚C. However, the carbonate is obvious in the case of fresh and 800˚C
calcined samples and these results supports the results of XRD, FTIR spectroscopy, as discussed in the previous
sections. Moreover, BET surface area studies also show that the activity and surface area of catalyst increases
while removing the carbonate and this supports the XPS results.
The survey and detailed XPS of calcined egg shell catalysts and fresh egg shell samples are shown in figure 4a-
d. The different atomic ratio for Ca/O, and CInorg./Ca was calculated and showed in table 3. The peaks in the
XPS spectra as expected were C, Ca and O. (Figure 4a-d). The inorganic C impurity had atomic percentages
varying from 0.68-0 for different samples as shown in table 3. The carbon comes from carbonate was observed
in the fresh and 800˚C calcination samples while being absent in the catalyst calcined at 900˚C (Figure 4a-d). C
1s spectra (Fig.4b) display a noticeable decrease in the relative intensity of spectra, indicating structural re-
arrangements and also removing the carbon causing the reduction in Cinorg/Ca, and Ca/O ratios, as given in
Table 3. A similar scenario was concluded for the XPS result of the egg shell catalyst, the carbonate was removed
from calcination process at 900˚C and 1000˚C, while the carbonate is obvious in the case of fresh and 800˚C
calcination samples and these results also supports the results of XRD, FTIR spectroscopy, as discussed in the
previous sections. Moreover, BET surface area studies show that the activity and surface area of catalyst
increases while removing the carbonate and this also supports the XPS results.
Catalyst Loading and methanol /oil ratio:
The reaction between the waste cooking oil and methanol without catalyst was not producing any product.
Increasing the catalyst loading (50:50) increased the yield of biodiesel as shown in the Figure 5. The yield of
biodiesel increased with increasing the catalyst loading until reaching 12 wt%. Increasing the catalyst loading
more than 12wt% increased the viscosity of the reaction mixture and forming a slurry. The process changed to
the mass transfer rate determining step due to mass transport limitation and any increase in catalyst loading will
not change the yield of biodiesel but will increase the requirement for mixing energy. That means the optimum
catalyst loading was 12 wt%. Searching for the optimal methanol/oil ratio, the experiments were conducted from
3:1-20:1 as shown in the Figure 6. Increasing the ratio of methanol/oil until reaching the ratio of 15:1 increased
the yield of biodiesel. Increase the ratio will increase the methoxy species on the surface of the catalyst and shift
the equilibrium in the forward direction. Increasing the ratio more than 15:1 will increase the chance of
dissolving the glycerol in the methanol and reverse the equilibrium to the backward direction. The same scenario
was concluded in a previous literature [21, 22, 23].
Reaction Temperature and Time:
The reaction temperature was checked in the range (30-70 ˚C) and was found that 60˚C the optimum temperature
where the highest yield of biodiesel was obtained (94%) as shown in Figure7. The maximum biodiesel yield
was obtained after 4 h reaction time and the test runs was conducted for the range (0-6 h) as shown in the Figure
8.
Cycling of Catalyst:
The mixture of catalyst (50 egg shell: 50 bone) after the first cycle was recycled by centrifuge and washing with
methanol. The mixture catalyst was tested for ten times without losing the activity. After that the activity of the
catalyst started to reduce until 17 cycles where the yield of biodiesel becomes zero due to the change of structure
as shown in the Figure9.

59
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

O (Auger)
300000
O (1s)
Ca (2p) P (2s)
O (2s)
Ca (2s) C (1s)
P (2p)

o
200000 1000 C
Intensity

o
900 C
100000
o
800 C

0 Fresh

1200 1050 900 750 600 450 300 150 0

Binding energy (eV)
Figure 3a XPS survey spectrum for fresh and calcined Bone at 800 ˚C, 900 ˚C, 1000 ˚C, respectively

40000

1000oC
30000
900oC
Intensity

800oC
20000

Fresh
10000

276 278 280 282 284 286 288 290 292 294 296 298 300 302
Binding energy (eV)
Figure3b XPS spectra of C1s for fresh and calcined Bone at 800 ˚C, 900 ˚C, 1000 ˚C, respectively

60
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1000oC

40000 900oC
Intensity

800oC

20000
Fresh

340 342 344 346 348 350 352 354

Binding energy (eV)
Figure3c XPS spectra of Ca for fresh and calcined Bone at 800 ˚C, 900 ˚C, 1000 ˚C, respectively

150000

1000oC

100000
Intensity

900oC

50000 800oC

Fresh

0
524 526 528 530 532 534 536 538 540
Binding energy (eV)
Figure3d XPS spectra of O for fresh and calcined Bone at 800 ˚C,900 ˚C, 1000 ˚C, respectively

61
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

12000
1000oC

9000 900oC
Intensity

800oC

6000
Fresh

126 128 130 132 134 136

Binding energy (eV)
Figure3e XPS spectra of P for fresh and calcined bone catalysts at 800 ˚C,900 ˚C, 1000 ˚C, respectively

300000 O (1s)
o
Ca (2p) 1000 C

o
200000 900 C
Intensity

o
800 C
100000 C (1s)

Fresh
0

1200 1050 900 750 600 450 300 150 0

Binding energy (eV)
Figure 4a XPS survey spectrum for a fresh and calcined egg shell at 800 ˚C, 900 ˚C, 1000 ˚C, respectively

62
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

30000 o
1000 C

o
900 C
Intensity
20000
o
800 C

10000 Fresh

276 279 282 285 288 291 294 297 300 303
Binding energy (eV)

60000
Figure4b XPS spectra of C1s for a fresh and calcined egg shell at 800 ˚C, oC˚C, 1000 ˚C, respectively
900
1000

40000 900oC
Intensity

800oC
20000
Fresh

0
340 342 344 346 348 350 352 354 356
Binding energy (eV)

Figure4c XPS spectra of Ca for a fresh and calcined egg shell at 800 ˚C, 900 ˚C, 1000 ˚C, respectively

120000
1000oC

900oC
80000
Intensity

800oC

40000
Fresh

524 526 528 530 532 534 536 538 540

Binding energy (eV)
Figure4d XPS spectra of O for a fresh and calcined egg shell at 800 ˚C , 900 ˚C, 1000 ˚C, respectively

63
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

100
95
90

Biodiesel yield, %
85
80
75
70
65
60
55
50
3 8 13 18 23
Catalyst amount, %

Figure5 Effect of catalyst concentration.

100

90
Biodiesel yield, %

80

70

60

50
3 8 13 18 23
Methanol to oil ratio

Figure6 Effect of methanol to oil molar ratio.

100

90

80
Biodiesel yield, %

70

60

50
30 40 50 60 70 80
Temperature, ˚C

Figure7. Effect of reaction temperature.

64
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

100
90
80

Biodiesel yield, %
70
60
50
40
30
20
10
0
0 1 2 3 4 5 6
Time, h

Figure8 Effect of reaction time.

100

80
Biodiesel Yield, %

60

40

20

0
1 6 11 16 21
Cycle number

Figure9 Effect of reusability of the catalyst.

Table 1. Crystallite sizes of bone and eggshell catalysts calcined at 800 oC, 900 oC and 1000 oC, respectively
S. No. Raw Material Calcination Temperature (˚C ) Crystallites Size (nm)
1 Bone 800 57.25
2 Bone 900 41.47
3 Bone 1000 95.00
4 Eggshell 800 25.47
5 Eggshell 900 25.39
6 Eggshell 1000 29.42

Table 2 BET surface area (SBET), total pore volume and cumulative pore volume of the bone samples calcined at 800 oC,
900 oC and 1000 oC, respectively
Raw Calc. Surface Total Pore Cumulative Pore volume (Å)
Material Temp. Area Volume a<20 20<a<50 50<a<100 100<a<500 a>500
Bone 800 4.1 0.0164 0.0002 0.0012 0.0009 0.0095 0.0046
(oC) (m²/g) (cm³/g)
Bone 900 90.6 0.0625 0.0451 0.0121 0.0003 0.0025 0.0025
Bone 1000 1.2 0.0026 0 0.0002 0.0003 0.0011 0.0010
Egg shell 800 3.4 0.0146 0 0 0 0.0082 0.0064
Egg shell 900 58.2 0.1098 0.0062 0.0715 0.0097 0.0162 0.0062
Egg shell 1000 4.8 0.0185 0.0007 0.0014 0.0012 0.0109 0.0044

65
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 3 Ca/O, Ca/P, P/O and CInorg./Ca Ratios for fresh and calcined bone catalysts at 800 oC,900 oC, 1000 oC,
respectively
S. No. Raw Calcination Ca/O Ca/P P/O CInorg./Ca
1 Bone
Material Fresh (˚C )
Temperature 0.35
Ratio 1.58
Ratio 0.22
Ratio 0.71
Ratio
2 Bone 800 0.37 1.55 0.24 0.24
3 Bone 900 0.39 1.54 0.25 0.03
4 Bone 1000 0.41 1.54 0.27 0.03
5 Egg shell Fresh 0.36 - - 0.68
6 Egg shell 800 0.37 - - 0.12
7 Egg shell 900 0.41 - - 0
8 Egg shell 1000 0.42 - - 0

4. CONCLUSIONS

The catalysts calcined at temperature range (800˚C-1000˚C) show a downward shift of the hysteresis ending
point as a result for the small inter-particle pore size. It can be concluded from the XPS results that the carbonate
was removed by calcination process at 900˚C and 1000˚C. However, the carbonate is obvious in the case of
fresh and 800˚C calcined samples and these results supports the results of XRD, FTIR spectroscopy. Moreover,
BET surface area studies also show that the activity and surface area of catalyst increases while removing the
carbonate and this supports the XPS results. Using a mixture of bone and egg shell calcined at (900˚C for 2h) as
a catalyst for the transesterification of waste cooking oil showed a higher activity and making the process
economically efficient and ecofriendly. The catalyst can be used 10 times without regeneration. The yield of the
process was 94% at the optimum conditions (12 wt % catalyst, 15:1 methanol/oil ratio, 60 ˚C, 4 h). The mixture
catalyst can be used industrially in a large-scale continuous packed bed reactor. The mixture catalyst can be used
as a base catalyst for other base catalyzed organic reactions. Also in the process the waste cooking oil will be
recycled and reused for producing biodiesel.

ACKNOWLEDGMENT

Authors would lick to thank Faculty of Engineering/ University of Al Qadisiyah for providing the laboratory
requirements.

REFERENCES

[1] Al-Dawody, M. F., Jazie, A. A., & Abbas, H. A. (2019). Experimental and simulation study for the effect of waste cooking
oil methyl ester blended with diesel fuel on the performance and emissions of diesel engine. Alexandria Engineering
Journal, 58(1), 9-17.
[2] Jazie, A. A., Abed, S. A., Nuhma, M. J., & Mutar, M. A. (2019). Continuous biodiesel production in a packed bed reactor
from microalgae Chlorella sp. using DBSA catalyst. Engineering Science and Technology, an International Journal.
[3] Jazie, A. A. (2019). DBSA-Catalyzed Sewage Sludge Conversion into Biodiesel in a CSTR: RSM Optimization and RTD
Study. Journal of Engineering & Technological Sciences, 51(4).
[4] Wei, Z., Xu, C., & Li, B. (2009). Application of waste eggshell as low-cost solid catalyst for biodiesel production.
Bioresource technology, 100(11), 2883-2885.
[5] Nakatani, N., Takamori, H., Takeda, K., & Sakugawa, H. (2009). Transesterification of soybean oil using combusted oyster
shell waste as a catalyst. Bioresource Technology, 100(3), 1510-1513.
[6] Boey, P. L., Maniam, G. P., & Hamid, S. A. (2009). Biodiesel production via transesterification of palm olein using waste
mud crab (Scylla serrata) shell as a heterogeneous catalyst. Bioresource technology, 100(24), 6362-6368.
[7] Viriya-empikul, N., Krasae, P., Puttasawat, B., Yoosuk, B., Chollacoop, N., & Faungnawakij, K. (2010). Waste shells of
mollusk and egg as biodiesel production catalysts. Bioresource technology, 101(10), 3765-3767.
[8] Viriya-Empikul, N., Krasae, P., Nualpaeng, W., Yoosuk, B., & Faungnawakij, K. (2012). Biodiesel production over Ca-
based solid catalysts derived from industrial wastes. Fuel, 92(1), 239-244.
[9] Jazie, A. A., Pramanik, H., & Sinha, A. S. K. (2013). Transesterification of peanut and rapeseed oils using waste of animal
bone as cost effective catalyst. Materials for Renewable and Sustainable Energy, 2(2), 11.
[10] Chojnacka, K. (2005). Biosorption of Cr (III) ions by eggshells. Journal of Hazardous Materials, 121(1-3), 167-173.
[11] Mori, K., Yamaguchi, K., Hara, T., Mizugaki, T., Ebitani, K., & Kaneda, K. (2002). Controlled synthesis of
hydroxyapatite-supported palladium complexes as highly efficient heterogeneous catalysts. Journal of the American
Chemical Society, 124(39), 11572-11573.

66
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[12] Engin, B., Demirtaş, H., & Eken, M. (2006). Temperature effects on egg shells investigated by XRD, IR and ESR
techniques. Radiation Physics and Chemistry, 75(2), 268-277.
[13] Rheinheimer, W., Lowing, D., & Blendell, J. E. (2020). Grain growth in Nio–MgO and its dependence on faceting and the
equilibrium crystal shape. Scripta Materialia, 178, 236-239.
[14] Jazie, A. A., Pramanik, H., & Sinha, A. S. K. (2012). Egg Shell Waste-Catalyzed Transesterification of Mustard Oil:
Optimization Using Response Surface Methodology(RSM). International Proceedings of Computer Science and
Information Technology, 56(52), 2012.
[15] Borgwardt, R. H. (1985). Calcination kinetics and surface area of dispersed limestone particles. AIChE Journal, 31(1), 103-
111.
[16] Kouzu, M., Kasuno, T., Tajika, M., Sugimoto, Y., Yamanaka, S., & Hidaka, J. (2008). Calcium oxide as a solid base
catalyst for transesterification of soybean oil and its application to biodiesel production. Fuel, 87(12), 2798-2806.
[17] Boynton, R. S. (1980). Chemistry and Technology of Lime and Limestone. John Wylie & Sons. Inc., New York, 380-486.
[18] Kameshima, Y., Kuramochi, S., Yasumori, A., & Okada, K. (1998). Analysis of surface state and stability during storage of
AlN powders by x-ray photoelectron spectroscopy. Journal of the Ceramic Society of Japan, 106(1236), 749-753.
[19] Pal’chik, N. A., Moroz, T. N., Maksimova, N. V., & Dar’in, A. V. (2006). Mineral and microelement compositions of
urinary stones. Russian journal of inorganic chemistry, 51(7), 1098-1105.
[20] LeGeros, R. Z. (2002). Properties of osteoconductive biomaterials: calcium phosphates. Clinical Orthopaedics and Related
Research®, 395, 81-98.
[21] Jazie, A. A., Pramanik, H., & Sinha, A. S. K. (2013). Transesterification of peanut and rapeseed oils using waste of animal
bone as cost effective catalyst. Materials for Renewable and Sustainable Energy, 2(2), 11.
[22] Jazie, A. A., Alshebaney, E. Jadeen and Abed, S. A., "In-Situ Dodecylbenzenesulfonic acid -Catalyzed Transesterification
of Micro Algae Chlorella Sp. for Biodiesel Production," 2019 International Conference on Power Generation Systems and
Renewable Energy Technologies (PGSRET), Istanbul, Turkey, 2019.
[23] Jazie, A. A., Abed, S. A. and H. Pramanik, "DBSA-Catalyzed Biodiesel Production From Sewage Sludge In A Micro-
Reactor: Box-Behnken Design Optimization," 2019 International Conference on Power Generation Systems and Renewable
Energy Technologies (PGSRET), Istanbul, Turkey, 2019.

67
 INFLUENCE OF ELECTROLESS COATING OF NANO
REINFORCEMENTS ON MAGNETIC PROPERTIES OF FECO
ALLOY COMPOSITES
Amar J. Albaaji
Materials Engineering Dep., Engineering College, University of Al-Qadisiyah, Al-Qadisiyah, Iraq , :
[email protected], 0000-0002-3551-5974

Ali A. Jazie
Chemical Engineering Dep., Engineering College, University of Al-Qadisiyah, Al-Qadisiyah, Iraq ,
[email protected], 0000-0002-0409-7001

Albaaji Amar J, Jazie Ali A.,. Influence of electroless coating of nano reinforcements on
Cite this paper as: magnetic properties of FeCo alloy composites. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
2020, Istanbul, Turkey

Abstract: Graphene nanoplatelets (GNPs) and carbon nanotubes (CNTs) were coated by electroless
process, which were then used as reinforcement for FeCo alloy. Planetary ball milling
was used for dispersion 1 vol.% of coated reinforcement. The powder samples were
sintered by spark plasma sintering (SPS) machine. Magnetic tests were performed on the
sintered material in order to evaluate the magnetic properties. Optical microscope was
used to study the variations in microstructure of the sintered samples. The results show
that the magnetic properties for both GNPs composite and CNT composite were improved
after coating.
Keywords: Magnetic properties, Electroless coating , FeCo alloy, GNP,CNT.
© 2020 Published by ECRES

Introduction

More electric aircraft applications require alloy with high magnetic properties. Soft magnetic FeCo alloy
satisfies such applications, however the mechanical properties for the alloy need to improve. In order to enhance
the mechanical properties of FeCo alloy, alloying with third element was used [1, 2, 3]. However, the
improvement in mechanical properties was achieved, the magnetic properties were deteriorated. Newly,
graphene and carbon nanotube are used to develop the different alloys and show significant improvement in
properties. However, achieving uniform dispersion for such reinforcement is still challenge [4]. Moreover, retain
the nanostructure through the dispersion and sintering processes is very difficult. Spark plasma sintering is novel
sintering process with many benefits over other sintering processes, which enable to maintain nanostructure
during sintering process [5]. Recently, the mechanical properties of FeCo alloy have been enhanced by spark
plasma sintering alone alloy and by adding bare graphene or carbon nanotube or coated carbonaceous
nanomaterials with assistance spark plasma sintering [6, 7, 8]. This is attributed to maintain the nanostructure
thanks to spark plasma sintering process [9]. However, to the best of our knowledge there is no such study on
influence of coating GNP on magnetic properties of FeCo alloy. The main aim of this research is to study the
influence of coated reinforcements on the magnetic properties in FeCo alloy, which were showed an
improvement in mechanical properties in the alloy.

Experimental work

0.06 g of the functionalised CNTs or GNPs was sensitized by mixing in sensitisation solution for 30 min. The
sensitisation solution was prepared by mixing 3.39 g of SnCl2. H2O in 1.5 ml of HCl, and then the volume was
made up 150 ml by de-ionised water and stirred magnetically for 30 min. The sensitized particles were washed
thoroughly in de-ionised water, and then the carbonaceous nanomaterials were separated from water by
centrifugal machine, and were dried on hot plate at 100 °C for 2 h. The activation solution was prepared by
mixing 0.075 g of PdCl2 and 3.9 ml of HCl in 150 ml of de-ionised water by magnetic stirrer for 30 min [10].

68
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The sensitised particles were activated by mixing in activation solution using magnetic stirrer for 30 minutes at
room temperature, then the particles were washed thoroughly and dried as above-mentioned procedure.

The chemical composition of 300 ml Ni-P coating bath is presented in Table 1 [11]. The chemicals were mixed
in 300 ml of de-ionised water; the solution was magnetically stirred to completely dissolve the salts and to
achieve the desired pH value. Ultrasonic probe was used to stirring the sensitised and activated carbonaceous
nanomaterials in plating bath for 10 min. Then, the coated carbonaceous nanomaterials were extracted from bath
by centrifuge machine, thoroughly washed and dried.
Table 1. Chemical composition of coating bath.
Chemical Amount
Nickel sulphate (NiSO4.7H2O) 7.5 g
Sodium phosphate (NaH2PO2.H2O) 4.5 g
Tri-sodium citrate (Na3C6H5O7.H2O) 1.5 g
Ammonium chloride NH4Cl 18 g
Ammonium hydroxide (2.0 M ammonia to adjust pH to
water) 8.25
Temperature 25 ºC

The flaked FeCo alloy powder was prepared by planetary ball milling machine under argon atmosphere. Fifty
grams of FeCo alloy was ball milled at BPR of ~ 2:1 for 6 h, with 8 ml of isopropanol alcohol as process control
agent (PCA), for operating conditions of a milling speed of 300 rpm, to avoid the overheating during process a
15 min work and 10 min pause were selected. To prepare the composites, twenty grams of flaky FeCo powder
was mixed with the 1 vol.% of coated carbonaceous nanomaterials by ball milling at ball to powder ratio of 1:1
for 1 h under argon atmosphere and using 6 ml of isopropanol alcohol as a PCA. The used operation parameters
for ball milling were 400 rpm for 30 min work and 15 min pause.

Results and discussion

The morphology of GNP after coating is presented in Fig.1, which shows particles of coating layer on the surface
of GNP.

Fig.1. Scanning electron microscope (SEM) of Ni-P coating GNP.

Shrinkage curves

The shrinkage curves of as a received FeCo alloy, ball milled monolithic FeCo alloy and composites with and
without coating for reinforcements are shown in Figure 2. No significant effect can be observed for both bare
and coated carbonaceous nano reinforcements on shrinkage curves. The only different is for ball milled with
respect to as received FeCo alloy, this can be rationalised to improve densification due to ball milling [7].

69
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig.2. Shrinkage curves of the indicated materials.

Microstructure

The microstructure of sintered composites for bare and Ni-P coated carbonaceous nanoreinforcements were
compared with the microstructure of as received and ball milled monolithic FeCo alloy in Fig.3. It seems that
the flaked powder is more compatible in dispersion GNP rather the CNT in FeCo alloy. The microstructure of
GNP composite shows less agglomeration for both coated and bared reinforcement. The size of agglomerated
islands was increased in CNT composite after coating with Ni-P. This can be attributed to agglomeration for
CNT during electroless coating processes. Moreover, the intensity of ball milling was not sufficient to
breakdown the agglomeration of coated reinforcements. In comparison to monolithic FeCo alloy the structure
was refined for all composite, this may be due to the role of carbonaceous reinforcement in inhibiting grain
growth during sintering processes.

70
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig.3. Optical microstructure of: (a) as received FeCo alloy; (b) ) ball milled FeCo alloy at same conditions of composite
(c) and (d) 1 vol.% bare CNT-FeCo alloy composite and 1 vol.% bare GNP-FeCo alloy composite respectively (e) and (f)
1 vol.% Ni-P coated CNT-FeCo alloy composite and 1 vol.% Ni-P coated GNP-FeCo alloy composite respectively.

Magnetic properties

The upper hysteresis curves for the as received and ball milled FeCo alloy as compared to CNT and GNP
composites of bare and coated reinforcements are presented in Fig.4. A summary for magnetic properties is
presented in Table 2. In comparison to the as received FeCo alloy the saturation induction was decreased after
embedding carbonaceous nanoreinforcement in FeCo alloy for both coated and bared fibres. The Ni-P was very
effective in reducing the coercivity value especially in GNP composite.

71
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig.4. Upper-halves of hysteresis curves for the indicated materials.

The relative density was increased in 1 vol. % bare CNT-FeCo composite and 1 vol. % Ni-P coated CNT-FeCo
composite in comparison to ball milled monolithic FeCo alloy, this leads to an increase in saturation induction
in composites with respect to monolithic FeCo alloy. The relative density was decreased in counterpart GNP
composite in comparison to ball milled FeCo alloy, therefore the lowest value for saturation induction was
obtained in these composites. The coercivity decreased as the reinforcement was coated with Ni-P, this can be
due to the ferromagnetic behaviour of coating layer which reduce the influence of inclusion on the coercivity.
Nickel additions to FeCo alloy decline the degree of ordering as well as it is one of effective element in refining
the microstructure in FeCo alloy [3]. Stable precipitates are formed from addition Ni to FeCo alloy, the presence
of stable precipitates combines with refinement of grain or sub-grain size due to nickel additions are responsible
for the change in both mechanical and magnetic properties [2, 8]. The decrease in saturation induction of Ni-P
coated reinforcement may also be attributed to the lower saturation magnetisation of Ni coated layer compared
to the base FeCo alloy. The coercivity was decreased after coating the reinforcement, this is attributed to
ferromagnetic behaviour of Ni which is coated the inclusion [2].
Table 2. Magnetic properties and density of FeCo alloy and composite of bare and coated CNT and GNP.
Induction at 25 kA/m Coercivity Relative
Sample details
(T) (A/m) density (%)

As received FeCo alloy

2.41 381 98.99
Ball milled FeCo alloy 2.22 512 98.62
1 vol.% bare CNT-FeCo composite 2.24 989 100.39
1 vol.% Ni-P coated CNT-FeCo 2.27 685 99.09
composite
1 vol.% bare GNP-FeCo composite 2.08 1100 98.37
1 vol.% Ni-P coated GNP-FeCo 2.13 547 97.06
composite

CONCLUSION

1- The influence of reinforcement on saturation induction can be reduced by coating the GNP and CNT with Ni-
P.

2-The increased coercivity value from including CNT or GNP in FeCo alloy can be reduced by coating the
carbonaceous nanomaterials by Ni-P.

72
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3-The granular bonding is reduced due to the agglomeration for the coated carbonaceous nanomaterials at grain
boundaries, however it is effect on coercivity is not significant.

REFERENCES
[1] Weimer J. A., Electrical power technology for the more electric aircraft, AIAA/IEEE Digital Avionics Systems
Conference, Fort Worth, TX, USA, 1993, pp. 445-450.
[2] Sundar RS, Deevi SC: Int Mater Rev, 2005, 3, vol 50, pp.157-192.
[3] Sourmail, T: Progress in Materials Science, 2005, 7, vol 50, pp.816-880.
[4] Kuzumaki, T., Miyazawa, K., Ichinose, H. and Ito, K. Processing of carbon nanotube reinforced aluminum composite.
Journal of Materials Research, 1998, 13(09), pp.2445-2449.
[5] Hungría, T., Galy, J., & Castro, A. Spark Plasma Sintering as a Useful Technique to the Nanostructuration of Piezo-
Ferroelectric Materials. Advanced Engineering Materials. 2009, 11, 615-631.
[6] Albaaji A.J., Hashim A.M., & Thamir H.Y. Electroless coating of carbon nanotube/graphene nanoplatelets for
reinforcing FeCo alloy. Periodicals of Engineering and Natural Sciences. 2019,7, 1879-1885.
[7] Albaaji, A. J., Castle, E. G., Reece, M. J., Hall, J. P., & Evans, S. L.: Materials & Design, 2017, vol 122, pp. 296-306.
[8] Mani, M.K., Viola, G., Reece, M.J., Hall, J.P. and Evans, S.L.: Materials Science and Engineering: B, 2014, vol 188,
pp. 94-101.
[9] Albaaji, A.J., Castle, E.G., Reece, M.J. et al. Synthesis and properties of graphene and graphene/carbon nanotube-
reinforced soft magnetic FeCo alloy composites by spark plasma sintering.2016, Journal of Materials
Science, 51, 7624–7635.
[10] Ang, L.M., Hor, T.A., Xu, G.Q., Tung, C.H., Zhao, S.P. and Wang, J.L.: Carbon, 2000, 3, vol 38, pp.363-372.
[11] Kong, F.Z., Zhang, X.B., Xiong, W.Q., Liu, F., Huang, W.Z., Sun, Y.L., Tu, J.P. and Chen, X.W.: Surface and Coatings
Technology, 2002, 1, vol 155, pp.33-36.

73
 EFFECT OF OPTIMIZED INTAKE MANIFOLD GEOMETRY ON
BEHAVIORS AND EMISSION LEVEL OF H2CNG FUELED
ENGINE

Rafaa Saaidia
Bishah Technical College, Department of Mechanical Engineering, Bishah, Saudi Arabia, rafaasaaidia@gmail,
ORCID: 0000-0002-3892-9726

Raid Oyouni
Bishah Technical College, Department of Mechanical Engineering, Bishah, Saudi Arabia, [email protected],
ORCID: 0000-0001-5217-8327

Haider Almuslam
Bishah Technical College, Department of Mechanical Engineering, Bishah, Saudi Arabia, [email protected],
ORCID: 0000-0002-6343-6018

Saaidia, R, Oyouni, R, Almuslam, H. Effect of optimized intake manifold geometry on

Cite this paper as: behaviors and emission level of H2CNG fueled engine. 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: The study aims to identify the effect of optimized intake manifold geometry on behaviors and
emission level of H2CNG fueled engine. For this purpose, a commercial Hyundai Sonata S.I.E
was modified to operate with a CNG and Hydrogen blend. In this paper, the optimal intake pipes
lengths will be predicted by using the analytical acoustic method. A new intake manifold is
designed and realized while taking the advantage of natural supercharging that is managed by the
acoustic propagation of overpressure waves. A series of tests were conducted on the engine,
equipped with the new manifold, within the speed range of 1000 to 5000 rpm. For various engine
speeds, the variation of brake torque (BT), NOx and CO emissions were investigated for gasoline,
CNG and hydrogen and CNG blend (HCNG) fueled engines via external mixtures. Measurements
showed that the novel geometry to improve the in-cylinder pressure by 10% at 3500 rpm. In
addition, 14 % reduction in NOx and 40% reduction in HC for high engine speeds and 40%
reduction in CO for speeds below 2000 rpm.
Keywords: Gas engine; Intake manifold; Acoustic; Natural Supercharging
© 2020 Published by ECRES

Nomenclature
ICE Internal combustion engines
f Frequency of rotation (Hz)
fr Resonance frequency (Hz)
k Vibration harmonic
Kt Isothermal compressibility coefficient
KS Isentropic compressibility coefficient
Mmel Mixture mass (kg)
P* Relative overpressure (Pa)
r Mass stoichiometric ratio
tf Engine vibration period (s)
Z Specific acoustic impedance (Pa.s.m-1)
ψ Displacement of a gas slice (mm)
φ Air fuel ratio
CNG Compressed natural gas
H2CNG Compressed natural gas and hydrogen mixture

74
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION

The intake system in the internal combustion engines (ICE) which have a great influence in the performance
and emissions characteristics. The performance of an internal combustion engine, whether diesel or spark
ignition, two or four strokes, naturally aspirated or supercharged, is directly conditioned by the mass of air
introduced into the cylinder [1]. This mass of air determines the maximum fuel load that can be introduced into
the cylinders and therefore the amount of energy which the engine capable of supplying. This energy is
transformed, during the engine cycle, into mechanical energy on a shaft, but also into unburnt, exhaust losses
and heat losses [2].
The intake manifold is fundamental for the optimal internal combustion engine operation [3].
Given the importance of this element of the intake system, the intake manifold has been the subject of study by
several research works [4,5,6,7,8]. The aim of the improvement of the intake part, is to reduce the pressure losses
between the intake manifold inlet and outlet through each cylinder as much as possible [4].
In the ICE technics, the intake manifold is inserted between the throttle air-fuel blend or air pure and the cylinder
head on the side of the intake ports [1]. The purpose of the intake manifold design is to ensure a uniform air or
air-fuel mixture quantity distribution to the engine cylinders. Accordingly, the choice of an optimal geometry of
the intake manifold is essential to improve engine performances.
Generally, the rationale for designing an intake manifold involves to multifaceted synthesis methods which are
commonly based on numerical and experimental examinations [9-14]. The intake manifold design is built in
order to ensure the best engine performances while sweeping engine speed range. The intervention can be done
at the manifold at the end of optimizing the amount of air introduced into the cylinder. The supercharging is
obtained due to the phenomenon of the acoustic wave propagation in the intake pipe [15]. This optimization
remains obtained without any modification or additional system, named acoustic or natural supercharging [16-
17].
According to the literature review, there are two basically techniques which are used to model the behavior of
the intake line; pulsed quarter-wave resonance [18-20] and Helmholtz resonator [21-22]. the quarter-wave
resonance method of cylinder natural supercharging is directly related to the phenomenon of pressure waves
propagation that move according to the acoustic speed [18]. The existence of this phenomenon is linked to the
unsteady operation conditions of engines. The periodic opening, closing of the intake and exhaust valves, as
well as the pistons movement, are the sources of unsteady flows that give rise to the pressure waves. Thus, they
can promote the cylinder filling at the intake valves level by pushing maximum air-fuel fresh charge into the
combustion chamber. Margary et al. [23] carried an experimental study at started and rated engine speed. They
proved that volumetric efficiency was significantly influenced by the acoustic phenomena in the intake manifold.
An improvement of 9% is obtained in volumetric efficiency when the intake pipes length is chosen by the
quarter-wave method. Maftouni et al. [24] investigated on the intake manifold runner’s length effect on the
engine volumetric efficiency. Their analyses were performed using CFD simulation for different engine
operating conditions. They observed that the engine volumetric efficiency enhances considerably when they
used the manifold geometry with 20% extended runners. Gustavo et al. have carried a numerical and
experimental studies to investigate different geometries of intake manifolds with the aim of improving its
efficiency. The results showed that engine volumetric efficiency increases by 6% at 3500 rpm, consequently
increasing the indicated power by 3.68 kW (4.93 hp).
The research carried out in this area generally deals with the choice of an optimal geometry without considering
the composition of the fuel mixture in the intake pipes. Particularly when using unconventional fuels after engine
conversion still not well achieved.
Thus, in the present paper, the optimal intake pipes lengths will be predicted using the analytical acoustic
method. Using the technique of natural supercharging is based on the impedance method, and while fixing the
engine speed at the value that corresponds to the maximum torque, it has been verified that the collector is
existing on the engine does not ensure good filling of the cylinders. Then we designed a new collector which
takes advantage of the natural supercharging which is generated by the acoustic propagation of the overpressure
waves. to achieve our objectives, a manifold was designed and manufactured while taking into consideration the
case of engine operation in air / H2-CNG carburetion. Moreover, the study of the effect of hydrogen enrichment
with CNG on the engine filling presents a second part of novelty.
The length of the pipes depends on the nature of the mixture and the engine operating speed. First, we will
choose a single length which corresponds to the frequency of 3000 rpm and this choice is justified by the fact

75
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

that most of the engine operating time is around this frequency also for this engine rotation frequency we notice
a clear difference between the couples delivered for each proportion of the mixtures [25].
For these reasons, this study purported to determine the geometry which benefits from natural supercharging at
the turn of this frequency while the contribution and on the environmental and mechanical level. The ultimate
objectives were to compare the operation of the engine subject of study with the old and the new geometry of
the collector. the comparisons are performed in terms of mechanical performance and exhaust emissions (Co,
HC and Nox).

2. ENGINE STUDIED GEOMETRY

The studied engine model is a Hyundai Sonata. The main characteristics of this engine are illustrated
in Table.1. Due to the gaze’s proprieties as shown in Tables 2 and 3, the studied engine was modified
and converted into CNG then, it was adapted to HCNG operation. To perform a series of experiments,
a test bench was designed and installed. This test bench allowed to determine the behavior of the
experimental engine with different intake manifold design.

Table 1. Engine characteristics.

Engine Parameters Value
Engine type Hyundai Sonata
Displacement 1998 cm3
Compression ratio 10.4:1
Cylinders L4
Power 106 KW @ 6000 RPM
Fuel Supply System Electronic Injection
Fuel Gasoline

Table 2. Fuels properties.

Characteristics Gasoline CNG Hydrogen
Chemical formula CH4 to CH12 Mix. Of mainly 40% C3H8 and 60% C4H10 H2
Molecular weight 100–105 50 2.016 g/mol
Lower heating value (MJ/kg) 44.45 47.14 120.21
Fire point (°C) 230–500 ~500 490
Ignition limits (% of volume) 1.0–7.6 1.5–9.5 4–75
Octane number 80–95 95–111.5 130

Table 3. Hydrogen properties.

Hydrogen Proprety Value
Minimum ignition energy(mJ) 0.02
Flame temperature (°C) 2045
Auto ignition temperature (°C) 585
Maximum velocity of flame (m/s) 3.46
Explosion range (vol %) 13–65
Diffusion coefficient (10−3 m2/s) 0.61

The diagram presented in Figure.2 shows the apparatus which is used in the experimental system consisting of
an internal combustion engine, a hydraulic brake, an acquisition card, an exhaust gas analyzer, a gas/petrol
switch, an emulator, a lambda analyzer, and a computer.

76
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1- Engine; 2- Hydraulic brake; 3-Vaporiser; 4- Switcher; 5- Mixer; 6- Gas analyzer; 7-CassyLab card; 8-
computer

Figure 2. Bench Parts.

3. GAS MODELING IN INTAKE PIPES

The macroscopic variables which characterize the thermodynamic equilibrium state of a fluid assumed to be an
ideal gas are related by its equation of state: pV = nRT, p is the pressure, V the volume and T the temperature
of the fluid.

We define two types of compressibility coefficient, one isothermal and the other isentropic
1 𝜕𝑉 1 𝜕𝑉
𝐾𝑇 = − 𝑉 (𝜕𝑝 )𝑇 𝑒𝑡 𝐾𝑠 = − 𝑉 (𝜕𝑝 )𝑆 . (1)

Longitudinal waves in intake pipe

We denote by ψ (x, t) the displacement, at time t, of the particles of section AB, with respect to their rest position
A0B0 of abscissa x ψ (x + dx, t) is, at the same time t. the displacement of the particles of the section CD relative
to its rest position C0D0 of abscissa x + dx.
The thickness of the cylindrical slice ABCD is dx at rest and dx + ψ (x + dx, t) -ψ (x, t) at time t. Its mass is also
that of the slice A0B0C0D0, it is worth ρ.S.dx, where S denotes the surface of a section as shown in figure.2.

Figure 3. Intake pipe model

The equation of motion of the cylindrical slice ABCD is obtained by applying the fundamental law of dynamics
to the cylindrical slice ABCD. The forces to be considered are the occasional Foc, the pressure forces exerted
on the AB and CD bases. So we have the equation of movement is written:

77
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝜕²𝑝∗ 𝜕²𝑝∗ 𝜕²𝑢∗ 𝜕²𝑢∗

𝜕𝑥²
= 𝜌𝐾𝑠 ( 𝜕𝑡² ) 𝑒𝑡 𝜕𝑥²
= 𝜌𝐾𝑠 ( 𝜕𝑡² ) (2)

Resolution with the impedance method

The section of the pipe has finite dimensions, one brings back to this section, that is to say to the variation of
flow 𝐴∗ 𝑈 ∗ which produces there the acoustic wave, the response of medium at a pressure disturbance 𝑝∗ . Hence
the concept of characteristic acoustic impedance:
𝑝∗
𝑍= . (3)
𝐴𝑈∗

The disturbances created by the acoustic wave being due to the combined effects of two opposite simple waves
are therefore written:
𝑃∗ = 𝑓𝛪 (𝑐0 𝑡 − 𝑥) + 𝑓𝛱 (𝑐0 𝑡 + 𝑥)
{ ∗ 1 (4)
𝑈 = [𝑓𝛪 (𝑐0 𝑡 − 𝑥) − 𝑓𝛱 (𝑐0 𝑡 + 𝑥)]
𝜌0 𝑐0

These disturbances are functions of the abscissa x of the section where they are observed at time t. the
characteristic impedance is then expressed by:
𝜌0 𝑐0 𝑓𝛪 (𝑐0 𝑡−𝑥)+𝑓𝛱 (𝑐0 𝑡+𝑥)
𝑍(𝑥. 𝑡) = . (5)
𝐴 𝑓𝛪 (𝑐0 𝑡−𝑥)−𝑓𝛱 (𝑐0 𝑡+𝑥)

In the case of a harmonic acoustic wave, the characteristic impedance of a pipe is independent of time;
it only depends on the abscissa of the section where, at any time, the acoustic disturbance is observed. Indeed:

𝑃 ∗ = 𝐵𝛪 𝑒 𝑗(𝜔𝑡−𝑘0 𝑥) + 𝐵𝛱 𝑒 𝑗(𝜔𝑡+𝑘0𝑥)
{ ∗ 1 (6)
𝑈 = [𝐵𝛪 𝑒 𝑗(𝜔𝑡−𝑘0 𝑥) − 𝐵𝛱 𝑒 𝑗(𝜔𝑡+𝑘0𝑥) ]
𝜌0 𝑐0

So the characteristic impedance takes the following form:

𝜌0 𝑐0 𝐵𝛪 𝑒 −𝑗𝑘0 𝑥 +𝐵𝛱 𝑒 𝑗𝑘0 𝑥
𝑍𝑥 = . (7)
𝐴 𝐵𝛪 𝑒 −𝑗𝑘0 𝑥 −𝐵𝛱 𝑒 𝑗𝑘0 𝑥

Optimal intake length

In case of a harmonic acoustic wave, the characteristic impedance of a pipe, is time independent; it
depends only on the abscissa of the section. In this study, the intake pipe is considered as a pipe (with a length
equal to “l”) closed at the first extremity (x = 0) and open at the second (x = l). Since the boundary conditions
of this tube, a system of equations will be held:

-at abscissa x=0 the pipe is open so: p*( x 0 )  0  Z0  0 ,

- at abscissa x=l the pipe is close so: U *( x l )  0  Zl   .

Based on the above conditions, the length of the optimum pipe for having the acoustic supercharging
phenomenon is:
(2𝑘 + 1)𝑐0
𝑓𝑟 =
4𝐿

78
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

 c
 L 
𝑐0 𝑐0  4 fr
=
4(
𝐿
)
= 4𝐿
2𝑘+1
 (8)
2𝑘+1
L L
2 k 1 

 2k  1

k is the harmonic vibration, that is an integer.

fr is the frequency of propagation of the wave of which it can be determined from the frequency of opening and
closing of the valve which is for its part a function of the engine speed (position on the angle of the crankshaft)
𝑚
We denote by r the mass stoichiometric ratio, defined as follows: 𝑟 = 𝑚𝑎
𝑐

𝑉𝑐
And by R ’, the following stoichiometric volume ratio: 𝑅′ = 𝑉
𝑎
1 1+𝑅 ′
We find that the density of the mixture is expressed by 𝜌𝑚é𝑙
=𝜌 ′
𝑐 ×𝑅 +𝜌𝑎
The specific heat at constant volume cv and the specific heat at constant pressure cp obey the laws of mixing.
They are determined by means of the elementary heat (cvi, cpi) of the various constituents and the quantities yi
(masses or volumes) of the latter, by the following formulas:

∑𝑖 𝑥𝑖 ×𝑐𝑣𝑖 ∑𝑖 𝑥𝑖 ×𝑐𝑝
𝑖
𝑐𝑣𝑚é𝑙 = ∑𝑖 𝑥𝑖
𝑎𝑛𝑑 𝑐𝑝𝑚é𝑙 = ∑𝑖 𝑥𝑖
(9)
The optimum length can be written in the form:
- When operating with gasoline:
1 1
𝑐𝑝 +𝜑×𝑟×𝑐𝑝 2 (𝑅𝑇)
𝐿 = ( 𝑐𝑣𝑒+𝜑×𝑟×𝑐𝑣𝑎) × ( 4𝑓 2 ) (10)
𝑒 𝑎 𝑟
-
- When operating with a gas mixture:
1
𝑦 𝑦 2 1
𝑐𝑝𝐶𝑁𝐺 +𝑐𝑝𝐻2 +𝜑×𝜌𝑎 ×𝑅 ′ ×( 𝐶𝑁𝐺 + 𝐻2 )×𝑐𝑝𝑎
𝜌𝐶𝑁𝐺 𝜌𝐻2 (𝑅𝑇)
𝐿=( 𝑦 𝑦 ) × ( 2) (11)
𝑐𝑣𝐶𝑁𝐺 +𝑐𝑝𝐶𝑁𝐺 +𝜑×𝜌𝑎 ×𝑅 ×( 𝐶𝑁𝐺 + 𝐻2 )×𝑐𝑣𝑎
′ 4𝑓 𝑟
𝜌𝐶𝑁𝐺 𝜌𝐻2

The formulas (10) and (11) allow us to calculate the optimum length of the tubing when operating with gasoline
and gas mixtures of CNG and hydrogen. This length depends on the two variables, the richness and the frequency.

AFR impact on optimal pipe length

First, we will try to identify the influence of the richness of the mixture on the length value of the mixture. From
the Eqs.[10-11] while fixing the engine at a speed well determine the only variable on which the length depends
and the richness (φ) by making a sweep of s richness level we can draw the curve presented in the figure 4. The
curve below shows the variation of the length of the tubing according to the variation of the richness and it can
be concluded that in the case of operation with gasoline the richness has no influence on the length. The richness
of the gasoline-air fuel mixture has a negligible influence on the length, on the contrary that of hydrogen-air as
shown in figure 4 can influence but not strongly and we can conclude that in the case of operation with hydrogen
and is in conditions or the mixture does not respect the stichometry the performance of the engine may deteriorate

79
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 4. Optimal pipe length variation versus AFR (gasoline and CNG40H2).

It is noted that the composition of the mixture can intervene in the choice of geometry, if the machine makes a
wide sweep of richness level during operation. The influence of this richness is clear in the range of low levels of
richness but if the mixture is in the turn of a unit of level of richness the influence is not important enough which
can count an advantage for the dimensioning pipes for our engine as long as a fixed geometry can be accepted
according to these criteria.

Engine speed impact on optimal pipe length

The intake length clearly depends on the engine speed for the two fuels as shown in the figures 5.

Figure 5. Optimal pipe length variation versus engine speed (gasoline and CNG40H2).

For very high speeds the lengths for hydrogen and petrol are close to values but for low speeds a difference in
geometry is clear and we cannot accept a common length for the two fuels on the contrary for large speeds like
shows the following curve which illustrates the difference in length for the two fuels according to the engine
speed. For the constraints just mentioned, it is obvious to design a geometry that respects them.

80
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. EXPERIMENTAL VALIDATION OF THE OPTIMIZED INTAKE MANIFOLD

GEOMETRY

First, we will choose a single length which corresponds to the frequency of 3000rpm and this choice is
justified by the fact that most of the engine operating time is around this frequency also, for this engine
rotation frequency we notice a clear difference between the couples delivered for each proportion of the
mixtures.
It is interesting to determine the geometry which benefits from natural supercharging at the turn of this
frequency while the contribution and economically and mechanically.

rpm
Figure 6. BT versus engine speed.

In our case the frequency which delivers the maximum torque is worth 3600 rpm which requires the
optimum lengths, for a unit richness level, as follows:

Tableau 4. Optimal intake pipe length

Harmonique k 0 1 2 3
Gasoline 2100 423 299 213
CNG-40H2 2090 418 298 213

Given the properties of hydrogen, particularly its low self-ignition energy, and to cushion the anomalies
relating to the friction of hydrogen with hot walls, a second solution has been proposed. The new
geometry has a cooling circuit as shown in the following figure.

81
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 7. cooled Intake manifold with different pipes lengths.

Effect of the optimized intake geometry on engine behavior

In this part we will deal with the behavior of the engine being studied with the new manifold which has
been designed and manufactured while respecting the conditions which have been mentioned above.
The manifold is mounted on the engine as shown in Figure 8. The collector is cooled by the engine
cooling water. For better cooling efficiency, the coolant inlet port at the manifold is directly coupled to
the radiator outlet line. The water flowing through the collector is discharged at the inlet of the pump.

3
2

1- Engine; 2- New intake manifold; 3-mixer Air/CNG-H2;

Figure 8. Banc d’essai avec le nouveau collecteur refroidi.

After the completion and adjustment of all elements of the test bench, a series of tests was carried out
in order to characterize the performance and emissions of the engine with the new manifold.

The volumetric efficiency is a parameter indicative of the performance of the engine. In addition, this
yield is totally linked to the quantity of charge which fills the cylinder. The higher air and fuel load in
the cylinder, the higher the pressure. Therefore, we will measure the maximum pressure in the cylinder,
when the piston reaches top dead center (TDC), for the old geometry of the manifold and compare it
with the new. The results are given and represented by the following histogram.
Figure 9 shows the maximum pressure in the cylinder for different engine speeds.

82
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 9. In-cylinder pressure at TBC.

From figure 9, we notice two important and main things. Firstly, it is clear to see that the corresponding
pressure to the optimized geometry of the new manifold (NG) is higher than the old geometry (OG) for
all engine speeds.

Secondly, it is noted that this difference is slight for the regimes far from that chosen for the
dimensioning of the pipes, which is a consequence of the pipes, which have become shorter. In fact, the
shorter pipes, the less friction and pressure drop will take place, which justifies the increase in pressure
for the new geometry. It should be noted that the pressure difference is maximum for the design regime,
which justifies the contribution generated by natural overeating.

Effects of the optimized intake manifold on engine emissions.

In this part we are interested in the engine behavior in terms of emissions while comparing that of the
old geometry with the new one. A series of tests is carried out with a mixture containing 40% of H2
with CNG (GNC/40H2). The three exhaust gases studied are NOx, CH HC and CO. Figures 10, 11 and
12 show the evolution of the emission as a function of different engine speeds for the old collector
(OG) and the new one (NG).

Figure 10. NOX concentration in function of engine speed.

83
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

We note that the NOx emission is practically identical for the geometries for regimes below 3200
rpm. This quasi equality is explained by the fact that combustion takes place at almost the same
temperature since it is the most favorable factor for the formation of NOx. Above 3200 rpm, we
notice that the new collector emits less NOx particles. This decrease is explained by the good
ventilation of the cylinders.

Figure 11. CO concentration in function of engine speed.

The previous figure illustrates the evolution of Co particles, as a function of the rotation frequencies.
The curve has two essential areas. The first zone includes the regimes lower than 3200 rpm in which
we notice that the new collector emits less Co. This is explained by the fact that the new collector
makes penetrate a better quantity of air than the old collector since the main factor responsible for the
formation of Co is the insufficient quantity of oxygen.
At high speeds the amount of CO has clearly decreased and is practically zero because the combustion
of hydrogen in this speed range is more stable and complete due to the physical properties of
combustion of hydrogen, particularly its flame propagation speed.

Figure 12. HC concentration in function of engine speed..

The existence of CH gases is almost linked to the insufficient air in the combustion chambers. Figure
12 shows that at low speeds the concentration of HC particles is high, is identical for the two geometries
of the collectors given the low speed of admission which generates a considerable effect of the pressure

84
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

drops. Also, the existence of hydrogen has an unfavorable role since at low speeds its incomplete
combustion as was discussed in the second chapter. On the other hand, at the high speeds, we notice a
gain in terms of HC particle emissions which results from the good filling provided by the optimized
collector.

5. CONCLUSION

An experimental study that has been carried out and dealt with the effect of the new manifold design
on engine behavior to achieve a better filling rate. By using the technique of natural supercharging
based on the impedance method, and while fixing the engine speed at the value which corresponds to
the maximum torque, it has been verified that the collector existing on the engine does not ensure good
filling of the cylinders. Then we designed a new collector which takes advantage of natural
supercharging which is managed by the acoustic propagation of overpressure waves. This study
confirms the choice of the length of the intake manifold pipes.

The collector chosen is dimensioned, so that the filling is better. Also, with this solution, the amount
of residual fuel in the pipes can be reduced and the problems of backlash and self-ignition can be
reduced in the case of operation with hydrogen.

The cooling of the collector seems beneficial and advantageous in terms of performance and safety
since the incoming load becomes less exposed to overheating from friction with the walls.

The new optimized collector offers a less polluting emission which improves combustion.

REFERENCES

[1] J.B. Heywood, Internal combustion engines fundamentals. New York: McGraw Hill, 1988.
[2] Sharma Vinod Kumar, Mohan Man, Mouli Chandra. Effect of intake swirl on the performance of single cylinder direct
injection diesel engine. IOP Conf. Series Materials Science and Engineering. 2017. 062077.
[3] Indira Priyadarsini Ch. Flow analysis of intake manifold using computational fluid dynamics. Int J Eng Adv Res Technol
(IJEART) 2016;2(1). ISSN: 2454-9290.
[4] Chaubey Abhishek, Prof AC, Tiwari.. Design and CFD analysis of the intake manifold for the Suzuki G13bb engine.
Int J Res Appl Sci Eng Technol (IJRASET) 2017;5(VI). ISSN: 2321-9653.
[5] G.K. Giannakopoulos, C.E. Frouzakis, K. Boulouchos, P.F. Fischer, A.G. Tomboulides. Direct numerical simulation of
the flow in the intake pipe of an internal combustion engine. Inter. J. of Heat and Fluid Flow. 68 (2017) 257-268.
[6] J. Benajes, E. Reyes, J. Galindo, J. Peidro, Predesign Model for Intake Manifolds in Internal Combustion Engines,
Society of Automotive Engineers, Inc., Paper SAE 970055, 1997
[7] D. Chalet, A. Mahe, J. Migaud, J.F. Hetet, A frequency modelling of the pressure waves in the inlet manifold of internal
combustion engine, Appl. Energy. 88 (2011) 2988–2994.
[8] D.E. Winterbone, R.J. Pearson, Design techniques for engine manifold: Wave action methods for IC engines, Society
of Automotive Engineers, Inc., Warrendale, 1999.
[9] Costa RC, Hanriot, SM & Sodré JR. Influence of intake pipe length and diameter on the performance of a spark ignition
engine. J Braz. Soc. Mech. Sci. Eng. 2014; 36: 29-35 https://doi.org/10.1007/s40430-013-0074-2.
[10] Royo R, Corberán J, and Pérez A. Optimal Design of the Intake System in 4-S I.C.E. SAE Technical Paper 940210,
1994.
https://doi.org/10.4271/940210.
[11] Sadeq AM, Bassiony MA, Elbashir AM, Ahmed SF, Khraisheh M. Combustion and emissions of a diesel engine
utilizing novel intake manifold designs and running on alternative fuels. Fuel 2019; 255: 115769.
https://doi.org/10.1016/j.fuel.2019.115769.
[12] Hennings Och S, Moura LM, Mariani VC, Coelho LDS, Velásquez JA, Domingues E. Volumetric efficiency
optimization of a single-cylinder D.I. diesel engine using differential evolution algorithm. Applied Thermal Engineering
2016; 108: 660–669.
https://doi.org/10.1016/j.applthermaleng.2016.07.042.25.
[13] Vaz J, Machado A, Martinuzzi R and Martins M. Design and Manufacture of a Formula SAE Variable Intake Manifold.
SAE Technical Paper 2017-36-0181, 2017.

85
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

https://doi.org/10.4271/2017-36-0181.
[14] Gocmen K, Soyhan HS. An intake manifold geometry for enhancement of pressure drop in a diesel engine. Fuel 2020;
261: 116193.
https://doi.org/10.1016/j.fuel.2019.116193
[15] M.F. Harrison, I. De Soto, P.L. Rubio Unzueta, A linear acoustic model for multicylinder IC engine intake manifolds
including the effects of the intake throttle, J. Sound Vib. 278 (2004) 975–1011
[16] Hadjkacem, S., Jemni, M.A. & Abid, M.S. Volumetric Efficiency Optimization of Manifold with Variable Geometry
Using Acoustic Vibration for Intake Manifold with Variable Geometry in Case of LPG-Enriched Hydrogen Engine. Arab
J Sci Eng 44, 731–738 (2019). https://doi.org/10.1007/s13369-018-3254-7
[17] Hiereth H, Prenninger P. Charging the Internal Combustion Engine. ISBN 978-3-211-33033-3, Springer, Wien, New
York; 2007.
[18] Borel M. Les phénomènes d’ondes dans les moteurs. Publications de l’Institut Français du Pétrole, éditions Technip,
France ; 2000.
[19] Desmond EW, Richard JP. Theory of Engine Manifold Design: Wave Action Methods for IC Engines. London :
Professional Engineering Pub. ISBN: 1860582095;2000.
[20] Jemni MA, Kantchev G, Abid MS. Influence of intake manifold design on in-cylinder flow and engine performances
in a bus diesel engine converted to LPG gas fuelled, using CFD analyses and experimental investigations. Energy 2011;
36: 2701-2715. https://doi.org/10.1016/j.energy.2011.02.011
[21] Samuel J, NS P, and Annamalai K. Effect of Variable Length Intake Manifold on a Turbocharged Multi-Cylinder
Diesel Engine. SAE Technical Paper 2013-01-2756, 2013. https://doi.org/10.4271/2013-01-2756.
[22] Bortoluzzi D, Cossalter V, and Doria A. The Effect of Tunable Resonators on the Volumetric Efficiency of an Engine.
SAE Technical Paper 983045, 1998.
https://doi.org/10.4271/983045.
[23] Margary R, Nino E and Vafidis C. The Effect of Intake Duct Length on the In-Cylinder Air Motion in a Motored
Diesel Engine. SAE Technical Paper 900057, 1990.
https://doi.org/10.4271/900057.
[24] Maftouni N, Ebrahimi R, and Hossein pour S, "The effect of Intake Manifold Runners Length on the Volumetric
Efficiency by 3-D CFD Model," SAE Technical Paper 2006-32-0118, 2006.
https://doi.org/10.4271/2006-32-0118.
[25] Saaidia R, Jemni MA, Abid MS. Simulation and Empirical Studies of the Commercial SI Engine Performance and Its
Emission Levels When Running on a CNG and Hydrogen Blend. Energies 2018; 11, 29; doi:10.3390/en11010029.

86
 AN FDTD-BASED SIMULATION STUDY ON DISTINGUISHING
CHARACTERISTICS OF INVERTER SWITCHING AND PULSE
FAULT DURING SUB-CYCLE TRANSIENT PERIOD

Guomin Ji
School of Elec. and Comp. Engr., The University of Oklahoma, Norman, Oklahoma, USA, [email protected],
ORCID: 0000-0001-9779-1723

Sudantha Perera, Yihru Huang, Yan Zhang

School of Elec. and Comp. Engr., The University of Oklahoma, Norman, Oklahoma, USA, [email protected],
ORCID: 0000-0002-0170-5112, 0000-0003-0868-2061, 0000-0002-3533-3484

Di Wu
Department of Elec. and Comp. Engr., North Dakota State University, Fargo, North Dakota, USA, [email protected],
ORCID: 0000-0002-4455-1504

Ji, G, Perera, S, Huang Y, Wu D. A FDTD-based Simulation Study on Distinguishing

Cite this paper as: Characteristics of Inverter Switching and Pulse Fault During Sub-cycle Transient Period. 8th
Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In this paper, we present a preliminary investigation on the characteristics of sub-transients induced
by fault and the switching transients of inverter-based resources (IBRs) from a physics perspective.
Instead of using the fundamental frequency-dependent circuit-based models, the sub-transients are
analyzed based on Maxwell's equations, which allows us to cover a much smaller time range, treat
nonlinear material properties and break the transients into the sub-cycle-or-less time frame.
Technically, Finite-Difference Time-Domain (FDTD) method is used to generate transient waves
in two domains for a fault circuit: spatial propagation of energy flow along the transmission line
and spot time evolution at any location of the transmission line. It is discovered that the energy
flow of the spatial propagation induced by incidental IBR switching could be very similar to the
one produced by fault pulse, but the spot time evolutions at each location of the transmission line
are distinctively different between fault pulse and incidental IBR switching during the sub-
transient period. These findings may be helpful for fault analysis and relay setting of power grids
with high penetration of IBRs.
Keywords: Sub-transients, electromagnetic wave propagation, finite-difference time-domain (FDTD) method,
inverter-based resources (IBRs), transient stability
© 2020 Published by ECRES

Nomenclature
IBRs inverter-based resources
FDTD Finite-Difference Time-Domain
NERC North American Electric Reliability Corporation
SNL Sandia National Laboratories
AVR automatic voltage regulators
CFL Courant-Friedrichs-Lewy

1. INTRODUCTION

Many bulk power grids that have a high penetration of utility-scale Inverter-Based Resources (IBRs) have
experienced difficulties in keeping IBRs online from unnecessary interruptions during the fault events. For
example, in Southern California, there are three major interruptions of solar PV generation at a thousand-MW-
level in 2016, 2017, and 2018 in a row. For these events, the North American Electric Reliability Corporation
(NERC) has assembled a joint task force consisting of experts in power generation, manufacturers, system
operators, and researchers to investigate the causes of the interruption and explore the solutions. In the recent
investigation reports [i][ii][iii], the task force has identified that the current methods have failed to predict responses

87
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

by protection and control systems of power grids and PV farms during the transient period of sequential or
cascading complex fault events. Particularly, most of the interruptions of PV generation were rooted in sub-
transient (sub-cycle) complexity.
In Fig. [1], we plot the time frames of the types of transient following a disturbance based on the definitions and
categorization of the joint task force of CIGRE Study Committee 38 and IEEE Power System Dynamic
Performance Committee provided [iv], as well as the overview of major models and methods for system stability
analysis at each time frame made by Sandia National Laboratories (SNL) [v]. Specifically, the time frame of fault
transient has three-time frames: 1) Sub-transient period, which lasts a few cycles after the fault, characterized
with very high but fast-falling; 2) Transient period which occurs after sub-transient period, featuring large
current that falls at a slower rate; 3) Steady-state period, in which current reaches its steady value. Particularly,
a sub-cycle transient is referred to as the electromagnetic process in the first cycle or so of the AC energy system
(16ms@60Hz).
Transient stability has been traditionally addressed focusing on the stability issues in the transient period by
controlling the electromagnetic torque of synchronous generators with fast, automatic voltage regulators (AVR)
[vi]
. The goal is either to improve transient stability, as presented in [vii], or to enhance the fault-ride-through
capability of electric machines, e.g. [viii].
It has been realized that most of the sub-transient analyses are for transient issues in conventional power systems
with low penetration of IBRs, and the impact of grid-connected IBRs has not been adequately considered as
stated in recent investigations, e.g., [ix] as well as emphasized by NREC. Indeed, despite the fact that the works
have been done based on fundamental-frequency-based transient models, few research studies have been
reported to investigate the sub-transient characteristics of IBRs and faults from a more fundamental perspective.
Specifically, in Section II, the challenge to sub-cycle transient stability analysis is firstly discussed through an
overview of the current methods and the fundamental limitations, followed by the presentation of the physics
models and derivations of the one-dimension finite-difference time-domain (FDTD) algorithm used in this study.
Moreover, the equation of spatial propagation tracing Poynting energy flow along the transmission line is
derived for numerical computation. In Section 3, the setup of the simulation-based experiment using the models
and algorithms derived is described, as well as the results of the investigation. Discoveries and concluding
remarks are provided in Section IV.

2. METHODOLOGY DESCRIPTION

In the investigation presented in this paper, the study of the sub-transient characteristics is based on
electrodynamics described by Maxwell's equations. Analysis from a physics perspective not only allows us to
avoid the limitation of fundamental-frequency-based circuit models, but the findings are also more general.
Technically, we use the Finite-Difference Time-Domain method introduced by Yee [x] for modeling and
computing electromagnetic waves to cover a wide time range, including the sub-transient (sub-cycle) period.

Limitation Of Current Methods:

Sub-transient is an ultra-short but complex process in an AC power grid (50/60Hz). Its characteristics cannot be
described as precisely as the later dynamics in the transient period marked in Fig. [1]. Current methods of fault
transient analysis are based on the classic electric circuit theories and models that are valid only for the period
after a stable fundamental frequency (of 50/60Hz) is established or recovered, thus, the model validation, and
state estimation theoretically require the information of at least one-cycle following a disturbance, to ensure an
appropriate response of the protection system.
The assumption on a stable estimation of the fundamental frequency is the modeling bottleneck for sub-cycle
transient analysis. Stable establishment or recovery of the fundamental frequency is critical in that, 1) it serves
as the reference frequency that defines the harmonics and noises used in protection models and filters; 2) it
consolidates the building block (e.g., operating inductance 𝐿 = 𝜆𝑡 /𝑖𝑡 ) for the normal working range of the AC
energy system.
In practice, the sub-cycle transient associated issues have been avoided by various over-sizing strategies such as
the requirement of having fault-ride-through capability or stability margins, which allows the fault protection
and stability control systems to gain an adequate amount of time to reduce the likelihood of misfires.
Today, most of the conventional methods for transient analysis rely on the validation of electric circuit models
under the fundamental frequency of 50/60Hz, including symmetrical components, traveling-wave-based EMTP,
etc. With the century-long developments and experiences, these tools seemed to work very well for most of the

88
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

stability problems for transitional power systems, except some complex fault transient or lightning strokes.
Transitional power systems are made of electric machines, transformers, and other conventional components.
Unlike the IBRs made of solid-state materials, the conventional components can ride-through faults for multiple
cycles or even longer.
Unfortunately, the conventional methods are not easily applicable for sub-transient analysis, especially for the
identification of the characteristics of sub-cycle transients. This is because the validation of electric circuit
models used for fault transient analysis requires at least 2 cycles of sampling to predict the system state correctly.
Lack of understanding and methods to deal with the sub-cycle transients makes it challenging to integrate a large
amount of IBRs into the power grid, particularly into the distribution grids [xi]. This is mainly due to the low
tolerance of solid-state material in IBRs, more switching transients caused by IBRs, as well as the lower degree
of compensation and instrumentation in distribution systems. In fact, this is one of the root reasons that caused
the major interruptions of PV generation in California.

The Model:
The electromagnetic field properties of the power system under a sub-transient condition can be modeled starting
from Maxwell's curl equations (Faraday's and Ampere-Maxwell's equations). From the curl equations, one can
extract six scalar equations, which are in the same composition as in Eq. [1], where Ψ, Φ ∈ (𝐸, 𝐻), 𝛼 ∈ (𝜀, 𝜇),
𝛽 ∈ (𝜎 𝑒 , 𝜎 𝑚 ), Γ ∈ (𝐽, 𝑀), and 𝜉 ∈ (−1,1). The 𝐸, 𝐻, 𝐽, and 𝑀 are the electric field strength, magnetic field
strength, electric current density, and magnetic current density. The material properties represented by 𝜖, 𝜇, 𝜎 𝑒 ,
and 𝜎 𝑚 are electric permittivity, magnetic permeability, electric conductivity, and magnetic conductivity. The
𝑥, 𝑦, and 𝑧 subscriptions represent the components in 𝑥, 𝑦, and 𝑧 directions, and 𝑖 subscriptions represent the
impressed components of current densities.
In order to generate all six scalar equations, Ψ and Φ are to take E or H alternatively. When Ψ is E and Φ is 𝐻,
𝜉 = 1. When we swap 𝐻 and 𝐸, 𝜉 = −1. The subscript 𝑥, 𝑦 or 𝑧 can be shifted (𝑥 → 𝑦 → 𝑧 → 𝑥) to generate
the equations for all the components.

Description Of A General Three-dimension FDTD Method:

Fundamentally, the theory of FDTD is the employment of approximation of derivatives for numerically updating
the space and time by finite differences in Maxwell's equations. The detailed discussion of the derivation and
presentation of the FDTD updating equation in various grids can be found in our previous work [xii] as well as
other literature [10] and [xiii].
Typically, a set of three-dimensional updating equations for a rectangular grid (derived by employing forward,
backward, central, and averaging difference equations) are used, which can be described in Eq. [2].

Denoting the indexes (𝑖𝑥 , 𝑖𝑦 , 𝑖𝑧 ) as 𝑖⃗, (𝑖𝑥 ± 1, 𝑖𝑦 , 𝑖𝑧 ) as 𝑖⃗ ± ⃗1⃗𝑥 , (𝑖𝑥 , 𝑖𝑦 ± 1, 𝑖𝑧 ) as 𝑖⃗ ± ⃗1⃗𝑦 , and (𝑖𝑥 , 𝑖𝑦 , 𝑖𝑧 ± 1) as
⃗⃗𝑧 , and the superscript 𝑛, 𝑛 ± 1, and 𝑛 ± 1 for representing the time indexes, we have. Moreover, let us take
𝑖⃗ ± 1 2
𝛿𝐻 = 1/2 and 𝛿𝐸 = 1. The updating coefficients 𝐶𝛹𝑥𝛹 (𝑖⃗), , 𝐶𝛹𝑥𝛷𝑧 (𝑖⃗), 𝐶𝛹𝑥𝛷𝑦 (𝑖⃗), and 𝐶𝛹𝑥𝑗 (𝑖⃗), of the Eq. [2]
depend on the material properties at index 𝑖⃗.

Note that n is the time index representing the time evolution of electric or magnetic wave at given mesh i for all
spots, i.e., n = 1,2, ⋯ , t, ⋯ , T.

The One-dimension FDTD Equations For Line Transient Analysis:

Considering the material and the geometry of transmission line in 𝑦 and 𝑧 directions are uniform without
magnetic current sources, thus a transient over the transmission line can be considered a spatial propagation of
energy as the result of time evolution at every spot in 𝑦 and 𝑧 directions of space with all the field components
in 𝑥 direction is zero. That is, the wave propagation mode of transverse electromagnetic (TEM) wave with the
field components and waves in 𝑦 and 𝑧 directions is uniform and the spatial derivation of field components in 𝑦
and 𝑧 directions (𝑀𝑖𝑥 , 𝑀𝑖𝑦 and 𝑀𝑖𝑧 ) are zeros. Then we can consider only the case of the 𝑧-polarized TEM mode.
So, the spatial wave propagation occurs only in 𝑥-direction, and Eq. [1] can be described by two simplified
scalar equations as Eqs. [3a,3b].
The one-dimension FDTD equations are illustrated in Fig. [2]. In Fig. [2] we show that a one-dimensional
electromagnetic field model where the fields are samples in space and time. The electric-field meshes are shown
as blue arrows oriented in the 𝑧 direction and magnetic-field meshes as red circles oriented in the 𝑦-direction.

89
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Δ𝑥 is the spatial offset between electromagnetic field sample meshes and Δ𝑡 is the time step width (Δ𝑥 ∼
cm, Δ𝑡 ∼ ns). The index 𝑖 corresponds to the spatial step, effectively the spatial location, while the index $n$
corresponds to the time step.

Derivation Of Equations For Energy Spatial Propagation And Spot Time Evolution:
It is known that the rates of energy flow in electromagnetic wave propagation can be expressed in terms of
energy flux vector, 𝑆⃗𝑡 (= 𝐸⃗⃗𝑡 × 𝐻
⃗⃗𝑡 ). 𝑆⃗𝑡 points in the direction of propagation with oscillating magnitude in a
propagating linearly polarized electromagnetic plane wave.
According to the relation of the electric and magnetic fields, the corresponding magnetic field is obtained by
dividing the electric field by the intrinsic impedance (𝜂 = √𝜇/𝜀) of the transmission material, i.e., 𝐻 ⃗⃗ =
(√𝜀/𝜇)𝐸⃗⃗ = (1/𝜂)𝐸⃗⃗. The energy flux magnitude is described as follows in Eq. [4].
In the one-dimensional case under consideration, we only consider one electric field component and one
magnetic field component. So the computation of the energy flux vector can be simplified to the scalar equation
as Eq. [5] and the direction of the energy flux vector is the wave propagation direction (𝑥). The 𝐸𝑧𝑛 (𝑖 ) and
1
𝑛−
𝐻𝑦 2 (𝑖 ) components can be calculated in the time domain using one-dimension FDTD equations.
In our study, we consider transient energy flow along the transmission line can be seen as the energy
accumulation in the electromagnetic wave propagation direction (𝑥) of the spatial distribution of electromagnetic
field components (𝐸𝑧 and 𝐻𝑦 ) of the wave. The energy density of the corresponding electromagnetic field model
for transient analysis of fault and switching of inverters can be constructed as following Eq. [6].
According to Eq. [4] and [6], the energy wave propagation along the transmission line can be simply represented
by Eq. [7], where 𝑘𝜂,𝛥𝑡 is the linear coefficient depending on the material of the transmission line and the time.
Note that the energy flow is in the direction of the electromagnetic field wave propagation along the transmission
line, and we call it spatial propagation.

3. SIMULATION STUDY ON SUB-TRANSIENTS

Experiment setup:
The transient property of the electric power distribution system with faults or IBRs can be analyzed by a one-
dimensional electromagnetic field model using multiple FDTD meshes. Fig. [3] is the electric circuit of a simple
transmission line model with faults or IBRs. 𝜀0 and 𝜇 are the representations of the transmission line properties
(relative permeability of material annealed in 𝐻 ∼ 105 ). The following assumptions are made in the one-
dimension electromagnetic field model: 1) We consider a one-dimensional space where there are only variations
in the 𝑥 direction; 2) We assume that the electric field only has a 𝑧 component; 3) Magnetic field only has 𝑦
non-zero component which is time-varying; 4) A homogeneous material is modeled (a constant impedance for
all meshes).
In the experiment, we consider the electromagnetic wave propagates in a homogeneous material where 50
meshes for the electric and magnetic field arrays are modeled. The electromagnetic field arrays are initialized as
zero. The total number of time steps in the experiment is set as 300. The fault pulse signal is modeled using a
Gaussian pulse function at mesh i = 48 that has a maximum peak value of unity at time step 20 shown in Fig.
[4a]. We consider that the fault is almost gone after the time step 𝑡𝑓 = 41. The switching of inverters signal is
modeled using a period square wave function at mesh 𝑖 = 48 with a maximum peak value of unity and a 20%
duty cycle, as shown in Fig. [4b]. The switching signal is disappeared after the time step 𝑡𝑠 = 41. For the stability
of a simulation, we assume the Courant number of half unity 𝑆𝑐 = 0.5, which satisfies the Courant-Friedrichs-
Lewy (CFL) condition (𝑐Δ𝑡 ≤ Δ𝑥) in our one-dimension electromagnetic field model. The constant
characteristic impedance is used. For other constant characteristic impedance materials, the results profiles are
the same. The short terminal boundary condition is employed.
Experiment Results:
In our experiment, we consider the transient waves from two perspectives. One is the spatial propagating wave
of energy along the transmission line, and the other is the time evolution of the spots.
The spatial propagation of the electric field with 300-time steps is shown in Fig. [5]. The space domain has 50
meshes with a short terminal boundary condition on the left and right sides. The fault happens at mesh 48. The
vertical axis gives the electric field magnitude with different colors. One can see that the electric field originates
from mesh 48 at the right side of the grid, and it is seen to be propagating away from the fault mesh 48 to either

90
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

side. The electric field wave has two peaks. One is traveling to the left (the negative 𝑥 direction) and the other
to the right (the positive 𝑥 direction). Once the going field encounters the end of the grid at mesh 1 and mesh
50, they are both reflected and inverted and then travel to the opposite as time progresses. The electric field
propagates back and forth, inverting its sign at the boundaries.
Based on the spatial propagation and time evolution of the electric field, the slowly propagating energy envelope
generated by fault and the switching of inverters can be obtained using Eq. [6] and Eq. [7]. Fig. [6] shows the
energy envelopes induced by fault ((a)-(d)) and the switching of inverters ((e)-(h)) at mesh 2, mesh 10, mesh 25
and mesh 46, respectively. It is noticed that the energy envelopes induced by fault and the switching of inverters
are not very distinguishable for each mesh.
However, from the other perspective, the time evolution in the 𝑧-direction at the spots of transmission line offers
different information for the system with fault and switching of inverters, which can be seen in Fig. [7]. In Fig.
[7], we show the time evolutions of the electric field for mesh 2, mesh 10, mesh 25 and mesh 46 induced by
fault ((a)-(d))) and switching of inverters ((e)-(h))) at mesh 46. The horizontal axis is the time steps and the
vertical axis is the electric field strength for different meshes. We notice that the difference in time evolution is
obvious. The time evolutions induced by fault are smooth, while the ones induced by the switching of inverters
are oscillating dramatically, which indicates that the different time evolutions of the electric field in the direction
that is perpendicular to the transmission line can provide information about the system even though the energy
envelopes along the transmission line are not distinguishable for the system with fault pulse and IBRs switching.

3. FINDINGS AND CONCLUDING REMARKS

In this paper, we presented a study on the characteristics of fault pulse and switching operation of IBRs during
a sub-cycle transient period. Different from many previous works of transient analysis, this study is based on
the models and algorithms from electromagnetic field theory. The equations of spatial propagation and time
evolution of electromagnetic waves were derived and presented. Through analysis and simulation experiments
on spatial propagation of energy wave along the transmission line and time evolutions at different spots of the
transmission line. It is found that the sub-transient spatial wave preparations generated by faults and switching
transients of IBRs are very similar and thus indistinguishable, while the spot time evolutions of two are very
different. These findings may be useful for the impact analysis of IBRs on protection in power systems with
high penetration of IBRs.

Figure 1. Transient periods of fault currents.

91
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. One-dimension electromagnetic field space-time setup.

Figure 3. The transmission line model for sub-transient analysis.

Figure 4. (a) Fault pulse and inverter switching, (b) Signals for the experiment.

Figure 5. Simulated envelopes of spatial propagation of electric field and spot time evolution at different meshes along
the line in a time-space wave.

92
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 6. Envelopes of spatial propagation of energy generated by fault (a)-(d) and the switching of inverters (e)-(h).

Figure 7. Time evolution of electric field at different mesh locations generated by fault (a)-(d) and the switching of
inverters (e)-(h) at mesh 48.
Equations:

𝜕Ψ𝑥 1 𝜕Φ𝑦 𝜕Φ𝑧

= (𝜉 −𝜉 − 𝛽𝑥 Ψ𝑥 − Γ𝑖𝑥 ) (1)
𝜕𝑡 𝛼𝑥 𝜕𝑧 𝜕𝑦
1
𝑛+𝛿 𝑛+𝜉
Ψ𝑥 Ψ (𝑖⃗) = 𝐶Ψ𝑥Ψ (𝑖⃗) ⋅ Ψ𝑥 (𝑖⃗) + 𝐶Ψ𝑥𝑗 (𝑖⃗) ⋅ Γ𝑥 2 (𝑖⃗) + 𝐶Ψ𝑥Φ𝑧 (𝑖⃗)
𝑛

𝑛+𝛿 𝑛+𝛿 ⃗⃗⃗⃗⃗

⋅ (𝜉Φ𝑧 Φ (𝑖⃗) − 𝜉Φ𝑧 Φ (𝑖⃗ − 𝜉1 𝑦 )) + 𝐶Ψ𝑥Φ𝑦 (𝑖⃗) (2)
𝑛+𝛿Φ
⋅ (𝜉Φ𝑦 (𝑖⃗) − 𝜉Φ𝑦𝑛+𝛿Φ (𝑖⃗ − 𝜉1
⃗⃗⃗⃗⃗𝑧 ))

1 1 1
𝑛+ 𝑛+ 𝑛+
𝐸𝑧𝑛+1 (𝑖 ) = 𝐶𝑒 (𝑖 ) ⋅ 𝐸𝑧𝑛 (𝑖 ) + 𝐶𝑗 (𝑖 ) ⋅ 𝐽𝑧 2(
𝑖 ) + 𝐶𝑒ℎ (𝑖 ) ⋅ (𝐻𝑦 2(
𝑖 ) − 𝐻𝑦 2(
𝑖 − 1)) (3a)

93
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1 1
𝑛+ 𝑛−
2( 2(
𝐻𝑦 𝑖 ) = 𝐶ℎ (𝑖 ) ⋅ 𝐻𝑦 𝑖 ) + 𝐶ℎ𝑒 (𝑖 ) ⋅ (𝐸𝑧𝑛 (𝑖 + 1) − 𝐸𝑧𝑛 (𝑖 )) (3b)

1 2
< 𝑆⃗ >= < 𝐸⃗⃗ × 𝐻
⃗⃗ >≈ 𝐸 (4)
𝜂
1
𝑛−
𝑆𝑥𝑛 (𝑖 ) = 𝐸𝑧𝑛 (𝑖 ) ⋅ 𝐻𝑦 2(
𝑖) (5)

ℰ𝓍 (𝑖 ) = Δ𝑡 ⋅ ∑ 𝑆𝑥𝑛 (𝑖 ) (6)
𝑛=1

ℰ𝓍 (𝑖 ) = 𝑘𝜂,Δ𝑡 𝐸𝑧2 (𝑖) (7)

ACKNOWLEDGMENT
This work was partially supported by the US Department of Energy, Solar Energy Technologies Office, under
contract No.DE-EE-0008772.

REFERENCES

[1] NERC. 1,200 MW Fault Induced Solar Photovoltaic Resource Interruption Disturbance Report. www.nerc.com, 2017.
[2] NERC. 900 MW Fault Induced Solar Photovoltaic Resource Interruption Disturbance Report. www.nerc.com, 2018.
[3] NERC. April and May 2018 Fault Induced Solar Photovoltaic Resource Interruption Disturbances Report.
www.nerc.com, 2019.
[4] Kundur P, Paserba J, Ajjarapu V, Andersson G, Bose A, Canizares C, Hatziargyriou N, Hill D, Stankovic A, Taylor C,
Van Cutsem T. Definition and classification of power system stability IEEE/CIGRE joint task force on stability terms and
definitions. IEEE transactions on Power Systems 2014;19(3):1387-401 < 10.1109/TPWRS.2004.825981>
[5] Vugrin E, Baca M. Electric Power System Modeling and Analysis. SANDIA REPORT: SAND, 2013.
[6] Kundur P, Balu NJ, Lauby MG. Power system stability and control. New York: McGraw-hill, 1994.
[7] Renedo J, Garcı A, Rouco L. Active power control strategies for transient stability enhancement of AC/DC grids with
VSC-HVDC multi-terminal systems. IEEE Transactions on Power Systems 2016; 31(6):4595-604 <
10.1109/TPWRS.2016.2517215>
[8] Dıez-Maroto L. Improvement of voltage ride through capability of synchronous generators with supplementary
excitation controllers. MSc, Universidad Pontificia Comillas, Madrid, Spain, 2013.
[9] Kou G, Chen L, VanSant P, Velez-Cedeno F, Liu Y. Fault Characteristics of Distributed Solar Generation. IEEE
Transactions on Power Delivery 2019; 1-1 < 10.1109/TPWRD.2019.2907462>
[10] Yee K. Numerical solution of initial boundary value problems involving Maxwell's equations in isotropic media. IEEE
Transactions on antennas and propagation 1996; 14(3):302-7 < 10.1109/TAP.1966.1138693>
[11] Chen PC, Malbasa V, Kezunovic M. Locating sub-cycle faults in distribution network applying half-cycle DFT
method. In: 2014 IEEE PES T&D Conference and Exposition; 14 Apr 2014: IEEE, pp. 1-5.
[12] Perera S. Physical Knowledge Based Scalable Phased Array Antenna Modeling for Radar Systems. PhD, the University
of Oklahoma, Norman, USA, 2016.
[13] Taflove A, Hagness SC. Computational electrodynamics: the finite-difference time-domain method. Norwood, MA:
Artech house, 2005.

94
 LIMITATION OF FUNDAMENTAL ANALYTICS FOR FAULT
TRANSIENTS ON PLANNING OF TRANSMISSION SYSTEMS
WITH HIGH PENETRATION OF SOLAR PV RESOURCES

Varun Perumalla
Enterprise Model Management, California ISO, Folsom, California, USA, [email protected],
ORCID: 0000-0001-9050-6379

Gopakumar Gopinathan
Operations Engineering Services, California ISO, Folsom, California, USA, [email protected],
ORCID: 0000-0001-8554-2148

Milad Javadi
Transmission Planning Division, Lower Colorado River Authority, Austin, Texas, USA, [email protected],
ORCID: 0002-4041-8204

Almir Ekic, Di Wu
Dept. of Elec. and Comp. Engr., North Dakota State University, Fargo, North Dakota, USA, [email protected],
ORCID: 0000-0002-8178-4569, 0000-0002-4455-1504

Rama Ramakumar
School of Elec. and Comp. Engr., Oklahoma State University, Stillwater, Oklahoma, USA,
[email protected], ORCID: 0000-0001-7275-4715

Dhruv Sharma
VIT Bhopal University, Bhopal, India, [email protected],
ORCID: 0000-0002-2191-2645

Guomin Ji, John Jiang

School of Elec. and Comp. Engr., Norman, Oklahoma, USA, [email protected]
ORCID: 0000-0001-9779-1723, 0000-0002-1590-8527

Perumalla, V, Gopakumar, G, Javadi, M, Ekic A, Wu D, Ramakumar, R, Ji G, Jiang, J.

Limitation of Fundamental Analytics for Fault Transients on Planning of Transmission
Cite this paper as:
Systems with High Penetration of Solar PV Resources. 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: In this paper, we provide feedback on the challenges to transmission system protection for
integration of high-penetration inverter-based resources, particularly the grid-connected photo
voltaic (PV) generations. We first put forth a brief overview about three-in-a-row events of major
interruptions of solar PV generations in Southern California. Then we present an investigation of
the conventional analytics for protection analysis and basic assumptions. By pointing out the
fundamental limitations of these basic methods in resolving the difficulties associated with major
interruptions of PV generation, we argue that 1) most of the emerging protection problems
associated with inverter-based resources are resurfacing of the long-standing sub-cycle
electromagnetic transient issue, which is not well-understood and has been avoided in practice by
enforcing ride-through requirement or over-sizing strategies, 2) the bottleneck in dealing with sub-
transient (sub-cycle) stability issues is identified with an overview of fundamental engineering
concepts/analytics of fault analysis, and 3) advancing the frontier of fundamental analytic should
be prioritized.
Keywords: Power system transient stability, sub-cycle transient, power system protection, inverter-based
resource, photovoltaic generation.
© 2020 Published by ECRES

95
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Nomenclature
PV photo voltaic
BES bulk energy system
NERC North American Electric Reliability Corporation
IBRs Inverter-based Resources
CAISO California Independent System Operator
SNL Sandia National Laboratories
CREZ Competitive Renewable Energy Zones
GTC Generic Transmission Constraints
TO transmission owner

1. INTRODUCTION

Over the past few decades, solid state inverter has become one of the major technologies that power homes and
businesses around the world. As the penetration of utility-scale inverter-based resources (e.g., solar photovoltaic
(PV) and wind power resources) directly connected to the bulk energy system (BES) increases, state-of-the-art
dynamic models have failed to predict responses by power systems and PV farms to extreme events. The North
American Electric Reliability Corporation (NERC), after investigating three major grid disturbances three years
in a row where large-scale inverter-based generation units went off-line, suggested the importance and urgency
of development of new control and protection technologies to increase grid reliability and resilience.

Major PV Generation Interruptions In-a-row And Lessons:

On August 16, 2016, the Blue Cut fire in Southern California created multiple line faults on the 500kV
transmission system of WECC. During these events, the inverters at multiple BES solar farms stopped producing
current, with the most significant event dropping 1200 MW of PV generation [xiv]. On October 9, 2017, about
three months after the 2016 NERC/WECC report, the second very similar BES disturbance occurred in Southern
California, due to Canyon 2 Fire [xv].
The third major disturbance (Angeles Forest and Palmdale Roost) caused two major cascading events, which
indicates that, unfortunately, the recommendation and actions based on the lessons learned from above two
events do not seem to be working very well. The two major interruptions occurred in 2018 in California as the
results of correct clearances of two normal faults. As described in [xvi], the first Angeles Forest disturbance
occurred on April 20 which was initiated by a phase-to-phase fault of 500 kV transmission line, which was
cleared corrected and timely in 2.6 cycles. The second Palmdale Roost disturbance occurred on May 11 which
was also initiated by a single-line-to-ground fault in 500 kV transmission system. The resulting was cleared by
line relay protection within 3 cycles.
In response to these three events, a NERC/WECC joint task force that was assembled by the NERC Operating
Committee to investigate this disturbance, determine the causes, and develop key findings and recommendations
to ensure that occurrences such as this one are mitigated throughout the North American BES. The reports of
investigation were published [1,2,3] in 2017, 2018 and 2019 respectively. The key findings of three reports are
summarized below:
 undesired trip or activate momentary cessation 1,
 incorrect response of protection or control systems,
 erroneous high and low frequency response of IBRs,
 undesired responses to over and under voltages and stops of gating the transistors,
 unnecessary tripping due to DC reverse current or AC overcurrent,
 uncorrelated ramp rate interactions with return from momentary cessation during the recovery period.

As reported in the NERC reports [1,2,3], there were no solar PV facilities de-energized as a direct consequence of
these fault events; rather, the facilities ceased output as a response to the electromagnetic induced by fault on
the system, in near instantaneous time constants for frequency or voltage signals,
 at the outset or during the transient,
 sub-cycle transient (milliseconds),
 something often referred as harmonics or noises.

11
As noted by NERC [xiv], the inverters did not “trip”, they went into so-called Momentary Cessation, which in inverter
language, does not equal trip unless the manual reset is needed.

96
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

All these certify that these events are associated with transient stability issues, and the recommendations were
made based on conventional wisdom about transient stability:
 enforcing ride-through requirements,
 developing better instantaneous response and control technology for Inverter-based Resources (IBRs),
 constructing better models of IBRs.

Further, NERC has stressed the importance of continued analyses of inverter-based resource performance under
existing and future penetration levels for determination of if there are any reliability risks on planning and
operating philosophies of bulk transmission system employed today. It is also highly recommended to perform
analysis to determine if potential resource loss events caused by momentary cessation or inverter tripping could
impose a reliability risk. Both California Independent System Operator (CAISO) and Lower Colorado River
Authority (a major TO in center Texas) have adopted more sophisticated dynamic models that are used in NERC
Inverter-Based Resource Performance Task Force for their Dynamic Security Assessment tool and offline
studies. At this stage, we have concerns about the limitation of fundamental analytics on generating good
solutions to the problems, other than typical strategies such as capacity over-sizing, elevating ride-through
requirements and limiting the output of PV generations.

2. STATE-OF-THE-ART: A PIECE OF CLOUD IN CLEAR

CIGRE Definition And Classification Of Power System Stability:

Power grid stability has been recognized as an important problem for secure system operation. Many blackouts
and major disturbances as the aforementioned ones caused by power system instability have illustrated the
importance of this phenomenon. A joint task force of the CIGRE Study Committee 38 and the IEEE Power
System Dynamic Performance Committee have investigated the issues of stability comprehensively and
provided definitions and categorization of major stability issues in power system operations. Further, more
detailed discussion on the categorization as well as the underlying models and methods for real-time system
stability analysis, including power flow stability, contingencies and short circuits, provided in a research report
of Sandia National Laboratories (SNL) [xvii].

Figure 1. The time horizons of various fault transients in power system analysis.

According to the discussions by CIGRE and SNL, the current state-of-the-art studies work quite well for most
of the stability problems, except those which may happen during the sub-transient period following a fault or
lightening stroke. Particularly in sub-cycle transient period, there are not enough data of fault currents to warrant
response of protection or control equipment as to be discussed in the next subsection. The sub-transient period
is shown with the gap between fault inception time and fault clearing time in Fig. 1.

97
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Current State-of-the-Art Of Transient Analysis For System Protection

Power system transient stability analysis for system protection and relay procurement is based on two major
types of methods: Symmetrical Components and Traveling Waves.
Unbalanced abnormal conditions may arise from faulty voltage sources or loads. It is possible to analyze some
simple types of unbalanced networks using straightforward solution techniques and wye-delta transformations.
However, power networks can become quite complex and many situations would be very difficult to handle
using ordinary network analysis. For this reason, a technique which has come to be called symmetrical
components has been developed based on conventional electric circuit theory. The concept of symmetrical
components has become acceptable as a prevailing way of describing the properties of many types of network
elements such as transmission lines, motors and generators.

(a)

(b)
Figure 2. (a) Symmetrical components, (b) Traveling waves method.

In symmetrical components, any set of unbalanced three-phase quantities are expressed as the sum of three
symmetrical sets of balanced phasors, as shown in Fig. 2(a). Using this tool, unbalanced system conditions, like
those caused by major fault types, may be visualized and analyzed. Additionally, most microprocessor-based
relays operate using symmetrical component quantities.
Similarly, traveling waves-based analytics uses voltage and current measurements to calculate incident and
reflected waves, traveling in opposite direction along a transmission line, as shown in Fig. 2(b). The traveling
wave has the same physics nature as that of positive and negative sequence components. By taking advantage
of the internal protection elements, the communications channel to the remote terminal, and Global Positioning
System-based time synchronization, traveling waves can be used to assess the electromagnetic transient
characteristics for relay protection.
With the century-long development and experiences, these tools seemed to work very well for most of the
stability problems for power systems consisting of conventional components such as electric machines, and
transformers. Unlike the energy systems made of solid-state material, e.g. IBRs, conventional components can
ride-through faults for multiple cycles or even longer.

A Cloud In A Clear Sky

We would like to propose that electromagnetic sub-transient, particularly the sub-cycle-transient, is a cloud that
obscures the clear sky depicted by the current wisdom about power system stability. As reported in the NERC
investigation and analysis on major grid-connected PV interruptions [1,2,3], none of the interruption involving
BES-connected solar PV resources across the regions was tripped as a direct consequence of line outages or
faults; rather, they are all caused by the electromagnetic transients induced by disturbance that occurred far
away. In fact, as stated in the NERC reports that all of these major disturbances are directly associated with sub-

98
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

cycle transient as the instantaneous responses, such as tripping, momentary cessation and ride-through, to
electromagnetic transient ac overvoltage, ac overcurrent, dc reverse current, instantaneous frequency excursion
are all in the order of milliseconds or less.
We have been avoiding the shadow areas under the cloud. Sub-cycle electromagnetic transients are not the
clouds that just formed. We've known the sub-cycle electromagnetic complications since the developing stages
of analytics for power system stability. Generally, our strategy is to manage the stability by not letting the system
enter in the operation zones, not even near the skeptical areas that may cause sub-cycle transient complications
or cannot be well explained by our understanding based on the theories and experiences.
This strategy is clearly reflected in ALL NERC recommendations to IBRs summarized below:
 equipment oversizing
 enforcement and enhancement of ride-through capability,
 expansion of ride-through curves (e.g. IEEE PRC-024-2),
 extension or elimination of momentary cessation of IBRs,
 reduction of sensitivity of IBRs to disturbance signals,
 modification of relay or control system setting to avoid tripping for "Unharmful" or "not-damaging" dc reverse
currents that could result during sub-cycle transient overvoltage conditions.

3. A DEVELOPING FRONTIER OF SCIENCE AND TECHOLOGY FOR POWER SYSTEM

PROTECTION

The major events of disturbances/interruptions of solar PV generation and system stability impact in California
as well as many other events brought to the forefront the limitation or even incompetence of the existing methods
and models in addressing sub-cycle transient issues.

Fundamental Challenges To Sub-cycle Transient Analysis With The Existing Methods

The symmetrical component concept was immediately seen as a milestone for electrical system studies, and
many applications have been developed since then [xviii]. For a fault transient condition, regular phase components
could be too complicated to determine, the symmetrical components method gives 3 decoupled combinations
assuming a stationary establishment of fundamental frequency-based energy flow.
If this assumption is not valid, there is no clear assessment for the symmetrical components, zero-sequence
components, other frequency components (the harmonics), and everything else that cannot be explained (the
noises). This intrinsic assumption limited the applicability of symmetrical component-based method in fault
protection during sub-cycle transient period when neither the system itself nor the sources are well aligned in
the frequency domain, sometimes which is known as both the system and sources are not symmetrical. This is
one of the main reasons that all devices are required to withstand the fault for several cycles to ensure enough
time to reveal the system states.
In other words, symmetrical components-based fault analysis heavily depends on the validation of impedance
models at the fundamental system frequency that characterize the steady-state, so we have to cast models at
system frequency, rate equipment at system frequency, and band-pass signals at system frequency. Although the
fundamental frequency indeed dominates the states, the states cannot be observed until the system is stationary
after some time. The characteristics state is unknown during sub-transient period when the domination of
fundamental frequency is unclear, and it is difficult to find the meaningful or true magnitudes for the key
frequency components, as well as other related variables such as impedance, voltages and currents.
Traveling waves is an advanced physics-based method for faster system protection as for examples described in
[xix][xx]
. Traveling wave-based fault locating functions in a protective relay take advantage of the internal
protection elements, the communication channel to the remote terminal, and Global Positioning System-based
time synchronization. This approach provides accurate fault location estimation for transmission lines
automatically and quickly after the fault. Thanks to the advances in instrumentation and communication
technologies [xxi][xxii], many other traveling wave-based methods were developed, where traveling wave methods
were combined with other technologies to facilitate the analysis and detection of faults (e.g., [xxiii]).
All these traveling wave-based methods feature a fast response and high accuracy at least in theory, demonstrated
simulation studies and some applications. However, as summarized in a recent review on traveling wave-based
methods [xxiv], they are also facing some insurmountable technical difficulties:
 the detection of the wave head is the key to traveling wave fault location. If the wave head cannot be
captured successfully, or the wave head does not exist at all at the occurrence of a fault, the fault location
will fail. For instance, when the line is grounded through a large resistance, the transient traveling wave

99
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

signals are too weak to be detected, disabling the fault location under these circumstances. Moreover, if
a fault is caused by a gradual change in the transition resistance, the traveling wave may also be too
weak to be discovered, resulting in the failure of the fault location,
 these methods rely on the time accumulated for the wave head to arrive at certain points to the fault
locations, or traveling distance of fault signal, which is the product of the time recorded and the wave
speed. Therefore, the accuracy of fault location is dependent, to a great extent, on the estimation
accuracy of wave speed. Knowing wave speed is 𝑣𝑡 = 1/√𝐿𝐶 (where 𝐿 is the inductance and 𝐶 is the
capacitance of the propagation medium), thus these methods depend on the actual line parameters and
models,
 accuracy in fault location depends upon sampling frequency. Since the speed at which the wave travels
over transmission lines is slightly lower than light speed, in order to achieve higher accuracy, a very
high sampling frequency has to be used in traveling wave fault location methods. Therefore, more
expensive and complex equipment are required,
 the wave head must be identified to locate the fault, which is often carried out by experienced
professionals and cannot be implemented automatically by computers,
 since the traveling wave fault location uses the signal with high frequency components for the analysis,
it is vulnerable to interference of external signals, which could be problematic for a high penetration of
IBRs or grid-connected solar PV farms,
 if the fault occurs near a line end, it is very difficult to identify the wave head with information of this
line end due to the high speed of the traveling wave,
 another difficulty arises for faults near the buses or for those faults occurring at near zero voltage
inception angle because, if the voltage is zero when the fault starts, there is not an abrupt change of the
line continuity and a wave head is not produced.

Most of these aforementioned problems of traveling wave-based methods become more significant when they
are applied to the system protection of the power grid with high penetration of grid-connected IBRs or solar
PVs.

The Issues With The Existing Methods And Why They Cannot Be Avoided-Some Comments From
CAISO
Continuous occurrence of large-scale interruptions of grid-connected renewable energy resources in many
countries, like the three major interruptions in-a-row in California, highlighted the limitation of our current
wisdom about power system protection and operation. In a recent presentation by the National Renewable
Energy Laboratory about the technical challenges of high levels of IBRs in power grids [xxv], as well as many
others, the new characteristics of future power grid is summarized which include:
 less synchronous generators,
 more variable, inverter-based generation,
 more distributed generation and controllable (active) loads,
 loads becoming more power electronics based (e.g. LEDs, VSD, inverters, converters),
 mobility - migrating towards electric vehicles.

Some of the definite needs are also highlighted below:

 normal and abnormal voltage and frequency control,
 protection schemes that work under high levels of IBR,
 ability to blackstart grids with high levels of IBR,
 accurate models of IBR controls for transient and dynamic analysis – “moving from equations that
describe physics to models that describe inverter controls”,
 grid codes and standards are needed that define response, characteristics for IBRs to transient and
dynamic events.
Given the tremendous success and experiences in the past century, we know how to do voltage and frequency
control in normal operating conditions and in contingencies, we know how to procure protection schemes, we
know how to blackstart grids, we know how to do transient and dynamic stability analysis, we have strict grid
codes and standards. We have been applying the current wisdom to address these integration problems. But why

100
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

it seems not working as expected? What are the root challenges that the new characteristics of the future grid
bring in?
Based on the aforementioned discussion and analysis on symmetrical components, and traveling waves, we infer
that the imitated applicability of the current wisdom root from an old problem, a problem that we have been
avoiding for over a century to sub-cycle transients or sub-transients. The physics during the sub-cycle or less
transient period is beyond the scope of conventional wisdom, meaning it is NOT very well understood: either
the assumption of the model becomes not valid, missing representation of the impacting variables, unable to
deal with high dimension of non-linearity, etc. Due to the intrinsic nature of semiconductor material used to
make inverters, the hair-triggering feature and few fault current contributions cannot be eliminated by simply
imposing the conventional avoiding strategies such as ride-through or over-sizing.
The professionals and researchers in the power industry are gaining better understanding about the issues, and
have been seeking solutions from technologies from other engineering disciplines such as Artificial Intelligence,
Machine Learning, Big Data, etc. However, maybe we haven't really spent enough time to think about the
limitation of the knowledge in our own domain. Is this something we can get around using strategies similar to
issues related to integration of power generation from wind? Or, have we hit the ceiling on the old wisdom?

Renewable Power Generation Curtailment Due To Sub-transient Stability Concerns-Texas Scenario

The fast-growth of utility-scale renewable resources in specific geographical areas in conjunction with
decommission of conventional power plants have introduced serious limitations in taking a full benefit of
renewable resources due to stability concerns. Over the last few decades, Competitive Renewable Energy Zones
(CREZ) are gradually formed in various (mostly distinct) areas with weak transmission systems due to suitable
regional weather condition, inexpensive real-estate, availability of an EHV (300-kV and higher) transmission
circuit and/or recommendations by Independent System Operators. Rapid growth of CREZ areas impelled ISOs
to define and enforce Generic Transmission Constraints (GTC) based on the outcome of planning assessments.
A stability GTC is defined to curtail the amount of renewable power export out of a specific CREZ area to a
certain MW in order to avert voltage transient instability under pre-contingency or any NERC P1, P1.2, P7.1
and/or P6 contingency events.
For instance, currently there are a number of active GTCs being enforced in the Texas transmission grid
(ERCOT), namely, Panhandle Stability, McCamey Stability and Nelson Sharpe – Rio Hondo Stability GTCs
[xxvi]
. These GTCs set a generation cap on CREZ areas under pre/post contingency operating conditions. In real-
time operation, a GTC will be added to the next run of SCED if a contingency/outage occurs on an identified
BES transmission element included in a so-called watch list table. Then, the total renewable power export will
have to be curtailed; and the generator operators will be penalized if they violate the generation limitation.
These limitations not only reduce flexibility of the system operation and resource adequacy of interconnected
bulk transmission systems like ERCOT or WECC, but also can cause difficulties in setting the protection scheme
at the POI of renewable resources and EHV transmission circuits for different operation scenarios. Development
of comprehensive and insightful tools for stability assessments and generation interconnection studies at the
stage of planning can effectively avoid unnecessary challenges/limitations such as GTCs.

4. CONCLUDING REMARKS

Three major power supply interruptions in California revealed the limitation of existing technologies for
transient stability (sub-cycle) issues. Here we surmise that the issue could be much larger than simply seeking
technical solutions to prevent similar events in the future. With an overview of fundamental engineering
concepts/analytics of fault analysis, we identify the bottleneck in dealing with sub-transient (sub-cycle) stability
issues.
The views and opinions of the ISO in California and a TO (transmission owner) in Texas are discussed to fortify
the basic questions raised by us. Having witnessed the impacts of solar PV generation interruptions, limitations
of present approaches and the lack of effective tools, we propose it will be prudent to limit entry of a very large
amount of grid-connected PV and other IBRs.

ACKNOWLEDGMENT
This work was partially supported by the US Department of Energy, Solar Energy Technologies Office, under
contract No.DE-EE-0008772.

101
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] NERC. 1,200 MW Fault Induced Solar Photovoltaic Resource Interruption Disturbance Report. www.nerc.com, 2017.
[2] NERC. 900 MW Fault Induced Solar Photovoltaic Resource Interruption Disturbance Report. www.nerc.com, 2018.
[3] NERC. April and May 2018 Fault Induced Solar Photovoltaic Resource Interruption Disturbances Report.
www.nerc.com, 2019.
[4] Vugrin E, Baca M. Electric Power System Modeling and Analysis. SANDIA REPORT: SAND, 2013.
[5] Chicco, G. and Mazza, A. 100 years of symmetrical components. Energies 2019; 12(3): 450 <10.3390/en12030450>
[6] Marx S, Johnson BK, Guzmán A, Skendzic V, Mynam MV. Traveling wave fault location in protective relays: Design,
testing, and results. In: Inproceedings of the 16th Annual Georgia Tech Fault and Disturbance Analysis Conference; 6
May 2013, Atlanta, GA.
[7] Schweitzer EO, Kasztenny B, Guzmán A, Skendzic V, Mynam MV. Speed of line protection-can we break free of
phasor limitations?. In: 2015 68th Annual Conference for Protective Relay Engineers; 30 Mar 2015, IEEE, College
Station, TX, USA: pp. 448-461.
[8] Ando M, Schweitzer EO, Baker RA. Development and field-data evaluation of single-eng fault locator for two-thermal
hvdv transmission lines part i: Data collection system and field data. IEEE Transactions on Power Apparatus and
Systems 1985; PAS-104 (12):3524-30 <10.1109/TPAS.1985.318905>
[9] Ando M, Schweitzer EO, Baker RA. Development and field-data evaluation of single-end fault locator for two-terminal
hvdv transmission lines-part 2: Algorithm and evaluation. IEEE transactions on power apparatus and systems 1985;
PAS-104 (12):3531-7 < 10.1109/TPAS.1985.318906 >
[10] Magnago FH, Abur A. Fault location using wavelets. IEEE Transactions on Power Delivery 1998; 13(4):1475-80 <
10.1109/61.714808>
[11] De Andrade L, De Leão TP. Travelling wave based fault location analysis for transmission lines. In: EPJ Web of
Conferences; 02 October 2012: EDP Sciences, pp. 04005.
[12] Kroposki BD. Summarizing the Technical Challenges of High Levels of Inverter-based Resources in Power Grids.
National Renewable Energy Lab.(NREL), Golden, CO (United States); 2019 May 22.
[13] ERCOT. Transmission and security desk operating procedure. 2019.

102
 .USING FLOATING-INTERLEAVING BOOST CONVERTERS
FOR SMALL WIND ENERGY CONVERSION

Moncef Justin Lalou

College of Engineering and Architecture Fribourg, University of Applied Sciences HES-SO, Fribourg, Switzerland,
[email protected], ORCID: 0000-0002-7058-6487

Capone Michele
Former colleague, College of Engineering and Architecture Fribourg, University of Applied Sciences HES-SO,
Fribourg, Switzerland, [email protected], ORCID: 0000-0002-5654-6972

Moncef J., Lalou. Using floating-interleaving boost converters for small wind energy
Cite this paper as:
conversion. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: Small Wind Energy Conversion System (SWECS), based on two back-to-back voltage source
converters (VSC), with transformerless power injection into the utility grid, prove to be quite
effective and reliable, but its efficiency and the compactness of its implementation can be
improved. For this purpose, a modified topology, by introducing a floating-interleaving boost
converter (FIBC), is proposed and evaluated.
In this paper, theoretical and practical aspects related to the development of a FIBC dedicated to
small wind power conversion, are presented. For a given rated values of power and voltage at the
injection point into the grid, and a range of generator voltage, a FIBC converter is selected and
sized, then realized and tested in stand alone operation and as element of a SWECS prototype.
Experimental results confirm the suitability of the modified topology to achieve a very high
efficiency of the wind energy conversion.
Keywords: Small wind energy, boost converter, FIBC, SWECS, efficiency.
© 2020 Published by ECRES

1. INTRODUCTION

Over the last years, wind energy has been pursuing its worldwide growth as primary source of the electrical grid.
With public incentive, massive investments in new plants and technologies made it possible to strongly increase
the installed power capacity. Whereas this performance results mainly from large on and off-shore wind farms,
we see a parallel development of small wind turbines – object of this study - whose power is rated up to
approximately 10 kW, and dedicated to off-grid electrification or to locally supply the distribution grid.
As it widely reported, for instance in [4,5], the standard solution of a SWECS consists in a three phase, multipolar
permanent magnet synchronous generator (PMSG), which is directly driven by the small wind turbine, and
whose stator phases are connected to two VSCs (Voltage Source Converters) in back to back topology. These
serves to adapt the PMSG three phase voltage with variable frequency and amplitude to those of the utility grid,
while allowing to operate at maximum power point (MPP). The grid side inverter (VSI) is normally connected
to the power injection point through a current harmonic filter, usually of LCL type [6].
This configuration enables full control over active/reactive powers in transit from the generator and into the grid,
with control decoupling of the two converters thank to the DC-link capacitor, but has the drawbacks to generate
pronounced losses with its numerous hard switching commutators (2 VSCs), and shows a heavy and
cumbersome DC-link capacitor which increases costs and reduces the lifetime of the system.
In an attempt to overcome these drawbacks, the idea is to replace the generator side rectifier (VSR) by a diode
rectifier in series with a boost converter (BC) of suitable properties (figure 1) in terms of efficiency, voltage
ratio and compactness. To this end, the main purpose of this study is to introduce and evaluate a BC of the family
of floating- interleaving boost converters (FIBCs). The same converter was successfully used, under similar
requirements, to elevate Fuel Cell output voltage in an electric vehicle drive [1].

103
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1: Structure of the SWECS.

In the following, the topology of the FIBC is first determined, and its electrical components sized by considering
the rated power and input/output voltage levels. Then the necessary operating controls are performed for the
FIBC and the VSI. The control of the FIBC output voltage and its inner phases currents controls, are first
performed, then the control of active (and reactive) power at the grid injection point. Finally, the experimental
data is collected for the FIBC operating in stand alone, and as part of the SWECS prototype, hence enabling to
assess real performance of the proposed topology.

2. FIBC SELECTION AND CONTROL

Basically, a BC from the FIBCs family consists of a set of N conventional BC, which are shared into two groups
of N/2 conventional BC mounted in parallel and whose outputs are connected in series. Figure 2 gives an
illustration of a quadriphase FIBC (N=4). This topology results naturally in a higher voltage amplification
capacity than the conventional BC, and in a reduced voltage and current of the passive components, as well as
switching devices, thus decreasing their size and weight. Moreover, the control signals of the N switchs,
according to PWM control method, are shifted by an angle of 360/N degree in respect to each other, which leads
to lower the input currents ripple.

Figure 2: Four phases FIBC topology with ohmic load.

But the most relevant advantage for wind energy harvesting is an improved intrinsic efficiency, as shown in [1]
through analytical comparison of power losses between the conventional BC and the FIBC for N=2,4,6, at a
rated power of 5 kW and an output voltage of 400 V. From this study, it comes notably that the FIBCs with
N=4, 6 remains at high levels of efficiency (above 90%) over a wide range of input current, while it decreases
sharply for N=2 and even more with the conventional converter. The FIBC converter is therefore selected to
elevate DC voltage delivered by the generator and a diode rectifier, with N=4 (see figure 2) as it’s proven that it
detains the best tradeoff between efficiency and number of phases i.e. the complexity [1].
For a proper operation of the VSI, the output of the FIBC (i.e. DC-link) is controlled at constant voltage,
depending on the rated AC voltage of the utility grid. In order to improve the dynamic behaviour of the control
loop and to guarantee a balanced power sharing between the four phases, the currents are also controlled,
obviously at the same reference. Due to the intrinsic switching operation of each phase, it’s appropriate to use a
sliding mode controller, which is known to be very robust in face of input voltage variations and output power
[1].

104
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The corresponding control scheme is given – for each phase - in figure 3, where the common current set value
(ilc) is delivered by the voltage controller.

Figure 3: Sliding mode current controller scheme (phase 1).

In a similar fashion to the classical state control, the overall control results from three contributions : feedforward
(gain kw), integral (gain kr) and feedback (gain ks). In sliding mode operation at constant PWM frequency, the
average switching control (𝑇𝑖 ) of a switch i (i=1 to 4) is derived from the general formula given in [2, p. 21] :
𝑢𝑜𝑢 − 𝑢𝑖𝑛 + 𝐴𝑘 (𝑖𝑟𝑒𝑓 − 𝑖𝑖 ) + 𝐵𝑘 𝑑(𝑖𝑙𝑟𝑒𝑓 )/𝑑𝑡
𝑇̅𝑖 =
𝑢𝑜𝑢 + 𝑢𝑖𝑛

This result involves the upper (uou) and lower (uin) values of the switching control Ti, the reference (iref) and the
real value (ii) of the phase current, as well as two parameters : Ak=2∙Li∙kr/ks, Bk=2∙Li∙kw/ks where Li denotes the
phase inductance. Normally, these two parameters are analytically determined through pole placement method,
in sliding mode operation. But in this case, both cannot be loosely assigned since the state model of a phase
current shows a nul state matrix. The two parameters were then adjusted in results-oriented manner using an
appropriate simulator in Matlab/Simulink environment (see Experimental setup section). Furthermore, the PI,
parallel form, DC-link voltage controller is designed using pole placement method on the basis of the transfer
functions of the output voltage with respect to PWM duty cycle [3].

3. CONTROL AND OPERATION OF THE SWECS

In order to enable conversion at maximum power point (MPP), the generator speed is controlled, with a reference
depending on the wind speed and the turbine blades features, in accordance with the tip speed ratio method. For
this purpose, the active power at the injection point serves as control variable, which in turn, is controlled through
the VSI (as well as the reactive power). Simultaneously, the DC-link voltage is also controlled at a sufficient
level, given the grid voltage to guarantee the normal operation of the VSI.
Figure 4 gives a comprehensive description of the SWECS control scheme, where the key elements are two dual
PI/P controllers that are operating simultaneously to control the grid currents (id, iq) in Park domain. The
references of these two currents are calculated according to the respective power references and grid voltage,
which is assumed to be constant. For each current, a parallel form PI controller outputs the reference of the
corresponding LCL filter capacitive current. Le latter is in turn subject to a P control with feedforward
compensations dedicated at dynamic decoupling of id and iq. The parameters of the two controllers are optimised
by shaping of the open loop transfer function of id, iq, according to the methodology described in [7]. The P gain
(kp2) is set to reach a loop damping factor of 0.95, and the PI parameters (kp1, ki1) to impose a phase margin of
30.

105
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 4: Control structure of the SWECS.

4. EXPERIMENTAL SETUP

A prototype is developed in the aim to assess the operating performance of the proposed SWECS. The objective
is to estimate efficiency of the power electronics conversion (from the generator to the grid), and to observe the
behaviour of FIBC and VSI controls. For this purpose, the subsystem from the wind turbine up to the FIBC input
is simply replaced by a variable DC voltage source.
The following rated values have been set : apparent injected power : 5 kVA, grid AC line voltage : 400 V, DC-
Link voltage : 700 V, and the DC input range of the FIBC is assumed to vary in a relatively wide interval
[100..400] V, which leads a transfer ratio within the range [1.75..7]. These settings will enable to specify the
FIBC, the VSI and the LCL filter, for PWM switching frequencies of 20 kHz (FIBC) and 10 kHz (VSI).
The passive elements of the FIBC are specified to largely comply with input/output settings, and also to enable
a low input current ripple to protect the generator windings. The actual specifications are optimised as trade-off
between the volume clutter, cost and performance, which leads to the following specifications : inductors :
4x5mH, 6 Adc, 1.5 App @ 20 kHz, and electrolytic capacitors : 4x 4.7 mF, 400 Vdc, 4x18kΩ voltage balancing
resistors. The active elements of the FIBC consists in two IGBT bridge modules 1200 V MiniSKiiP IPM with
their evaluation boards. Figure 5 illustrate the FIBC assembly, which is tightened in a 16-inch rack in order to
cut electromagnetic radiation due to lead wires.
Furthermore, the passive elements of the LCL filter are specified according to the design methodology detailed
in [6], dedicated to low scale energy conversion systems, which leads to the following specifications : L1=10
mH (VSI side); L2=2.5 mH, C=2 F in series with a damping resistance Rf=10.7 .
The remaining elements of the SWECS prototype consist in a fully integrated VSI (Semikron SkiP 25AC), two
proprietary TMS 320 based DSP boards for the FIBC and the VSI controls implementation, current and voltage
sensors with related signal acquisition electronics (see figure 4).

Figure 5: Four-phase FIBC prototype : A: inductor (5 mH), B: electrolytic capacitors (4.7 mF), C: IGBT module with
cooler.

106
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

5. EXPERIMENTAL RESULTS

In a first step, the FIBC was tested in stand-alone mode, with a voltage source of 10 A/400 V as input, and as
load a variable power resistor (20..400 ). Figure 6 displays the four phases currents responses (il1 to il4) to a
common reference value of 3 A (Ilref).

Figure 6: Responses of the FIBC’s four phases currents.

The parameters Ak and Bk of the sliding mode controller were adjusted in numerical simulation using
Matlab/Simulink, and thereafter fine tuned in the real setup, which results in the following values : A k=20, Bk=0.
It can be observed a settling time of ca 1.3 ms, and almost no overshoot, while a moderate current ripple of ca
8% is recorded in steady state.
The same setup serves to test the PI control of the FIBC’s output (DC-link) voltage, whose proportional
(kp=0.047) and integral (ki=7.2) gains were first computed using pole placement method, then experimentally
refined in results-oriented approach.
Figure 7 displays I/O measurements of the FIBC, with DC-link reference at 700 V and Vin at 145 V. In this case,
corresponding to a relatively high transfer ratio, an output mean value of 685 V with 15 V ripple is recorded.
Transitorily, the three output signals show a settling time of ca 200 ms, with almost no overshoot. In addition,
FIBC experimental efficiency is determined for several input voltage values, covering the range [100..400] V.
For this, input and output powers are calculated in a scope (LeCroy 44Xi) as rms values over 200 ms of
instantaneous power measurements, which leads to efficiency ranging from 96% to 98 %. This result, which
stresses extremely low power losses in the FIBC, is slightly higher than the one obtained in [1] with the same
rated power (5 kW) and the same switching frequency (20 kHz), which is clearly expected since the duty cycle
in [1] (0.7) is toward the upper limit in this case (0.75, corresponding Vin=400 V), and the input current in [1] is
much higher (70 versus 10 A).

Figure 7: I/O signals of FIBC in stand alone mode : voltages: Vin, V0; currents: Iin, I0; powers: Pin, P0. 3 dB filtering option
set at each scope channel.

107
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In a second step, the FIBC converter was tested as part of the SWECS test bench, piloted by active and reactive
P/PI currents controllers, which were first theoretically optimised as previously shown, then fine tuned in the
real setup. This results in the following gains values : kp2=300, ki1=474 and kp1=1.5. In parallel, the reference
value of the DC-Link voltage (V0) is set at 700 V so as to enable power injection at 400 Vrms. Figure 8 shows
the DC-Link voltage behaviour, along with the injected active power (the reactive power reference value being
nul), this at varying input voltage (Vin). One can observe the good stability of V0 with respect to Vin and to the
injected active power into the grid, regardless some gaps that appear between real and reference powers due to
the hardware.

Figure 8: SWECS dynamic behaviour. Pgrid/Qrid : injected active/reactive power; Vin/V0 : input/output voltages of the
FIBC.

6. CONCLUSION

In this paper has been proposed an alternative topology for small wind power conversion systems, in which a
four phases FIBC boost converter has been used along with a diode rectifier, in order to replace the VSR of the
conventional back-to-back topology. The FIBC converter has been designed, prototyped and tested.
Experimental results confirms the good efficacy of output voltage and phases currents controls, and just as
important, a very high efficiency. An experimental study has also been successfully conducted at the SWECS
level, with power injection into the utility grid. The proposed topology bears a good potential to improve
efficiency and compactness of the conventional one.

ACKNOWLEDGMENT
The authors would like to thank those students of the High school of engineering and architecture, Fribourg, who actively
participated to the laboratory work related to this paper.

REFERENCES

[1] Kabalo, M & Co. Experimental validation of high-voltage-ratio low-input-current-ripple converters for hybrid fuel cell
supercapacitor systems. IEEE Transactions on Vehicular Technology 2012, 61:3430-3440.
[2] Bühler, H. Réglage par mode de glissement. Lausanne, Switzerland : Presses polytechniques romandes, 1986.
[3] Capone, M. Chaîne électromécanique à haut rendement pour la petite éolienne. MSc Deepening Project, College of
Engineering and Architecture Fribourg, University of Applied Sciences HES-SO, Fribourg, Switzerland, 2018.
[4] Islam, Md & Co. Power converters for wind turbines: current and future development. In: Mendez-Vilas A, editor.
Materials and Processes for Energy: Communicating Current Research and Technological Developments. Spain: Formatex
Research Centre, 2013, pp.559–571.
[5] El Hawary, M R & Co. Principles of Electric Machines with Power Electronic Applications. IEEE Press : A John Wiley
and sons Inc;2002.
[6] Reznik, A & Co. LCL Filter design and performance analysis for grid-interconnected systems. IEEE transactions on
industry applications 2014, 50(2):1225-1232.

108
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[7] Xuetao P. & Co. Analysis and evaluation of the decoupling control strategies for the design of grid-connected inverter
with LCL filter. In: Proceedings of the International Conference on Renewable Power Generation (RPG 2015), January
2015, Beiing: pp.1244-1249.
[8] Kabalo, M & Co. Experimental evaluation of four-phase floating interleaved boost converter design and control for fuel
cell applications. IET power electronics 2013, 6:215-226.
[9] Kabalo, M & Co. Advanced hybrid dual loop control for multi-phases interleaved floating DC-DC converter for fuel
cell applications. In : IEEE Industry Applications Society Annual Meeting, 7-11 Oct. 2012, IEEE, pp. 18.
[10] Li, S & Co. Optimal and direct-current vector control of direct-driven PMSG wind turbines. IEEE transactions on
power electronics 2012, 27(5):2325-2337.

109
COMPARISON AND ANALYSIS OF PERFORMANCE VALUES OF
FIXED ANGLED SYSTEMS AND RADIATION TRACKING
SYSTEMS IN SOLAR POWER PLANTS

Kaplan Yusuf Alper

Department of Energy Engineering, Osmaniye Korkut Ata University, Osmaniye, Turkey, [email protected]

Saldamli Mehmet Ali

EMTA Cable & Conductor, R&D Center, Osmaniye / Turkey, [email protected]

Kaplan Yusuf Alper, Saldamlı Mehmet Ali. Comparison And Analysis Of Performance
Cite this paper as: Values Of Fixed Angled Systems And Radiation Tracking Systems In Solar Power Plants. 8th
Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Renewable energy sources such as solar energy have great potential to alleviate some of the
negative environmental problems, including climate change caused by intensive fossil fuel
use.Solar energy will play an important part of future energy systems because of its rapid
installation, technology development and lowering costs. Accordingly, in order to use solar
radiation intensity in a wide range in the most efficient way, the plants must either be at an optimum
angle in the area where they are installed or they should be designed with moving systems in such
a way that they close almost all of the rays coming from the sun.The most common types of power
plants in our country and even in the world consist of systems that do not follow the fixed angle
sun. There is a certain angle in the construction of these plants by making use of certain
simulations. In this study, we will compare the performance and efficiency values of the solar
tracking system as well as the performance and efficiency values of the equivalent power plants
operating under real site conditions that are designed in a stable manner. price performance ratios
can be clearly defined and decision making will be easier for the type of power plant to be
constructed.Solar tracking systems are designed according to the most efficient position they can
get from the sun by means of radiation sensors, and are also designed with elements such as simple
computer and PLC using open-loop solar data based on meteorological data. In our study, using
three different power plant types, all possible criteria were evaluated with open loop controlled
PLC software and production differences were evaluated.According to this study, although the
efficiency rate of 4-axis tracking systems is 21% higher than that of a 2-axis system on an autumn
months, it is concluded that it produces approximately 24% more energy than fixed-angle systems
Keywords: Solar radiation intensity, simulation, open loop, PLC, 4-axis tracking system
© 2020 Published by ECRES

1.INTRODUCTION

Energy is the most important resource and strategical tool influencing the shape of society/nations for all
countries. The availability and cost of energy significantly effects our quality of life, the structure of national
economies, the relationships between nations, and the stability of our environment [1].

Renewable energy is a current issue in the world as well as in Turkey. Turkey is located at the meeting point of
Asia, Europe and Africa continents and it stands as a bridge between Europe and Asia. Its size is 779,452km2.
Turkey's population is about 83 million in 2020. As a results of the social and economic development of the
country, the energy demand of country is increasing continuously [2].

Renewable energy sources; Considering the environmental friendliness with the carbon emission rates it releases
to nature and reducing dependency due to its natural origin, they are very important resources. Renewable energy
sources are mainly solar energy; It can be diversified as hydraulic, wind, geothermal, biomass and wave energies.
Although solar energy is the most important of the energy types in this list, it can be said that it has an indirect
or direct effect on other energy types [3].

110
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Turkey has developing policies in the renewable energy field. Although it is a beneficial mean of obtaining
energy, there are barriers on renewable energy production. In order to develop renewable energies, those barriers
must be removed and suitable implementations should be adopted to overcome them. Turkey, in terms of fossil
fuel sources, can be considered as a poor country; this situation creates economical and political obstacles on
the development of Turkey. Besides these problems, Turkey has a large renewable energy potential but lacks
implementation policy and essential instruments, so the resources remain unutilized [4].

As of the end of September 2019, according to the sources of our power distribution board is shown in figure
1(a). [5] and the unlicensesd installed power by recources in figure-1(b). [6].

Figure 1(a). Energy distribution by source Figure-1(b). Distribution of unlicensed installed power
by recources

According to the data in the chart above, the share of Solar Power Plants across the country is around 6.2%. And
also, the unlisenced power plant rate from the photovoltaic is 94,7%. This rate has reached these levels with the
facilities built within the scope of the “Regulation on Unlicensed Electricity Generation in Electricity Market”
published in the official newspaper number 28783 dated 02.10.2013. Similarly; Within the scope of the
regulation published in the official newspaper dated 12.05.2019 and numbered 30772 [7, 8,9]. this rate will
increase much more with the new facilities to be applied within the scope of the establishment of Solar Power
Plants for roof and facade applications, and the share of the sun will gain more importance in the production of
electricity [10,11].

The yearly solar irradiation and monthly variation of average daily insolation duration of Turkey are given
Figure 2 and Figure 3, respectively [9].

Figure 2. The Solar Potential map of Turkey (GEPA)

111
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Monthly variation of average daily insolation duration [10]

Considering the above figures 2 and 3, the importance of Turkey's solar energy potential is understood more
clearly. Turkey's annual theoretical solar energy potential is estimated at 6150 TWh, and the 305 TWh of this
potential is estimated to be economically available [12].

2. HIGHLIGHTS OF RESEARCH

As a result of irregular field layouts / installations, existing roof layouts and inexperienced applications, the
efficiency of solar panels, which are already inefficient, decreases more and the return time of the investments
increases. In order to eliminate the negative effects of these and similar situations, or to obtain the maximum
value that can be obtained from the inefficient panels, it is important to set up the power plant installations under
optimum conditions before designing them. The power plants built in our country are optimally designed with
fixed angle power plants resulting from the simulation applied by the private companies within the framework
of mathematical modeling criteria prepared according to the past radiation values prepared by the computer
environment. The purpose of our study is to perform efficiency analysis and production calculation under real
field conditions and with real values close to 100%.

3.METHODOLOGY

In this study, a comparative efficiency analysis was carried out in Osmaniye, Kadirli, with the same power and
location, with three different operating systems. Each power plant; These variables are considered constant and
only production values are compared because they are affected by the same temperature, the same temperature,
the same contamination and the same environmental conditions. In each power plant; 2KW Goodwe brand
inverter and 3 series connected Yingli Brand 255 Wp solar panel was used. Cable cross sections and lengths are
also kept equal.

Electrical Properties of Materials;

Details including electrical properties of Inverter and PV panels, which are the basic materials in the installation
of power plants, are available in the tables below.

Table 1. GW 2000-NS Electrical Properties, YINGLI Solar Panel Electrical Properties [13,14]

112
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

First Santral Type;

By examining the types of plants in our location, the plants were installed at the same value as the facilities
established at an angle of 250 degrees to the sun in the south direction. With the adjustable bolt mechanism on
the same table, the facility was brought to a fixed angle and tightened to remain the same throughout the year.
Here, during the working period, only the production data of a facility with 25 0 fixed views to the south was
obtained.

Picture-1. Fixed angle plant (GES-1), Figure-4. GES-1 Production values

GES-1 Production Values

120 kWh 106.2 kWh
100 kWh 88.5 kWh
80 kWh 79.6 kWh
60 kWh 56.8 kWh 72.7 kWh 69.6 kWh
40 kWh
20 kWh
0 kWh
Sep-19 Oct-19
Nov-19 Dec-19
Jan-20
Feb-20

Picture-1. Fixed angle plant (GES-1) Figure-4. GES-1 Production values

Second Santral Type;

According to the annual movement of the sun, a power plant following the sun was established in the North -
South direction. East-West position is fixed on the table with adjustable bolt mechanism. The North-South
position was designed to follow the annual solar position automatically with a linear motor driven by a 24V DC.
In this facility, it receives sunlight at a 900 degree angle during the noon. Therefore, a more efficient power plant
was built compared to fixed facilities. The production differences between the fixed power plants and the
production data received during the study period were evaluated. Picture-2: Double Axis Tracking System
(GES-2), Figure-5. GES-2 Production Values

GES-2 Production Values

120 kWh 107.8 kWh
100 kWh 90.7 kWh
84.44 kWh 77.8 kWh
80 kWh
60.9 kWh 71 kWh
60 kWh
40 kWh
20 kWh
0 kWh

Picture-2: Double Axis Tracking System (GES-2) Figure-5. GES-2 Production Values

Third Santral Type

A power plant, which positions itself according to the daily and monthly movements of the sun, has been
established. A table operating with 24V DC linear motors following both the East-West position and the North-
South position has been installed. Daily action; the daily movement of the sun was monitored with the S7-1200
PLC using the sunrise-sunset data of the Boğaziçi University Kandilli Observatory and Earthquake Research
Institute ASTRONOMY LABRATORY [15]. Solar tracking software is 100% written with our own algorithms
and was created to follow the sun in the most efficient way. A very stable algorithm has been established by
eliminating the conditions such as sensor, faults, and light scattering caused by the cloud of the air. Seasonal

113
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

movement, on the other hand, was calculated proportionally by the annual movement of the sun in the north-
south direction, and was controlled by the PLC to follow the sun. In this way, it is ensured that the sunlight
always comes to the panels at an angle of 900 at any time of the day and maximum efficiency is obtained from
the sun. The programming written for the annual movement was also used for the control of the GES-2 plant.
The data received during the production period were compared with the production data of other power plant
types and the production differences between them were evaluated. Picture-3. Four Axis Tracking Santral (GES-
3), Figure-6. GES-3 Production Values

GES-3 Production Values

134,1 kWh
150 kWh 109 kWh
98.93 kWh
88.7 kWh
100 kWh 83.6 kWh
69.4 kWh
50 kWh
0 kWh

Picture-3. Four Axis Tracking Santral (GES-3) Figure-6. GES-3 Production Values

4.RESULTS

The data of production values can be made on-line by following the SEMS PORTAL page of GoodWe Inverters
or historical production reporting can be obtained. A second monitoring is also done using KAEL energy
analyzers to prevent data loss. The above production values are graphs formed as a result of data from both
monitoring modules. In the following Image-4 and Figure-7, general views of the three plants and 6-month
comparative production values are given.

GES Production Values Comparision

500 kWh 411.50678

331.1 kWh kWh
400 kWh 343.84 kWh
300 kWh
200 kWh
100 kWh
0 kWh
GES-1
GES-2
GES-3

Picture-4. Solar Plant Overview Figure-7. Plants Production Values

Looking at the above data, there are differences in the amount of energy produced by three different systems in
the same time periods. The figüre-5 above, shows the amount of energy produced by all three plants over a 6-
month period. When we examine the above data, we can reach the following results..

6- month production period summary;

Total energy production of GES-1 Plant: 331,1 kWh
Total energy production of GES-2 Plant: 343,8 kWh
Total energy production of GES-3 Plant: 411,5 kWh
If we calculate the rates in between;
((411,5 kWh - 331,1 kWh) / 331,1 kWh)*100 = %24.28 (GES-3 versus GES-1)
((343,8 kWh – 331,1 kWh) / 331,1 kWh)*100 = %3,85 (GES-2 versus GES-1)

114
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

While the production difference between a four-axis tracking system and a fixed system is about 24%, the
production difference between a two-axis tracking system and a fixed system is about 3.8%.

5.CONCLUSION

In this study, three different types of power plants under the same climate and ambient conditions were evaluated
for the Kadirli district of Osmaniye, considering the shadow, pollution, temperature and other conditions as
equal. Production data were collected for each plant separately. Data collection processes were collected with
the help of Sems Portal and Kael Energy analyzers. According to the data obtained, it is concluded that a 4-axis
power plant is 24.71% more efficient than a fixed power plant in a 3-month period and a 2-axis power plant
produces 3.13% more energy than a fixed power plant..
REFERENCES

[1] Kaplan Y. A., “Overview Of Wind Energy In The World And Assesment of Current Wind Energy Policies in Turkey”
Renewable and Sustainable Energy Reviews, (2015) pp. 562-568
[2] Benli, Hüseyin. "Potential application of solar water heaters for hot water production in Turkey." Renewable and
Sustainable Energy Reviews 54 (2016): 99-109
[3] E. T. Karagöl, İ. Kavaz "Renewable energy in the world and in Turkey in April 2017 Number: 197 SETA
[4] Demirdizen, Hasan Gence. “market development of renewable energy in turkey. diss. middle east technical university”,
2013
[5] https://www.teias.gov.tr/ Teiaş Electric Statistics Working Group 07.07.2019, p.1
[6] http://epdk.gov.tr/ Electricity Market 2018 Market Progress Report, p.4
[7] Rincon, L., Puri, M., Kojakovic, A., & Maltsoglou, I. (2019). The contribution of sustainable bioenergy to renewable
electricity generation in Turkey: Evidence based policy from an integrated energy and agriculture approach. Energy
policy, 130, 69-88.
[8] https://www.resmigazete.gov.tr/eskiler/2019/05/20190512-1.htm Sunday, May 12, 2019, Official Gazette No. 28783,
Regulation on the production of unlicensed electricity in the electricity market
[9] “Photovoltaic Technology: Turkey and the world situation, General Photovoltaic Applications and Textiles”,
Electronic Journal of Textile Technologies 4(2), 2010 (43-58)
[10] Topkaya S.O., “A discussion on recent developments in Turkey’s emerging solar power market”, Renewable and
Sustainable Energy Reviews 16 (2012) 3754– 3765
[11] Kaygusuz, K., & Avci, A. C. (2019). Potential and utilization of solar energy policies in Turkey. Journal of Engineering
Research and Applied Science, 8(1), 1087-1098.
[12] Melikoglu, M. (2017). Pumped hydroelectric energy storage: Analysing global development and assessing potential
applications in Turkey based on Vision 2023 hydroelectricity wind and solar energy targets. Renewable and Sustainable
Energy Reviews, 72, 146-153.
[13] http://www.goodwe.com.tr/media/files/ns%20seri%20datasheet.pdf
[14]https://www.merkasol.com/WebRoot/StoreLES/Shops/62387086/4F37/C141/6C65/6629/D59B/C0A8/28BC/EC37/P
anel_Solar_Yingli_235_EN.pdf
[15] Bogazici University Kandilli Observatory and Earthquake Research Institue, Annual Solar Times for Osmaniye City

115
 ON RENEWABLES AND THE RESILIENCE OF SOCIO-
TECHNOLOGICAL ENERGY SYSTEMS

Amadi-Echendu Joe
Dept of Eng & Technology Management, Univ of Pretoria, South Africa, [email protected]
ORCID: 000-0001-9839-7621

Thopil George Alex

Dept of Eng & Technology Management, Univ of Pretoria, South Africa, [email protected]
ORCID: 0000-0002-4196-2455

Amadi-Echendu, JE, Thopil, GA. On renewables and the resilience of socio-technological

Cite this paper as:
energy systems. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The era of Society 5.0 cum fourth industrial revolution is characterised by the fusing of biological,
ecological, physical, technological and socio-political worlds, and leads to the notion of ‘socio-
technological cyber physical systems’ where cyber physical and social systems are interdependent
and inextricably interwoven. Socio-technological energy systems represent an intriguing class of
interconnected systems and the resilience of such systems is an imperative for sustainability. This
paper discusses the resilience and vulnerability of a case study socio-technological energy system.
It is intriguing that robustness inherent in the cyber physical subsystems imputes absorptive
resilience to the case study socio-technological energy system towards the introduction of
renewable source technologies.
Keywords: Energy systems, Socio-technological energy systems, Resilience, Vulnerability
© 2020 Published by ECRES

Nomenclature
STCPS Socio-technological cyber physical system
STES Socio-technological energy system
CPS Cyber physical system
4IR Fourth industrial revolution
VUCA Volatile, uncertain, complex, ambiguous
ESI Energy systems infrastructure
KIC Key industrial customers

1. INTRODUCTION

The Society 5.0 concept epitomizes the complex consequence of 4IR fusing of biological, ecological, physical,
and socio-political worlds. The Society 5.0 concept represents a society where we can resolve various social
challenges by incorporating innovative 4IR technologies into every human endeavour in a manner that
continuously creates new values, making “humanity more conformable and sustainable” [1] [2]. Society 5.0
encompasses four preceding societies, viz – (i) hunting and gathering existence, (ii) agriculture-driven
subsistence, (iii) industrialised commerce, (iv) information driven globalisation, and (v) fusing of worlds [3].
The era of Society 5.0 cum 4IR is further characterised by the trendy management acronym: VUCA – that
is, “volatility, uncertainty, complexity, and ambiguity” [4] featuring (i) unexpected or unstable change, (ii) cause
and effect uncertainty, (iii) countless and vagarious interconnectivity and inter-communicability, and (iv)
“unknown unknowns”.
The era of Society 5.0 cum 4IR is also characterised by hyper-connectivity such that “information about
everything” has become a pervading cliché. This cliché may be attributable to the exponential possibilities
proffered via 4IR technologies. The technological possibilities increasingly manifest unprecedented paradigm
shifts in societal transformations, implying that technological possibilities and societal transformations are
intertwined [5]. For instance, 4IR technologies not only build upon the interconnectivity and inter-
communicability already established in the preceding era but also, the uncanny behaviour of humanoid robots

116
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

tantalizingly extends Society 5.0 human intelligence. In fact, the increase in human-machine interfaces and
deployment of humanoid robots are manifestations of the rapid and high degree of convergence between the
virtual (cyber) and real (physical) realms. Inadvertently, the convergence of these realms raises the notion of
socio-technological cyber physical systems (STCPS)2 as a complex consequence of the era of Society 5.0 cum
4IR.
The VUCA challenge is also an unintentional consequence of “information about everything” cliché where
extraordinary levels of interconnectivity, inter-communicability, and interdependency are interwoven such that
the effects of uncertainty readily propagate between the constituents and subsystems of STCPS [6]. For example,
a change in a constituent 3 or subsystem could trigger a massive change in other constituents and subsystems as
well as the overall STCPS. In this regard, resilience and vulnerability have taken on renewed significance for
sustainable management of STCPS, and especially, our intriguing category of socio-technological energy
systems (STES). In this regard, the discourse in this paper deals in part with the broader question of how to
sustainably manage the underlying cyber physical infrastructure so as to meet the ever-increasing demand for
energy to serve humanity’s appetite for instant gratification. Hence, we focus on describing the resilience and
vulnerability of a case study socio-technological energy system. The notion of STCPS and the STES derivative
is discussed in section 2 together with the concepts of resilience and vulnerability. A concise discourse on the
case study STES is presented in section 3. Further discussion on the resilience and vulnerabilities associated
with the case study STES is included in section 4 with concluding remarks in section 5.

2. SOCIO-TECHNOLOGICAL SYSTEMS, RESILIENCE AND VULNERABILITY

The notion of STCPS and the STES derivative is discussed in this section together with the concepts of resilience
and vulnerability.

Socio-Technological Energy System

Figure 1. The notion of Society 5.0 cum 4IR socio-technological energy system.

For the purposes of our discourse, we posit that a STCPS arises from digitalized technology fusion of social
systems with cyber physical systems (see Figure 1). This means that infrastructure systems for education, energy,
health, security, and transportation et cetera are inextricably interwoven with social systems in the service of
humanity [7]. It is not surprising that technological revolutions in infrastructure influence economic [8] and
societal transformations [5] and vice versa. As illustrated in Figure 1, a social system consists of interdependent
and interrelated cultural, as well as communicative and intelligent structural elements that constitute a logically
coherent whole. The digitalized technology driven interdependencies and interrelationships accentuate the
behaviour and performance of the overall system to be better or greater than the sum of the individual behaviours
and performances of the respective constituents of the system [9]. Similarly, a cyber physical system (CPS)

2 (see more detail in section 2)

3 Social systems incorporate geopolitical and/or special interest constituencies

117
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

consists of interconnected and interdependent engineered elements that incorporate computational algorithms
and communication capabilities. Fusion between the social and cyber physical subsystems is underpinned by
digitalized technology platforms of the 4IR era. Furthermore, 4IR technologies (e.g., blockchain plus artificial
intelligence manipulations of big data) facilitate coordination, control, and monitoring of the complex operations
of the fused STCPS.
The 4IR parlance of internet-of-things (IoT) presupposes that every component, constituent, element, and
subsystem of a STCPS can interconnect and inter-communicate (cf: IoT enabled device). This means that every
IoT enabled component, constituent, element, and subsystem represents a node, and the linking of the nodes
establishes networks within the STCPS. Assuming that each node can be linked to another, then, the linking of
𝑛(𝑛−1)
all the nodes establishes a meshed network topology which comprises 𝑙 = 2 links; where n is the number
of nodes and l is the number of links. Each link will have a weight 𝑤𝑖,𝑖=1…𝑙 that determines the strength of the
relative interrelationships and interdependencies between the respective nodes in the network. Theoretically,
many network topologies are possible depending upon the number of nodes and the composition of the links
between the respective nodes in a STCPS. For the case where the CPS comprises energy infrastructure, the
corresponding STES encompasses the pertinent physical assets, organizational constituents or stakeholders.
Given the hyper contexts of information about everything, interdependency, interconnectivity, inter-
communicability, it is conceivable that the linkages and networks within the STES will evolve and transform in
response to changes or stresses that emanate from sources endogenous or exogenous to the system.

Resilience and Vulnerability

Acknowledging that a STES can be subjected to VUCA-like challenges from endogenous or exogenous sources,
then the resiliency of the constituents and subsystems (and even that of the overall system), takes on pivotal
importance for sustainable management of STES. Resilience is often defined in the context of CPS (e.g., [10]
[11] [12]), however, given the notion of STES, we adopt a broader definition of resilience as systemic processes
and capacities to accommodate transient, intransigent, or evolutionary stress from endogenous or exogenous
sources.
This implies that resilience encompasses the authoritative ability with which a system ‘absorbs’, ‘recovers’, and
‘restores’, as well as ‘adapts’ and ‘evolves’ when subjected to vagarious combinations of transient, intransigent
or evolutionary stressors. This broader definition takes into consideration that the overall behaviour and
performance of the STES should be better or greater than the sum of the behaviours and performances of the
constituents and subsystems. Furthermore, the normative functionality of the STES only describes one of the
many behaviours and performances that the STES could exhibit. The extent to which the STES absorbs the
transient stressors before recovery establishes the robustness of the system, whereas the resourcefulness of the
STES is extent to which the system restores, adapts and transforms in conformance to the intransigent or
evolutionary nature of the stressors. It is pertinent to consider that the robustness of the system may be distributed
between the social constituents and the cyber physical elements, while the resourcefulness will invariably be a
predominant feature of the social constituents of the STES. Therefore, resilience may not only be regarded in
terms of the normative functionality desired from the cyber physical elements, but also, resilience must
encompass the behaviours and performances of the social constituents of the STES.
On the one hand, resilience is generally regarded in terms of how a system responds to stressors, but, on the
other hand, vulnerability is about intrinsic or inherent weaknesses that expose a system, or make the system
readily susceptible to stressors [cf: 13]. A system is regarded as being vulnerable if the intrinsic or inherent
weaknesses are without protection against endogenous and exogenous stressors. Given the extraordinary levels
of interconnectivity, inter-communicability and interdependency evident in the era of Society 5.0 cum 4IR, it is
implicit that the effects of uncertainty can readily propagate between the constituents and subsystems of STES.
For instance, a change in the behaviour of a social constituent or an attack on a cyber physical element could
trigger a massive change in other constituents and subsystems as well as the overall STES. From a management
viewpoint, resilience and vulnerability constitute a dicephalous pair in the sense that the identification of a
system’s inherent weakness or susceptibility is concomitant to the identification of the system’s robustness.

3. CASE STUDY OF A SOCIO-TECHNOLOGICAL ENERGY SYSTEM

As mentioned earlier, the discourse in this paper deals in part with the broader question of how to sustainably
manage the functional and technical performance of the underlying cyber physical infrastructure so as to meet

118
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the ever-increasing demand for affordable and reliable energy supply. Our approach to tackling this question
begins with the recognition that increasing concerns about the environment and increased demand are drastically
influencing global energy mix (see, for example [14] [15]). In particular, emerging economies are expected to
experience the highest growth in energy demand, and it is acknowledged that renewable source technologies
will increasingly be introduced and deployed towards meeting the anticipated growth in energy demand. It is
remarkable that modern power grids are being designed and built with the capacity to integrate renewable and
legacy technologies. For developing countries and/or emerging economies, the challenge of integrating
renewable and legacy technologies is exacerbated by the complex conflation of issues of governance, markets,
regulation, social and economic inclusivity, et cetera. Given this context, we have narrowed our discourse here
to a desktop examination of the resilience cum vulnerability of the STES in a chosen developing/emerging
economy.
For brevity, the nexus and central constituent of our case study STES is a state-owned enterprise that
monopolistically controls ownership, operation and regulation of the energy infrastructure and utility in the
country. Interestingly, the nexus of our case study STES is the vertically-integrated energy systems infrastructure
(ESI) created circa 1920s to meet the then needs of an emerging industrialisation premised upon the extraction
of mineral resources. Over time, an exclusive political regime more or less ‘strengthened’ the ESI monopoly
through increased state control mechanisms that guaranteed security of supply whilst mitigating economic
sabotage. Within the past twenty-five years, a post-transformational regime leveraged the increased
monopolistic robustness of the vertically-integrated ESI enterprise towards addressing the pressing needs for
socio-economic transformation of the polity. Indications are that the global financial crisis that occurred in 2007-
8 prompted preliminary but serious questions about the need to reform the case study STES [16]. Subsequently,
increasing concerns about the environment, significant increases in electricity tariffs coupled with poor
reliability of supply and ageing infrastructure, plus other issues of inclusivity have fuelled zealous debates
around the structure, legislation, regulation, and management of the STES [16]. Currently, there are many
vociferous arguments for and against how the vertically-integrated monopoly enterprise and its composite cyber
physical infrastructure should be restructured. It is remarkable that informal arguments and debates are
accentuated by the challenge of how to integrate renewable source technologies within the legacy ESI. Given
our STCPS notion, a brief interpretation of the case study STES is depicted in Figure 2.

Figure 2. Case study socio-technological energy system.

The core of the STES is a tightly connected generation, transmission, and distribution/reticulation infrastructure
which is completely controlled by a vertically-integrated business enterprise. Revenue for the enterprise is
primarily derived from a variety of customers, especially the historically key industrial consumers (KICs) of
electricity. The customers who desire a normative behaviour and performance of the STES interface directly
with the ESI. Here, normative behaviour and performance include, inter alia, affordable and stable tariffs,
reliable supply and quality service. It is worth mentioning that a significant portion of the tariffs is bound in long
term agreements with historical KICs of electricity. The legacy long term contracts were established to guarantee
security of supply whilst mitigating economic sabotage in the era preceding Society 5.0 cum 4IR.
The vertically-integrated business enterprise is owned by a shareholder that is represented by a number of public
institutions. Invariably, the multifarious representation of the shareholder has created perplexing accountabilities
and responsibilities within the STES. For instance, the state as the sole shareholder to the STES maintains a

119
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

strong ideological stance of utilizing the ESI business enterprise to deliver on short term as well as long term
socio-economic objectives. This strong ideological stance tends to be diminished due to political wrangling
between the governance institutions. Thus, such ambiguity and vagueness of the governance institutions
inadvertently accentuate the monopolistic control of the business enterprise as the respective governance
institutions seek to remain tightly coupled to the ESI enterprise well beyond the cybernetic layer.
The STES also features a variety of interest groups among its stakeholders. Some of the interest groups interface
with the ESI enterprise only at the cybernetic layer. Others (e.g., trade unions) and also the governance
institutions interface with the business enterprise beyond the cybernetic layer. Inextricably, several of the social
constituents of the STES interface directly with the ESI. For example, special customers such as municipalities
and key industrial customers directly interface with the cyber physical elements of the ESI.

4. DISCUSSION

Acknowledging that the context is the era of Society 5.0 cum 4IR, then our depiction in Figure 2 intrinsically
implies that there are linkages between the social constituents and cyber physical elements of the STES. In
essence, the social constituents or behavioural components, that is, the governance institutions, interest groups,
customers, and ESI business organization are multifaceted and digitally interconnected constituents, and they
are also interconnected to the cyber physical elements of the STES through various digitalized technology
platforms. Some of the links and networks between the STES components have been established in cyberspace
through more formal means, i.e., legislation, regulations and standards, other links and networks have been
established via cultural or traditional conventions. For example, ownership and control of the ESI physical assets
have been established by law while relationships between trade unions and the physical assets are more or less
conventionally defined. The networks of interrelationships and interdependencies between the respective social
and cyber physical components, as well as the inter-communicability among and between all the STES
components are all facilitated by virtual and real linkages.
From the viewpoint of resilience, there are various networks between the components of the STES and the
linkages determine both the robustness and vulnerabilities associated with the system. The goal for each
behavioural component of the STES is to achieve its own objectives and in the course of doing so, each
constituent introduces change into the STES. At certain periods in time, the behavioural components may align
towards particular objectives, and such alignment introduces endogenous change that not only creates new
networks and linkages but also sustains existing networks by strengthening the necessary linkages. As reported
in [16] and mentioned earlier, during a previously exclusive economic and socio-political dispensation,
specialized linkages and networks were established between the sole shareholder of the ESI enterprise, the ESI
business organization, and the KICs.
It can be argued that over time, these specialized linkages and networks implanted a certain level of robustness
and absorptive resistance of the STES toward certain kinds of exogenous change. The ongoing perplexities in
accountabilities, decisions, and responsibilities of the governance institutions inadvertently stretch Δtabsorption (see
Figure 3) thereby accentuating the absorptive resistance of the STES. Also the strong ideological stance of the
current democratic dispensation to usurp the existing structure of the ESI business enterprise (in order to deliver
on short term as well as long term socio-economic objectives) further serves to sustain the inherent robustness
and absorptive resistance of the STES. The ongoing dilemma of usurping the inherent robustness and absorptive
resistance within the STES is manifesting as sacrificial deterioration in the functional and technical performance
of the legacy energy infrastructure, as evidenced by increasing load shedding over longer periods of time [17].

120
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Absorptive resilience phase of case study STES.

In conjunction with the current democratic dispensation, pervasive globalisation and erstwhile 2008 financial
crises, the global trend towards the introduction of renewable energy sources (in response to increasing concerns
about the environment) combine to induce transient, intransigent, and evolutionary changes throughout the
energy sector. As illustrated in Figure 3, we narrowly define normative functionality and performance of the
STES in terms of (i) stable and reliable supply of electricity by the ESI enterprise at (ii) affordable tariffs (prices)
to all categories of customers, aggregated with (iii) an assessment of the current condition of the ESI. The shapes
and lines depicted in Figure 3 are derived from our manipulation of publicly available data and information (e.g.,
[17]).
In our case study STES, we posit that the inherent robustness and absorptive resistance hitherto established in
terms of the existing interdependencies between the social constituents are making it difficult for the STES to
recover so as to restore the ESI to acceptable levels of performance. Most of the publicly available data about
the STES support the degrading trend in the overall behavior and performance of the STES as illustrated in
Figure 3. The degradation is magnified by a more rapid deterioration in the functional performance of the ESI
coupled with puzzling behavior of the incumbent ESI enterprise. For our desktop study, we surmise that
irrespective of the availability of renewable energy technologies, the case study STES currently exhibits
absorptive resilience in the sense that the inherent robustness is evidently being usurped since current
interventions (resourcefulness) are seemingly absorbed without concurrent recovery or restoration of the ESI.
The corollary is that the usurping of inherent resilience makes the STES more vulnerable to increasing VUCA
challenges (cf: [18]).

5. CONCLUDING REMARKS

In conclusion, our resilience approach to examining the case study STES complements views that have been
expressed by others (e.g., [16] [17]). Many formal and informal views continue to vociferously indicate that the
case study STES should be urgently reformed to take advantage of renewable technologies to stimulate and
bolster recovery and restoration of the legacy ESI in the first instance; then in the second instance, to cause the
ESI in particular, and by implication the overall STES to adapt and transform through increased and substantive
integration of renewable source technologies. As an inference, we postulate that in the era of Society 5.0 cum
4IR, the tendency exists for a STCPS to exhibit absorptive resilience by usurping inherent robustness in the
composite CPS. The manifestation of absorptive resilience tendency could be more conspicuous especially
where there is ambiguity in the respective behaviours and performances of the social constituents.

REFERENCES

[1] Fukuyama, M., “Society 5.0: Aiming for a New Human-Centered Society” Japan SPOTLIGHT • July / August 2018

121
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[2] Őnday, Ő, Japan's Society 5.0: Going Beyond Industry 4.0 https://www.researchgate.net/publication/330500307 Jan
2019.
[3] Salgues, B., Society 5.0: Industry of the Future, Technologies, Methods and Tools. ISTE Ltd and John Wiley & Sons
Inc, 2018. ISBN 978-1-78630-301-1
[4] Bennet, N., Lemoine, GJ. What VUCA really means for you. Harvard Business Review, Jan – Feb 2014
[5] Amadi-Echendu, J.E., (2018). The role of technology in societal transformation. In Innovation and Technology for
Societal Transformation - Proceedings of the 4th Technical University of Kenya International Conference, Mbai, F &
Kwanya, T (eds). ISBN 978-9966-132-24-8 pp1-3
[6] Whiting, K., Coronavirus isn't an outlier, it's part of our interconnected viral age. World Economic Forum. 4 March
2020.
[7] Mansoor, Z and Williams, M.J. Systems Approaches to Public Service Delivery: Lessons from Health, Education, and
Infrastructure. “Systems of Public Service Delivery in Developing Countries” on 14-15 May, 2018, at the Blavatnik
School of Government, University of Oxford
[8] Carlota P., 2009. "Technological revolutions and techno-economic paradigms," The Other Canon Foundation and
Tallinn University of Technology Working Papers in Technology Governance and Economic Dynamics 20, TUT
Ragnar Nurkse Department of Innovation and Governance.
[9] Crossman, A. "Social System." ThoughtCo, Jun. 22, 2018, thoughtco.com/social-system-302659.
[10] Arghandeh, R., von Meier, A., Mehrmanesh, L., & Mili, L., (2016). On the definition of cyber physical resilience in
power systems. Renewable and Sustainable Energy Reviews, 58: pp1060–1069
[11] Southwick, S.M., Bonanno, G.A., Masten, A.S., Panter-Brick, C., & Yehuda, R., (2014). Resilience definitions, theory,
and challenges: interdisciplinary perspectives. European Journal of Psychotraumatology 5:25338
[12] Fiksel, J., (2006). Sustainability and resilience: toward a systems approach, Sustainability: Science, Practice and
Policy, 2:2, pp14-21
[13] Moret, W., (2014). Vulnerability assessment methodologies: A review of the literature. A Report for USAID ASPIRES
PEPFAR fhi360
[14] Wood, J., Fukushima gets new life as a renewable powerhouse after the 2011 meltdown. World Economic Forum, 25
Feb 2020.
[15] IRENA, 2019. International Renewable Energy Agency database dashboard.
http://resourceirena.irena.org/gateway/dashboard/?topic=4&subTopic=19
[16] Khan, M., Thopil, G.A., & Lalk, J., (2016). Review of proposals for practical power sector restructuring and reforms
in a dynamic electricity supply industry. Renewable and Sustainable Energy Reviews 62 (2016) pp326–335
[17] Wright, J.R., & Calitz, J., Setting up for the 2020s: Addressing South Africa’s electricity crisis and getting ready for
the next decade. A CSIR Energy Centre Presentation, Pretoria January 2020.
[18] Global Risks Report 2020, World Economic Forum, 15th Edition Insight Report

122
 CFD INVESTIGATION ON PISTON HEAD GEOMETRY EFFECT
ON IN-CYLINDER FLOW FOR HYDROGEN FUELED ENGINE

Raid Oyouni
Bishah Technical College, Department of Mechanical Engineering, Bishah, Saudi Arabia, [email protected],
ORCID: 0000-0001-5217-8327

Rafaa Saaidia
Bishah Technical College, Department of Mechanical Engineering, Bishah, Saudi Arabia, rafaasaaidia@gmail,
ORCID: 0000-0002-3892-9726

Oyouni, R, Saaidia, R. CFD Investigation on piston head geometry effect on in-cylinder flow
Cite this paper as: for hydrogen-fueled engine. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,
Turkey

Abstract: In this paper, a CFD simulation is carried out. The numerical investigation aims to compare the
effect of the pistons heads geometries on the in-cylinder flow when the hydrogen is used as fuel.
The in-cylinder flow velocity and turbulent kinetic energy were studied during the intake phase.
The calculation was investigated in two plans to describe the swirl and tumble aspect. To reach
this goal, a 3D CFD code was adopted. For a fixed engine speed, the simulation was performed
for hydrogen fueled engine via external mixtures. Whereas, three pistons configurations were
determined in three shapes (concave, plate, convex). SWFS code was used to perform a 3D
simulation at specific angles of the Crank Shaft (100, 600, 1000 ). The reveled results showed that
the choice of piston is fundamental, particularly its geometry of head which has a great influence
on the flows that occur in the cylinder. Referring to the cartography which presents the evolution
of velocity and turbulence we concluded that better performance have appeared when the convex
piston was used, particularly that of filling in-cylinder and vortices which are far from the walls of
the cylinder.
Keywords: Hydrogen engine; Piston Geometry; Flow Simulation; CFD
© 2020 Published by ECRES

Nomenclature
IC Internal combustion engines
CFD Computational Fluid Dynamic
H2ICE Hydrogen internal combustion engine
H2FC Hydrogen fuel cells

1. INTRODUCTION

The present worldwide society faces major issues. Among the most significant through them, that of the rise of
the total population, the increase of the demand for energy in various fields of the industry and the accessibility
of energy sources with the lowest cost. The petroleum derivatives, in particular, fossil fuel is rapidly run out. In
fact, it is considered one of the causes of environmental pollution and are accordingly not classified as a
permanent solution of energy [1]. Furthermore, any deficiency that occurs to these energy sources will
negatively poses to the global economy [2]. In addition, the increased use of fossil fuels leads to air pollution, a
lack of quality and expands the amount of greenhouse gases (GHG). The importance of hydrogen as a fuel in
vehicles (cars, trucks, buses, motorcycles) has become a subject of interest because of the severe external factors
generated via the large-scale impacts of air pollution and global warming. Hydrogen is a reliable fuel, this is due
to the high potential because of its multiple properties. Hydrogen as a fuel has a lot of particular characteristics
That makes it better than conventional fuels. Simultaneously, similar properties have been found to be liable for
a group of undesirable combustion issues, for instance, pre-ignition, flashback, and High NOx emissions in High
load situations [3].

123
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Properties such as density, diffusion, minimum ignition energy, auto-ignition temperature, flame speed,
flammability limit, and deactivation distance have been found to be helpful for the operation of the Internal
Combustion engine.
The lightest element is Hydrogen; It has an extremely low density and a low specific volume. There is, therefore,
a low efficiency when hydrogen is utilized at the intake of an IC engine. One of the positive characteristics of
low energy density is at times Being useful for causing the fast scattering in the air. High diffusivity features an
inclination for quicker spillage through openings and micro-cracks, which additionally permits leaked hydrogen
to diffuse quickly in an incombustible proportion [4,5].
The minimum ignition energy of the hydrogen-air mixture is around one-tenth of that of the methane-air mixture
[6]. This property for the hydrogen engine facilitates the ignition with a very low energy spark. However, the
concentration of the remaining energy (suffix of combustion stroke) and might happen as a pre-ignition issue in
the form of a hot spot. The hydrogen extinction distance is much less than that of gasoline and methane. The
hydrogen extinction distance is much less than that of gasoline and methane. Due to properties, for instance, the
lower quenching distance and the minimum ignition energy, the hydrogen combustion is not only begun with a
low energy spark, but it is likewise increasingly hard to extinguish and therefore easier to preserve it.
The spacious of hydrogen flammability range permits to drop for the rich mixture in the engine. The output
power will decrease because of the reduction of equivalence ratio that Accompanying to low fuel injected. The
hydrogen flame of high speed has a clear impact in the performance of the engine.
There is a many research carried out to decide the best kind of fuel as an energy bearer for output power in the
engine. Whilst, there are constantly many aspects that should be taken into consideration, for example, cost,
comfort, greenhouse gas emissions, and client acceptance which isn't easy to achieve.
Most of these studies are comparing hydrogen internal combustion engine vehicles (H2ICE) and fuel cells
(H2FC) with electric and gasoline vehicles. The comparisons depend on carbon dioxide (CO2) emanations and
essential energy consumption. The results justify the hydrogen advantages compared to other fuels [7]. In 1924,
Ricardo accomplished some works on the performance of the engine running on hydrogen. Erren et al. had a
work-study which treats the advantage of utilizing hydrogen as a fuel concerning to non-polluting aspects during
the 1930s [8, 9]. Hydrogen has long been recognized as a fuel with highly desirable properties for application
in IC engines.
At present, hydrogen is utilized in Several applications. Adt and Swain tested the emissions and efficiency of
the partially loaded four-cylinder engine changed over to work with hydrogen and methanol [10]. They obtained
results which demonstrate that the NOx emissions from engine powered by hydrogen were lower than from the
engine powered by methanol; and that the thermal efficiency of a hydrogen engine was higher than the methanol
engine. Tianshen et al. considered the impact of hydrogen methanol mixture as an additive for a spark-ignition
engine to improve the internal combustion process [11]. Their outcomes demonstrated that a limited quantity of
hydrogen provided to the methanol-air mixture would increase the velocity of flame propagation, minimize the
time of the ignition delay and improve efficient thermal efficiency.
Given the assortment of IC engines, hydrogen has been attempted a few times as an alternative fuel, by and
large, it is utilized as a secondary additive to fossil fuels. Because of its physical properties, the utilization of
hydrogen as the fundamental and sole fuel is limited.
The utilization of hydrogen could add to the accomplishment of environmental protection objectives. In fact,
hydrogen can be gotten from water and, by the time of its combustion, produces the water only. From an
environmental standpoint, the most invaluable method of production is that of electrolysis when the electricity
utilized in this process is of renewable origin (wind energy, hydraulic energy, solar energy).
In scientific terms, the use of hydrogen in thermal engines is helpful from an environmental side of view, but it
is constantly joined by a deterioration in performance. Moreover, the in-cylinder flows in S.I engines have far
greater influence on the engine behaviors more than just the effective filling of the cylinder.
The design of the intake parts in the engine itself significantly impacts the flow in the cylinders. In particular,
the piston which is the driving force behind the intake. Many researches in this context have been done to talk
about the impact of the geometry of the head on the flow and movement development of fuel mixtures in
cylinders [12-14].
Kacem et al. [15] as well as Hamzehloo and Aleiferis [16], Harshavardhan and Mallikarjuna [17], Rahiman et
al. [18], Schmitt et al. [19] and Giannakopoulos et al. [20] recommended the numerical simulation approach as
a significant instrument to investigate flow inside the cylinder. The SolidWorks Flow Simulation (SWFS) code
was utilized in many studies treating various subjects in Computational Fluid Dynamics (CFD) [15,21–26]. n
this study, the SolidWorks Flow Simulation (SWFS) code was utilized to analyze the in-cylinder flow

124
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

characteristics. Similar to other CFD codes, “SWFS” associates of accuracy and functionality with ease-of-
utilization. This code utilizes Finite Volume, multi grid, Multi block time averaged Navier-Stokes equations.
The flow solver subsists methodology in splitting the domain around each node in the grid into sub-volumes; to
ensure the flow continuity betwixt the nodes. The spatial discretization is acquired by following a method for
the tetrahedral interpolation scheme. With respect to the temporal discretization, the implicit formulation is
adopted [25]. The transport equation is integrated over the control volume [25,26].
For these reasons, this study purported to investigate the effects piston head shape on in-cylinder flow when the
engine is fueled with hydrogen. The SWFS code used to characterize the in-cylinder flow for different piston
geometries (convex, flat and concave). The simulation was done for different crank angle during the intake
stroke.

2. FLOW IN INTAKE PIPES OF THE MANIFOLD

In this section we aim to characterize the flow along the manifold pipes and at the seat of the intake valve. The
simulation just treats the intake stroke therefore the interaction with the cylinder walls and the combustion
process are not part of the study. Consequently, the model k-ε was considered more adequate for our objective
than other models such as k-w and the random number generator (RNG) k-ε. In addition, the k-ε model has been
widely used in studies of gas flows through the manifold and cylinders in heat engines as mentioned in the
literature.
This model has been adopted to describe the flow behavior in intake manifold and cylinder.

Mathematical formulation

Our study was limited to the engine intake phase. To predict turbulent flows, we used Navier-Stokes equations
at Favre average. The mass conservation, the momentum conservation and the energy conservation are described
according to the following equations:
𝜕𝜌 𝜕
+ (𝜌𝑢𝑖 ) = 0 (1)
𝜕𝑡 𝜕𝑥𝑖

𝜕𝜌𝑢𝑖 𝜕 𝜕𝑃 𝜕 𝑅
+ (𝜌𝑢𝑖 𝑢𝑗 ) + = (𝜏 + 𝜏𝑖𝑗 ) + 𝑆𝑖 𝑖 = 1; 2; 3 (2)
𝜕𝑡 𝜕𝑥𝑗 𝜕𝑥𝑖 𝜕𝑥𝑗 𝑖𝑗

𝜕𝑝𝐻 𝜕 𝜕 𝑅
𝜕𝑝 𝑅
𝜕𝑢𝑖
+ (𝜌𝑢𝑖 𝐻) = (𝑢𝑖 (𝜏𝑖𝑗 + 𝜏𝑖𝑗 ) + 𝑞𝑖 ) + − 𝜏𝑖𝑗 + 𝜌𝜀 + 𝑆𝑖 𝑢𝑖 + 𝑄𝐻 ; (3)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖 𝜕𝑡 𝜕𝑥𝑗

𝑢² (4)
𝐻=ℎ+
2
Where ‘‘ ρ ”is the density of the fluid; ‘‘ U ”is the speed of the fluid; ‘‘ H ”is the thermal enthalpy; ‘‘ Τij ”is the
viscous tensor of shear stress; ‘‘ Si ”is an external force of distributed mass; ‘‘ Qi ”is the diffusive heat flow
and‘ ‘QH” is a heat source. The indices were used to indicate a summation on the three coordinate directions.
The energy equation is presented as follows:

𝜕𝜌𝐸 𝜕 𝑝 𝜕 𝑅 𝑅
𝜕𝑢𝑖
+ (𝜌𝑢𝑖 (𝐸 + )) = (𝑢𝑖 (𝜏𝑖𝑗 + 𝜏𝑖𝑗 ) + 𝑞𝑖 ) − 𝜏𝑖𝑗 + 𝜌𝜀 + 𝑆𝑖 𝑢𝑖 + 𝑄𝐻 ; (5)
𝜕𝑡 𝜕𝑥𝑗 𝜌 𝜕𝑥𝑖 𝜕𝑥𝑗

𝑒2
𝐸 =ℎ+ (6)
2

Where "e" is internal energy.

At a constant specific heat ratio γ = Cp / Cv, the pressure is given by the ideal gas law:

125
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑃 = 𝜌𝑟𝑇 = (𝛾 − 1)𝜌𝑒 (7)

Turbulence

The description of the turbulence in the intake flow allowed a good understanding of the behavior of the engine.
To reduce emissions and improve performance, turbulence modeling was essential to characterize the mixture
and combustion in an engine. The turbulence model focuses on the calculation of the turbulent kinetic energy k
and the turbulent dissipation function ε.
The turbulent kinetic energy and its dissipation rate are described by the following equations:

𝜕𝜌𝐾 𝜕 𝜕 𝜕𝐾 𝜇𝑡
+ (𝜌𝑢𝑖 𝐾) = ( (𝜇 + )) + 𝑆𝐾 (8)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖 𝜕𝑥𝑖 𝜎𝑘

𝜕𝜌𝜀 𝜕 𝜕 𝜕𝜀 𝜇𝑡
+ (𝜌𝑢𝑖 𝜀) = ( (𝜇 + )) + 𝑆𝜀 (9)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖 𝜕𝑥𝑖 𝜎𝜀

𝑆𝐾 𝑎𝑛𝑑 𝑆𝜀 𝑎𝑟𝑒 𝑡ℎ𝑒 𝑠𝑜𝑢𝑟𝑐𝑒 𝑡𝑒𝑟𝑚𝑠 𝑤ℎ𝑖𝑐ℎ 𝑎𝑟𝑒 𝑑𝑒𝑓𝑖𝑛𝑒𝑑 𝑎𝑠 𝑓𝑜𝑙𝑙𝑜𝑤𝑠;

(10)
𝑅
𝜕𝑢𝑖
𝑆𝐾 = 𝜏𝑖𝑗 − 𝜌𝜀 + 𝜇𝑡 𝐹𝑏
𝜕𝑥𝑗

𝑅
𝜕𝑢𝑖 𝜌𝜀 2
𝑆𝜀 = 𝐶𝜀1 (𝑓1 𝜏𝑖𝑗 + 𝜇𝑡 𝐶𝑏 𝐹𝑏 ) + 𝐶𝜀2𝑓2 (11)
𝜕𝑥𝑗 𝐾

𝐹𝑏 𝑖𝑠 𝑡ℎ𝑒 𝑡𝑢𝑟𝑏𝑢𝑙𝑒𝑛𝑐𝑒 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑏𝑦 𝑡ℎ𝑒 𝑏𝑢𝑜𝑦𝑎𝑛𝑐𝑦 𝑓𝑜𝑟𝑐𝑒𝑠.

𝐶 = 1 𝑠𝑖 𝐹𝑏 > 1
{ 𝑏 (12)
𝐶𝑏 = 0 𝑠𝑖 𝐹𝑏 ≤ 1

The empirical constants in the SWFS code have the following values:

𝐶𝜀1 = 1.44 and 𝐶𝜀2 = 1.92 (13)

CFD Simulation

The geometry of the intake system and the dynamics of fluids in internal combustion engines are complex and
considered unstable. CAD and CFD software are essential for describing the temporal and spatial evolution of
velocity fields, turbulence characteristics and mass concentrations of reactive species. We used Soliworks for
the design of the geometry of the model and the Flowsimulation for the simulation. One of the main advantages
of this code was its ability to import and mesh the required geometry directly from the CAD software. In

126
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

addition, SWFS has made it possible to study a wide range of heat and mass transfer phenomena for complex
geometries.

CFD simulation approach

The model studied included the elements of the Hyundai Sonata engine intake system (intake manifold, intake
valve, cylinder head, cylinder and engine piston).
The commercial software “SolidWorks Flow Simulation” (SWFS) was used for 3D simulation. This software
gives us the possibility to study aerospace flows in the intake circuit. The standard k-ε turbulence model is used,
based on the resolution of the Navier-Stokes equations. SWFS has the advantage of importing geometry directly
from SolidWorks (SW) software. However, the files already created by SW are imported into SWFS in order to
build the mesh necessary for the final calculation of the simulation.
The datagram of the three-dimensional simulation given in figure 2.1 describes the specific and chronological
approach of the SWFS code during the simulation. The simulation configuration included the 3D design of the
whole model, the choice of the resolution of the mesh, the definition of the properties of the fluid and the
imposition of the boundary conditions. As we studied internal flow, the geometry of the solid part of the model
will define the limits of the field of computation.

Figure 1. SWFS Datagramme process.

Geometric pattern

The model studied is a Hyundai Sonata intake manifold 2000cc with four cylinders in line. Due to the properties
of the hydrogen as shown in Tables 1, the engine is studied due to being converted to a bi-fuel hydrogen and
gosoline.

127
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2: The intake system configuration.

In our case we will choose three different geometries as shown in figure 3. the first geometry concerns a piston
with a convex head. the second geometry is that of a piston with a plate head and the third geometry and that of
a piston with a concave head. The three simulations are made in order to see the effect of the geometry of the
piston head on the in-cylinders flow.

Figure 3: Pistons heads configurations.

The CFD study aimed to discover the effect of the piston geometry on the in-cylinder through the intake manifold
and during the intake stroke when the engine is fueled with hydrogen pure.

Table 1. Properties of hydrogen.

Properties of hydrogen Value
Minimum ignition energy (MJ) 0.02
Flame temperature (° C) 2045
Auto-ignition temperature (° C) 585
Maximum flame speed (m.s-1) 3.46
Explosion range (vol%) 13–65
−3 2 -1
Diffusion coefficient (10 m .s ) 0.61

Bounderies conditions

The boundary and initial conditions were defined according to each operating case. Certain simplifications have
been taken for the calculation. The walls of the cylinders were assumed to be adiabatic walls. This study only
took into account the critical cylinder N ° 4 in the intake phase. Figures 4 and 5 show the condition at the limits

128
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

and the initial configuration. The surface of the piston head was considered a movable wall and its speed was
considered as an end condition. The displacement of the piston calculated according to the following equation:

𝑟 2 sin 𝐶𝐴 cos 𝐶𝐴 (
𝑉𝑝 = −𝑟 sin 𝐶𝐴 − (𝑚. 𝑠 −1 )
√𝑙2 − 𝑟 2 sin2 𝐶𝐴 (14)

The intake valves were considered as real walls with a dependence translational speed.

Figure 4. Boundary conditions imposed for the piston and the valves.

Since the manifold end is assumed to be in contact with the atmosphere, the pressure at the inlet of the intake
manifold was maintained at atmospheric pressure (i.e. 1.013 bar).

Figure 5. Boundary conditions imposed at the cylinder entry.

The monitoring of physical parameters was recorded for the plans P1 and P2 as shown in Figure 3. The first
plan, noted P1, was recorded for horizontal section that is chosen to highlight the flows which describes the
swirls happening in the cylinder. The second plan, called P2, was recorded in the combustion chamber for
vertical section that is chosen to highlight the flows which describes the tumbles happening in the cylinder.

Figure 6. Calculation plans (Cut plot surfaces).

129
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. RESULTS AND DISCUSSION

Those sections present the results of the numerical simulation linked to the effect of hydrogen use as fuel on the
flow behavior. The study deals with quasi-stationary regimes. The quasi-stationary results are presented at the
level of a planes "P1" and "P2" in the cylinder which P1 coincides with the top dead center (TDC) and the P2 is
a plane which passes vertically while cutting the plane of the valves in order to visualize the flow in all the
cylinder as shown in the figure 6.

Quasi-stationary gas flow

The fields of velocity, swirling motion, turbulent kinetic energy are determined.

In-cylinder velocity
Controlling the circulation of the hydrogen in the cylinder improves the quality of the flame and directly affects
the volumetric efficiency of the engine. The velocity in the cylinder is generated during the intake phase. The
studies are developed for an engine rotation frequency of 3000 rpm. The contours are taken for crankshaft
positions at: ɵ = 10 °, ɵ = 60 ° and ɵ = 100 ° . The crankshaft angles are taken after top dead center towards the
intake phase. The results are shown in Figures 7 and 8.
the results draw up in figure 7 visualize the swirl fields in the plane P1.

Figure 7. Velocity contours (P1 plot surfaces).

For a crank angle of 10 degrees, we notice that the concave piston has more intense vortices than the other
geometries. This fact and undesirable when operating with hydrogen since the friction with the walls of the
cylinder will cause overheating which on its part can be the cause of self-ignition of the combustible mixture

130
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

given the low ignition energy of the hydrogen. However, this phenomenon is reduced since during admission
the penetration of the rapid and the small amount of air-hydrogen mixture generally prevents the occurrence of
phenomenon. The more the angle increases the more these vortices disappear. We also note that for the piston 3
with concave head the field of velocity is more important which justifies a better filling in the cylinders and
therefore an improvement in volumetric efficiency.

Figure 8 illustrates the fields of the tumble's movement in the cylinder during the intake phase. These fields are
presented in the "P2" plan. The results are taken for an engine speed equal to 3000 rpm, for the angles 10 °, 60
° and 100° of the CA. For the angle ɵ = 10 °, the tumble's movements are relatively weak. This behavior indicates
that the air-H2 flow is stable. the vortices generated by the tumbles are more numerous in the case of a piston
with a flat head. mas the intensity of the speed is better for the case of a concave piston. For the 100° CA we
notice that for the case of convex and flat pistons the air-hydrogen mixture has dead zones. These zones where
the fuel mixture is immobilized can be places where a pre-ignition phenomenon is triggered. Unlike the piston,
which carries the flat geometry, the mixture is practically in motion entirely.

Figure 8. Velocity contours (P2 plot surfaces).

Turbulent energy
The turbulent kinetic energy is related to the intensity of the turbulence. Turbulence created in the cylinder can
be considered an advantage to optimize the preparation of the air-fuel mixture and combustion. During the intake
phase, the distribution of the turbulent kinetic energy is presented in plane "P1" and "P2" (Figure 9 and 10). For
the crankshaft angle ɵ = 10° towards the admission time, the turbulent kinetic energy has relatively low values
given the low piston speed but a dominance is noted for the concave piston.

131
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 9. Turbulent energy contours (P1 plot surfaces).

From figure 8. it is noted that at ɵ = 60 °, extreme values of turbulent energy are observed, in particular around
the intake valve and near the walls of the cylinder. This increase promotes the homogenization of the air-fuel
mixture. In addition, it may be advantageous to have an optimal preparation of the air-hydrogen mixture before
the compression phase. For the angle ɵ = 100°, it can be seen that the values of the turbulent kinetic energy are
markedly reduced. This decrease is linked to the reduction in the piston speed. The hydrogen enrichment causes
a significant acceleration of the particles in the mixture. This behavior affects the agitation between the particles
and their friction with the walls. An increase in kinetic energy values, especially in the intermediate zones of the
intake phase, is produced.

132
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 10. Turbulent energy contours (P2 plot surfaces).

From Figure 10 we notice that the intensity of the turbulent kinetic energy is greater in the case of the flat piston.
Also we notice that the agitation is stronger in the areas far from the walls of the cylinders which brings us to
advise to use a piston with concave head when operating with hydrogen as an alternative fuel.

3. CONCLUSION

In this paper we have conducted a numerical simulation which deals with the effect of piston geometry on the
flow of the fuel mixture in the cylinder when operating with hydrogen. Three configurations of the piston head
geometry have been chosen in order to characterize the flows. The first geometry of the piston is chosen with a
domed head (a convex shape), the second piston has a flat head and the third is that of a piston with a concave
head. The results were mainly:

- The geometry of the piston head has a huge effect on the in-cylinder flow characteristics.
- Whatever the geometry of the piston the area below the valves always carries the highest values
- The in-cylinder filling is better when using a concave piston.
- The piston with convex head brings us back to strong agitation of combustible mixture with the walls of the
cylinders which is disadvantageous in the case of the use of hydrogen.

ACKNOWLEDGMENT
The authors would like to thank Engines and Vehicles department the administration of Bishah Technical
College (TVTC) for their help and financial assistance.

133
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] Alazemi J, Andrews J. Automotive hydrogen fuelling stations: an international review. Renew Sustain Energy Rev
2015;48:483–99.
[2] Midilli A, Dincer I. Key strategies of hydrogen energy systems for sustainability. Int J Hydrog Energy 2007;32:511–
24.
[3] Wallner, T., Scarcelli, R., Nande, A., and Naber, J., "Assessment of Multiple Injection Strategies in a Direct-Injection
Hydrogen Research Engine," SAE Int J.Engines 2(1):1701-1709, 2009
[4] Adt, R.R., Hershberger, D.L., Kartage, T., Swain, M.R., 1973. The hydrogen air fueled automobile engine, Part I.
[5] Adt, R.R., Greenwell Jr., H., Swain,M.R., 1974. The hydrogen and methanol fueled air breathing automobile engine.
In: Paper Presented at the THEME Conference, University of Miami.
[6] Lewis, B., Elbe, G.V., 1987. Combustion, Flames, and Explosions of Gases. Academic Press.
dx.doi.org/10.1016/B978-0-12-446751-4.50001-9.
[7] Delorme A, Rousseau A, Sharer P, Pagerit S, Wallner T. Evolution of hydrogen fueled vehicles compared to
conventional vehicles from 2010 to 2045. SAE Paper No. 2009-01-1008 (2009).
[8] Sebastian Verhelst, Thomas Wallner. Hydrogen-fueled internal combustion engines. Progress in Energy and
Combustion Science 35 (2009) 490 – 527.
[9] Aslan E, Ergeneman M, Sorusbay C. Use of hydrogen in internal combustion engine as fuel. Istanbul Technical
University: Istanbul; 1991.
[10]Adt RR, Swain MR. The hydrogen/methanol–air breathing automobile engine. The Hydrogen Economy Miami
Energy Conference, 18–20 March, Miami Beach, USA, 1974. p.S10-38–48.
[11]Luis LEMOYNE, Contribution à l'étude de la formation du mélange des moteurs à allumage commandé à injection
multipoint application à la réduction des excursions de richesse en fonctionnement transitoire, Thèse de doctorat en
Sciences appliquées1997, ENSAM, Paris
[12]Rafaa Saaidia , Mohamed Ali Jemni and Mohamed Salah Abid, Simulation and Empirical Studies of the Commercial
SI Engine Performance and Its Emission Levels When Running on a CNG and Hydrogen Blend, Energies 2018,
11(1), 29; https://doi.org/10.3390/en11010029
[13]M.El-Adawy, M.R.Heikal, A. Rashid A.Aziz M.I, SiddiquiHasanain A, Abdul Wahhab, Experimental study on an
IC engine in-cylinder flow using different steady-state flow benches. Alexandria Engineering Journal Volume 56,
Issue 4, December 2017, Pages 727-736. https://doi.org/10.1016/j.aej.2017.08.015
[14]Jacques Boree and Paul C. Miles. In-Cylinder Flow. Encyclopedia of Automotive Engineering. 2014 John Wiley &
Sons, Ltd. DOI: 10.1002/9781118354179.auto119.
[15]Kacem, S.H.; Jemni, M.A.; Driss, Z.; Abid, M.S. The effect of H2 enrichment on in-cylinder flow behavior, engine
performances and exhaust emissions: Case of LPG-hydrogen engine. Appl. Energy 2016, 179, 961–971.
[16] Hamzehloo, A.; Aleiferis, P.G. Large eddy simulation of highly turbulent under-expandedhydrogen and methane
jets for gaseous-fuelled internal combustionengines. Int. J. Hydrog. Energy 2014, 39, 21275–21296.
[17]Harshavardhan, B.; Mallikarjuna, J.M. Effect of piston shape on in-cylinder flowsand air-fuel interaction in a direct
injection spark ignition engine—A CFD analysis. Energy 2015, 81, 361–372.
[18]Rahiman, A.; RK, A.R.; AD, M.S.; Ramis, M.K. CFD Analysis of Flow Field Development in a Direct Injection
Diesel Engine with Different Manifolds. Am. J. Fluid Dyn. 2014, 4, 102–113.
[19]Schmitt, M.; Frouzakis, C.E.; Tomboulides, A.G.; Wright, Y.M.; Boulouchos, K. Direct numerical simulation of
multiple cycles in a valve/piston assembly. Phys. Fluids 2014, 26, 035105
[20]Giannakopoulos, G.K.; Frouzakisa, C.E.; Boulouchosa, K.; Fischerb, P.F.; Tomboulides, A.G. Direct numerical
simulation of the flow in the intake pipe of an internal combustion engine. Int. J. Heat Fluid Flow 2017, 68, 257–268.
[21]Genc, O.; Toros, S.; Timurkutluk, B. Determination of optimum ejector operating pressures for anodic recirculation
in SOFC systems. Int. J. Hydrog. Energy 2017, 43, 20249–20259.
[22]Amaral, C.; Brandão, C.; Sempels, É.V.; Lesage, F.J. Thermoelectric power enhancement by way of flow
impedance for fixed thermal input conditions. J. Power Sources 2014, 272, 672–680.
[23]Driss, Z.; Mlayeh, O.; Driss, D.; Maaloul, M.; Abid, M.S. Numerical simulation and experimental validation of the
turbulent flow around a small incurved Savonius wind rotor. Energy 2014, 74, 506–517.
[24]Bellos, E.; Tzivanidis, C.; Daniil, I.; Antonopoulos, K.A. The impact of internal longitudinal fins in parabolic
trough collectors operating with gases. Energy Convers. Manag. 2017, 135, 35–54.
[25]Driss, Z.; Bouzgarrou, G.; Chtourou, W.; Kchaou, H.; Abid, M.S. Computational studies of the pitched blade
turbines design effect on the stirred tank flow characteristics. Eur. J. Mech. B Fluids 2010, 29, 236–245.
[26]Driss, Z.; Abid, M.S. Use of the NaviereStokes equations to study of the flow generated by turbines impellers. Navier-
Stokes Equ. Prop. Descr. Appl. 2012, 3, 51–138.

134
 RENEWABLE ENERGY UTILIZATION AND POLICIES OF
TURKEY
Ünaldı Gülizar Gizem
Osmaniye Korkut Ata University, Osmaniye, Türkiye, [email protected], ORCID: 0000-0003-
1876-9283

Kaplan Yusuf Alper

Osmaniye Korkut Ata University, Osmaniye, Türkiye, [email protected], ORCID: 0000-0003-1067-
110X

Ünaldı, Gülizar Gizem, Kaplan, Yusuf Alper. Renewable Energy Utilization and Policies of
Cite this paper as:
Turkey. 8th Eur. Conf. Ren. Energy Sys.24-25 August 2020, Istanbul, Turkey

Abstract: In all developing countries, the demand for renewable energy has increased as a conclusion of
diminishing on the usage of fossil fuels. Aside from facilitating to the development of humankind,
environmental and atmospheric damage caused by fossil fuels and becoming dependent to the
other countries for energy lead the way for replacing them with renewable energy sources. In
former times, mining and thermal power plants were thought as the exact solution. However, as
the time progresses, hydroelectric power plants had become much more preferred. Today, in aim
to decrease foreign dependency and have a more balanced energy installation, public and private
investments in the field of renewable energy take place in the light of legal frameworks and
incentives. Despite the fact that the country is far behind the rational levels in terms of installed
capacity of renewable energy, currently legal framework as well as financial infrastructure is
settled up for investments. Turkey is much more prosperous on the renewable resources which
should be used efficiently translated into an advantage compared to Europe. Consequently, the
usage of renewable energy which has already started to accelerate is expected to reach to desired
levels in the near future. In this study, the utilization of renewable energy resources and
implementation of the polices are explained in detail.
Keywords: Renewable Energy, Hydroelectric Energy, Wind Energy, Solar Energy, Geothermal Energy,
Biomass Energy, Energy Policy
© 2020 Published by ECRES

Nomenclature
HEP Hydroelectric Power Plant
LNG Liquefied Natural Gas
MENR Ministry of Energy and National Resources
RES Renewable Energy Source
OECD Organization for Economic Co-operation and Development

1. INTRODUCTION

Technological developments, industrialization and the increment in the world population are rapidly increasing
the demand for energy. Energy which is the basic input in production process is an essential element in
improving the welfare level of societies and is used in almost every field in daily life. Depending on the
increasing population, the improvement in the prosperity situation of the societies and technological
developments, the energy requirement of the world augments. For the purpose of satisfy this need, decline of
the non-renewable resources which has limited reserves popularize the utilization of renewable resources, so the
assessment of energy sources should be accomplished in Turkey and all over the world.

Nowadays, many developed countries require to find new energy sources and evaluate these energy sources
economically both in domestic and foreign markets [1]. In this regard, energy independency and environmental
factors which harmed by fossil fuel resources have decreased the preferability of these resources. The fact that

135
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

reserves of the fossil fuel used in energy generation are rapidly draining away has been influential on energy
policies for concentrating on renewable energy resources. Concerns about the exhaustion of fossil energy
sources, energy security problems and increasing environmental problems have led policy makers to pay greater
attention to renewable energy sources all over the world [2]. On this basis, the importance of reaching the
renewable energy sources and using them efficiently has become undeniable for the countries who aim to control
the energy production and consumption. The possible problems which can be caused from this process can be
solved through sustainability efforts made by concentrating on the renewable energy sources policies which
directs many international political events today. Continuity of industrialization and more energy generation and
consumption can be provided through correctly implemented policies which contribute the economical
development and level of social welfare of the society.

2. RENEWABLE ENERGY

The demand for energy continues to rise day by day in whole world. The amount of energy needed derives
largely from fossil fuels. In today's world, due to the unavoidable increases in the prices of fossil fuels, the
possibility of exhaustion in the future and the permanent damage has caused to the environment, a great trend
towards alternative energy sources has started in recent years. In this field, solar energy, wind energy, wave
energy, tidal energy etc. can be count as examples of alternative energy sources. Especially solar and wind
energy are the leading energy types in the use of alternative energy sources in the world and in our country.
There exist also increases in the number of different sized wind and solar farms established in different regions
of our country. Along with these, technological developments in our developing world contribute directly to the
energy sector, as in many other fields. The efficient use of alternative energies with the developing technologies
enhance the interest to this type of energy resources.

There exist significant potential for renewable energy resources in geographically diversified areas in Turkey
and utilization of such resources for electrical energy generation has begun recently. For instance, Aegean
Region can be described productive location regarding geothermal water basins, usage area of solar resources is
abundant in Central Anatolia, Mediterranean and Southeastern Anatolia Regions and Marmara Region has an
extensive capacity in terms of wind resources. However, while these resources had not been utilized until
recently, solar power has only been used for heating water for domestic purposes, and likewise, geothermal
water basins have been utilized in different fields such as heating and tourism. The ratios for electrical energy
installed power of renewable and non-renewable resources in Turkey for 2019 are shown in Figure 1.

Figure 1. Installed Power in Turkey [7]

Energy generation with renewable sources is increasing globally. By the year 2030, Turkey’s energy demands
are expected to increase more than 100 percent compared to today. Thus, Turkey’s passionate 2023 vision
declares especially attractive goals for the renewable energy sector [4]. For this reason, the Ministry of Energy
and National Resources (MENR) encourages to increase the share of renewable energy sources (RES) in
electricity generation and it is striving to improve the whole capacity of renewable sources to 61,000 MW by
2023. 34,000 MW of this total installed generation will be composed of hydropower; 20,000 MW of wind power,

136
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1000 MW of geothermal, 5000 MW of solar, and 1000 MW of biomass energy. Table 1 shows the estimated
resource-based electricity generation rates in the 2023 MENR strategic plan [5].
Table 1. The estimated resource-based electricity generation amounts in the 2023 MENR strategic plan [5].
Renewable Energy Resources 2015 2017 2019 2023
Hydropower 25,526 28,763 32,000 34,000
Wind 5660 9549 13,308 20,000
Geothermal 412 559 706 1000
Solar 300 1800 3000 5000
Biomass 377 530 683 1000
Total 32,275 41,241 49,697 61,000

Hydroelectric Energy

Hydroelectric power (hydropower) is a renewable energy source where electrical power is derived from the
energy of water moving from higher to lower elevations [6]. Hydroelectric energy uses water as the source and
is shown as a renewable energy source that does not directly contribute air pollution. Considering historically,
the first settlements were established on the riverside and various pedals and wheels were used to make use of
the water. With the development of technology, hydroelectric energy is considered as an important energy source
in today's developed and developing countries in terms of producing their own energy and providing energy
security. The number of activated hydroelectric power plant is 603 and the installed power of hydroelectric
power plant is 27.017 MWe. The produced electrical energy is approximately 71.082GWh/year with
hydroelectric power plants [8].

Wind Energy

The wind is used to convert kinetic energy created by the air in motion into electrical energy. This process is
described as conversion into electrical energy using wind turbines or wind energy converter systems. First, the
wind strikes the blades of the turbines, causing their rotation and connection with the turbine. It converts kinetic
energy into return energy by moving it through the shafts connected to the generator and electromagnetism.
Electrical energy is obtained from generators, either stored by batteries or delivered directly to the receivers.
Although wind turbines vary in size and types, generally, they are classified as horizontal axis wind turbines and
vertical axis wind turbines based on the rotation axis. Turbine axis structure and location are two important
factors in wind power technologies. There exist two different ways of installation, onshore and offshore.

Due to the geographical characteristics of Turkey, the country has abundant potential for wind energy. In Figure
2,50 meters high measured wind energy potential of Turkey is observed [9]. The number of activated wind
power plant is 171 and the installed power wind power plant is 6.328 MWe. The produced electrical energy is
approximately 15.838GWh/year with wind power plants [8].

137
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. Technical Wind Energy Potential of Turkey

Solar Energy

Solar energy is highly effective energy that emerges as a result of the fusion reaction that converts the hydrogen
gas in the core of the sun into helium. Technologies such as solar collectors, solar power plants and solar cells
(photovoltaic batteries) have been developed to take advantage of this energy coming to our world through solar
rays. Thanks to these technologies, solar energy can be used directly as heat energy or indirectly by converting
it into electrical energy. In order to reduce CO2 emissions, energy production from renewable resources is one
of the major tools. Solar energy has become an efficient solution to the world's energy challenges which is one
of the most important renewable sources. Solar energy is a huge alternative source of energy both in terms of
safety and the environment, also the fact that it provides energy to the earth more than humankind needs. These
properties make it the most significant alternative source of energy in the face of consuming energies [10]. In
addition, the crucial part of the solar energy in our daily lives is non-negligible. Thermal energy is used for water
heating, distillation, drying of agricultural and animal products or electric power generation; these examples can
be enlargeable. The number of activated solar power plant is 520 and the installed power wind power plant is
802 MWe. The produced electrical energy is approximately 1.184GWh/year with wind power plants [8].

Geothermal Energy

Geothermal energy is the natural heat of the earth produced and stored underground. The temperature in the
center of the Earth is on average 5500 ° C and this is approximately as much as the temperature emitted by the
Sun. This temperature arises from the actual formation of planets and the breakdown of radioactive elements in
the earth's crust. Geothermal energy systems are created by making use of hot reservoirs, usually in the form of
hot water and steam, to provide heating and electricity generation. Geothermal energy technologies such as
double and triple flash devices, dry steam and dual steam cycles are used in electrical energy production.
Although geothermal energy is far above the sufficient energy source in satisfying the energy needs of humanity,
its utilization is limited due to the high investment cost of geothermal energy systems that can be used in heating
and electricity generation.

Turkey’s geothermal potential is 31,500 MW. 78% of these geothermal areas are located in Western Anatolia,
9% in Central Anatolia, 7% in the Marmara Region, 5% in Eastern Anatolia and 1% in the other regions. 90%
of geothermal sources are low and moderate temperature and are convenient for heating, thermal tourism,
minerals production, etc. and 10% is appropriate for electric energy generation. 55% of the geothermal fields in
Turkey are appropriate for heating applications [4]. The number of activated geothermal power plant is 31 and
the installed power geothermal power plant is 992 MWe. The produced electrical energy is approximately
4.213GWh/year with geothermal power plants [8].

Biomass Energy

Biomass energy is the energy source obtained from substances of vegetable and animal origin which contain
carbohydrate compounds. Today, fuels such as bioethanol, biodiesel and biogas are obtained using biomass
energy sources. Bioethanol and biodiesel are a type of fuel produced using a variety of plants or animal oils.
Biogas is a predominantly methane and carbon dioxide gas formed by fermentation of organic substances
(vegetable and animal wastes, urban and industrial wastes) in an oxygen-free environment.

In our country, biomass energy is mostly used in biogas production. In 2017, there exist 71 active power plants
and 444 MWe installed power in biogas production in Turkey [8]. The biogas production potential of our country
is around 4500 MWe. If all of this capacity is used, 20.000 GWh of electricity can be produced and 7.3% of the
electricity demand of our country can be obtained from biogas plants [8].

3. ENERGY POLICIY

The efforts of developing countries in the field of industry enhance the demand for energy. With the
advancement in technology and rising welfare, individual energy consumption has increased steadily over the
years. From the electricity demand forms prepared, it is understood that the energy requirement will increase

138
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

gradually over the years. The everlasting demand of the humankind is electrical power which can be produced
from oil, natural gas, hydraulics, thermal energy and nuclear energy as well as the renewable resources such as
wind, solar, geothermal and biomass energy etc. Choosing the most proper resources for the countries are based
on factors such as their geographical structure, underground and ground resources. For developing countries
such as Turkey, investment programs are organized according to the existing sources as efficiently as possible
by adhering to the thermal energy, hydraulic, natural gas or liquid fuel plants, it is clear that a balanced energy
distribution and harmless environment for the future cannot be reachable. While electricity generation of our
country had been supplied completely through thermal resources during the first years of the Republic of Turkey,
distribution of our installed power by resources as of the end of September 2019; 31.4 percent hydraulic energy,
28.6 percent natural gas, 22.4 percent coal, 8.1 percent wind, 6.2 percent solar, 1.6 percent geothermal and 1.7
percent is in the form of other sources [3].

Energy Policies of Turkey

The demand of Turkey for energy and natural resources is increasing day by day due to economic and population
growth. Since 2002, Turkey in electricity demand in Organization for Economic Co-operation and
Development (OECD) annual growth rate of 5.5% has been the country with the fastest growth. Our country
has an installed power of 90.4 GW as of the end of July 2019. This figure reveals that the installed power of our
country has increased 3 times in the last 15 years [11].

The two main characteristics of the Turkish energy markets are the ever-increasing demand and foreign
dependence in the supply of energy resources. Turkey aims to reduce this dependency by taking steps within the
framework of its domestic and national energy strategy. It is possible to summarize the basic elements of our
country's energy strategy as follows: Prioritizing activities related to energy security, taking into account the
increasing demand and import dependency, considering environmental concerns at every stage of the energy
chain in the context of sustainable development, creating a transparent and competitive market by ensuring
necessary reform and liberalization, increasing productivity and efficiency and enhancing research and
development (R&D) studies related to energy technologies.

Figure 3. Electrical energy gross demand and increment rate [11]

4. RESULTS

Energy is one of the most critical elements in terms of global political economy for ensuring security of
countries. Due to the differences in the geographical distribution of fossil fuels and their proven reserves, many
countries started to lead globally towards renewable energy sources and develop their technological
infrastructure in this respect. The rapid use of renewable energy sources in the world has created competitive
markets and caused a decrease in costs. In all this process, the supports and technological advances which
contribute to improvement of renewable energy sources increase energy efficiency and create a more
environmentally sensitive process. More efficient utilization of energy is demonstrated by the studies carried
out which will make a great contribution to the reduction of an important cost factor for the national economies.

The renewable energy policies of Turkey must be improved and some specific incentives must be done such as
long-term credits and tax reductions. The abundance of Turkey in terms of renewable energy, policy options

139
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

converted into a benefit in terms of exploiting the capacity is of considerable significance. In this process, it will
be very important to follow the ever-increasing technological developments in the field of renewable energy in
global framework and to support technological developments in order to provide renewable energy with
domestic production. When these mentioned fundamental challenges are addressed, Turkey will be well poised
for various investment in renewable energy in the future. This study is very important in terms of general
evaluation of renewable energy and emphasizing the importance of this issue.

REFERENCES

[1] Zafar, Usman, et al. "An overview of implemented renewable energy policy of Pakistan." Renewable and Sustainable
Energy Reviews 82 (2018): 654-665.
[2] BULUT, Umit; MURATOGLU, Gonul. Renewable energy in Turkey: Great potential, low but increasing utilization,
and an empirical analysis on renewable energy-growth nexus. Energy Policy, 2018, 123: 240-250.
[3] Republic of Turkey Ministry of Energy and Natural Resources
[4] ERDIN, Ceren; OZKAYA, Gokhan. Turkey’s 2023 energy strategies and investment opportunities for renewable
energy sources: Site selection based on electre. Sustainability, 2019, 11.7: 2136.
[5] MENR. Strategic Plan. Available online: www.enerji.gov.tr/tr-TR/Stratejik-Plan (accessed on 13 November 2018).
[6] Ånund Killingtveit, Future Energy (Third Edition) Improved, Sustainable and Clean Options for our Planet 2020,
Pages 315-330
[7] TEİAŞ - Electrical Statistics, 30 June 2019
[8] KAYİŞOĞLU, Birol; DİKEN, Bahar. Current Status and Problems of Renewable Energy in Turkey. Journal of
Agricultural Machinery Science, 15.2: 61-65.
[9] Kaplan, Y. A. (2015). Overview of wind energy in the world and assessment of current wind energy policies in
Turkey. Renewable and Sustainable Energy Reviews, 43, 562-568.
[10] ŞAHAN, Muhittin; TOKAT, Özgür; YÜKSEL, O. K. U. R. Daily Total Solar Radiation Measurements in Osmaniye.
SüleymanDemirel University Faculty of Science and Letters, Journal of Science, 2015, 10.2: 97-105.
[11] Republic of Turkey- Ministry of Foreign Affairs

140
 CFD ANALYSIS OF THE EFFECT OF H2CNG BLEND NATURE
ON IN-CYLINDER FLOW

Haider Almusallm
Bishah Technical College, Department of Mechanical Engineering, Bishah, Saudi Arabia, [email protected],
ORCID: 0000-0002-6343-6018

Rafaa Saaidia
Bishah Technical College, Department of Mechanical Engineering, Bishah, Saudi Arabia, rafaasaaidia@gmail,
ORCID: 0000-0002-3892-9726

Almusallm, H, Saaidia, R. CFD analysis of the effect of H2CNG blend nature on in-cylinder
Cite this paper as:
flow. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: In this study, a CFD simulation is carried out. The numerical investigation aims to compare the
effect of hydrogen enrichment with CNG on the in-cylinder flow when the H2CNG is used as
alternative fuel. The in-cylinder flow velocity, mach number and turbulent kinetic energy were
studied during the intake stroke when the engine is turned at 3000 rpm. The calculation was
investigated in two plans to describe the swirl and tumble aspect. For the fixed engine speed, the
simulation was performed when fuel supply is via external mixtures. Whereas, SWFS code was
used to perform a 3D simulation at specific angles of the Crank Shaft (10°, 60°, 120° and 175°).
The reveled results showed that the addition of H2 to CNG influenced the in-cylinder flow. The
hydrogen enrichment causes a significant acceleration of the particles in the mixture. This behavior
affects the agitation between the particles and their friction with the walls. An increase in kinetic
energy values, in particular in the intermediate zones of the intake phase, is produced.
Keywords: Engine; H2CNG; Conversion; CFD

Nomenclature
GHG Greenhouse Gases
BTDC Before Top Dead Center
SWFS SolidWorks Flow Simulation
CFD Computational Fluid Dynamics
RNG Random Number Generator
CAD Computer-Aided Design
TDC Top Dead Center
BDC Bottom Dead Center

1. INTRODUCTION

Today's global society is facing serious issues. The most important among those issues, that of the increase of
the world population, the increase of the request for energy in all industry fields and the availability of energy
sources at a low cost. Traditional fossil fuels are running out fast. Additionally, they have been hurting the
environment in a destructive way and therefore not seen as standing and permanent solutions to the energy needs
in the world [1]. Furthermore, oil prices could oscillate by any lack in these types of energy sources and be
considered a threat to the world energy security and the global economy [2]. More than that, growing the global
in the use of fossil fuels causes reducing the local air quality and increases the quantity of greenhouse gases
(GHG). It clears that the number of vehicles out there are getting higher from one year to another in the entire
world’s countries. As a result, the rate of pollutant emissions is going up, and air pollution problems are getting
worse overnight. The order of these emissions fluctuates depending on the type of vehicle.
Nowadays, the significance of hydrogen as a fuel in vehicles has become an interesting topic due to the severe
externalizes generated by the large-scale effects of global warming and air pollution. Hydrogen has a huge power
due to the many properties that makes it a trustworthy fuel. Hydrogen as a fuel has a set of extraordinary
characteristics that sets it aside from traditional fuels. Hydrogen can be used as a fuel in different ways. It can
be used as a single fuel or as an addition in a mixture of hydrocarbons. Hydrogen can be used in mixtures of

141
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

low-density hydrocarbons which gives us a minimum need for agreements and adjustments in internal
combustion engines [3-9]. The enrichment of hydrogen with hydrocarbons has freshly become well supported
as a strategy to increase engine’s performance while reducing emissions. Likewise, several studies have been
carried out to test the performance of AC motors using H2-CNG and H2-petrol fuels [10-25].
The filling process in the cylinder of an internal combustion engine undergoes cyclical variations unfavorable
for the operation of the engine since they involve the initiation of abnormal combustion from different
conditions. In particular, the instabilities of coherent structures which essentially contain the majority of the
kinetic energy of the flow seem to be the main agent responsible for cycle-to-cycle variations in the cylinder
[26]. However, the origins of VCCs are not fully known, but their effects have been demonstrated in several
research studies and are reflected in an increase in polluting emissions and a decrease in engine performance
[27]. Hence the need to analyze more deeply the causes of cycle-to-cycle variations in order to achieve
reproducibility of the aerodynamic structure in the engine cylinder and therefore to optimize the initial
conditions of combustion. In this context, a better spatial-temporal overview of the dynamics of the flow in the
combustion chamber is necessary. However, the effect of the CNG-Hydrogen mixture of the direct filling flow
in the cylinder during the intake stroke on engine performance has been under-studied. The identification of the
flow field in a cylinder of internal combustion engines during the four stages is an important step. Indeed, it
helps to understand the physical phenomenon that occurs during engine cycles. Computational fluid mechanics
(CFD), which has evolved in recent years, has facilitated the calculation of complex flows around objects. The
solution of the flow problems can be achieved shorter time and cheaper than the analytical method by the
numerical solution method. For better understanding of the flow phenomena CFD techniques are commonly
used and today almost unavoidable to include them in the machine design and testing [28-29]. Several works
[30-37] recommended the CFD approach as a fundamental tool to analyze the engine in-cylinder flow. The
SWFS code was adopted in many studies treating different analysis in CFD [38–43]. In this study, the
SolidWorks Flow Simulation (SWFS) code was used to characterize the in-cylinder flow behaviors.
Therefore, a numerical simulation based on the fluid dynamics (CFD) is conducted in this paper. A three-
dimensional study, the objective of which is to study the characteristics of the flow of the gas mixture upon
admission. The focus of the study is identifying the effect of the amount of hydrogen added to the CNG-H2
mixture on in-cylinder flow.

2. FLOW IN INTAKE PIPES OF THE MANIFOLD

The turbulence model k-ε is the generally used. The k-ε model spotlights the evolution of kinetic energy
regarding to turbulence and considers that there is a harmony between the Reynolds constraints and the viscous
constraints [37]. The standard k-ε turbulence model is used in this study. In this section we aim to characterize
the tumbles and swirl evolution during the intake stroke. The simulation just treats the intake stroke therefore
the interaction with the cylinder walls and the combustion process are not part of the study. Consequently, the
model k-ε was considered more adequate for our objective than other models such as k-w and the random number
generator (RNG) k-ε.

Mathematical formulation

the conservation of mass equation applied for the volumes of control can be written as follows:
𝜕𝜌 𝜕
+ (𝜌𝑢𝑖 ) = 0 (1)
𝜕𝑡 𝜕𝑥𝑖

𝜕𝜌𝑢𝑖 𝜕 𝜕𝑃 𝜕 𝑅
+ (𝜌𝑢𝑖 𝑢𝑗 ) + = (𝜏 + 𝜏𝑖𝑗 ) + 𝑆𝑖 𝑖 = 1; 2; 3 (2)
𝜕𝑡 𝜕𝑥𝑗 𝜕𝑥𝑖 𝜕𝑥𝑗 𝑖𝑗

𝜕𝑝𝐻 𝜕 𝜕 𝑅
𝜕𝑝 𝑅
𝜕𝑢𝑖
+ (𝜌𝑢𝑖 𝐻) = (𝑢 (𝜏 + 𝜏𝑖𝑗 ) + 𝑞𝑖 ) + − 𝜏𝑖𝑗 + 𝜌𝜀 + 𝑆𝑖 𝑢𝑖 + 𝑄𝐻 ; (1)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖 𝑖 𝑖𝑗 𝜕𝑡 𝜕𝑥𝑗

𝑢² (2)
𝐻=ℎ+
2

142
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Where ‘‘ ρ, U, H, 𝜏𝑖𝑗 , Qi , Si, QH ” are respectively the density of the fluid; the speed of the fluid; the thermal
enthalpy; the viscous tensor of shear stress; external force of distributed mass; the diffusive heat flow and heat
source. The indices were used to indicate a summation on the three coordinate directions. The energy equation
is presented as follows:

𝜕𝜌𝐸 𝜕 𝑝 𝜕 𝑅 𝑅
𝜕𝑢𝑖
+ (𝜌𝑢𝑖 (𝐸 + )) = (𝑢𝑖 (𝜏𝑖𝑗 + 𝜏𝑖𝑗 ) + 𝑞𝑖 ) − 𝜏𝑖𝑗 + 𝜌𝜀 + 𝑆𝑖 𝑢𝑖 + 𝑄𝐻 ; (3)
𝜕𝑡 𝜕𝑥𝑗 𝜌 𝜕𝑥𝑖 𝜕𝑥𝑗

𝑒2
𝐸 =ℎ+ (6)
2

Where "e" is internal energy.

At a constant specific heat ratio γ = Cp / Cv, the pressure is given by the ideal gas law:

𝑃 = 𝜌𝑟𝑇 = (𝛾 − 1)𝜌𝑒 (7)

Turbulence

A fully turbulent flow is observed during the intake stoke on the engines. Thereby to reduce emissions and
improve engine behaviors, turbulence flow modeling is fundamental to describe the mixture and combustion
process in an engine. The model contains mainly two transport equations; one describes the turbulence kinetic
energy (k) and the other one describes the turbulence dissipation rate (ε). The k-ε model has various formations
as Standard, Realizable, and RNG models. This model gives good results in the free shear layers, near-wall zone
as in our case that near the cylinders wall is studied (boundary layer), and wake structure flows [37].
The turbulent kinetic energy and its dissipation rate are given as:

𝜕𝜌𝐾 𝜕 𝜕 𝜕𝐾 𝜇𝑡
+ (𝜌𝑢𝑖 𝐾) = ( (𝜇 + )) + 𝑆𝐾 (8)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖 𝜕𝑥𝑖 𝜎𝑘

𝜕𝜌𝜀 𝜕 𝜕 𝜕𝜀 𝜇𝑡
+ (𝜌𝑢𝑖 𝜀) = ( (𝜇 + )) + 𝑆𝜀 (9)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖 𝜕𝑥𝑖 𝜎𝜀

𝑆𝐾 𝑎𝑛𝑑 𝑆𝜀 𝑎𝑟𝑒 𝑡ℎ𝑒 𝑠𝑜𝑢𝑟𝑐𝑒 𝑡𝑒𝑟𝑚𝑠 𝑤ℎ𝑖𝑐ℎ 𝑎𝑟𝑒 𝑑𝑒𝑓𝑖𝑛𝑒𝑑 𝑎𝑠 𝑓𝑜𝑙𝑙𝑜𝑤𝑠;

(10)
𝑅
𝜕𝑢𝑖
𝑆𝐾 = 𝜏𝑖𝑗 − 𝜌𝜀 + 𝜇𝑡 𝐹𝑏
𝜕𝑥𝑗

𝑅
𝜕𝑢𝑖 𝜌𝜀 2
𝑆𝜀 = 𝐶𝜀1 (𝑓1 𝜏𝑖𝑗 + 𝜇𝑡 𝐶𝑏 𝐹𝑏 ) + 𝐶𝜀2𝑓2 (11)
𝜕𝑥𝑗 𝐾

𝐹𝑏 𝑖𝑠 𝑡ℎ𝑒 𝑡𝑢𝑟𝑏𝑢𝑙𝑒𝑛𝑐𝑒 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑏𝑦 𝑡ℎ𝑒 𝑏𝑢𝑜𝑦𝑎𝑛𝑐𝑦 𝑓𝑜𝑟𝑐𝑒𝑠.

143
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

CFD Simulation

The geometry of the intake system and the dynamics of fluids in internal combustion engines are complex and
considered unstable. CAD and CFD software are essential for describing the temporal and spatial evolution of
velocity fields, turbulence characteristics and mass concentrations of reactive species. We used Soliworks to
create the geometry entities of the intake system and the SWFS for the simulation. One of the main advantages
of this code was its ability to import and mesh the required geometry directly from the CAD software. In
addition, SWFS has made it possible to study a wide range of heat and mass transfer phenomena for complex
geometries.

CFD simulation approach

The model studied is the Hyundai Sonata engine intake system (intake manifold, intake valve, cylinder head,
cylinder and engine piston).
The datagram of the simulation process given in figure 1 describes the specific and chronological approach of
the code during the simulation. The simulation configuration included the geometry design of the whole model,
the definition of the properties of the fluid and the imposition of the boundary conditions. As we studied internal
flow, the geometry of the solid part of the model will define the limits of the field of computation.

Figure 1. SWFS simulation Datagram.

Geometric pattern

The model studied is a Hyundai Sonata intake manifold. Table 1 illustrates the main characteristics of this
engine. Due to the properties of the gases as shown in Tables 1 and 2, the engine is studied due to being converted
to a bi-fuel H2GNC engine.

144
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2: the intake system geometry.

The CFD study aimed to discover the effect of the volume fraction of hydrogen on the flow in the cylinder
through the intake manifold and around the seats of the intake valves.
Table 1. Fuel properties.
Characteristics CNG Hydrogen
Chemical formula Mix of C3H8 and C4H10 H2
Molecular weight (g/mol) 50 2
Lower calorific value (MJ / kg) 47 120
Flash point (°C) 500 490
Ignition limits (% of volume) 1.4–10 4–75
Octane number 94–112 130

Table 2. Hydrogen properties.

Properties of hydrogen Value
Minimum ignition energy (MJ) 0.02
Auto-ignition temperature (° C) 585
Maximum flame speed (m.s-1) 3.46
Diffusion coefficient (10−3 m2.s-1) 0.61

The initial conditions imposed

The initial conditions were defined and imposed according to the engine operating case. The walls of the
cylinders were assumed to be adiabatic walls. This study only took into account the critical cylinder N ° 4 in the
admission phase. Figures 2 and 3 show the limits and the initial configuration of the conditions. The surface of
the upper piston was considered a movable wall and its velocity was considered as an exit condition. The
displacement of the piston calculated according to the following equation:

𝑟 2 sin 𝐶𝐴 cos 𝐶𝐴 (
𝑉𝑝 = −𝑟 sin 𝐶𝐴 − (𝑚. 𝑠 −1 )
√𝑙2 − 𝑟 2 sin2 𝐶𝐴 (14)

The intake valves were considered as real walls with a dependence translational speed.

145
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Boundary conditions for the piston and the valves.

The manifold inlet port is assumed to be in contact with the atmosphere.

Figure 4. Boundary conditions at the collector entry.

The monitoring of physical parameters was recorded for two critical planes (P1) and (P2) as shown in Figure 5.
The first plan, noted P1, was recorded just under the valves coinciding with the TDC. The second plan, noted
P2, was recorded vertically to the combustion chamber while passing through the valves.

Figure 5. Calculation plans.

146
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. RESULTS AND DISCUSSION

This section presents the results of the numerical simulation linked to the effect of hydrogen enrichment on the
flow behavior of the H2GNC mixture. The study deals with quasi-stationary regimes. The quasi-stationary
results are presented at the level of a planes "P1" and "P2" in the cylinder as shown in the figure 5.

In-cylinder velocity
The good Controlling of the circulation of the gas mixture in the cylinder improves the flame quality and
improved the volumetric efficiency of the engine. The in-cylinder velocity occurred during the intake phase.
The simulation is conducted for a 3000-rpm. The results are logged in simultaneously for crankshaft positions
at: ɵ = 10°, 60°, 120° and 175°.
The results are illustrated in the figure 6. We can see that the flow structure exhibits a large turbulence aspect.
For the starting angle, we notice that the flow is stable and is similar for all types of mixtures. The velocity
begins to increase gradually and reaches maximum at angles ɵ = 60 ° and ɵ = 120 °. It's totally driven by the
movement of the piston which reaches its maximum acceleration between these angles. For the final limit angles
that corresponds to ɵ = 175°, the velocity degrades and becomes relatively low. A swirling "tumble" structure
forms at the start of admission into the upper part of the combustion chamber, under the exhaust valves. As the
admission progresses, the vortices move towards the center of the chamber and then continue until the bottom
dead center (BDC) going towards the center of the measurement plane and merge.

Figure 6. Velocity fields for different crankshaft angles in the P2 plane.

On the other hand, we notice that the higher the hydrogen content the higher the velocity is high, and it is due
to the low density and diffusibility of the hydrogen molecule.

147
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 7 illustrates the fields of the swirling swirl movement in the cylinder during the intake phase. These fields
are presented in the "P" plan. The results are taken for an engine velocity equal to 3000 rpm, for the angles 10°,
60°, 120° and 175° of the crankshaft. For the angle ɵ = 10°, the swirl movements are relatively weak for the
case of GNC and GNC20H2. This behavior indicates that the air-fuel flow is stable. But for the GNC40H2
mixture these swirl movements are more important. For the angle ɵ = 60° the vortex movement becomes more
intense. Indeed, the tangential components of the velocity are more important, in particular, near the walls of the
cylinder and far from the intake valves. This helps homogenize the air-fuel mixture. This vortex movement
eliminates dead zones in the cylinder and subsequently avoids detonation problems and ensures homogeneous
combustion of the mixture. For the final angle (120°), the vortex movement becomes weaker and smaller and
this is due to the maximum opening of the intake valves which allows a smoother passage. This behavior helps
stabilize the mixture to avoid the backfire problem. The intensity of the "swirl" increases for 175° angles and
becomes more numerous due to the restriction of the passage surface through the valves.

Figure 7. In-cylinder streamlines.

Mach number
Figure 8 expresses the number of mach for the different crankshaft angles: ɵ = 10 °, ɵ = 10 °, ɵ = 60 °, ɵ = 120
° and ɵ = 175 ° after TDC while scanning the admission time. The mach number is the ratio between the velocity
of the fluid and the velocity of sound. It is an adimensional number which expresses the compressibility effect
of a flowing fluid. It is noted that the number of mach is always low during the entire admission phase. Therefore,
the fluid in the combustion chamber is considered by approximation to be incompressible. Thereafter, the
pressure is not taken into account where the density of the fluid is taken constant.

148
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 8. Mach number fields for different crankshaft angles in the P2 plane

Turbulent energy
Turbulence created in the cylinder can be considered an advantage to optimize the preparation of the air-fuel
mixture and combustion. During the intake phase, the distribution of the turbulent kinetic energy is presented
in plane "P2" (Figure 9). first and foremost, we notice that the evolution of the turbulence with several
configuration all depends on the positions of the piston and the admission valve. These flow movements are
fundamental for the characterization of the filling procedure in the cylinders. For the starting phase which
corresponds to the crankshaft angle ɵ = 10°, the turbulent kinetic energy has relatively low values given the
low piston velocity.

149
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 9. Turbulent kinetic energy fields for different crankshaft angles in the P plane.

For the crank angle of 10° which corresponds to the starting of the piston movement, from the the previous map
we note that the turbulence is subdivided into two zones. The first zone is located before the valves and the
second zone after the valves and in the combustion chamber. We notice that the agitation level which is
characterized by the level of turbulence is important in zone 2 and particularly under and in the vicinity of the
valves. this phenomenon is explained by the fact that the movement of the fluid is driven by the piston, also the
zone of restriction of passage at the level of the valves is weak, therefore the gas mixture is driven in movement
in the combustion chamber in the first place.
For the angle of 60° of crankshaft, the piston reaches a significant speed. From figure 9, it is noted that at ɵ =
60°, extreme values of turbulent energy are observed, around the intake valve seats and near the cylinder's walls.
This increase promotes the mixing and the homogenization of the air-H2CNG mixture. In addition, it may be
advantageous to have an optimal preparation of the air-fuel mixture before the compression stroke. We notice
that the turbulence is very strong just before the valves and it is proportional to the faction of the hydrogen
present in the mixture. This can be annoying since the valves are greatly heated and the high level of agitation
with its walls can generate self-ignition and the backfire phenomena will take place.
For the crank shaft angles corresponding to ɵ = 120° and ɵ = 175°, the speed of the piston begins to decrease
gradually. The values of the turbulent kinetic energy are markedly reduced. This decrease is linked to the
reduction in the piston speed. The hydrogen enrichment causes a significant acceleration of the particles in the
mixture. This behavior affects the agitation between the particles and their friction with the walls. An increase
in kinetic energy values, especially in the intermediate zones of the intake phase, is produced.

150
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

5. CONCLUSION

In his paper, a CFD simulation is carried out. The simulation essentially deals with the effect of hydrogen
enrichment with the CNG on the flows in the cylinders. The results obtained were mainly:
- The addition of hydrogen improves the filling of Issa cylinders and the physical properties of the hydrogen
molecule, particularly its density and diffusibility.
- Adding more hydrogen to CNG would improve the filling of the cylinder and thus improve the volumetric
efficiency of the engine.
- Above a certain level of hydrogen in the mixture, agitation has a high critical level around the valve. This justifies
the generation of an abnormal combustion which in turn leads to backfire problems.

ACKNOWLEDGMENT

The authors would like to thank Engines and Vehicles department the administration of Bishah Technical
College (TVTC) for their help and financial assistance.

REFERENCES

[1] Alazemi J, Andrews J. Automotive hydrogen fuelling stations: an international review. Renew Sustain Energy Rev
2015;48:483–99.
[2] Midilli A, Dincer I. Key strategies of hydrogen energy systems for sustainability. Int J Hydrog Energy 2007;32:511–
24.
[3] Bysveen M. Engine characteristics of emissions and performance using mixtures of natural gas and hydrogen. Energy
2007;32:482–9.
[4] Saravanan N, Nagarajan G, Narayanasamy S. An experimental investigation on DI diesel engine with hydrogen fuel.
Renew Energy 2008;33:415–21.
[5] Ganesh RH, Subramanian V, Balasubramanian V, Mallikarjuna J, Ramesh A, Sharma R. Hydrogen fueled spark ignition
engine with electronically controlled manifold injection: an experimental study. Renew Energy 2008;33:1324–33.
[6] Mohammadi A, Shioji M, Nakai Y, Ishikura W, Tabo E. Performance and combustion characteristics of a direct injection
SI hydrogen engine. Int J Hydrog Energy 2007;32:296–304.
[7] Martinez-Martinez S, Sanchez-Cruz F, Riesco-Avila J, Gallegos-Munoz A, Aceves S. Liquid penetration length in direct
diesel fuel injection. Appl Therm Eng 2008;28:1756–62.
[8] Masood M, Mehdi S, Reddy PR. Experimental investigations on a hydrogen-diesel dual fuel engine at different
compression ratios. J Eng Gas Turbines Power 2007;129:572–8.
[9] Masood M, Ishrat M, Reddy A. Computational combustion and emission analysis of hydrogen–diesel blends with
experimental verification. Int J Hydrog Energy 2007;32:2539–47.
[10] Szwaja S, Bhandary K, Naber J. Com+parisons of hydrogen and gasoline combustion knock in a spark ignition engine.
Int J Hydrog Energy 2007;32:5076–87.
[11] Ghazal OH. A comparative evaluation of the performance of different fuel induction techniques for blends hydrogen–
methane SI engine. Int J Hydrog Energy 2013;38:6848–56.
[12] Das L. Utilization of hydrogen-cng blend in internal combustion engine. Hydrog Energy Prog 1996;2:1513–36.
[13] Rakopoulos C, Giakoumis E. Second-law analyses applied to internal combustion engines operation. Prog Energy
Combust Sci 2006;32:2–47.
[14] Bauer C, Forest T. Effect of hydrogen addition on the performance of methane fueled vehicles. Part I: effect on SI
engine performance. Int J Hydrog Energy 2001;26:55–70.
[15] Biffiger H, Soltic P. Effects of split port/direct injection of methane and hydrogen in a spark ignition engine. Int J
Hydrog Energy 2015;40:1994–2003.
[16] Nagalim B, Duebel F, Schmillen K. Performance study using natural gas, hydrogen-supplemented natural gas and
hydrogen in AVL research engine. Int J Hydrogen Energy 1983;9(8):715–20.
[17] Dulger Z. Numerical modeling of heat release and flame propagation for methane fueled internal combustion engines
with hydrogen addition. PhD thesis, University of Miami,1991.
[18] Swain MR, Yusuf MJ, Dulger Z, Swain MN. The effects of hydrogen addition on natural gas engine operation. SAE
paper 932775, 1993.
[19] Yusuf MJ. Lean Burn natural gas fueled engines: engine modi1cation versus hydrogen blending. PhD thesis, University
of Miami, 1993.
[20] Wallace JS, Cattelan AI. Hythane and CNG fueled engine exhaust emission comparison. Proceedings 10th World
hydrogen Energy Conference, Cocoa Beach, USA, June 20–24, 1994. p. 1761–70.
[21] Raman V, Hansel J, Fulton J, Brudery D. Hythane an ultraclean transport fuel. Proceedings, 10th Worlh hydrogen
Energy Conference, Cocoa Beach, USA, June 20–24, 1994.p. 1797–806.

151
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[22] Hoekstra RL, Collier K, Mulligan N, Demonstration of hydrogen mixed gas vehicles, 10th World hydrogen Energy
Conference, Cocoa Beach, USA, June 20–24, 1994.
[23] Hoekstra RL, Collier K, Mulligan N, Chew L. Experimental study of clean burning vehicle fuel. Int J Hydrogen Energy
1995;20:737–45.
[24] Liu Z, Karim GA. Knock characteristics of dual-fuel engines fueled with hydrogen fuel. Int J Hydrogen Energy
1995;20: 919–24.
[25] Das LM. Utilization of hydrogen-cng blend in internal combustion engine, 11th World Hydrogen Energy Conference,
Stuttgart, Germany, June 23–28, 1996. p. 1513–35.
[26] Jacques Borée and Paul C Miles. “In-Cylinder Flow.” Institut Pprime, Cnrs, Université De Poitiers.
[27] Goryntsev, D., A. Sadiki, M. Klein, and J. Janicka. 2009. “Large Eddy Simulation Based Analysis of the Effects of
Cycle-to-Cycle Variations on Air-Fuel Mixing in Realistic DISI IC-Engines.” Proceedings of the Combustion Institute 32
II(2): 2759–66.
[28] Fernandez-Gamiz, U., Demirci, M., İlbaş, M., Zulueta, E., Ramos, J. A., Lopez-Guede, J. M., Kurt, E.
Computational characterization of an axial rotor fan. Journal of Energy Systems 2017; 1(4): 129-137
[29] Yildiz, A, Dandil, B. Investigation of effect of vehicle grilles on aerodynamic energy loss and drag coefficient.
Journal of Energy Systems, 2018; 2(4): 190-203, DOI: 10.30521/jes.461133
[30] Kacem, S.H.; Jemni, M.A.; Driss, Z.; Abid, M.S. The effect of H2 enrichment on in-cylinder flow behavior, engine
performances and exhaust emissions: Case of LPG-hydrogen engine. Appl. Energy 2016, 179, 961–971.
[31] Hamzehloo, A.; Aleiferis, P.G. Large eddy simulation of highly turbulent under-expandedhydrogen and methane jets
for gaseous-fuelled internal combustionengines. Int. J. Hydrog. Energy 2014, 39, 21275–21296.
[32] Harshavardhan, B.; Mallikarjuna, J.M. Effect of piston shape on in-cylinder flowsand air-fuel interaction in a direct
injection spark ignition engine—A CFD analysis. Energy 2015, 81, 361–372.
[33] Rahiman, A.; RK, A.R.; AD, M.S.; Ramis, M.K. CFD Analysis of Flow Field Development in a Direct Injection Diesel
Engine with Different Manifolds. Am. J. Fluid Dyn. 2014, 4, 102–113.
[34] Schmitt, M.; Frouzakis, C.E.; Tomboulides, A.G.; Wright, Y.M.; Boulouchos, K. Direct numerical simulation of
multiple cycles in a valve/piston assembly. Phys. Fluids 2014, 26, 035105
[35] Giannakopoulos, G.K.; Frouzakisa, C.E.; Boulouchosa, K.; Fischerb, P.F.; Tomboulides, A.G. Direct numerical
simulation of the flow in the intake pipe of an internal combustion engine. Int. J. Heat Fluid Flow 2017, 68, 257–268.
[36] Rafaa Saaidia , Mohamed Ali Jemni and Mohamed Salah Abid, Simulation and Empirical Studies of the Commercial
SI Engine Performance and Its Emission Levels When Running on a CNG and Hydrogen Blend, Energies 2018, 11(1),
29; https://doi.org/10.3390/en11010029
[37] Shufat, S, Kurt, E, Elhadad, K, Hancerliogullari, A. A Numerical model for a Stirling engine, Journal of Energy
Systems 2018; 2(1), 1-12, DOI: 10.30521/jes.379164
[38] Genc, O.; Toros, S.; Timurkutluk, B. Determination of optimum ejector operating pressures for anodic recirculation in
SOFC systems. Int. J. Hydrog. Energy 2017, 43, 20249–20259.
[39] Amaral, C.; Brandão, C.; Sempels, É.V.; Lesage, F.J. Thermoelectric power enhancement by way of flow impedance
for fixed thermal input conditions. J. Power Sources 2014, 272, 672–680.
[40] Driss, Z.; Mlayeh, O.; Driss, D.; Maaloul, M.; Abid, M.S. Numerical simulation and experimental validation of the
turbulent flow around a small incurved Savonius wind rotor. Energy 2014, 74, 506–517.
[41] Bellos, E.; Tzivanidis, C.; Daniil, I.; Antonopoulos, K.A. The impact of internal longitudinal fins in parabolic trough
collectors operating with gases. Energy Convers. Manag. 2017, 135, 35–54.
[42] Driss, Z.; Bouzgarrou, G.; Chtourou, W.; Kchaou, H.; Abid, M.S. Computational studies of the pitched blade turbines
design effect on the stirred tank flow characteristics. Eur. J. Mech. B Fluids 2010, 29, 236–245.
[43] Driss, Z.; Abid, M.S. Use of the NaviereStokes equations to study of the flow generated by turbines impellers. Navier-
Stokes Equ. Prop. Descr. Appl. 2012, 3, 51–138.

152
 EFFECT OF MOLAR FRACTION OF THE INTRINSIC LAYER
ON THE PIN STRUCTURE FOR SOLAR CELL BASED AN
INDIUM GALLIUM NITRIDE USING AMPS-1D

Mousli Latifa
Laboratory Smart Grid and Renewable Energy, University Tahri Mohammed Béchar, Algeria,
[email protected], ORCID: 0000-0003-2153-7782

Benmoussa Dennai
Laboratory of Renewable Energy development and their applications in the Saharan areas University Tahri
Mohammed, Béchar, Algeria, [email protected], ORCID 0000-0002-1281-6926

Azeddine Bouziane
Laboratory of Physics and Semiconductors devises, University Tahri Mohammed, Béchar, Algeria,
[email protected], ORCID: 0000-0002-5892-7138

Latifa, M, Dennai, B, Bouziane, Effect of Molar Fraction of the Intrinsic layer on the PIN
Cite this paper as: Structure for Solar Cell Based an Indium Gallium Nitride Using AMPS-1D. 8th Eur. Conf.
Ren. Energy Sys.24-25 August 2020, Istanbul, Turkey

Abstract: In this work, we studied the influence of the mole fraction x of the intrinsic layer in a PIN structure
on the efficiency of a homo-junction cell based on InGaN.
First, we studied the performance of a solar cell based on a PN junction, which we will take as a
reference. In this work we have chosen a gap Eg = 1.4 eV, which is considered as the best
absorption gap of the solar spectrum and which gives a efficiency of 27.46%, with the following
parameters of the solar cell ICC = 29.127mA / cm2, Vco = 1.0623V and form factor of 0.887%.
Second, we introduced an intrinsic (undoped) layer between the P and N layers to form a PIN solar
cell, to improve performance compared to the PN cell. The improvement of the efficiency of
10.42% of the PIN cell compared to the PN cell based on InGaN, while studying the influence of
the molar fraction x and the thickness of the intrinsic layer on the performances of the cell, which
gives a efficiency of 37.88%. We used the AMPS-1D software for the simulation and the obtaining
of the characteristics in the cases mentioned above.
Keywords: InGaN, PIN, AMPS-1D, efficiency, molar fraction.
© 2020 Published by ECRES

1. INTRODUCTION

These days, homo-junction or multi-junction solar cells exist, but in laboratories, because of their very high costs
for their design. However, the development of the manufacture of these solar cells and the improvement of their
photovoltaic conversion efficiency is the primary problem for this technology replacing the use of fossil fuels.
The material used in this work is Indium Gallium Nitride, this last is a semiconductor belonging to the category
of nitride III, which represents a new category of materials having important properties for optoelectronics with
a yield and coefficients of very interesting absorption.
The advantage of this alloy “InGaN” as is its forbidden band (gap) [1], which has the characteristic of being a
direct gap, and allows the absorption of a fairly large part of the solar spectrum, such as InN established with a
prohibited energy band of 0.7ev [2]. For GaN, the width of the forbidden band is 3.42ev [2].
In this work, the introduction of an intrinsic layer with a thickness chosen for better absorption of the solar
spectrum, which is considered an important factor for the design of a photovoltaic cell based on a PIN structure.
The objective of the simulation of the PIN structure solar cell is to compare the performance of this PN junction
cell, and see the importance of the intrinsic layer that we introduce with a well-defined thickness.

153
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2. SIMULATION MODEL

The simulation model used in this work is the AMPS-1D, which is a versatile computer simulation tool that
allows analyzing the physical concepts of diodes, sensors, photodiodes and photovoltaic devices.
This model can calculate the parameters of a solar cell, such as conversion efficiency, short circuit current
density (Jsc), open circuit voltage (Vco), form factor (FF), spectral response, it makes it possible to obtain the
curvatures of the energy, of the position of the level of Fermi in the structure etc.., its use makes it possible to
take into account parameters like the recombination on the surface of the charge carriers. The software is based
on the resolution of the continuity equations for electrons and holes [4], thus the resolution of the Poisson
equation for obtaining the electric field.
The simulations were carried out under standardized conditions (at a temperature of 300 k and an AM1.5
illumination (ASTM G 173-03), with 1 sun and an anti-reflection layer considered to be perfect without loss of
reflection), it should be noted that the software makes it possible to take into account possible reflections by
introducing the reflection and transmission coefficients at the interfaces.
The structure of the solar cell is composed of a P-type emitter, an N-type base and an intrinsic intermediate layer
as illustrated in Figure 1.

3. BASIC EQUATION

The material parameter equations used for the simulation of InGaN's solar cells are:
Band Gap (Eg)

The material has a direct energy band [1], which makes it possible to adapt a forbidden band of InN of 0.7ev to
the forbidden band of GaN of 3.42ev, ranging as a function of x, the molar fraction of indium in the In xGa1-xN,
the gap of InGaN is given by vegard law with a parameter of curvature [2-3]:
𝐸𝑔 𝐼𝑛𝐺𝑎𝑁 = (1 − 𝑥)𝐸𝑔 𝐺𝑎𝑁 +x𝐸𝑔 𝐼𝑛𝑁 − 𝑏𝑥(1 − 𝑥) (1)

Where EgInGaN is the forbidden energy band of InGaN, EgInN, EgGaN are the forbidden energy bands of InN and
GaN, respectively, x is the concentration of indium in InxGa1-xN and b = 1.43ev [5] the curvature parameter.
Absorption Coefficient

For solar cells, the absorption coefficients are important parameters which will determine the quantity of photons
absorbed by the material, are defined by the following relationship [6]:
1,24
∝(𝜇𝑚−1) = A√ − 𝐸𝑔 (𝑥) (2)
𝜆

Where λ is the wavelength of the incident photon in (μm) and A the average equilibrium factor of the absorption
coefficient which is equal to 2,281005 μm−1 ev −1
Electronic Affinity

The electronic affinity is given by the following relation [6, 7]:

χInGaN =[χInN x + χGaN (1 − x)] − [0.8x(1 − x)] (3)

Where χ InGaN the affinity of the material, χGaN = 4,1ev [8] is the affinities of GaN, χInN = 5.6 ev [15] The affinity
of InN.

Relative Permittivity

Dielectric permittivity or permittivity is physical properties which measure the optical properties of the alloy,
given by the following relation as a function of x:
ℇ𝐼𝑛𝐺𝑎𝑁 = (1 − 𝑥)ℇ𝐺𝑎𝑁 + 𝑥 ℇ𝐼𝑛𝑁 (4)

154
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1: the values of the permittivity used [9-10-11-12]

GaN InN
Interval 8,5 - 10,4 11,0 – 15,3
Used 8,9 15,3

The Effective Density of States

The conduction band (Nc) It is given by the relation [7, 13]:

Nc (InxGa1-xN) = [0,9x + 2,3(1-x)] 1018 (5)

The valence band (Nv) It is given by the relation [7, 13]:

Nv(InxGa1-xN) = [5,3 x + 1,8 (1-x)]1019 (6)

Electron and hole Mobility (𝝁𝒏 , 𝝁𝒑 ) :

The mobility model used to study the behavior of holes and the mobility of electrons in an In xGa1-xN alloy as a
function of the composition of the material.
The mobility of electrons and holes is given by the following relation [14]:

𝜇𝑚𝑎𝑥 𝑖,ℎ −𝜇min 𝑖,ℎ

𝜇𝑖,ℎ (𝑁) = 𝜇min 𝑖,ℎ + 𝛾𝑖,ℎ (7)
𝑁
1+( )
𝑁𝑔 𝑖,ℎ

Where i represent the electrons, h represents the holes, N the doping concentration and the specific parameters
μmin i,h , μmax i,h , Ng i,h , γi,h are gives table 2:

Table 2: Parameters used in the calculation of mobility of electron and hole

μmin i μmax i μmin h μmax h γi γh Ng i Ng h
GaN 55 1000 3 170 1 2 2 e17 3 e17
InN 30 1100 3 340 1 2 2 e17 3 e17

3. STRUCTURE

Figure 1: Schematic diagram of solar cells (a) p-n and (b) p-i-n

155
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 3: Parameters of the models used in the simulation for cells PN and PIN
Layers In0.62Ga0.38N In0.71Ga0.29N In0.62Ga0.38N In0.62Ga0.38N In0.75Ga0.25N In0.62Ga0.38N
P I N P I N
Eg (ev) 1.4 1.2 1.4 1.4 1.1 1.4
𝜀𝑟 12.87 13.44 12.87 12.87 13.70 12.87
𝑁𝐴 (𝑐𝑚−3 ) 1016 / 1016 /
𝑁𝐷 (𝑐𝑚−3 ) / 1018 / 1018
𝜇𝑛 (𝑐𝑚2 /𝑣. 𝑠) 212.5 1000 212.5 212.5 1000 212.5
𝜇𝑝 (𝑐𝑚2 /𝑣. 𝑠) 169.82 170 169.82 169.82 170 169.82
𝑁𝐶 (𝑐𝑚−3 ) 1.43x1018 1.31x1018 1.43x1018 1.43x1018 1.25x1018 1.43x1018
𝑁𝑉 (𝑐𝑚−3 ) 3.97x1019 4.28x1019 3.97x1019 3.97x1019 4.25x1019 3.97x1019
𝜒 (𝑒𝑣) 4.84 5.00 4.84 4.84 5.07 4.84
𝑑 (𝑛𝑚) 500 700-750 1000 500 400-450 1000
Layers In0.62Ga0.38N In0.81Ga0.19N In0.62Ga0.38N In0.62Ga0.38N In0.85Ga0.15N In0.62Ga0.38N
P I N P I N
Eg (ev) 1.4 0.99 1.4 1.4 0.93 1.4
𝜀𝑟 12.87 14.08 12.87 12.87 14.34 12.87
𝑁𝐴 (𝑐𝑚−3 ) 1016 / 1016 /
𝑁𝐷 (𝑐𝑚−3 ) / 1018 / 1018
𝜇𝑛 (𝑐𝑚2 /𝑣. 𝑠) 212.5 1000 212.5 212.5 1000 212.5
𝜇𝑝 (𝑐𝑚2 /𝑣. 𝑠) 169.82 170 169.82 169.82 170 169.82
𝑁𝐶 (𝑐𝑚−3 ) 1.43x1018 1.16x1018 1.43x1018 1.43x1018 1.11x1018 1.43x1018
𝑁𝑉 (𝑐𝑚−3 ) 3.97x1019 4.63x1019 3.97x1019 3.97x1019 4.77x1019 3.97x1019
𝜒 (𝑒𝑣) 4.84 5.19 4.84 4.84 5.27 4.84
𝑑 (𝑛𝑚) 500 250 1000 500 200 1000

4. RESULTS AND DISCUSSIONS

It is noted that the efficiency increases with the thickness of the intrinsic zone up to a certain maximum then
begins to decrease; this can be explained by the fact that the electric field begins to decrease with thence of the
value of the thickness, which gives a maximum efficiency. The decrease in y efficiency due to the decrease in
the field in the intrinsic layer is the consequence of recombination in this layer.

Eff=32.949%,Egi=1.2ev,(725nm)
38,0 Eff=36.207%,Egi=1.1ev,(400nm)
Eff=37.202%,Egi=0.99ev,(250nm)
37,5
Eff=37.876%,Egi=0.93ev,(200nm)
37,0
36,5
36,0
35,5
Efficiency (%)

35,0
34,5
34,0
33,5
33,0
32,5
32,0
31,5
31,0

0 100 200 300 400 500 600 700 800 900

Intrinsic layer thickness (nm)

Figure 2: we notice the best performance for the structure Eg=1.4-1.2-1.4 is 32.949% at 725nm, Eg=1.4-1.1-1.4 is
36.207% at 425nm, Eg=1.4-0.99-1.4 is 37.202% at 250nm end Eg=1.4-0.93-1.4 is 37.876% at 200nm

The absorption of InGaN-based homo-junction solar cells under AM1.5 light shows that when the gap of the
intrinsic layer (Eg = 1.2ev, 1.1ev, 0.99ev, 0.93 ev) is lower than the active layer (Eg = 1.4 ev), the intrinsic layer
can absorb more photons. The photons absorbed in the intrinsic layer generate electron-all pairs, which are
separated by the internal field. The photocurrent depends on the photons absorbed and the collection efficiency
in the intrinsic layer.

156
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The increase in the molar fraction of different structure (In 0.71Ga0.29N, In0.75Ga0.25N, In0.81Ga0.19N,
In0.85Ga0.15N) decreases the gap of the intrinsic layer, by increasing the absorption, which positively influences
the efficiency. The increase in efficiency [1] figure shows a passage of a large number of photons; the intrinsic
layer behaves like a tunnel layer up to a critical thickness when we notice a decrease in efficiency due to blocking
on the passage of photons. The simulation results show that the efficiency of the collection strongly depends on
the molar fraction of indium and the critical thickness matches to this molar fraction.

46
1,0685
44
1,0680
42
1,0675
40

Current density(mA/cm )
Open circuit voltage(V)

2
1,0670
38

1,0665
36

1,0660 34

1,0655 32

1,0650 30

1,0645 28

0,60 0,65 0,70 0,75 0,80 0,85 0,60 0,65 0,70 0,75 0,80 0,85

Mole Fraction intrinsic layer (b) Mole Fraction intrinsic layer

(a)
(a) (b): the density of the short-circuit current as a
Figure 3: (a) the open circuit voltage as a function of the molar fraction,
function of the molar fraction

Figure [3]: (a) shows the open circuit voltage Vco decreases with the increase in the molar fraction (x = 0.71,
0.75, 0.81, 0.85). For each thickness a specific value matches to the maximum Vco for each molar fraction, which
can be explained by large saturation currents in the thick layers.
(b) shows the short-circuit current density as a function of the molar fraction of a homo-junction solar cell.
According to the curve Jsc strongly depends on the molar fraction and the thickness of the intrinsic layer. When
x = 0.71 corresponds to a current density Jsc = 35.025mA / cm2, x = 0.85 a current density Jsc = 44.292mA
2
/ cm ; the increase in the current density due to absorption increases the photons in the intrinsic layer which
would generate hole electron pairs which will be collected.

38

36

34
Efficiency(%)

32

30

28

26
0,60 0,65 0,70 0,75 0,80 0,85
Mole fraction intrinsic layer

Figure 4: Efficiency as a function of the molar fraction

Figure [4] shows the positive influence of the molar fraction on the efficiency, when x increases the efficiency
increase, which is explained by the reduction of the gap, which favors a large number of photons, which will be
absorbed.

157
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Optimal Structure

From the results below, obtained using the AMPS-1D software, we can determine which solar cell has the best
performance while changing the molar fraction x = 0.71, 0.75, 0.81, 0.85 for the PIN junction. The current-
voltage characteristic of solar cells having different molar fraction with the optimal efficiency for the PIN and
PN structures are illustrated in Figure 5, and the corresponding PV parameters (V co, Jsc, form factor and
efficiency) are summarized in the table 4.

Table 4: optimal PV parameter of the solar cell

Type of cell Jsc (mA/cm2) Vco (V) Fill Factor Efficiency (%)
PN (x=0.62) 29.127 1.0683 0.887 27.458
PIN (x=0.71) 35.025 1.0678 0.883 32.949
PIN (x=0.75) 39.028 1.0664 0.870 36.207
PIN (x=0.81) 42.054 1.0648 0.831 37.202
PIN (x=0.85) 44.292 1.0647 0.803 37.876

50
50
45
45
40
40
35

Current density(mA/cm )
2
35
30
Curent density(mA/cm )
2

30
25
25
20
20
15
15
10
10 PN(x=0.62)
PIN(x=0.71)
5
5 PIN(x=0.75)
Junction(PN) PIN(x=0.81)
0 0 PIN(x=0.85)
Junction(PIN)
-5 -5
0,0 0,2 0,4 0,6 0,8 1,0 1,2 0,0 0,2 0,4 0,6 0,8 1,0 1,2

Open Circuit Vltage(V) Open Circuit voltage(V)

Figure 5: characteristic IV of the PN and PIN solar cell

5. CONCLUSION

The study of the influence of the intrinsic layer on the performance of PN and PIN solar cells is simulated by
AMPS-1D, the intrinsic layer influences by its thickness and the nature of the material (the molar fraction). Our
study has given us that the best performance is obtained by the following values using 200nm as the thickness
of the intrinsic layer with a molar fraction x = 0.85 for the PIN cell (In0.62Ga0.38N /In0.85Ga0.15N/ In0.62Ga0.38N)
for an efficiency of 37.876%, a structure which increases the efficiency by 10.42% compared to an equivalent
PN cell (In0.62Ga0.38N / In0.62Ga0.38N) with an efficiency of 27.458%. In our simulation, we considered that only
70% of the incident photons are transmitted to the surface, and 10% of the transmitted photons are reflected at
back contact and a recombination speed of 10-3 cm / s under AM1.5 illumination.

REFERENCES

[1] Junqiao Wu. When group-III nitrides go infrared: New properties and perspectives. Journal of Applied Physics,
106(1):5, 2009. 25, 32 .
[2] GF Brown, JW Ager III,W Walukiewicz, and J Wu. Finite element simulations of compositionally graded InGaN solar
cells. Solar Energy Materials and Solar Cells, 94(3) :478–483, 2010. 00089. 25, 26, 49, 50, 79 .
[3] G. Franssen, I. Gorczyca, T. Suski, A. Kamin ́ska, J. Pereiro, E. Munoz, E. Iliopoulos, A. Georgakilas, S. B. Che, Y.
Ishitani, and others. Bowing of the band gap pressure coefficient in InGaN alloys. Journal of Applied Physics, 103(3)
:033514, 2008. 26, 52
[4] M.Boukais, B.Dennai, A. Ould Abbas, J. Nano Electron. Phys. 7 N 3, 03015 (2015).
[5] M . Nawaz, A.Ashfaq, semicond. Sci. Technol. 27, 035019 (2012)
[6] N. Li, Photoelectron. Dev. 3 (2005)
[7] M.E . Levinshtein, S.L. Rumyantsev, M.S.Shur, (Wiley: Chichester, UK: 2001).

158
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[8] J.S.Foresi and T.D.Moustakas, ‘Metal contact to gallium nitride’ Appl. Pys, let. 642859(1993).
[9] V. Chin, T. Tansley, and T. Osotchan, ‘Electron Mobilites in Gallium, Indium, and Aluminum Nitrides,’ J. Appl. Phys,
vol. 75,no. 11, pp. 7365-7372, (1994).
[10] B. Nag, ‘Electron Mobility in Indium Nitride,’J. Cryst. Growth, vol. 269,pp. 35-40 (2004).
[11] S. Davydov, ‘Estimates of Spontaneous Polarization and Permitivities of AlN, GaN, InN, and SiC Crystals,’ Phys.
Sol.Stat, vol. 51, no. 6, pp.1231-1235 (2009).
[12] A. Barker and M. Ilegems, ‘ Infrared Lattice and Free-Electron Dispersion in GaN,’ Phys., vol.53. no. 12, pp. 8775-
8792 (1982).
[13] J. Li, S. Lin et al., ‘Theoretical stidy on InxGa1-xN/Si hetero-junction solar cells,’ in Thin Film Solar Technology, A.
E. Delahoy and L. A. Eldada, Eds, vol. 7409 of Procedings of SPIE, August (2009).
[14] T. T. Mnatsakanov, M.E, L.I.Ppmortseva, S.N. Yurkov. Simin and M. Asif Khan,’Carrier Mobility Mmdel for GaN’.
Solid Electronics, vol, no 1, pp. 111-115, (2003).
[15] J. Wu and W. Walukiewicz. Band gaps of InN and group III nitride alloys. Superlat- tices and Microstructures, 34(1)
:63–75, 2003. 52, 55

159
 UPGRADING OF PERFORMANCE OF AIR TO AIR HEAT PIPE
HEAT EXCHANGER BY USING CUO+ZNO HYBRID NANO
FLUID

Kerim Martin
Gazi University, Ankara, Turkey, [email protected], ORCID: 0000-0002-1960-8070

İpek Aytaç
University of Turkish Aeronautical Association, Ankara, Turkey, [email protected], ORCID:0000-0003-1213-8325

Çağdaş Filiz
Kilis 7 Aralık University, Kilis, Turkey, [email protected], ORCID: 0000-0002-5438-5853

Adnan Sözen
Gazi University, Ankara, Turkey, [email protected], ORCID: 0000-0002-8373-2674

Ümit İskender
Gazi University, Ankara, Turkey, ORCID: 0000-0001-8604-882X

Martin, K, Aytaç, İ, Filiz, Ç, Sözen, A, İskender, Ü. Upgrading of Performance of Air to Air

Cite this paper as: Heat Pipe Heat Exchanger By Using CuO+ZnO Hybrid Nano Fluid. 8th Eur. Conf. Ren.
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In this study, improvement of thermal performance of a heat pipe heat recovery unit is aimed by
using CuO+ZnO nano fluid. Heat recovery unit used in the experiments has 5 heat pipes. Each
heat pipe is made of copper and 1 m long. Their inner and outer diameters are 23.4 mm and 25.4
mm respectively. Heat pipes consist of 3 sections whose names are evaporator, condenser and
adiabatic sections. The length of these sections are 450 mm, 400mm and 150 mm respectively.
The unit also has two air ducts. One of them is cold air duct. The other is hot air duct. The
experiments were performed in 2 different cold air flow rates, 3 different hot air flow rates and 2
different heating powers. 1/3 of evaporator sections of heat pipes are filled with working fluid in
the experiments. Both pure water and nano fluid were used in the experiments separately and
results were compared. The best improvement in efficiency was obtained in the experiments used
CuO+ZnO hybrid nano fluid when Re number was equal to 6700 in cold air duct. The improvement
rate was 73 % when compared to pure water result at same conditions. In that experiment, heating
power and cold air flow rate were 2000 W and 30 g/s respectively.
Keywords: Heat pipe, Heat recovery, Nano fluid, CuO, ZnO
© 2020 Published by ECRES

1. INTRODUCTION

Energy demand of humans is rising day by day with improving technology and rising population. Fossil fuels
are mostly used to meet this energy demand. Using of fossil fuels causes lots of environmental problem. Some
precautions have to be taken to protect human life. Energy efficiency applications are one of the most effective
methods in this issue. Although it is predicted that industry 4.0 will decrease energy use, energy efficiency will
be a popular topic forever. Because life spans of energy sources were decreased. If not fınd new reserves they
will soon be used up.
If it considered that a big amount of energy is used in industrial area, energy efficiency applications in this area
will be more effective. In the industry, heat energy is used directly or indirectly in lots of proses. So, large
amounts of waste heat are arisen. Emission of waste heat to the atmosphere causes both global warming and
serious energy loss. Heat recovery units can decrease effects of this negative situation.
Heat exchangers have a big role in heat recovery. Heat pipe heat exchangers are one of these heat exchangers.
The main purpose in heat pipes is to transfer the heat from one system to another thanks to evaporating and

160
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

condensing [1]. Heat pipe is a closed pipe that is vacuumed. They can work in different temperature range and
be produced in different geometries. They have a great heat transfer potential in even small volumes. Heat pipes
have wide range of uses thanks to some properties like absence of moving part, silent working, short operation
time, working different temperature range, convenience of designing and manufacturing [2].
The working principle of heat pipe is evaporating and condensing of the working fluid. Heat pipes generally
consist of tree section. These sections are evaporator, condenser and adiabatic sections. There is a heat flux to
heat pipe in evaporator section. Working fluid evaporates here. Adiabatic section is a isolated area that there is
no heat flux. The fluid condenses in condenser section by losing heat. So, cycle is completed when the fluid
came bact to evaporator section. This cycle goes on as long as there is a temperature difference between
evaporator and condenser sections. Fluid evaporates at a lower temperature than that of atmospheric pressure
because of negative pressure in heat pipe. This situation enables heat pipe to work lower temperatures.
Pure water and water based fluids are widely used in heat pipes. Water has high latent heat of evaporation.
However high boiling point and low heat convection capacity limit the usage of water in heat pipe [3]. Improving
of thermal performance of heat pipe is possible by using nano fluid instead of pure water. It was the first time
that nano fluid term was used in a conference by Choi in 1995. Choi and Eastman mentioned that adding same
metal particles to water could enhance heat transfer. After that time, using of nano fluid became popular in
various applications. Nowadays, both single and hybrid nano fluids are being used in heat transfer units.
In this study, hybrid nano fluid which contains CuO and ZnO nano particles was used. These two particle was
added to pure water at the same rates by weight. This hybrid nano fluid was used in heat pipe heat exchanger.
Heat transfer performances pure water and nano fluid were compared.

2. MATERIAL AND METHOD

Preparation of nano fluid

The base fluid which was used in experiments was pure water. The mixing rate of two nano particles was 1:1
by weight. Nano particles were got form Nanografi Nano Teknoloji Bil İml ve Dan Ltd Şti firm. CuO particles
have 99.99 % purity and 38 nm average particle size. ZnO particles have 99.99 % purity and 18 nm average
particle size. The other properties of nano powders were given at table 1. Prepared nano fluid has 2 %
concentration by weight.
Section titles can be Theory, Experimental Setup, Methodology, Model, Simulation, Results and Discussions,
Experimental Findings, Analytical Study, etc. But other titles are also welcome. Uppercase letters should be
used in all section headings.
Table 1: Nano particle properties
Nano Particle Density (kg/m3) Thermal Specific Heat(J/kgK)
Copper Oxide 6320 32.9
Conductivity(W/mK) 531
Zinc
(CuO) Oxide 5500 23.4 494
(ZnO)
To avoid settlement and to make fluid more stable, 0,2 % Triton x 100 was used as surface activator. Pure water,
nanoparticle and Triton X-100 are mixed in certain ratios and kept in ultrasonic bath for 6 hours. This process
ensures that the nanoparticles are homogeneously dispersed into the liquid.

Experimental setup

The experimental setup consists of two air ducts, heat pipe group that has 5 heat pipe, 2 frequency controlled
fans and 4 1000 watt resistances. Figure 1 schematically shows the experimental setup.

161
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. Schematic representation of the experimental setup.

As can be seen in Figure 1, two air ducts are placed on top of each other. Hot and cold air does not mix with
each other. This indicates that there will be no problem in using the system in heat recovery units. The heat pipe
group consists of 5 heat pipes. The material of these heat pipes is copper. Their outer and inner diameters are
25.4 mm and 23.4 mm respectively. Heat pipes are 1 m long. There are 2 thermocouples in each heat pipe.
Temperature measurement is taken with 10 thermocouples. Air temperatures are measured with 4
thermocouples, both at the inlet and outlet of both channels. These measurements are recorded on the computer
with a data logger. The location of the heat pipes can be seen in figure 2. The figure isn’t on scale.

Figure 2. Top view of heat pipes(unscaled)

Experiments were carried out at 3 different hot air speeds, 2 different cold air speeds and 2 different heating
power to see the performance of the system under different conditions. Experiments were repeated using both
pure water and CuO + ZnO hybrid nano fluid. The results were compared. Each heat pipe was vacuumed for 30
minutes before the liquid was charged. The amount of vacuum in the pipe can be read instantly from the
manometers installed on it. Liquid charge amount is determined as 65 ml for each heat pipe. This value
corresponds to 1/3 of the evaporator part of the heat pipe.

3. RESULTS AND DISCUSSION

The efficiency of the heat pipe was calculated by the ratio of the heat ((Qc) ̇ ) taken by the air in the condenser
area to the heat (Q ̇e) given by the air in the evaporator area (Equation 1). The heat received by the air passing

162
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

through the cold duct was calculated by the air flow, the specific heat of the air at that temperature and the inlet
and outlet temperatures (Equation 2). The heat given by the air passing through the hot duct was determined in
a similar way (Equation 3). The thermal resistance of the heat pipe was obtained by dividing the average
temperature difference in the evaporator and condenser regions by the heat received by the heat pipe in the
evaporator section (Equation 4). The Reynolds numbers in the channels were calculated by equation 5.

𝑄̇
𝜂 = 𝑄̇𝑘 (1)
𝑒
𝑄̇𝑐 = 𝑚̇𝑐𝑜𝑙𝑑 𝑎𝑖𝑟 𝑐𝑝,𝑐𝑜𝑙𝑑 𝑎𝑖𝑟 (𝑇𝑐𝑜𝑙𝑑 𝑜𝑢𝑡 − 𝑇𝑐𝑜𝑙𝑑 𝑖𝑛 ) (2)
𝑄̇𝑒 = 𝑚̇ℎ𝑜𝑡 𝑎𝑖𝑟 𝑐𝑝,ℎ𝑜𝑡 𝑎𝑖𝑟 (𝑇ℎ𝑜𝑡 𝑜𝑢𝑡 − 𝑇ℎ𝑜𝑡 𝑖𝑛 ) (3)
∆𝑇
𝑅= (4)
𝑄𝑒
𝜌𝑎𝑖𝑟 𝑉𝑎𝑣𝑔 𝐷ℎ
𝑅𝑒 = 𝜇
(5)

Figure 3. Comparison of efficiency and improvement rates of working fluids in calculated Re numbers (for cold air duct)

Reynolds numbers were calculated separately for both air ducts and experiments were performed in turbulent
flow type. The graph in Figure 3 shows the thermal efficiencies obtained at different Reynolds numbers in the
cold air duct. As can be seen from the Figure 3, CuO + ZnO nanofluid has better thermal efficiency values than
pure water in all experiments. The efficiency improvement rates in the heat pipe were between 14% and 73%.
The highest improvement rate was achieved when Reynolds number was 6700. Here, the air velocity value in
the cold air duct is 0.44 m / s. While the efficiency of pure water was 11%, the efficiency of CuO + ZnO nano
fluid was 19% and the improvement rate was 73%.

Figure 4 Temperature dispersion along the heat pipe (Recold = 6700)

The temperature dispersion graph along the heat pipe of the experiment, where the Reynolds number with the
highest improvement rate is 6700, is shown in Figure 4. The low temperature difference between the evaporator
and the condenser zone means that the heat pipe's thermal resistance is low. When the graphic in Figure 4 is

163
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

examined, it is seen that the temperature difference between the condenser region and the evaporator region is
lower in the experiments using CuO + ZnO hybrid nanofluid. Therefore, it can be said that a reduction in thermal
resistance is achieved by using nanofluid in the heat pipe.

Figure 5 Change of thermal resistances according to Reynolds numbers in the cold air channel

The heat resistance values and the percentages of improvement in heat resistance varying according to Reynolds
numbers are shown in the graphic in Figure 5. It is understood from the graph that when using nanofluid instead
of pure water, some decrease in thermal resistances has been achieved for all experiments. Improvement in heat
transfer was achieved by decreasing thermal resistance. The biggest decrease in thermal resistance was 40.4%
while the cold air channel Reynolds number was 12400.

4. RESULTS

In this study, the effect of using hybrid nano fluid in air-to-air heat pipe systems on heat pipe performance was
investigated. The nano fluid used is a hybrid nano fluid containing equal amounts of CuO and ZnO nanoparticles.
Nano fluid has a concentration of 2% by weight. In the experiments, both pure water and nano fluid were used
and the thermal performances of these two fluids were compared. This comparison was made using the thermal
efficiency, thermal resistance values and average temperature dispersion along the heat pipe obtained in different
Reynolds numbers in the cold air duct. The following findings were obtained from the experimental results;
• The use of nano-fluid instead of pure water provided an increase in efficiency in all Reynolds numbers.
However, the turbulence of the flow in the air channel prevented a linear effect between Reynolds numbers and
efficiency.
• The system can operate in low temperature ranges.
• The fact that the air passing through the channels does not mix with each other indicates the usability of
the system in clean air heating systems.
• Different studies can be done using different base fluids or nano fluids.

REFERENCES

[1] Kaya, M, Gürel, A,E, Ağbulut , Ü, Ceylan, İ, Çelik, S, Ergün, A, Acar, B,. Performance analysis of using CuO-Methanol
nanofluid in a hybrid system with concentrated air collector and vacuum tube heat pipe. 2019, 199, 1-9. DOI:
https://doi.org/10.1016/j.enconman.2019.111936
[2] Öztürk, A. Experımental And Theoretıcal Investıgatıon Of The Performance Of Heat Pıpe Heat Recovery Unıt Used
Hvac Unıts. PhD, Karabük University, 2018.
[3] Teng, T, P, Hsu, H, G, Mo, H, E, Chen, C, C,. Thermal efficiency of heat pipe with alumina nanofluid. 2010, 504, 380-
384. DOI: 10.1016/j.jallcom.2010.02.046
[4] Choi, S,U,S, Eastman A, J,. Enhancing thermal conductivity of fluids with nanoparticles. In: Conference: 1995
International mechanical engineering congress and exhibition: 12-17 Nov 1995: San Francisco, CA (United States).

164
 EXPERIMENTAL INVESTIGATION OF THE USE OF
MGO+ZNO MIXTURE IN THERMOSIPHON TYPE HEAT PIPES
WITHIN THE SCOPE OF AIR-TO-AIR HEAT EXCHANGER
DESIGN

Kerim Martin
Gazi University, Ankara, Turkey, [email protected], ORCID: 0000-0002-1960-8070

İpek Aytaç
University of Turkish Aeronautical Association, Ankara, Turkey, [email protected], ORCID: 0000-0003-1213-8325

Çağdaş Filiz
Kilis 7 Aralık University, Kilis, Turkey, [email protected]; ORCID: 0000-0002-5438-5853

Adnan Sözen
Gazi University, Ankara, Turkey, [email protected], ORCID: 0000-0002-8373-2674

Cuma Kılınç
Gazi University, Ankara, Turkey, ORCID: 0000-0002-6258-617X

Martin, K, Aytaç, İ, Filiz, Ç, Sözen, A, Kılınç, C. Experimental Investigation of the Use of

Cite this paper as: MgO+ZnO Mixture in Thermosiphon Type Heat Pipes within the Scope of Air-to-Air Heat
Exchanger Design. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In this study, performance analysis of air-to-air heat pipe heat recovery systems was performed
using hybrid nanofluid and pure water containing MgO+ZnO (Magnesium oxide+Zinc oxide)
particles. It is aimed to improve the heat performance of the heat pipe heat recovery system. In the
experimental setup, 5 heat pipe groups with a diameter of 25.4 mm and a diameter of 23.4 and a
diameter of 100 cm were considered. It is manufactured in a way that the evaporator region of the
heat pipes is 450 mm, the adiabatic region is 150 mm and the condenser region is 400 mm. In
experimental studies, working liquids were injected into 1/3 of the evaporator region volumes of
each heat pipe. 65 ml of work fluid was charged to each heat pipe. Experiments 2 different cooling
air flows (30 g/s, 60 g/s) in order to detect the heat drawn from the condenser region; made in 3
different heating air flows (50 g/s, 70 g/s, 90 g/s). 2 different heater power (1 kW and 2 kW) was
used for the evaporator region. A total of 24 experiments were carried out using 12 pure water and
MgO+ZnO mixture hybrid nanofluid, and then the results were compared. It has been observed
that MgO+ZnO hybrid nanofluid provides some improvement in thermal performance compared
to pure water as a result of the experiments conducted in the cold air duct. Best improvement rate
in thermal performance; while Reynolds number is 11050 in the cold air duct, it is the moment that
48% recovery is achieved and nanofluid is used as working fluid.
Keywords: Heat pipe, Heat recovery, nanofluid, thermal performance, MgO, ZnO
© 2020 Published by ECRES

1. INTRODUCTION

Energy is vital for life to continue on earth. The vast majority of energy needs are traditionally provided from
fossil fuels such as coal and oil, and nuclear power. However, increased environmental awareness and current
energy policies have led to the emphasis on working towards reducing fossil fuels and nuclear energy use [1].
Alternative energy sources have an important share in reducing the use of fossil fuels. The most important factors
that accelerate the increase in energy consumption; the rapid increase in the world population is the intense
involvement of technological tools and equipment in human life and industrialization. These factors increase
energy consumption by 4-5% every year. For this reason, applications aiming to increase energy efficiency have
started to be given importance in order to avoid an energy bottleneck today. In this sense, heat recovery samples

165
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

have an important place. The importance of heat recovery units is increasing every year. Because these heat
recovery units save energy and are environmentally friendly products.
Heat pipes are among the commonly used heat recovery units. Heat pipes; simple structure, easy installation,
not high initial investment cost and high amount of heat transfer, etc. thanks to the features, they are very useful
equipment.
Heat pipe; It consists of 3 parts: evaporator region, adiabatic region and condenser region. (Figure 1). Heat pipes
were sealed and vacuumed inside. In this way, the fluid inside evaporates at a lower temperature than
atmospheric pressure. And the evaporated fluid moves to the upper part of the heat pipe, the fluid reaches the
condenser region. The steam coming to the upper part of the pipe, where it gives its heat out of the pipe,
condenses and changes phase and passes into the liquid phase. The fluid, which passes into the liquid phase,
returns to the evaporator region at the bottom of the pipe again under the effect of gravity. Thus, the cycle is
completed. The upper part condenser region, where heat is given to the outdoor environment by condensation,
is the lower part evaporator region, where heat is drawn from the pipe external environment by evaporation [2].

Figure 1 Heat Pipe and Parts

2. EXPERIMENTAL METHOD

Preparation of Nanofluid

Increasing the amount of heat transfer of the working fluid; nanoparticles are provided by hanging in the
prepared liquid. Magnesium oxide+Zinc oxide (MgO+ZnO) mixture is a highly useful semiconductor metal
oxide mixture and has many advantages such as being resistant to photochemical and chemical erosion, no
toxicity, easy handling and low cost. MgO+ZnO properties are largely dependent on the crystal phase. The MgO
and ZnO particles were ground as nano particles. Nano particles are 40-60 nm in size. The prepared nanofluids
were adjusted to contain 2% by weight of nanoparticles. In order to prevent precipitation, Triton-X-100 chemical
was added as a surfactant in 0.2%. The nanofluid was prepared for the experiment by waiting 5 hours in the
ultrasonic bath for homogeneous dispersion of MgO and ZnO nanoparticles in pure water. There is an issue to
be considered during the preparation of nanofluids; To prevent fluid evaporation during preparation, the
container must be tightly closed.
Heat Exchanger

The heat exchanger bundle consisting of 5 designed heat pipes is shown in Figure 2. All heat pipes are made of
copper material, and they are all manufactured independently of each other. The inner diameter of the
manufactured copper pipes is 23.4 mm, the outer diameter is 25.4 mm and the length is 100 cm. The condenser
region of the heat pipes is manufactured as 400 mm, adiabatic region 150 mm and the evaporator region 450
mm. Filling and unloading can be done easily by placing valves on the upper and lower ends of each heat pipe.
In addition, thanks to the manometer connected to the upper part of the heat pipes, the pressure inside the pipe
was monitored instantaneously.
10 K-type thermocouples were placed on a total of 5 heat pipes in the heat exchanger bundle. 5 of these 10 K-
type thermocouples are fixed with thermal tape in the middle of the evaporator region and 5 in the middle of the
condenser region. Other ends of these thermocouples are connected to a 16 channel datalogger to transmit the
temperatures to the computer.

166
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Heat transfer takes place from the heat pipe walls to the fluid inside the evaporator region. With the air coming
from the hot air duct at the bottom, the fluid in the evaporator region at the bottom of the heat pipes becomes
hot and evaporates over time. Then the evaporated fluid rises and moves to the condenser region. This heat,
which is carried up through the heat pipe, is then transferred to the cold source. That is, the steam in the condenser
region is condensed by releasing its heat to the air coming from the above cold air duct and returns to the
evaporator region below with the help of gravity from the heat pipe walls. This cycle continues continuously
when the heat input into the system is continuous. In short, it is possible to say that heat is transferred from one
region to another by using the phase change of the work fluid.

Experimental Study

In Figure 2, schematic view of the experimental setup where the experiments are carried out are shown. The
parts of the test apparatus are formed as follows: 2 air ducts with 45 cm x 260 cm, 2 40 cm x 260 cm area and
1.3 m length, two radial fans and one body tube heat exchanger. One of the two radial fans sends cold air to the
condenser region and the other sends hot air to the evaporator region. The speed of the air flow passing through
the upper and lower air ducts and the mass flow rate were measured with the help of an anemometer.
In the experimental setup prepared, 14 temperature measurement points were determined. Data from these 14
points were recorded on the computer by means of temperature pairs and a dataloggar. One of the thermocouples
is placed in the condenser inlet-outlet channels, one in the evaporator inlet-outlet channels, 5 in the condenser
region of heat pipes, 5 in the evaporator region of heat pipes. Temperatures were measured with K-type
thermocouples and temperatures were recorded with a tolerance of 0.1 0C.
65 ml of working fluid was placed in each heat pipe. This value corresponds to 1/3 of the evaporator volume in
the heat pipe. As can be seen from the schematic view of the test apparatus, the heat pipe bundle is positioned
at an angle of 90 0C horizontally during the tests. 2 different heater power (1 kW and 2 kW) was used in the
experiments. Experiments were carried out at 2 different cold air speeds (30 g/s, 60 g/s) and 3 different hot air
speeds (50 g/s, 70 g/s, 90 g/s).
After vacuuming the 5 heat pipes in the experimental setup, 2% by weight of MgO+ZnO mixture was added to
each of them as a working liquid, first pure water and then nanofluid. 12 experiments were done with both
working liquids (pure water and MgO+ZnO mixture). Each heat pipe was vacuumed for 30 minutes before liquid
charge. During the experiment, the amount of vacuum in the pipe could be read and checked instantly thanks to
the manometer installed on it.

Figure 2 Schematic view of the experimental setup

3. RESULTS AND DISCUSSION

Efficiency values were calculated for both working fluids and the results were compared. The improvement rate
obtained was determined as "%".
Condenser zone at different air flow rates and different inlet temperatures (1 kW and 2 kW) to the evaporator
zone; The efficiency and improvement rates of pure water and MgO + ZnO / water nanofluid working fluids of
the heat pipe heat exchanger were investigated.
The heat transferred from the air to the heat pipe in the evaporator region and the heat transferred from the heat
pipe to the air in the condenser region are calculated by equations 1 and 2. Here 𝑚̇𝐶 and 𝑚̇𝑒 are the mass air
flow rates of cold and hot channels, respectively. cp shows the specific heat of the air. Tin and Tout are the inlet
and outlet temperatures of the air in the duct.

167
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Heat transfer amounts transferred to the inlet and outlet air; air duct mass flow rate and specific heat are
calculated using the following 1 and 2 formulas using different inlet and outlet temperatures of the inlet air:

𝑄̇𝐶 = 𝑚̇𝑐𝑜𝑙𝑑 𝑎𝑖𝑟 . 𝑐𝑝,𝑐𝑜𝑙𝑑 𝑎𝑖𝑟 . (𝑇𝑐𝑜𝑙𝑑,𝑜𝑢𝑡 − 𝑇𝑐𝑜𝑙𝑑,𝑖𝑛 ) (1)

𝑄̇𝑒 = 𝑚̇ℎ𝑜𝑡 𝑎𝑖𝑟 . 𝑐𝑝,ℎ𝑜𝑡 𝑎𝑖𝑟 . (𝑇ℎ𝑜𝑡,𝑜𝑢𝑡 − 𝑇ℎ𝑜𝑡,𝑖𝑛 ) (2)

The ratio of the condenser zone heat output to the evaporator zone heat input is defined as the thermal
performance of the heat pipe and is calculated by the following formula [3]:
𝑄̇𝑐 𝑄̇𝑜𝑢𝑡
ɳ (%) = 𝑄̇𝑒
= 𝑄̇𝑖𝑛
(3)

∆𝑇
𝑅= 𝑄𝑒
(4)

𝜌𝑎𝑖𝑟 .𝑉𝑎𝑣𝑔 .𝐷ℎ

𝑅𝑒 = (5)
𝜇

The % improvement rate of the heat pipe heat exchanger is calculated with the following equation [4]:
(ɳ𝑤𝑎𝑡𝑒𝑟 − ɳ𝑛𝑎𝑛𝑜 𝑙𝑖𝑞𝑢𝑖𝑑 ) 𝑋 100
% İ𝑚𝑝𝑟𝑜𝑣𝑒𝑚𝑒𝑛𝑡 𝑟𝑎𝑡𝑒 = ɳ𝑤𝑎𝑡𝑒𝑟
(6)

In Figure 3, the efficiency and improvement rate comparisons between the liquids used in the experimental
studies are given in the calculated Reynolds numbers for the cold air duct. As a result of 30-minute studies at
each speed stage, experimental results data were obtained. During the experiments, when the Reynolds number
is 12500 in the cold air duct, the efficiency of the working fluids is greatest. In the experiments, it was concluded
that the maximum efficiency values obtained in the hot air duct and cold air duct were approximately 27% for
pure water and 34% for the mixture of MgO + ZnO nanofluid.

Figure 3. Comparison of efficiency and improvement rates of working fluids in calculated Re numbers (for cold air duct)

As a result of the experiments, when the improvement rate is highest in the cold air channel, the Reynolds
number is 11050. According to the graph of the data obtained in the cold air channel (Figure 5), a maximum
recovery rate of 48% was obtained.
The main reason for these improvements is the presence of MgO+ZnO particles in the nanofluid used as working
fluid. These nanoparticles increase the heat transfer amount of the working fluid they contain. It has been
demonstrated in these experiments that heat transfer increases with the use of nanofluids. With the increase of
the heating power, the heat transfer rate of the working fluids has increased and also the increase in the return
speed of the working fluid to the evaporator area has been observed after the experiments.
As a result of the experiments, the efficiency of the heat pipe; It was determined that it increased with the use of
nanofluid as working fluid. To the highest efficiency; With η = 34%, the Reynolds number in the hot air duct
was 17000 and the Reynolds number in the cold air duct was 12500, while the experiment was performed using
MgO+ZnO nanofluid.

168
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 4 Temperature dispersion along the heat pipe (Recold = 11050)

The temperature dispersion graph along the heat pipe of the experiment, where the Reynolds number with the
highest improvement rate is 11050, is shown in Figure 4. Due to the low temperature difference between the
condenser and the evaporator, it can be concluded that the thermal resistance of the heat pipe is low. When the
graphic in Figure 6 is examined, it is seen that the temperature difference between the condenser and the
evaporator is lower in the experiments using MgO+ZnO hybrid nanofluid. Therefore, it can be said that a
reduction in thermal resistance is achieved by using nanofluid in the heat pipe.

Figure 5 Change of thermal resistances according to Reynolds numbers in the cold channel

The heat resistance values and the percentages of improvement in heat resistance varying according to Re
numbers are shown in the graphic in Figure 5. It is understood from the graph that when using nanofluid instead
of water, some decrease in thermal resistances has been achieved for all experiments. Improvement in heat
transfer was achieved by decreasing thermal resistance. The biggest decrease in thermal resistance was 70.7%
while the cold channel Reynolds number was 6700.

4. RESULTS

The purpose of this experimental study was made to determine the performance of the heat pipe heat exchanger.
MgO+ZnO hybrid nanofluid with a concentration of 2% by weight and pure water were chosen as working
fluids. The following findings have been reached on the performance of the heat pipe heat recovery unit:

i. During operation at 1 kW and 2 kW, no relationship (correct or inverse ratio) was found between the increase
in Reynolds number and efficiency in general. However, the use of nanofluid instead of water caused an increase
in efficiency in all Re numbers.
ii. Maximum improvement rates have been achieved in Reynolds numbers, where the efficiencies are the lowest,
in both the hot air and cold air ducts.
iii. It has been determined that the system can operate in low temperature ranges.
iv. As the Reynolds number increased in the hot air duct in the heat pipe heat exchanger, the improvement rate
varied between 13-71.7%.
v. Heat pipes were produced without wicks and heat pipes were installed in the test facility. The heat pipe is
positioned at an angle of 90 ° C with the ground, so that the gravitational force is utilized to the maximum return
of the fluid in the heat pipe.

169
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

vi. Working fluids; Compared to thermal and heat recovery performance, it has been observed that MgO+ZnO
nanofluidic enhances system performance compared to pure water.

REFERENCES

[1] Aksangör, N. N. et al. 2019. PV syst Simülasyon Aracı Kullanarak Ankara’da Fotovoltaik Güneş Sistemlerin
Performans Analizi. 3rd International Symposium on Innovative Approaches in Scientific Studies. ISAS 2019. 4 (1):
217-220.
[2] Tosun, H., E. Söylemez, and M. Özşen. 2013. Isı Borulu Isı Geri Kazanım ve Nem Alma Ünitesi. 11. ULUSAL
TESİSAT MÜHENDİSLİĞİ KONGRESİ: 373-391.
[3] Sözen, A. et al. 2016. A comparative investigation on the effect of fly-ash and alumina nanofluids on the thermal
performance of two-phase closed thermo-syphon heat pipes. Applied Thermal Engineering 96: 330-337.
[4] Zhang, J. et al. 2013. Experimental study on the heat recovery characteristics of a new-type flat micro-heat pipe array
heat exchanger using nanofluid. Energy Conversion and Management 75: 609-616.

170
 METHOD OF DETERMINING A RATIONAL DESIGN
REFRIGERATION CAPACITY OF OUTDOOR AIR
CONDITIONING SYSTEMS

Mykola Radchenko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0002-1596-6508

Eugeniy Trushliakov
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: [0000-
0001-6171-5438

Andrii Radchenko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0002-8735-9205

Dmytro Konovalov
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID:
0000-0001-7127-0487

Veniamin Tkachenko
PJSC "Zavod "Ekvator", Mykolaiv, Ukraine, [email protected],n ORCID: [0000-0003-0790-8698]

Oleksii Zielikov
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: [0000-
0001-8051-6063

Radchenko, M., Trushliakov, E, Radchenko, A, Konovalov, D, Tkachenko, V, Zielikov, O.

Cite this paper as: Method of determining a rational design refrigeration capacity of outdoor air conditioning
systems. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Outdoor air conditioning systems (OACS) found wide applications as independent centralized air
conditioning systems (ACS) as well as the outdoor air preliminary processing subsystems to run
the combined outdoor and indoor air conditioning systems of buildings more efficiently due to air
feeding to indoor environments at the target temperature level to avoid overloading the indoor air
conditioning systems. One of the most reasonable reserves to improve the efficiency of OACS is
to enable their operation to provide covering yearly conditioning duties according to the actual
variable climatic conditions for a long operation period as possible by selecting a rational design
refrigeration capacity without the OACS oversizing. A simplified improved method for
determining a rational design refrigeration capacity to match current air conditioning duties
without oversizing OACS, quite easy and quick to use and appropriate in design practice, has been
developed. The proposed method is quite applicable for designing the traditional ACS as well as
Variable Refrigerant Flow (VRF) systems.
Keywords: outdoor air, current thermal load, annual refrigeration production
© 2020 Published by ECRES

INTRODUCTION

Outdoor air conditioning systems (OACS) found wide applications as independent central air conditioning
systems (ACS) [1, 2] as well as the outdoor air processing subsystems, named outdoor air processing (OAP)
unit [3, 4], to run the combined outdoor (OACS) and indoor (IACS) air conditioning systems of buildings [5, 6]
more efficiently due to air feeding to indoor environments at the target temperature level to avoid overloading
the IACS [7, 8]. Since the amount of refrigeration ∑(Q0∙τ), spent on the air procession in OACS, depends on the

171
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

current ambient air temperature ta reduction Δta = ta – ta2 during its cooling to a target air temperature ta2 , as well
as on the duration τ of the ACS operation, the efficiency of using the installed (design) refrigeration capacity
Q0∙of the refrigeration machine (RM) of OACS can be estimated by its total value ∑(Q0∙τ) spent for air
conditioning over a certain period of the operation of RM. Issuing from this a design refrigeration capacity Q0
is to be determined according to annual refrigeration production ∑(Q0∙τ) matching current air conditioning duties
without oversizing the RM [9, 10].

LITERATURE REVIEW

Many studies are devoted to increase the energy efficiency of OACS by improving air procession [11, 12],
application of modern methods of analyses [13, 14] and monitoring [15, 16] to reduce energy consumption.
Some of such methods involve cooling degree-hour (CDH) calculations [17] and their modified versions [18,
19, 20] to match current cooling demands according to actual climatic conditions as well as thermal demand
management (TDM) [21, 22] based on different criteria [23, 24] to save energy while building conditioning [25,
26], engine intake air cooling (EIAC) [27, 28] and combined cooling, heat and power [29, 30].
Some of technical innovations [12] and methodological approaches in waste heat recovery refrigeration [28]
might be successfully applied to enhance energy efficiency of the Variable Refrigerant Flow (VRF) systems [7,
8].
In spite of existence of a lot of control systems to minimize power consumption and support indoor thermal-
humidity environments, the problem of choosing a rational design refrigeration capacity to provide annually
efficient operation at changeable current loadings and to exclude oversizing a refrigeration machine (RM) needs
further solution.
The aim of the study is to develop a method to determine a rational design refrigeration capacity of OACS in
accordance with current climatic conditions to avoid oversizing.

RESEARCH METHODOLOGY

The efficiency of OACS and their RM performance depends on their loading and a duration of their yearly
operation. Therefore, the annual refrigeration output (refrigeration production) in response to air conditioning
duties is considered as a primary criterion for the choice of a rational design load of OACS [9, 10]. For this the
current refrigeration capacities, generated by RM at any time period in response to (the air conditioning duties
for) outdoor air cooling down to the target air temperature, have been summarized over the year to determine
the annual refrigeration production.
The authors develop a novel method of rational designing based on the yearly loading characteristic curve of
annual refrigeration production dependence on the design specific refrigeration capacity to choose its value, that
provides closed to maximum annual refrigeration production without oversizing the RM. It was done by adding
to well known methods [17, 25] some stages aimed to determine a more precise value of rational design
refrigeration capacity.
In order to generalize the results and simplify calculations for any total refrigeration capacity Q0 , it is convenient
to present it in relative (specific) value per unit air flow rate (Ga = 1 kg/s) – in the form of specific refrigeration
capacity: q0 = Q0 /Ga , kW/(kg/s), or kJ/kg, where Q0 is the total refrigeration capacity when cooling the air with
the flow rate Ga: Q0 = сa ξ∙Δta ∙Ga , where Δta = ta – ta2 – decrease (depression) of outdoor air temperature; ta2 –
target air temperature at the air cooler outlet; ξ – specific heat ratio of total (full) heat, inclooding sensible and
latent heat, to sensible heat of humid air; сa –specific heat of humid (moist) air [kJ/(kg·K)].
The specific annual refrigeration production ∑(q0∙τ) = ∑(сa ξ∙Δta ∙τ), [kg∙h/kW]; τ – the hours lapsed in a year
at a outdoor temperature of tamb , cooled to a target temperature ta2 .
There are two addition stages added to calculate an optimal specific refrigeration capacity q0.opt , providing a
maximum rate of annual refrigeration production increment ∑(q0∙τ)/q0 and corresponding annual refrigeration
production ∑(q0∙τ)opt , as the second stage of calculation procedure, and determining a precise value of rational
specific refrigeration capacity q0.rat , providing the annual refrigeration production ∑(q0∙τ)rat closed to its
maximum value, but without unproductive expenses of refrigeration capacity q0 caused by oversizing RM
without obtaining a sensible increasing the annual refrigeration production ∑(q0∙τ), as the third stage. With this
a rational specific refrigeration capacity q0.rat is associated with second maximum rate of annual refrigeration
production increment just beyond the optimal annual refrigeration production ∑(q0∙τ)opt : [∑(q0∙τ) –
∑(q0∙τ)opt ]/q0 , where ∑(q0∙τ)>∑(q0∙τ)opt .

172
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

RESEARCH METHODOLOGY

Great yearly variations of specific refrigeration capacities q0.10 , q0.15 and q0.20 , needed for cooling outdoor air
from its actual current temperature ta to ta2 = 10, 15 and 20 °С (2017, Mykolayiv, Ukraine, Fig. 1) show the
existence of real problem to choose a design refrigeration capacity of the RM to cover actual conditioning duties
without its oversizing.

q0, kW/(kg/s)
45
40
35
ta2= 10 C
30
25 ta2= 15 C
20
ta2= 20 C
15
10
5
0
0 30 60 90 120 150 180 210 240 270 300 330 τ,
360days
390
Figure 1. Variations of specific refrigeration capacities q 0.10 , q0.15 and q0.20 , needed for cooling outdoor air from its
actual temperature ta to ta2 = 10, 15 and 20 °С during 2017 in Mykolayiv region

A rational design refrigeration capacity may be determined from the yearly loading characteristic curve built up
by summarizing the refrigeration production ∑(q0∙τ) for various design specific (at Ga = 1 kg/s) refrigeration
capacity q0 at different target temperature of cooled air ta2 = 10, 15 and 20 °С at the outdoor air cooler outlet
(Fig. 2).

Σ(q0τ), kWh/(kg/s)
70000
t2= 10 C
60000
50000
40000
t2= 15 C
30000
20000
t2= 20 C
10000
q0, kW/(kg/s)
0
0 5 10 15 20 25 30 35 40 45
Figure 2. Specific annual refrigeration production ∑(q0∙τ) required for cooling the outdoor air to the temperatures
ta2 = 10, 15 and 20 °С versus a design specific refrigeration capacity q0

As it is seen, a design specific refrigeration capacity q0.10 = 33-35 kW/(kg/s) for cooling the ambient air from
the current temperatures ta to ta2 = 10 °C provides annual refrigeration production ∑(q0∙τ)= 65-66 MWh/(kg/s)
closed to a maximum value 67 MWh/(kg/s) within sensible high rate of its increment and could.
It is evident, that with increasing the target temperature of cooled air ta2 to 15 and 20 °C it is more problematic
to choose quite precise values of rational specific refrigeration capacity q0.15rat and q0.20rat so as behaviour of
characteristic curve ∑(q0∙τ) = f(q0) becomes more gentle (Fig. 2). To solve this task the original method of
determining a rational design refrigeration capacity has been improved by adding addition stages aimed for
calculation of optimal design refrigeration capacity q0.opt, corresponding to a maximum rate of annual
refrigeration capacity increment ∑(q0∙τ)/q0∙(Fig.3), and more precise value of rational specific refrigeration
capacity q0.rat (Fig.4 and 5).

173
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Σ(q0τ)/q0, hours Σ(q0τ), kWh/(kg/s)

2500 100000
Σ(q0τ)/q0.10
t2= 10 C
2000 80000
Σ(q0τ)/q0.15 Σ(q0τ)10
t2= 15 C
1500 60000
Σ(q0τ)/q0.20 Σ(q0τ)10opt
t2= 20 C
1000 Σ(q0τ)15 40000
Σ(q0τ)15opt
500 Σ(q0τ)20 Σ(q0τ)20opt 20000

0 q0.opt q0.opt q0.opt 0

0 5 10 15 20 25 30 35 q0, kW/(kg/s)
40 45
Figure 3. Relative increment ∑(q0∙τ)/q0 of annual refrigeration production and annual refrigeration production ∑(q0∙τ)
required for cooling outdoor air to target temperatures ta2 = 10, 15 and 20 °C versus a design specific refrigeration
capacity q0

As Fig. 3 shows, a maximum rate of annual refrigeration capacity increment ∑(q0∙τ)/q0∙for cooling outdoor air
to the temperature ta2 = 15 °C takes place at the optimal design refrigeration capacity q0.opt=18 kW/(kg/s) and
provides annual refrigeration production ∑(q0∙τ)opt = 28 MWh/(kg/s) significantly less than its maximum value
34 MWh/(kg/s).
To achieve closed to maximum annual refrigeration production ∑(q0∙τ) at more precise value of rational specific
refrigeration capacity q0.rat without oversizing the refrigeration machine it is proposed to determine the maximum
rate of annual refrigeration increment ∑(q0∙τ)/q0∙within the range beyond its value ∑(q0∙τ)opt = 28 MWh/(kg/s)
corresponding to optimal design refrigeration capacity q0.opt=18 kW/(kg/s), as (∑(q0∙τ)– ∑(q0∙τ)opt )/q0∙(Fig. 4
and 5).

Σ(q0τ)/q0, hours (Σ(q0τ)-Σ(q0τ)opt)/q0, hours

2500 300
t2= 10 C
2000 250
t2= 15 C
200
1500
Σ(q0τ)/q0 150
1000
(Σ(q0τ)-Σ(q0τ)opt)/q0 100
500 50
t2= 20 C
rat20
q0, kW/(kg/s)
0 opt20 opt15 rat15 opt10 rat10 0
0 5 10 15 20 25 30 35 40 45
Figure 4. Relative increments of annual refrigeration production ∑(q0∙τ)/q0 and (∑(q0∙τ)– ∑(q0∙τ)opt )/q0∙within the range
beyond the optimal value ∑(q0∙τ)opt (Fig.3) needed for cooling outdoor air to target temperatures ta2 = 10, 15 and 20 °C
versus a design specific refrigeration capacity q0

174
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

As Fig. 4 and 5 shows, a maximum rate of annual refrigeration increment (∑(q0∙τ) – ∑(q0∙τ)opt )/q0 within the
range beyond its value ∑(q0∙τ)opt = 28 MWh/(kg/s) (in Fig.3) corresponding to optimal design refrigeration
capacity q0.opt=18 kW/(kg/s) for cooling outdoor air to the temperature ta2 = 15 °C takes place at the rational
design refrigeration capacity q0.rat=24 kW/(kg/s) and provides quite precise annual refrigeration production
∑(q0∙τ)rat = 33 MWh/(kg/s) that is very closed to its maximum value 34 MWh/(kg/s) but at a design refrigeration
capacity q0.rat=24 kW/(kg/s) much less than oversized its value q0 = 32 kW/(kg/s).

10(Σ(q0τ)-Σ(q0τ)opt)/q0, hours Σ(q0τ), kWh/(kg/s)

3000 (Σ(q τ)-Σ(q τ) )/q 80000
0 0 opt 0.15 t2= 10 C
2500 70000
Σ(q0τ)0.10rat 60000
t2= 15 C
2000 50000
1500 Σ(q0τ)0.15 Σ(q0τ)0.15rat 40000
1000 t2= 20 C 30000
Σ(q0τ)0.20rat 20000
500 10000
0 rat rat rat 0
0 5 10 15 20 25 30 35 q40 0, kW/(kg/s)
45
Figure 5. Annual refrigeration production ∑(q0∙τ) and relative increment of annual refrigeration production (∑(q0∙τ)–
∑(q0∙τ)opt )/q0∙within the range beyond the optimal value ∑(q0∙τ)opt (Fig.3) needed for cooling outdoor air to target
temperatures ta2 = 10, 15 and 20 °C versus a design specific refrigeration capacity q0

Thus, a developed new method enables to determine a rational design refrigeration capacity that provides to
cover current air conditioning duties according to actual climatic conditions and avoid oversizing a refrigeration
machine.

CONCLUSION

A novel methodology to determine a rational design refrigeration capacity provides covering the current air
conditioning duties according to actual climatic conditions without oversizing a refrigeration machine of OACS.
An improved method includes two addition stages added to the basic method to calculate an optimal specific
refrigeration capacity q0.opt, providing a maximum rate of annual refrigeration production increment ∑(q0∙τ)/q0
and corresponding annual refrigeration production ∑(q0∙τ)opt, as the second stage of calculation procedure, and
determining a precise value of rational specific refrigeration capacity q 0.rat, providing the annual refrigeration
production ∑(q0∙τ)rat closed to its maximum value, but without unproductive expenses of refrigeration capacity
q0 caused by oversizing RM, as the third stage. With this a rational specific refrigeration capacity q 0.rat is
associated with second maximum rate of annual refrigeration production increment within the range beyond the
optimal annual refrigeration production ∑(q0∙τ)opt: [∑(q0∙τ) – ∑(q0∙τ)opt ]/q0, where ∑(q0∙τ)>∑(q0∙τ)opt.
A proposed method is quite suitable to determine a rational design refrigeration capacity of any ambient air
cooling system for traditional comfort air conditioning (central and combined type as VRF system) as well as
for combustion engine inlet air cooling to enhance their efficiency, including trigeneration systems.
The method to calculate an optimal specific refrigeration capacity q 0.opt, providing a maximum rate of annual
refrigeration production increment ∑(q0∙τ)/q0 (the second stage of the full scale method) is very effective energy
saving method of maintenance of refrigeration capacity, in particular, with accumulation of excessive
refrigeration capacity at lowered current thermal loads to cover increased thermal loads, as well as to improve
the application of compressors with regulation of refrigeration capacity due to decrease their sizes.

175
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] Southard, L.E., Liu, X., Spitler, J.D.: Performance of HVAC systems at ASHRAE HQ ASHRAE Journal, 2014, vol.
56(9), pp. 14-24.
[2] Chua K. J., Chou S. K., Yang W. M., Yan J.: Achieving better energy-efficient air conditioning – A review of
technologies and strategies. Applied Energy, 2013, vol. 104, pp. 87–104.
[3] Liu, C., Zhao, T., Zhang, J.: Operational Electricity Consumption Analyze of VRF Air Conditioning System and Air
Conditioning System Based on Building Energy Monitoring and Management System. Procedia Engineering, 2015,
vol. 121, pp. 1856–1863.
[4] Zhu Y., Jin X., Du Z., Fang X., Fan B.: Control and energy simulation of variable refrigerant flow air conditioning
system combined with outdoor air processing unit. Appl. Therm. Eng., 2014, vol. 64, pp. 385–395.
[5] Enteria N., Yamaguchi H., Miyata M., Sawachi T., Kuwasawa Y.: Performance Evaluation of the Variable Refrigerant
Flow (VRF) Air-Conditioning System Subjected to Partial and Unbalanced Thermal Loadings. Journal of Thermal
Science and Technology, 2016, vol. 11(1), pp. 1–11.
[6] Im, P., Malhotra, M., Munk, J.D., Lee, J.: Cooling Season Full and Part Load Performance Evaluation of Variable
Refrigerant Flow (VRF) System Using an Occupancy Simulated Research Building. Proceedings of the 16th
International Refrigeration and Air Conditioning Conference at Purdue, July 11–14 2016: West Lafayette, USA.
[7] Lee, J. H., Yoon, H. J., Im, P., Song, Y.–H.: Verification of Energy Reduction Effect Through Control Optimization of
Supply Air Temperature in VRF-OAP System. Energies, 2018, vol. 11(1).
[8] Zhang, L., Wang,Y., Meng, X.: Qualitative analysis of the cooling load in the typical room under continuous and
intermittent runnings of air-conditioning. Procedia Engineering, 2017, vol. 205, pp. 405–409.
[9] Radchenko, A., Radchenko, M., Trushliakov, E., Kantor, S., Tkachenko, V.: Statistical method to define rational heat
loads on railway air conditioning system for changeable climatic conditions. In: 5th International Conference on
Systems and Informatics: ICSAI 2018, 2018: Jiangsu, Nanjing, China, pp. 1308–1312.
[10] Radchenko, M., Radchenko, R., Tkachenko, V., Kantor, S., Smolyanoy, E.: Increasing the Operation Efficiency of
Railway Air Conditioning System on the Base of Its Simulation Along the Route Line. In: Nechyporuk M. et al. (eds)
Integrated Computer Technologies in Mechanical Engineering. Advances in Intelligent Systems and Computing,
Springer, Cham, 2020, vol. 1113, pp. 461-467.
[11] Chen, J., Xie, W.: Analysis on load-undertaking of fan coil unit with fresh air system. Advanced Materials Research,
2013, vol. 614, pp. 678-681.
[12] Trushliakov, E., Radchenko, M., Bohdal, T., Radchenko, R., Kantor, S.: An innovative air conditioning system for
changeable heat loads. In: Tonkonogyi V. et al. (eds.) Grabchenko’s International Conference on Advanced
Manufacturing Processes. InterPartner-2019. Lecture Notes in Mechanical Engineering. 2020: Springer, Cham, 10 p.
[13] Khatri, R., Joshi, A.: Energy performance comparison of inverter based variable refrigerant flow unitary AC with
constant volume unitary AC. Energy Procedia, 2017, vol. 109, pp. 18–26.
[14] Park, D.Y., Yun, G., Kim, K.S.: Experimental evaluation and simulation of a variable refrigerant-flow (VRF) air-
conditioning system with outdoor air processing unit. Energy Build, 2017, vol. 146, pp. 122–140.
[15] Radchenko, A., Mikielewicz, D., Forduy, S., Radchenko, M., Zubarev, A.: Monitoring the Fuel Efficiency of Gas
Engine in Integrated Energy System. In: Nechyporuk M. et al. (eds) Integrated Computer Technologies in Mechanical
Engineering. Advances in Intelligent Systems and Computing, 2020: Springer, Cham, vol. 1113, pp. 361-370.
[16] Radchenko, М., Radchenko, R., Ostapenko, O., Zubarev, A., Hrych, A. Enhancing the Utilization of Gas Engine
Module Exhaust Heat by Two-Stage Chillers for Combined Electricity, Heat and Refrigeration, Proceedings, 5th
International Conference on Systems and Informatics, ICSAI 2018, 2019: Jiangsu, Nanjing, China, pp. 240-244.
[17] Forsyth, J.L. Gas turbine inlet air chilling for LNG, 2013: IGT International Liquefied Natural Gas Conference
Proceedings, 3, pp. 1763-1778. https://www.scopus.com/record/display.uri?eid=2-s2.0-84904725071&origin=
inward&txGid=b9d07662f250f4a4511cbfc9c242297a
[18] Coskun, C., Demiral, D., Ertürk, M., Oktay, Z. Modified Degree-Hour Calculation Method, Solar Power, Prof. Radu
Rugescu (Ed.), 2012, InTech, Available from: http://www.intechopen.com/books/solar-power/modified-degree-hour-
calculation-method
[19] Coskun C. A novel approach to degree-hour calculation: Indoor and outdoor reference temperature based degree-hour
calculation. Energy, 2010; vol. 35, pp. 2455-2460.
[20] Oktay Z, Coskun C, Dincer I. A new approach for predicting cooling degree-hours and energy requirements in
buildings. Energy, 2011; pp. 4855-4863.
[21] Canova A, Cavallero C, Freschi F, Giaccone L, Repetto M, Tartaglia M.: Optimal energy management. IEEE Industry
Applications Magazine, 2009, vol. 15, pp. 62–65.
[22] Suamir I. N., Tassou S. A.: Performance evaluation of integrated trigeneration and CO2 refrigeration systems. Applied
Thermal Engineering, 2013, vol. 50, pp. 1487-1495.
[23] Rocha M.S., Andreos R., Simões-Moreira J.R.: Performance tests of two small trigeneration pilot plants. Applied
Thermal Engineering, 2012, vol. 41, pp. 84–91.
[24] Freschi F., Giaccone L., Lazzeroni P., Repetto M. Economic and environmental analysis of a trigeneration system for
food-industry: A case study Applied Energy 2013, vol. 107, pp. 157–172.

176
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[25] Cardona E., Piacentino A.: A methodology for sizing a trigeneration plant in mediterranean areas. Applied Thermal
Engineering, 2003, vol. 23, 15 р.
[26] Rodriguez-Aumente P.A., Rodriguez-Hidalgo M. C., Nogueira J. I., Lecuona A., Venegas M. C.: District heating and
cooling for business buildings in Madrid. Applied Thermal Engineering, 2013, vol. 50, pp. 1496–1503.
[27] Farouk, N., Sheng L., Hayat Q.: Effect of Ambient Temperature on the Performance of Gas Turbines Power Plant.
International Journal of Computer Science, 2013, Vol. 10, Issue 1, No 3, pp. 439–442.
[28] Radchenko, R., Radchenko, A., Serbin, S., Kantor, S., Portnoi, B. Gas Turbine Unite Inlet Air Cooling by Using an
Excessive Refrigeration Capacity of Absorption-Ejector Chiller in Booster Air Cooler. In: HTRSE-2018, 2018: E3S
Web of Conferences 70, 03012, 6 p.
[29] Marques R. P., Hacon D, Tessarollo A., Parise J.A.R.: Thermodynamic analysis of trigeneration systems taking into
account refrigeration, heating and electricity load demands. Energy and Buildings, 2010, vol. 42, pp. 2323–2330.
[30] Ortiga, J. Bruno, J.C., Coronas, A.: Operational optimization of a complex trigeneration system connected to a district
heating and cooling network. Applied Thermal Engineering, 2013, vol. 50, pp. 1536–1542.

177
 THE IMPACT OF EXTREME WEATHER ON ELECTRICITY
DEMAND FROM HOMES HEATED BY AIR SOURCE HEAT
PUMPS

Chesser Michael
UCD, Dublin, Ireland, [email protected], ORCID: 0000-0002-8548-5339

Lyons Pádraig
ESB Networks, Dublin, Ireland, [email protected], ORCID: 0000-0001-6508-539X

O’Reilly Padraic
LIT, Limerick, Ireland, ORCID:

Carroll Paula
UCD, Dublin, Ireland, [email protected], ORCID: 0000-0003-1029-1668

Chesser, M., Lyons, P., O’Reilly, P., Carroll, P., The Impact of Extreme Weather on
Cite this paper as: Electricity demand from homes heated by Air Source Heat Pumps. 8th Eur. Conf. Ren.
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The Irish government is implementing policies to transition Ireland to a low carbon and
environmentally sustainable economy by 2050. Ireland has sectoral targets of 600,000 installed
heat pumps by 2030, currently roughly 28,000 are installed. Such a high target of heat pumps will
not only have a significant effect on electricity demand but also on the management and operation
of the grid. In this paper we explore residential electricity demand using an innovative dataset from
a field monitoring trial of deep retrofitted homes heated by an air source heat pumps (ASHPs). To
assess the effect of extreme weather events on heat pump demand and the impact of large-scale
adoption of heat pumps, this paper constructs the after diversity maximum demand per home. In
particular we address the research questions: what is the impact of extreme weather on heat pump
demand, and which statistical distributions beat model the after diversity maximum demand per
home heated by an ASHP events. The software package RStudio was used to model several
different distributions. Based on goodness-of-fit statistics and criteria, a Gamma distribution best
models average coincident electricity demand for homes heated by ASHPs.
Keywords: air source heat pump; probability density distributions; home electricity use; Ireland.
© 2020 Published by ECRES
Nomenclature
ADMD After Diversity Maximum Demand
ASHP Air Source Heat Pumps
BER Building Energy Rating
CDF Cumulative Distribution Function
DSO Distribution System Operators
DWH Domestic Hot Water
HP Heat Pump
kW kilowatt
RES Renewable Energy System
SH SuperHomes
TSO Transmissions System Operators

1. INTRODUCTION

Energy efficiency and renewable energy is of paramount importance in the European Union's strategy for
sustainable growth. Ambitions for low-carbon economies and competitive secure energy systems were endorsed

178
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

and enhanced in ‘Council directive 2018/2001’ [1]. Heating must be substantially decarbonised if goals for the
reduction of carbon dioxide and other greenhouse gas emissions are to be achieved.
Heat Pumps (HPs) are promoted as an attractive alternative for residential consumers to replace fossil fuel
heating systems. HPs are energy recovery systems that use some electricity to transfer higher temperature heat
from the external ground or air to the space heating and hot water systems of a building. They are advertised as
low maintenance systems offering lower costs and contributing to lowering carbon emissions and improving
local air quality. Additional benefits accrue when the electricity used to drive the HPs comes from renewable
energy systems (RES) [2].
The Irish government is implementing policies to transition Ireland to a low carbon, climate resilient and
environmentally sustainable economy by 2050. Ireland has sectoral targets of 70% for renewable electricity and
600,000 installed HPs of which 400,000 will be retrofitted by 2030 [3]. Ireland lies in the north Atlantic Ocean
on the European continental shelf and has a temperate oceanic climate. Ireland benefits from the North Atlantic
Current and Gulf Stream airflows. Additional heat is carried by the airflows giving Ireland a milder climate than
other places at similar latitudes. There is demand for space heating and domestic hot water (DWH) heating. Heat
demand is low density because of the dispersal patterns of the population of 4.8 million people, the nature of the
housing stock, and Ireland's temperate climate. Hence district heating has very limited potential, but HPs,
particularly air sourced heat pumps (ASHPs), have significant potential to decarbonise heating. Currently
domestic heating in Ireland relies heavily on conventional gas or oil boilers with hydronic radiators for space
and water heating. Renewable heat accounts for just 6.9% [4]. An estimated 28,000 HPs are currently installed
in Ireland [5].
Such a high target of HPs will not only have a significant effect on electricity demand but also on the
management and operation of the grid. According to [6], there are four potential problems; at the Transmission
System Operators (TSO) level, problems are peak demand and ramp rate increase and at the Distribution System
Operators (DSO) level, excessive voltage drops and the reinforcement of low voltage feeder and transformers.
The metric after diversity max demand (ADMD) is used by DSOs in the design of distribution networks where
coincident demand is aggregated over a large number of customers [7]. This metric allows the DSO to dimension
the low voltage network in anticipation of the average concurrent maximum demand.
In achieving national climate action plans, we need to estimate the impact of the adoption of low carbon
technologies on ADMD. In this paper we explore the demand from ASHP demand using an innovative dataset
from a field monitoring trial over the duration of an extreme weather event, ‘Storm Emma’ in 2018. The planned
uptake of HPs could have a significant effective on electricity demand not only during the winter in temperate
climates but under extreme conditions. Love et al. [6] use a winter period when calculating the ADMD for HPs
in the UK, because during these times heating energy demand is at its highest. In particular we address the
following research questions:
1. What impact does an extreme cold weather event have on electricity usage of homes heated by
ASHPs when compared to a typical winter day?
2. Which statistical distributions best models coincident demand from homes heated by ASHPs
demand?
Electricity use from homes with high time resolution is difficult to model due to seasonal and diurnal variations
as a result of appliance use, lighting and heating [8,9]. Bottom-up models of electricity usage data can provide
opportunities to investigate changes in technologies and usage patterns [10]. As noted by Munkhammar, et al.
note that there are many studies on various levels of time resolution using different distributions for home load
profiles such as Normal, Log-Normal, Gamma, Gumbel, Inverse Normal, Beta, Exponential, Rayleigh and
Weibull [11-18]. There is generally no standard distribution type for modelling home electricity use [8].
McLoughlin et al. used a Weibull and log – logistic probability distribution function when characterising
domestic electricity consumption in Ireland [9].

2. METHODOLOGY

We explore statistical distributions for the additional ADMD per ASHP heated home using a unique dataset of
the electricity usage of deep retrofitted typical Irish homes located in the countryside.
ADMD is used by DSO’s in the design of distribution networks where coincident demand is averaged over a
large number of customers [7]. To assess the impact HP’s will have we first smooth the electricity use data by
calculating the average usage per customer in the sample of N customers for each time step over the sample time
horizon T. Eq. 1. shows how we create the smoothed time series TSt.

179
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑁
1
𝑇𝑆𝑡 = ∑ 𝑑𝑖𝑡 ∀ 𝑡 = 1 … 𝑇 (1)
𝑁
𝑖=1

We used the software package RStudio for data analysis and modelling. The R package ‘fittdistrplus’ was used
as it provides functions for fitting univariate distributions to different types of data and allowing different
estimation methods. We fit and evaluate distributions for the data in TSt.
We calculate the ADMD per home, ADMD,
using Eq.2. to find the maximum average electricity demand over the sample time horizon:
𝐴𝐷𝑀𝐷 = 𝑚𝑎𝑥𝑡 ∈𝑇 (𝑇𝑆𝑡 ) (2)

2.1 The Superhomes Projects

Superhomes 1.0 (SH1.0) was a concept developed by the Tipperary Energy Agency. It combined a range of deep
retrofit measures along with the installation of ASHPs, Wood Stoves and Photovoltaic (PV) technologies. It
helped homeowners upgrade their homes to a very high building energy rating (BER). As part of the (SH 1.0)
electricity meters were installed in the homes and the electricity usage was monitored in 19 homes.

2.2 Data Collection

The SH1.0 dataset includes the average electricity use in kilowatts (kW) at 15 minutes resolution and several
metrological readings. The electricity metering recorded total home electricity usage including the ASHP and
DWH hot water immersion heating. Space Heating and DWH account for roughly 80% of energy use in the
average Irish Home [19]. Temperature data for the dataset was sourced from Gurteen weather station in
Tipperary which is the nearest weather station to Superhomes.
Due to privacy concerns, information about the size and make of the HP as well as socioeconomics and
dwelling characteristics are not made available for this analysis. Recording of the homes began at different times
in 2018 with the majority starting in January. All recordings finished on 28 th February 2019. There were some
instrumentation failures during the field trial. As a result, not all homes have data for the selected time period.

2.3 Sample Period – Storm Emma

Irish winters generally do not experience major snowstorms, because of their rarity when they do occur, they
cause serious disruption. Sometimes during a winter, significant snowfall occurs with blocking high pressure to
the north of Ireland pushing the North Atlantic jet stream and associated storm systems south, bringing an
easterly Polar Continental airflow. This type of pattern can follow a phenomenon called a “Sudden Stratospheric
Warming (SSW)” [20]. There have been 25 SWW between the years 1958 to 2013. The SSW on the 12 th of
February 2018 resulted in a polar continental air mass, named “The Beast from the East”. Resulting in snow fall
across the country which accumulated due to temperatures staying below freezing even during the day. Storm
Emma arrived in Ireland on the 28th February 2018, combining with “The Beast from the East” resulting in
stronger winds creating blizzard like conditions driving snow further inland. The extreme weather event lasted
until 4th of March 2018, resulting in the cancellation of public transport across the country, the closure of schools
and businesses and 100,000 homes and businesses losing power [21].
The replacement of the traditional methods (natural gas, solid fuels & oil) of home heating with electric sources
such as HPs will increase demand for electricity, especially during cold winters. According to Leahy & Foley
[21], “the daily total wintertime electricity demand in Ireland has been shown to increase by around 1.6 GWh
with every 1°C drop in air temperature”.

3. RESULTS

Focusing on research question 1, the impact of an extreme weather event has on electricity usage of homes
heated by ASHPs when compared to a typical winter day. We chose the 28 th February 2018 from the extreme
weather event and the same day the following year as a typical winter day. The 28 th February was chosen as Met
Eireann [22] notes Februarys maximum and minimum air temperatures for the country were recorded at the
Gurteen weather station on this day in 2018. During this time frame only eight of the SH1.0 recorded complete
data.

180
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

To illustrate the impact an extreme weather event had on average electricity usage, table 1 provides statistics
from a day during the event (2th column) with the same day the following year (3rd column) 28th February.
Comparing the temperature first, none of the descriptive statistics are above freezing temperature during 28 th
February 2018 while the same day the following year had an average temperature of 8.4 °C, a minimum
temperature of nearly 6°C and a maximum of roughly 12°C. Next, viewing the average electricity usage per
home we see the major impact an extreme weather event can have. The ADMD was 3.839 kW on the 28th
February 2018 while the following year it was only 2.331 kW. The magnitude for all the descriptive statistics is
considerably lower for electricity usage for the 28th February 2019. T-test for the difference of the means was
conducted for temperature and average electricity usage, in both tests we reject the null hypothesis of no
significant difference between the means. Also, from the empirical probability distribution of average electricity
usage per home for the two dates, the Wasserstein distance was calculated to be 1.43

Table 1: Descriptive Statistics for average electricity use per home and temperature for Storm Emma, A Day
during Storm Emma & The same day the following year.
28th February 2018 28th February 2019 Storm Emma, 28th February - 4th
March 2018
Temperature Average Temperature Average Temperature Average
(°C) electricity (°C) electricity (°C) electricity usage
usage (kW) usage (kW) (kW)
Min -4.90 1.086 5.80 0.349 -5.60 1.086
Median -2.75 2.554 7.85 0.993 -0.55 2.528
Mean -2.77 2.503 8.43 1.075 -0.67 2.529
Max -1.50 3.839 12.10 2.331 3.60 4.361
n 768 768 3,840

Figure 1. shows the temperature during the first three days of Storm Emma were below freezing and only rose
above 2.5 °C in the latter hours of 4th March. The minimum temperature of -5.6°C was recorded at 4:00 on 1st
March and the average temperature was -0.67°C over the duration of Storm Emma (Table 1, 4th column). The
ADMD value, i.e. the maximum value is 4.361 kW (Table 1, 4th column). This is the maximum coincident
demand the DSO can expect per es. The min average coincident demand use during Storm Emma was 1.086
kW.

Fig. 1. The averaged SH2.0 electricity usage TSt and Temperature, 28th February – 4th March 2018

Focusing on research question 2, the probability distributions that best models the average coincident demand
of homes heated by ASHPs, the histogram in Figure 2 shows the average electricity usage during Storm Emma
(28th February – 4th March 2018), the skewness of the data is 0.28 and the kurtosis is 2.78. The data is slightly
skewed to the right.

181
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig. 2. Coincident averaged electricity usage Histogram and Empirical Density, Storm Emma:28th February – 4th
March 2018.

In the next step we use curve fitting to model the average electricity use per home. As part of the ‘fittdistrplus’
package in RStudio, we use the function ‘descdist’ to produce a Cullen and Frey graph which is used to reduce
the set of possible candidate distributions. Based on this information, we consider; Gamma, log-normal, normal
and Weibull distributions, but not logistic or exponential distributions. The parameters of the distributions are
estimated using maximum likelihood estimates (MLE).
Table 2 presents the results of the ‘gofstat’ function which computes the goodness-of-fit statistics for parametric
distributions. The three goodness-of-fit statistics produced by this function are; Kolmogorov-Smirnov, Cramer-
von Mises and Anderson-Darling. All goodness-of-fit statistics are in favour of the Gamma distribution over the
other distributions. The Gamma fit measures are highlighted in bold font.

Table 2: Goodness-of-fit Statistics & Criteria

Gamma Log Normal Weibull
Normal
Goodness- Kolmogorov-Smirnov statistic 0.03 0.05 0.05 0.05
of-fit Cramer-von Mises statistic 0.14 0.22 0.18 0.27
statistics Anderson-Darling statistic 0.83 1.28 1.20 2.18

Goodness- Akaike's Information Criterion 804.08 811.27 811.93 831.40

of-fit criteria Bayesian Information Criterion 812.43 819.62 820.27 839.75

The MLE estimates for the Gamma distribution shape and rate parameters are shape  19.93, rate  7.88. We
use these estimates to explore the tails of the distribution for the extreme weather event. The 99, 95 and 90 %
cut off points are 3.86 kW, 3.47 kW and 3.27 kW per customer respectively. This allows the DSO to estimate
the probability of electricity demand for its portfolio of customers being above these cut off points during an
extreme weather event which occur once every two years.

4. DICUSSION & CONCLUSION

Ireland has set an ambitious target to grow from a base of less than 30,000 to 600,000 HP installations by 2030
in an effort to electrify and decarbonise the residential heating sector. This will have significant effect on
electricity demand and on the management and operation of the grid. The coincident average demand per homes
allows for network planning. Using the SH1.0 data we calculated the ADMD per home heated by an ASHP. In
our first research question we assessed what impact does an extreme cold weather event have on electricity
usage of homes heated by ASHPs when compared to a typical winter day. The ADMD was much lower 3.84
kW and 2.33 kW respectively. However, when the temperature differences between the two days are taken into
account, there would have been an expectation of a much higher ADMD on the day during the storm. This may
be due to the sample size of 8 homes, in future research we plan to explore detailed models to estimate ADMD.
In our second research question, we found that a Gamma distribution is best to model coincident demand from
homes heated by ASHPs demand after assessing several distributions on the data, and analysis of the tails of the
distribution provides useful information for network planning purposes. Also, the ADMD per home heated by

182
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

an ASHP as 4.361 kW during the extreme weather event of Storm Emma and The Beast from the East., and
analysis of the tails of the distribution provides useful information for network planning purposes.

ACKNOWLEDGMENT

Michael Chesser is funded by the Sustainable Energy Authority of Ireland RD&D 2018 programme,
“Exploration of Air Source Heat Pumps for Ireland's Residential Heating Needs", grant agreement RDD331.
Padraic O’Reilly and Limerick Institute of Technology’s participation in the Superhomes 2.0 project was funded
under Enterprise Ireland Hosting Grant Agreement Ref: TC-2013-0002B and IERC funding ref:
IERC_2017_001
an acknowledgment and thank to institutions or real people.

REFERENCES

[1]COM, Directive (EU) 2018/2001 of the European parliament and of the council of 11 December 2018 on the promotion of
the use of energy from renewable sources, Official Journal of the European Union (2001), accessed August 2019 (2018).
URL https://eur-lex.europa.eu/legalcontent/
EN/TXT/PDF/?uri=CELEX:32018L2001&from=EN
[2] Carroll, Paula, Michael Chesser, Padraig Lyons (2020). “Air Source Heat Pumps Field Studies: A Systematic Literature
Review” (submitted to Renewable and Sustainable Energy Reviews)
[4] SEAI, Energy in Ireland 2018 report, Technical report, Sustainable Energy Authority of Ireland SEAI, accessed Nov. 2019
(2018). URL https://www.seai.ie/publications/Energy-in-Ireland-2018.pdf
[5] EHPA, European heat pump market and statistics, online, accessed August 2019 (2018). URL
http://www.stats.ehpa.org/hp_sales/country_cards/
[6] Love, Jenny, Andrew ZP Smith, Stephen Watson, Eleni Oikonomou, Alex Summerfield, Colin Gleeson, Phillip Biddulph,
Lai Fong Chiu, Jez Wingfield, Chris Martin, Andy Stone, Robert Love. (2017) "The addition of heat pump electricity load
profiles to GB electricity demand: Evidence from a heat pump field trial." Applied Energy 204: 332-342.
[7] Barteczko-Hibbert, Christian. (2015) “After Diversity Maximum Demand (ADMD) report.” Report for the ‘Customer-Led
Network Revolution’ project: Durham University. http://www.networkrevolution.co.uk/project-library/diversity-maximum-
demand-admd-report/ accessed March 2020.
[8] Munkhammar, Joakim, Jesper Rydén, and Joakim Widén (2014) “Characterizing probability density distributions for
household electricity load profiles from high-resolution electricity use data” Applied Energy, 135: 382-390
[9] McLoughlin, Fintan, Aidan Duffy, and Michael Conlon. (2012) “Characterising domestic electricity consumption patterns
by dwelling and occupant socio-economic variables: An Irish case study” Energy and Buildings, 48: 240-248
[10] Fischer, Davis, Arne Surmann and Karen Byskov Lindberg (2020) “Impact of emerging technologies on the electricity load
profile of residential areas.” Energy & Buildings 208: 109614
[11] Gonzalez-Longatt, Francisco, José Rueda, István Erlich, Walter Villa, and Dimitar Bogdanov (2012) “Mean variance
mapping optimization for the identification of Gaussian mixture model: test case, intelligent systems (IS).” In 6th IEEE
international conference 158–163.
[12] Herman, R, and JJ. Kritzinger (1993) “The statistical description of grouped domestic electrical load currents”. Electric
Power Systems Research 27(1): 43–48.
[13] Heunis, Schalk W., and Ron Herman (2002) “A probabilistic model for residential consumer loads.” IEEE Transactions on
Power Systems 17(3) :621–625.
[14] Irwin GW, W. Monteith, and W.C., Beattie (1986) “Statistical electricity demand modelling from consumer billing data.”
IEE Proceedings C (Generation, Transmission and Distribution) 133: 328–335.
[15] Lin, Hsin-Hui, Kaung-Hwa Chen, and Rong-Tsorng Wang. (1993) “A multivariate exponential shared-load model.” IEEE
Transactions on Reliability 42:165–171.
[16] McQueen, Dougal.H., Patrick R. Hyland, and Simon J.,Watson. (2004) “Monte Carlo Simulation of residential electricity
demand for forecasting maximum demand on distribution networks.” IEEE Transactions on power systems 19(3):1685–
1689.
[17]Seppälä, Anssi. (1995) “Statistical distribution of customer load profiles.” In Proceedings 1995 International Conference on
Energy Management and Power Delivery EMPD’95 2: 696–701.
[18] Walker, Charles. F., and John L. Pokoski (1985) “Residential load shape modelling based on customer behaviour.” IEEE
Transactions on Power Apparatus and Systems 7:1703–1711.
[19]SEAI (2018) Energy in the Residential Sector 2018 URL https://www.seai.ie/publications/Energy-in-the-Residential-
Sector-2018-Final.pdf
[20] Met Eireann (2019) Storm Emma URL https://www.met.ie/cms/assets/uploads/2019/02/EmmaReport2019.pdf
[21] Leahy P.G. and Aoife M. Foley (2012) “Wind generation output during cold weather-driven electricity demand peaks in
Ireland.” Energy 39(1):48-53
[22] Met Eireann (2018) Weather Statement for February 2018 URL https://www.met.ie/climate/past-weather-statements#

183
 RESEARCH OF CHARACTERISTICS OF THE FLOW PART OF
AN AEROTHERMOPRESSOR FOR GAS TURBINE
INTERCOOLING AIR
Dmytro Konovalov
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected],
ORCID: 0000-0001-7127-0487

Mykola Radchenko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected],
ORCID: 0000-0002-1596-6508

Halina Kobalava
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0002-0634-5814

Victoria Kornienko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID:
0000-0003-3524-2045

Roman Radchenko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0002-8099-7327

Vitaliy Maksymov
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID:
0000-0001-6301-5857

Konovalov, D., Radchenko, M., Kobalava, H., Kornienko, V., Radchenko, R., Maksymov, V.
Cite this paper as: Research of characteristics of the flow part of an aerothermopressor for gas turbine
intercooling air. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: One of the promising areas to increase the fuel and energy efficiency of gas turbines is contact
cooling of cyclic air by using a two-phase jet apparatus – an aerothermopressor. The organization
rational parameters of work processes with the corresponding development of the
aerothermopressor design were studied. An experimental setup was developed to conduct an
experimental study for determining pressure losses in the aerothermopressor flow part. The
experimental setup is designed to simulate the aerothermopressor operation in the cooling cyclic
air of the gas turbine. Based on the obtained experimental data, an empirical equation was found
to determine the hydraulic resistance coefficient of the aerothermopressor flow part, depending on
the initial pressure and the amount of water injected. The deviation of the calculated hydraulic
resistance coefficient from the experimental ones is ± 25 %. The obtained results can be used in
the practice of designing the low-flow aerothermopressor for gas turbine cyclic air cooling.
Keywords: Cyclic air cooling, Flow part, Pressure losses
© 2020 Published by ECRES

1. INTRODUCTION

Complex schemes with cyclic air cooling are usually used to bring close the process of compressing the working
fluid in gas turbine compressors to isothermal. It leads to increase the fuel and energy efficiency [1, 2]. One of
the promising areas is contact cooling of cyclic air by using a two-phase jet apparatus – an aerothermopressor
[3]. This device provides efficient evaporative cooling of gas turbine without loss of total pressure. Due to
evaporative cooling in the aerothermopressor, an effect of thermogasdynamic compression is taken place – gas
pressure is increased in the process of instantaneous evaporation of water injected into the gas (air) flow
accelerated to a speed close to sound [4, 5]. An effective water dispersed atomization in the air flow occurs in

184
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the aerothermopressor. This ensures efficient evaporation of the water dispersed flow in the flow part of the gas
turbine compressor bringing the compression process closer to isothermal.

2. LITERATURE REVIEW

It is important to determine the organization rational parameters of work processes with the corresponding
development of the design of the aerothermopressor flow part at designing jet devices.
The aerothermopressor allows to compensate for pressure loss and reduce the work on compression in the
compressor and to increase the working fluid flow and, accordingly, to increase the gas turbine capacity [3]. In
addition, the aerothermopressor is compact and structurally and simple technologically to manufacture in
comparison with surface air coolers [4, 6, 7].
A significant influence on the working processes in the aerothermopressor is carried out by design factors. The
total pressure of the gas flow in the aerothermopressor is significantly affected by pressure losses due to
hydraulic and local resistance in the aerothermopressor flow part (confuser, evaporation chamber and diffuser)
[5, 8]. The operation of the aerothermopressor on the gas turbine exhaust gases was investigated [4]. Data were
obtained on the pressure loss for the experimental aerothermopressor (gas flow rate Gg = 11.5 kg/s) through
local and hydraulic resistance in different areas. Moreover, the total pressure was decreased up to 14 %.
The pressure losses in the aerothermopressor flow part depend primarily on the operating mode nature. The
process of accelerating the gas flow to a transonic speed (0.5-0.9 M) occurs in the confuser. In the receiving
chamber located before the confuser, liquid is supplied using a nozzle or other special devices [4, 8, 9]. In the
evaporation section, these flows interact, and, as a result, droplets are accelerated, crushed, heated and
evaporated, and gas is also cooled [5]. There are three main operating modes of the aerothermopressor [4, 5, 9]:
1) The influence of the water droplets resistance prevails over the positive effect of evaporation and determines
the gas flow behaviour. The Mach number is increased, the static pressure of the flow is decreased, and the water
temperature has risen approaching the saturation temperature; 2) Water evaporation predominates. The Mach
number is decreased, and the total and static pressure is increased; 3) Surface friction (in the first two modes
was relatively insignificant) becomes the predominant factor. This mode is taken place when the liquid
completely evaporates.
All three modes determine the effect of hydraulic resistance on pressure losses. One of the ways to increase the
aerothermopressor efficiency is to increase the total pressure by reducing losses due to hydraulic resistance of
the flow part. It might be achieved by providing liquid incomplete evaporation – dispersed mode. Incomplete
evaporation should provide a greater pressure increase, due to a friction losses decrease in the two-phase flow
and it should provide effective fine atomization of water, due to a drop reduction during partial evaporation.
It is known that pressure losses in the two-phase flow (liquid-gas) may be less than pressure losses in the single-
phase flow (gas) [10, 11, 12]. When providing liquid injection through the nozzle in an amount more necessary
for evaporation, mode 3 is absent (in this mode, surface friction of the single-phase flow predominates). It
positively affects the increase in total flow pressure as a result of the thermogasdynamic effect.
To determine pressure losses in the aerothermopressor, classical methods are used. These techniques are based
on the analytical determination of the local and friction resistance coefficients of each of the aerothermopressor
elements [5, 9]. Pressure losses due to resistance (acceleration or deceleration of the droplet) can be determined
by the aerodynamic resistance coefficient of the droplet in the flow w in accordance with the technique proposed
in [13]. Pressure losses in the evaporation chamber ch might be determined by the classical Blasius equation for
flow in the channels [14] or by the Lockhart-Martinelli phases slip model [11]. To calculate losses from the total
resistance for the confuser and diffuser are used equations for the local loss coefficients с, d [14].
Thus, the general equation for determining pressure losses in the aerothermopressor:
2
𝑤𝑎𝑖𝑟 ∙ 𝜌𝑎𝑖𝑟
Δ𝑃𝑟 = 𝜁𝐴𝑇𝑃 ∙ (1)
2

where ATP – total coefficient of hydraulic resistance in the aerothermopressor flow part; wair – average air
velocity in the evaporation chamber; air – air density.
The total coefficient of hydraulic resistance is determined by:

𝜁𝐴𝑇𝑃 = 𝜁𝑤 + 𝜁𝑐 + 𝜁𝑑 + 𝜁𝑐ℎ (2)

185
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The application of the indicated methods for calculating losses by the elements of the aerothermopressor flow
part often gives incorrect results. First of all, it is due to the complexity of accounting for the operating modes
of the flow in the aerothermopressor. Even more difficult is to determinate the losses due to the resistance of the
droplet [15] in the confuser and in the evaporation chamber initial part. This is due to the difficulty in
determining the initial velocity of the droplet, its diameter, etc. It has been experimentally established that such
losses can reach 10–20% [4, 5], but there are no equations for the exact determination of such losses for the
aerothermopressor. Determination of pressure losses in the aerothermopressor flow part experimentally, taking
into account water injection, will clarify the design methodology for such jet devices. This will provide an
accurate determination of the aerothermopressor effective use as a part of the gas turbine for cooling cyclic air
under various operating conditions, as well as in various climatic conditions.

3. RESEARCH METHODOLOGY

The installation of the aerothermopressor in the gas turbine cyclic air cooling system is proposed in accordance
with the scheme (Fig. 1, a) [3]. This arrangement between the compressors will reduce the operation of the gas
turbine high-pressure compressor due to lower initial temperature of the compression process. This will ensure
isothermal compression during the evaporation of the finely dispersed mixture in the high-pressure compressor
flow part.
To conduct an experimental study of determining pressure losses in the aerothermopressor flow part, an
experimental setup was developed (Fig. 1, b). The experimental setup is designed to simulate the
aerothermopressor operation in the cooling cyclic air of the gas turbine. The simulation of the aerothermopressor
operation to cool cyclic air of the gas turbine WR-21 from Rolls Royce (Ne = 25.250 kW, ge = 0.190 kg/(kW∙h),
е = 41.2%) were considered.
The principle of experimental setup operation (Fig. 1, b): after cleaning in the air filter (2) (Caterpillar - 4N-
0015 CAT), the air was supplied for compression to the screw compressor (3) (Atlas Copco ХА 85), and pumped
at 0.6 MPa into the air receiver. After deep cleaning in the three-section moisture separator and oil separator (7),
the air was heated in the ducted gas heater (11) to t1 = 50-180 °C. After preparation (approximation to gas turbine
cyclic air parameters), the air entered into the experimental aerothermopressor (19). Water for injection came
from a distilled water reserve tank (14) and it was injected by a high-pressure pump (18) (STIHL RE 98). Water
spray was carried out by nozzles of type FMT. At the pressure of 7.5 MPa, the nozzles provided a water flow
rate at the receiving chamber inlet of the aerothermopressor: FMT-43.0 – gw = 0.0175 kg/s; FMT-100.0 –
gw = 0.0407 kg/s; FMT-120.0 – gw = 0.0487 kg/s. The spraying angle was 70-90 ° and the average droplets
diameter was w = 18–20 μm with a maximum diameter of w = 50 μm.
The experimental aerothermopressor consists of the following elements: a receiving chamber with a nozzle and
system for injecting water into the flow; confuser; evaporation chamber; diffuser; nozzles for installing
temperature and pressure sensors. All elements of the aerothermopressor are removable allowed to carry out
studies for the different geometric characteristics: receiving chamber – diameter D1 = 65 mm; length L1 = 200
mm; confuser – inlet diameter Dc1 = 65 mm, outlet diameter Dc2 = 25 mm, length Lc = 34 mm, convergent angle
α = 30°; evaporation chamber – diameter Dch = 25 mm; length Lch = 125 and 175 mm (relative length lch = 5; 7);
diffuser – inlet diameter Dd1 = 25 mm, outlet diameter Dd2 = 65 mm, length Lch = 192 mm, divergent angle β =
6°; nozzle – the distance between the water exit point to the receiving chamber inlet L f = 5; 65; 125 mm (Lf =
125 mm corresponds to the evaporation chamber inlet).
To determine the local resistance coefficients for the diffuser and confuser, classical dependences of fluid
dynamics [5, 6] were used. The energy equation (Bernoulli equation for air flow, taking into account mechanical
specific losses):

𝜌𝑤12 𝜌𝑤22
𝜌𝑔𝑧1 + 𝑝1 + 𝑁1 = 𝜌𝑔𝑧2 + 𝑝2 + 𝑁2 + Δ𝑝𝑡 (3)
2 2

where ρgz1, ρgz2 – geometric pressure; p1, p2 – static pressure; Δpt – total losses of total pressure; N1 w1 ,
2

2
w 2
N 2 2 – dynamic pressure.
2
Dividing by ρg:

186
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑝1 𝑤12 𝑝2 𝑤22
𝑧1 + + 𝑁1 = 𝑧2 + + 𝑁2 + ∆ℎ𝑡 (4)
𝑔𝜌 2𝑔 𝑔𝜌 2𝑔
where Δht – total head losses.
The equation for total pressure is:

𝜌𝑤12
𝑝𝑡 = 𝑝1 + 𝑁1 (5)
2

Taking into account the fact that z1 = z2 and Eq. [4], the total head losses Δht:
𝑝𝑡1 − 𝑝𝑡2
∆ℎ𝑡 = (6)
𝑔𝜌

The local resistance coefficient ATP is:

2Δ𝑝𝑡
𝜁𝐴𝑇𝑃 = (7)
𝜌𝑤 2

Figure 1. Scheme of the gas turbine with cyclic air cooling by the aerothermopressor, (b) Scheme of the experimental
setup:
ATP – aerothermopressor; LPC, HPC – low and high pressure compressors; CC – combustion chamber; HPT, LPT – low
and high pressure turbines; PT – power turbine;
1 – inlet pipeline; 2 – air filter; 3 – screw compressor; 4 – safety valve; 5 – air receiver; 6, 9 – control valves; 7 – three-
section moisture separator; 8 – drain oil and moisture; 10 – temperature compensator; 11 – gas air heater; 12, 13 – inlet
and outlet of hot gas; 14 – water reserve tank; 15 – control valve of the injection system; 16 – return pipeline; 17 – pipeline
for supplying water to the injection system; 18 – high pressure pump; 19 – experimental aerothermopressor; 20 – control
valve at the outlet; 21 – output pipeline; P – pressure sensors; T – temperature sensors.

4. RESULTS

To determine the hydraulic resistance coefficient of the aerothermopressor flow part, a number of experimental
measurements of pressure losses Pr were carried out under various conditions (Fig. 2): the pressure at the inlet
to the receiving chamber – Р1 = 150; 200; 250; 300 kPa; injected water flow rate relative to the flow rate of the
two-phase flow – gw = 5-35% (up to 0.0487 kg/s). The measurements are showed: at the inlet pressure P1 = 300
kPa pressure losses are Pr = 55-75 kPa (20-24%), at P1 = 250 kPa pressure losses are Pr = 40-45 kPa (16-
18% ), at Р1 = 200 kPa pressure losses are Pr = 22-24 kPa (12-16%). In this case, with an increase in water
flow rate gw, pressure losses increase by 2-4%, and when the initial pressure decreases from 300 to 150 kPa, the
effect of the amount of injected water gradually decreases. At Р1 = 150 kPa, pressure losses are Pr = 18-20 kPa
(11-12%), but with an increase gw to 35% pressure losses remain practically unchanged, the effect of water flow
rate on pressure losses ceases. Pressure losses increase Pr with injected water flow rate increase gw at inlet
pressures P1 = 200–300 kPa is explained by the greater difference between the initial droplet velocity at the
nozzle exit and the air velocity – (ww / wair) = 1.1–1.3. Hence, there is a greater influence of the droplet resistance.

187
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

It can reach up to 20% of the total pressure loss in the aerothermopressor. At the pressure P1 = 150 kPa, the
droplet and air velocities are almost equalized and the droplet resistance decreases significantly.
Determination of the aerothermopressor hydraulic resistance coefficient ATP.E based on the data (Fig. 3): at P1
= 300 kPa – ATP.E = 1.15-1.30, at P1 = 250 kPa – ATP.E = 0.90-1.00 and at P1 = 200 kPa – ATP.E = 0.70-0.75.
The tendency of the hydraulic resistance coefficient behavior is the same as for absolute pressure losses Pr,
that is, with a water flow rate increase, the hydraulic resistance coefficient ATP.E increases by 0.05-0.15. At Р1
= 150 kPa – the value of ATP.E practically does not change and it is – 0.35-0.40. This behavior is explained by
the fact that there are no influences of the Re and the relative velocity M = (wair/wsound) on the hydraulic resistance
coefficient. This indicates there is a self-similar flow regime in the elements of the aerothermopressor flow part.
The value of the hydraulic resistance coefficient depends only on the geometric parameters of the channel:
confuser convergent angle α and diffuser divergent angle β; the degree of narrowing of confuser nс, the degree
expansion of diffuser nd; the relative length of the evaporation chamber (caliber) lch = (Lch/Dch).
Pr, kPa
Pr, %
80 30
P1 = 300 kPa
P1 = 300 kPa
60
20
250 kPa
250 kPa 200 kPa
40
150 kPa
200 kPa
10
20 150 kPa

0 0
0 10 20 30 gw,40
% 0 10 20 30 gw,40
%
(a) (b)
Figure 2. (a) Experimental dependences of pressure losses Pr (b) Pr = 100∙(Р1–Р2)/Р1 (%) on the relative flow rate of
the injected water gw at various initial pressures: P1 = 150; 200; 250; 300 kPa

The equations of the dependence of the hydraulic resistance coefficient in the aerothermopressor flow part ATP
based on the obtained experimental data were determined by the approximation method according to a number
of equations and depending on the initial pressure at the receiving chamber inlet P1 and the relative water injected
flow rate gw. In this case, the plane equations were selected:

𝜁𝐴𝑇𝑃 = 0.0983 ∙ 𝑔𝑤 + 5.8633 ∙ 10−6 ∙ 𝑃1 − 0.5004 (8)

 ATP.E
 ATP.E
1,75 1,4
+25% 0
1,50 1,2
P1 = 300 kPa
1,25 1,0
-25%
1,00 250 kPa 0,8

0,75 200 kPa 0,6

0,50 150 kPa 0,4

0,25 0,2

0,00 0,0
0 10 20 30 gw,40
% 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4
ATP.C
(a) (b)
Figure 3. (a) Experimental dependences of the hydraulic resistance coefficient ATP.E on the relative flow rate of the
injected water gw at various initial pressures and (b) Comparison of the hydraulic resistance coefficient calculated values
ATP.C in the aerothermopressor flow part (equation (8)) with the experimental ATP.Е: P1 = 150; 200; 250; 300 kPa.

This equation (regression coefficient – R = 0.8847; R2 = 0.7827) is obtained for the following characteristics of
the aerothermopressor operation: 1.2 · 105 < Re < 3.4 · 105; M = 0.25–0.65; gw = 0.05–0.35; Р1 = 150–300 kPa.

188
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The deviation of the calculated values of the hydraulic resistance coefficient ATP.C (equation [8]) from those
obtained during the experimental study ATP.E is  = ± 25 % (Fig. 3 (b)). Based on the calculated values ATP.C,
the absolute pressure loss in the aerothermopressor Pr.С was determined (equation [1]) and their comparison
with experimental Pr.E was carried out. In this case, the deviation of the calculated data with the experimental
ones was no more than P = ± 25 %. This indicates that the empirical dependence [8] for determining ATP can
be applied to the considered conditions when designing aerothermopressor for gas turbine cyclic air cooling.
But it should be noted that experimental data were obtained only at air flow rates to 0.52 kg/s, i.e., it is possible
to apply the obtained dependence only for the low-flow aerothermopressor with small diameters of the
evaporation chamber.

5. CONCLUSION
During an experimental study of the low-flow aerothermopressor the hydraulic resistance coefficients of its flow
part were determined. The experimental hydraulic resistance coefficient is ATP.E = 0.35-1.30, depending on the
initial pressure P1 and the injected water relative flow rate into the receiving chamber gw. The absolute values
of pressure losses are Pr = 18-75 kPa (11-24%).
An empirical equation was found to determine the hydraulic resistance coefficient of the aerothermopressor flow
part ATP, depending on the amount of water injected. This coefficient takes into account hydraulic losses: local
losses in the confuser and diffuser; friction losses in the evaporation chamber; losses due to resistance of water
droplets. The deviation of the calculated hydraulic resistance coefficient ATP.C from the experimental ones ATP.E
is  = ± 25 %. The deviation of the calculated pressure loss Pr from experimental is P = ± 25 %. This
indicates the possibility of applying this dependence under the conditions: 1.2 · 105 < Re < 3.4 · 105; M = 0.25–
0.65.
The results can be used in the design practice of the low-flow aerothermopressor for gas turbine air intercooling.

REFERENCES
[1] Jonsson, M., Yan, J.: Humidified gas turbines – a review of proposed and implemented cycles. Energy, 2005, vol. 30,
1013–1078.
[2] Boyce, M. P.: Gas Turbine Engineering Hand-book. Gulf Publishing Company, Houston, Texas, 2002.
[3] Konovalov, D., Kobalava, H.: Efficiency Analysis of Gas Turbine Plant Cycles with Water Injection by the
Aerothermopressor. In: Ivanov V. et al. (eds.) Advances in Design, Simulation and Manufacturing II. DSMIE 2019.
Lecture Notes in Mechanical Engineering, 2020, pp. 581–591. Springer, Cham.
[4] Fowle, A.: An experimental investigation of an aerothermopressor having a gas flow capacity of 25 pounds per second.
Massachusetts Institute of Technology, USA, 1972.
[5] Stepanov, I. R., Chudinov, V. I.: Some Problems of Gas and Liquid Motion in the Channels and Pipelines of Power
Plants. The Science Publ., Leningrad, 1977.
[6] Zhivitsa, V. I.: Intercoolers with thermopressor for two-stage ammonia refrigeration machines. Refrigeration
Engineering 2002, vol. 5, pp. 18–20.
[7] Konovalov D., Trushliakov, E., Radchenko, M., Kobalava, H., Maksymov, V.: Research of the aerothermopressor
cooling system of charge air of a marine internal combustion engine under variable climatic conditions of operation. In:
Tonkonogyi, V. et al. (eds.) Grabchenko’s International Conference on Advanced Manufacturing Processes.
InterPartner-2019. Lecture Notes in Mechanical Engineering, p. 10. Springer, Cham.
[8] Shapiro, A. H., et al.: The Aerothermopressor – a Device for Improving the Performance of a Gas-Turbine Power Plant.
In: Proceedings of the Trans. ASME, Cambridge, USA, 1956, pp. 617-653.
[9] Erickson, J.: A theoretical and experimental investigation of the aerothermopressor process. Massachusetts Institute of
Technology, USA, 1958.
[10] Radchenko, N.: A concept of the design and operation of heat exchangers with change of phase. Archives of
Thermodynamics: Polish Academy of Sciences 2004, vol. 4 (25), 3–19.
[11] Lockhart, R. W., Martinelli, R. C.: Proposed correlation of data for isothermal two-phase, two-component flow in
pipes. Chem. Eng. Progr. 1949, vol. 45(1), 39–48.
[12] Martinelli, R. C., Nelson, D. B.: Prediction of pressure drop during forced – circulation boiling of water. Trans. ASME
1948, vol. 70, 695–702.
[13] Dolinsky, A. A., Ivanitsky, G. K.: Heat and mass transfer in vapor-liquid dispersed flow. Naukova Dumka, Kyiv, 2008.
[14] Kulinchenko, V. R.: Hydraulics, hydraulic machines and hydraulic drive. Tsentr navchalnoi literatury Publ., Kyiv,
2006.
[15] Lin, A., et al.: Evaluation of Mass Injection Cooling on Flow and Heat Transfer Characteristics for High-Temperature
Inlet Air in a MIPCC Engine. International Journal of Heat and Mass Transfer 2019, vol. 135, 620–630.

189
 DYNAMIC ANALYSIS OF TRI-FLOATER WITH VERTICAL AXIS WIND
TURBINE SUPPORTED AT ITS CENTROID

Asadulla Thoppil
MES College of Architecture, Kozhikode, Kerala, India,[email protected], ORCID: 0000-0003-1799-
0015

M. Abdul Akbar
Department of Civil Engineering, Dr B R Ambedkar National Institute of Technology, Jalandhar (Punjab),
India,[email protected], ORCID: 0000-0002-4221-4395

V. Mustafa
Department of Civil Engineering, National Institute of Technology, NIT Campus (P.O), Calicut, Kerala-
673 601, India,[email protected]

DadiRamBabu
Department of Civil Engineering, Dr B R Ambedkar National Institute of Technology, Jalandhar (Punjab),
India, [email protected], ORCID: 0000-0001-9606-4233

Asadulla, Thoppil, M, Abdul Akbar, V,Mustafa. Dadi, RamBabu. Dynamic Analysis of

Cite this paper as: Tri-Floater with Vertical Axis Wind Turbine Supported at its Centroid. 8th Eur. Conf.
Ren. Energy Sys.24-25 August 2020, Istanbul, Turkey

Abstract: The paper deals with the dynamic analysis of a Tri-floater with 1MW offshore
Vertical Axis Wind Turbine (VAWT) placed at its centroid. Six line catenaries
mooring system was used for controlling the horizontal movement of the floater.
The floater was modeled as a rigid body with six degrees of freedom. Mass, damping
and hydrostatic stiffness were calculated using hydrostatic stability condition.
Aerodynamic load on vertical axis wind turbine was calculated using Streamtube
theory. Wave profile was calculated using Airy’s wave theory followed by the use
of Morison’s equation to determine the inertial and drag forces on the floater. A
computer program was developed for force calculation due to wind and wave and
for dynamic analysis. Dynamic analysis was carried out using Newmark-beta
method. The equation of motion for the floater was solved in time domain. Restoring
force by mooring lines at each instant of time was calculated based on the cable
profile. Responses of the Tri-floater with VAWT in different sea conditions were
analyzed and it was found thatsurge, heave and pitch are the predominant motions
for a straight (00) wave. These motions were also analyzed for waves of different
inclinations and the responses were studied and compared.
Keywords: Offshore Wind Turbine, Tri-floater, Wind Energy
© 2020 Published by ECRES

1. INTRODUCTION

Although the vision for large-scale offshore floating wind turbines was introduced by Professor William E.
Heronemus at the University of Massachusetts in 1972, it was not until the mid 1990s, after the commercial
wind industry was well established, that the topic was taken up again by the mainstream research community.
The need to exploit enhanced wind resources far offshore as well as in deep waters requires the use of floating
support structures. Various concepts of floating offshore wind turbines (FOWTs) have been proposed in the
past. It was not until recently that technological advances and policy supports have made FOWTsa viable
solution for harnessing ocean energy resources. Existing design concepts of floating support structures and
station keeping systems for FOWTsare mostly developed based on experience from the offshore oil and gas
industry, which has witnessed almost 60 years of design and operation.

190
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Review of Past Works

Hooft(1972) studied the hydrodynamic aspects of semi-submersible platforms, by subdividing the structure into
slender and simpler elements. The forces on each member were calculated using Morison’s approach and the
forces were summed up assuming that the hydrodynamic properties of each element of the submerged
construction were not influenced by neighboring elements. Tong et al. (1998) investigated the conceptual design
of a float consisting of a steel tripod space frame with the wind turbine mounted on it. The tower is bolted onto
the deck of a concrete cylindrical buoy hull with a wide bottom disk consisting of 8 line catenary mooring system
to control the motion. A 1.4MW offshore wind turbine is provided for locations between 100m and 300m water
depth. Bulder et al. (2002) presented the technical and economical feasibility of floating wind energy systems
for water depths greater than 50m. Different design concepts of the floating offshore wind turbine such as
cylindrical floater, cylindrical floater with tension leg, inverted spar, Tri-floater, quadruple floater, jack-up, etc.
were compared. Based on the cost analysis, it was concluded that the Tri-floater concept was technically and
economically the most feasible option. Ruoyu Zhang et al. (2013) studied the dynamic behavior of a semi-
submersible floating foundation of 600kW capacity wind turbine in 60m deep water considering the coupled
load effects of wind turbine-tower-floating foundation and mooring lines and the ocean environment. Borg et
al. (2014) presented an in-depth review of different aerodynamic engineering models for VAWTs and their
suitability for floating applications. Approaches to adequately model mooring systems and the structural
behavior of floating offshore wind turbines (FOWTs) during preliminary design stages were discussed with
particular emphasis on floating VAWT characteristics and its computational aspects.

2. MODEL DESCRIPTION

The different parameters of the model are provided in Table 2.1 and Table 2.2.

Table 2.1 Basic floater dimensions

Column diameter 7.7m
Column height 24m
Column draft 12m
Distance b/w column centre 68m
Pontoon diameter 3.3m
Heave plate diameter 15.4m
Total volume displacement 3078m3

Table 2.2 Turbine and tower dimensions

Machine rated power 1MW
Rotor diameter 38m
NACA0012(Standard
Blade cross section nomenclature for
aerofoil section)
Blade chord 0.5334m
Length of blade 17.01m
Height of hub 83m above SWL
*Cut-in, Rated, Cut-out
6m/s, 15m/s, 25m/s
wind speed
Rated rotor speed 27rpm
Tower base diameter 7.5m (hollow)
Tower top diameter 4.5m
Length of tower 70m
Centre of gravity of 31.8m above from
tower tower base
Tower weight 3528kN
Total machine weight 3352kN

191
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. AERODYNAMIC ANALYSIS

Wind Speed

External wind conditions in the offshore regime is defined in guidelines by Det Norske Veritas (DNV),
International Electro technical Commission (IEC) and Germanischer Lloyd (GL). The wind speed at 10m is
often used as the reference height in all the standards. The wind profile, V(z) denotes the average wind speed as
a function of height z above the ground. In the GL standards, the normal wind speed profile is assumed to be
given by the power law. The wind velocity at hub height is determined equation 4.1.

V ( z )  V z r 
 z  (4.1)
 zr 
Where zr and α is the power law exponent which takes a value of 0.11 for all wind speeds in offshore conditions
(as per GL standards). Wind velocity at the hub height will vary with time. Design wind velocity obtained from
the gust velocity is as follows:
 3t   2t  
V z, t   V z   0.37 * V gust * sin  1  cos   for, 0<t<T
 T   T 
= 𝑉(𝑧) for, t>T (4.2)
Where, T is the gust characteristics time period and Vgust is the largest extreme gust magnitude.

V gust  1.35 * V e1  V hub  (4.3)
Where,Ve1is the one-year extreme wind speed and Vhub is the wind speed at the hub height.

z
V e1  1.12 * V z r   (4.4)
 zr 
Wind Force on Tower

Wind force acting on the tower produces considerable effect on the floater motion. Wind velocity increases with
height whereas the tower projected area decreases with height. The tower is divided into a number of sections
and force acting on each of them is determined and added to calculate the total force on the structure. To find
the projected area of any section, the diameter of tower at any height z is required and it is given by equation
4.5.
D z  Db 1  z H   Dt  H 
z (4.5)
 
where, Dzis thediameter at any height z from the tower base, Dbis the tower base diameter, Dtis tower top
diameter and H is the height of the tower. From the design wind velocity (V), the static wind force (Fw) acting
perpendicular to an exposed area (A) can be computed as follows:
1
Fw   aV 2 Cs A (4.6)
2
Where, ρa is the density of air, Cs is the shape coefficient (Cs = 0.5 for cylindrical sections). Based on these
values and the dimensions given in Table 2.2, total force acting on the tower is calculated as 24.795kN.

Loading on Wind Turbine

Power generated by wind turbine depends on the interaction between the rotor blades and the wind. The analysis
of force on the turbine is carried out by assuming a control volume, in which its boundaries are the surface and
two cross-sections of the streamtube. The analysis is based on linear momentum theory. From the conservation
of linear momentum for a one dimensional, incompressible, time-invariant flow, the thrust is equal and opposite
to the change in momentum of air stream [Habtamu Beri et al.(2011)].

192
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

 a AV 1 2 4a1  a 
1
T (4.7)
2
Where,A is the projected area of turbine blade.
V V 2
a 1 (4.8)
V1
The thrust force on wind turbine can be characterized by a non-dimensional thrust coefficient as:
T
Ct  1 (4.9)
 A V2
2 a
Angle of attack can be expressed as
V a sin 
tan   (4.10)
V a cos   R
C N  C L cos   C D sin  (4.11)
C T  C L sin   C D cos  (4.12)
Where, CL is the lift coefficient and CD is the drag coefficient. The instantaneous thrust force (Ti) on a single
aerofoil is given by[Habtamu Beri et al.(2011)].

 V R 2 hc C t cos   C n sin  

1
Ti  (4.13)
2 a
Where h is the blade height and c is the chord length of blade.The aerodynamic efficiency of a wind turbine is
indicated by the power coefficient (Cp) given by
P max
Cp  (4.14)
1
 a AV 2
2
Where, Pmax is the maximum rated power of the wind turbine.The shape of aerofoil blade chosen for analysis is
NACA0012. For low values of angle of incidence, the aerofoil has a high lift to drag ratio. Axial induction factor
‘a’ for the calculation of thrust force and power produced by wind turbine is determined by an iterative procedure
developed by MATLAB.

4. DYNAMIC ANALYSIS OF FLOATER

Dynamic analysis of the semi-submersible floater which supports the wind turbine involves formulation of
equation of motion for analyzing the behavior of floater. The forces and moments due to wind loading on the
turbine blades and tower are transferred to the floater.
.
Hydrodynamic Loading
In this study, wave description is assumed to be regular. This simplification is based on certain arguments.
Irregular wave description is used for fatigue analysis where the lifetime load is required One is the operational
sea condition and another is rough sea condition.

Table 4.1 Sea states used for analysis

Wind velocity (m/s) Wave height (m) Time period (s)
Sea state 1(Operational) 12.2 4.2 9.2
Sea state 2(Rough) 21.1 8.78 11

5. RESULTS AND DISCUSSIONS

General
The system is modelled as a rigid body with six degrees of freedom. For the mathematical model developed, the
hydrostatic stability check gives a feasible result. All the forces were transferred to the centre of gravity of the
system as forces and moments. Input parameters used for validation are:column diameter of 10.7m, pontoon
diameter of 1.8m, pontoon length of 51.2m, heave plate diameter of 23.7m, column draft of 22.9m, wave height
of 13.5m, time period of 17s, mass of floater of 69629kN, mass of turbine and tower of 8369kN and tether
diameter of 0.15m.

193
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Validation
To validate the computer program developed, experimental study on similar platform conducted by Dominique
Roddier, Christian Cermelli, Alexia Aubault and Alla Weinstein was used [Dominique Roddier e. al. (2010)].
Input to the program includes the floater dimensions, waveparameters, wind thrust force and statics of tether
lines. The comparison of maximum, minimum and root-mean-square values of the responses are shown in Table
5.1.

Table 5.1 Comparison of results

Degree of Root-mean-
Maximum Minimum
freedom square values
Present work(Analytical) 13.32 -6.95 3.40
Surge (m) As per Dominique Roddier e. al.
13.26 -6.75 3.21
(2010)(Experimental study)
% Difference 0.45 2.96 5.91
Present work(Analytical) 4.52 -5.89 1.95
Heave (m) As per Dominique Roddier e. al.
4.86 -5.84 1.99
(2010)(Experimental study)
% Difference 6.99 0.85 2.01
Present work(Analytical) 5.07 -3.19 1.36
Pitch
As per refDominiqueRoddier e. al.
(degrees) 4.87 -3.87 1.27
(2010)(Experimental study)
% Difference 4.10 17.57 7.08

Table 5.2Responsesfor operational and rough sea conditions

Operational sea Rough sea
RMS RMS RMS RMS
Degree of freedom Maximum Minimum (transient+st (steady Maximum Minimum (transient+st (steady
eady) state) eady) state)
Surge (m) 1.1739 -0.8571 0.2688 0.1964 5.3509 -5.6929 1.3146 0.915
Heave (m) 0.6391 -0.808 0.2353 0.2268 3.2543 -3.847 1.2604 1.2243
Pitch (degrees) 0.2967 -0.0157 0.1613 0.1611 1.0436 -0.4777 0.3216 0.3070

Dynamic analysis was carried out separately for both operational and rough sea conditions. Only the responses
in surge, heave and pitch directions are predominant and responses in other directions are negligible. Table 5.2
gives the maximum, minimum and root-mean-square (RMS) values of surge, heave and pitch motions of floater
for operational sea conditions and rough sea conditions. Due to large wave force acting on the structure in rough
sea conditions, the floater responses in surge, heave and pitch have larger variation to that in operational sea
conditions. For rough sea conditions, the maximum response value of surge, pitch and heave increased by 356%,
409% and 252% respectively when compared with operational sea conditions. The maximum of heave response
in rough sea is only 3.2543m which indicates that sufficient free board will be available. For both sea conditions,
surge response is less on comparison with the overall dimensions of the Tri-floater indicating its high stability.

Response for Various Inclinations of Wave

Surge, heave and pitch responses of Tri-floater were computed for 00, 300, 450 and 600inclinations of wave with
x-axis in operational sea conditions are given in Table 5.3.

Table 5.3 Maximum response values for 00, 300, 450 and 600 inclinations of wave
Maximum Response Values
DOF
0o wave 30o wave 45o wave 60o wave
Surge (m) 1.1739 1.0008 0.8343 0.6199
Heave (m) 0.6391 0.6327 0.5765 0.4842
Pitch (deg) 0.2967 0.2663 0.2383 0.2030

Surge, heave and pitch degrees of freedom have maximum response values for 00 inclination of wave. The values
are found to be reducing as the inclination of wave varies from 0 0 to 600. The percentage reduction of maximum

194
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

response value of surge from 0 0 to 600 is high. A decrease of 15% is found in the maximum response value of
surge as the inclination of wave in operational sea conditions changes from 0 0 to 300. Between 00 and 450,
maximum response value decreases by 29% and between 0 0 and 600, it decreases by 47%. The percentage
reduction of maximum response values of heave is less compared to that of surge as the inclination of wave
changes from 00 to 600. Between 00 and 300, a decrease of 1% is observed which decreases up-to 24% for values
between 00 and 600. Hence, the percentage reduction of maximum response values for surge is twice as that for
heave between 0 0 and 600 inclinations of wave. Pitch response also shows considerable percentage reduction in
maximum response values for a change of wave inclination from 0 0 to 600. For pitch response, the percentage
reduction of maximum response value lies between the variations observed for heave and surge.

6. CONCLUSION
The behavior of Tri-floater for waves hitting from different incident directions were studied. The effect of wave
is more significant compared to wind in the platform heave motion of Tri-floater. For rough sea conditions, the
pitch response attains a maximum value of 1.040, which is less than the limiting value of 100, ensuring turbine
efficiency even under extreme conditions. With a change of sea condition from operational to rough, the
maximum response value of surge, heave and pitch increases by 356%, 409% and 252% respectively. For the
Tri-floater, the maximum response values of surge, heave and pitch responses decreases as the inclination of
wave changes from 00 to 600. The percentage reduction of maximum response values is more for surge and less
for heave. The percentage reduction of maximum response values of surge is two times as that for heave between
00 and 600 inclinations of wave.

REFERENCES
1. Bulder, B. H., Henderson A., Huijsmans and Peeringa J.M., “Floating offshore wind turbines for Shallow
waters”,Proceedings of the International Conference on Ocean, Offshore and Arctic Engineering, 2002.
2. Dominique Roddier, Christian Cermelli, Alexia Aubault, Alla Weinstein, “WindFloat: A floating foundation for offshore
wind turbines”, Journal of renewable and sustainable energy 2, 2010.
3. HabtamuBeri, Yingxue Yao “Double Multiple Stream Tube Model and Numerical Analysis of Vertical Axis Wind
Turbine”, Journal of Energy and Power Engineering, 262-270, 2011.
4. Hooft J. P., “Hydrodynamic Aspects of Semisubmersible Platform”, Publication No. 400 Netherlands ship Model Basin,
1972.
5. Michael Borg, Esteban UtreraOrtigado, Maurizio Collu, Feargal P. Brennan “Passive Damping Systems for Floating
Vertical Axis Wind Turbines Analysis”, 2013.
6. Michael Borg, Maurizio Collu, Andrew Shires.“Offshore Floating Vertical Axis Wind Turbines, Dynamics Modelling
State of the Art Part I: Aerodynamics”,2014.
7. Ruoyu Zhang, Yougang Tang, JunHua, ShengfuRuan, ChaoheChen “Dynamic Response in Frequency and Time
Domains of a Floating Foundation for Offshore Wind Turbines”, Ocean Engineering,Vol 60, 2013.
8. Tong K.C., “Technical and Economic Aspects of a Floating Offshore Wind Farm”, Journal of Wind
EngineeringandIndustrialAerodynamics, 74-76, 1998.

195
 AN APPLICATION OF A WIRELESS POWER TRANSFER AT
LOW FREQUENCY RANGE

Erol Kurt
Department of Electrical and Electronics Engineering, Technology Faculty, Gazi University, Ankara, Turkey, E-mail:
[email protected], ORCID: 0000-0002-3615-6926

Fatemeh Zafarmand
Branch of Electrical and Electronics Engineering, Institute of Natural Sciences, Gazi University, Ankara, Turkey, E-
mail: [email protected], ORCID: 0000-0003-2215-8207

Hikmet Fidanboy
Department of Electrics, Technical Sciences Vocational School, Gazi University, Ankara, Turkey, E-mail:
[email protected], ORCID: 0000-0003-1143-5678

Şevki Demirbaş
Department of Electrical and Electronics Engineering, Technology Faculty, Gazi University, Ankara,Turkey, Email:
[email protected], ORCID: 0000-0001-9111-684X

Kurt, E., Zafarmand, F., Fidanboy, H. Demirbas, S., An application of a wireless power
Cite this paper as: transfer at low frequency range. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,
Turkey

Abstract: In this study, a low frequency wireless power transfer (LFWPT) device has been designed and
implemented. The aim of investigation of low frequency system is to eliminate the un-healthy
situation due to non-ionized radiation of the high frequency regimes. The transmitter uses a special
winding to transmit the electromagnetic signals to the environment, especially for a few
centimeters distance. The receiver winding, which is the same with the transmitter winding catches
the signal and stores it in capacitor. The system uses maximum 17649 Hz to transmit the signal,
that is fairly low compared to the other studies in the literature. A DC-AC inverter is used to obtain
high frequency signals for the transmitter circuit. The experiments have proven that the receiving
antenna can get 1 V amplitude from the transmitted peak-to-peak amplitude of 140V at the
operation frequency.
Keywords: wireless power transfer, power, IGBT, low frequency
© 2020 Published by ECRES

1. INTRODUCTION

Electrical form of energy has a wide application area in our daily-life such as transportation, medicine,
agriculture, communication, industry, illumination. Almost all devices use the electricity, thereby the electrical
devices require a power transmission and that transmission is mainly made from the network connection via a
cable [1-3]. Due to Joule heating and damages on the cables, there may be accidents in the houses and industrıes
in addition to the cable costs. Therefore, to eliminate these artifacts, there exist a continuous trend to transmit
the electrical power without a cable or any other conducting material between the network connection point and
device. This goal has caused many attempts to design and implement different wireless power transmitting
systems, which are mainly made for frequencies higher than 100 KHz, however such a high frequency regime
for the transmission causes a non-ionizing electromagnetic pollution [4]. That is especially a problem for health.
According to the international standards, there exist a number of available regulations on the thresholds of the
non-ionizing radiation [5]. To fulfill these regulations, the main task is to make energy transmission at low
frequency band. The present work mainly focuses on this idea to test the achievements of low frequency
transmission [6,7].

According to the literature, there exist many applications on wireless transfer systems [8-11]. The main idea has
become the transmission of electricity in near distance economically and efficiently. Strictly speaking, the

196
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

transmission is provided between a transmitter and receiver by using a resonance frequency. For instance, a
three phase ac voltage can be converted to a dc voltage via a rectifier first, then this dc voltage can be converted
to a higher frequency ac voltage and sent to an antenna. The receiver antenna would get the signal at the identical
resonant frequency. Afterwards, it can be stored into a capacitor or a battery via a dc conversion.

Historically, Tesla invented some different wireless power transmission components at the beginning of 1900
[12]. After one century, the wireless power transmission gets much interest due to the raising importance of
electricity in all sectors from mobile to electric vehicles [13]. For instance, wireless chargers are nothing else
than the application of wireless power transfer. Indeed, the transfer is possible for short distances as for mobile
phones [14-17].

Other issue related to the power transfer is the antenna structure. The antennas which operate well for
information, cannot be used for the energy transfer. Because their modulation is capable to send/receive the
information, indeed the large amount of energy is lost in the air media. Apart from such antenna structures,
multi-directional antennas with directed radiation modes can be used for the wireless transfer, efficiently.

In many recent applications such as laptops, mobile phones, etc., very close distances with low powers (i.e.
mW) are encountered [18,19]. Besides, a more efficient, radiation- free and medium distance wireless solutions
are focused with long life oscillatory electromagnetic modes.

The methodology of power transfer is related to the phenomena of resonant coupled circuits principle. This
principle indeed provides a connection between two circuits, which are not far away from each other. For medium
and long distances, the nature of these couplings has not been fully understood and that is an on-going work.
For instance, the works in MIT proved that the power can be transferred with this principle in 2007. The power
P=60 W was transferred in a wireless manner to 2 meters away with 60% efficiency [14- 20]. However, the
efficiency still stays as main problem. If this principle can be applied for larger scales in power and distance,
there will be so many advantages such as;

1) High voltage transmission lines will be removed.

2) The connection between the electrical plants will be easier globally.
3) High expenditures on transmission lines and related components will be removed.
4) Freedom of preferences for transmitting institution and clients will be obtained.
5) The electricity cost for the clients will decrease.
6) The power will be delivered via air media to the harsh environmental locations easily.
7) There will be no unplanned electricity cuttings and short circuits caused by transmission lines and transformers.
8) The forest fires caused by transmission lines will be prevented.
9) The power cable accidents involved by the workers and technicians will be prevented.

In the present work, initially the wireless power transmission systems are briefly introduced. Then, a theoretical
study with Matlab application will be presented in Sec. 3. The novel part of the work is to use low frequency
circuit elements to generate and recover the power signal to our knowledge. The experiments results and main
findings are given in Section 4. Finally, the paper ends with a conclusions section.

2. WIRELESS POWER TRANSMISSION

In recent years, wireless power transmission systems have become so important for human beings, especially for
charging the electrical vehicles and mobile phones. For electrical vehicles, wireless power transmission instead
of charging with a power cable, has become more secure for drivers and passengers since most of the batteries
are charged in a short time with high amounts of charge capacity. Since the inductive wireless power transmission
is mostly suitable for such a charging application, many of the literature studies use the inductive systems as in
Refs. [21,22]. In the inductive unit, there exist two types of antennas: Circular coil system with a centered core
and a toroidal coil system.
Another important issue is to decrease the charging time, since the customers prefer the short charging
durations in the products. However, the smaller coils can cause some thermal problems. therefore, serial
connected resonant capacitors are preferred in many circuits. In Ref. [23], a 7 kW wireless system was explored

197
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

and the effects of coil geometries, core losses and connection coefficients were examined. The efficiency was
also found for these geometries. Kudo and coworkers [24] suggested a more efficient system with lower voltage
scales and higher electrical charge amounts, thereby a double layer capacitor system with low voltage and high
currents which was considered for low electrical load.

In the literature, changing the full-bridge rectifier with a current doubler rectifier is not an appropriate selection,
because the inductors used in the current doubler rectifier are physically bigger and heavier than the diodes of
full-bridge rectifiers. Apart from that, current doubler rectifiers can be preferred especially for high power systems
due to the fact that the secondary coils are heavier than the ones in the rectifier. In another work, Chigira, et. al.,
[25] practiced a current doubler rectifier system and found that the solenoidal coils exhibit better performance
than the circular ones in the directional power transmission. They also offered a new H-shaped coil system and
tested that in their work.

3. SIMULATIONS

The simulations are entirely performed in MatLab package program. In the simulations, a dc-ac converter and
controller units have been designed for the energy transmitter. An appropriate PWM signal is applied into the
converter via a STM32F407VG Controller. The block diagram of overall system simulated in MatLab Simulink
is given in Fig.1. Components used in simulation program are presented in Table 1. Fig. 2(a) gives a sample of
pulse waveforms with a periodic output. The antenna’s transmitter and receiver voltage waveforms are given by
Fig. 2(b). As it is obvious from the signals, they show ideal sinusoidal signal characteristic with low THD values.
Load current and voltage waveforms are shown in Fig. 3 (a,b).
Table 1. Components used in simulation
Item no Item Description
1 Transmitter Inductor 0.7 mH
2 Receiver Inductor 0.7 mH
3 Capacitor 900 μF
4 Resonance frequency (Hz) 17649

At the terminals of receiver unit, an electrical load has been connected (Fig. 1). The current is found as 11 mA
(Fig. 3(a)). In the case of voltage, amplitude is observed as 0.11 V (Fig. 3(b)). It is obvious from Fig. 3 that the
output over the load is ideal sinusoidal, too.

Figure. 1. System simulated in MatLab Simulink

198
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)
Figure. 2. (a) PWM signals of single phase full bridge inverter, (b) transmitter and receiver antenna voltage waveforms

(a) (b)
Figure 3. (a) Load current and (b) voltage waveforms

The maximum THD values of the voltages from the transmitter and receiver antennas are given in Fig. 4(a,b).
It is clear that the THD values differ for the transmitting and receiving units. Strictly speaking, the lowest THD
value is obtained at the receiver antenna unit with 0.87%. However, the THD value increases for the transmitting
antenna unit due to the switching process of IGBTs and transmitting unit filter. The THD value of the transmitter
antenna voltage is measured as 3.59%, which is appropriate for such low frequency wireless power transfer
applications. Note also that both THD plots indicate the superharmonics of the fundamental frequency of 17.649
kHz. For the transmitting antenna, almost four superharmonics appear, however, in the case of receiver antenna,
only two superharmonics exist due to the low THD value.

(a) (b)
Figure 4. The THD values of (a) transmitter antenna voltage and (b) receiver antenna voltage.

4. EXPERIMENTAL RESULTS AND DISCUSSION

The experimental setup of the work is given in Fig. 5(a). The upper left corner is the controller unit and the
lower left corner shows the IGBT units for low frequency excitation. Lower right corner is the feeding
transformers of controllers. The energy storage is shown in upper right corner with capacitors.

In the experimental work, the diodes DSEI 12-20 are used. 2MBI100U-4A IGBTs are used for the resonance
frequency adjustment and dc-ac conversion in converter. The output signal of the converter has been directly

199
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

applied into the antenna. While Fig. 5(b) gives the larger form of controller unit, Figs 5(c,d) show the antenna
types, which are used in this work. Components used in experiments are given in Table 2.

(a) (b)

(c)
Figure 5. (a) The overall experimental setup, (b) Spiral type transmitter and receiver antennas (Antenna A), (c) Toroidal
transmitter and receiver antennas (Antenna B)

Table 2. Experımental Components

Item no Item Description
1 Diodes DSEI 12-20
2 IGBT Switches 2MBI100U-4A
3 Transmitter Inductor 0.7 mH
4 Receiver Inductor 0.7 mH
5 Capacitors 0.2 µF
6 Resonance frequency 13.33 kHz

During the operation of the power transmitter, received output signal of IGBTs for the operation of circuit are
presented in Fig. 6.

Fig. 6. IGBTs output signals.

200
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig. 7(a) gives the signal from the transmitter circuit antenna A. The signal is highly affected by the radio
frequencies at the surrounding media. Note that it is in the order of 10 Volts, therefore, it can be said that the
spiral type antennas are not appropriate for such low frequency applications (i.e. 13.3 kHz). However, Antenna
B (i.e. toroidal antenna) gives more clear results as shown in Fig. 7(b). with a 2 cm distance, the signal from the
transmitting circuit decreases to 1.2 V in amplitude from the value of 140 V.

(a) (b)
Fig. 7. (a) Transmitted signal from the Antenna A, (b) Transmitted and received signals from the Antenna B with 2 cm
distance

5. CONCLUSIONS

A wireless power transfer system has been designed and implemented in this paper. The system operates at
17649 Hz with a low frequency compared to the other microwave frequency wireless power transferring systems
theoretically. Besides, the experimental setup enables t otransmit the power at lower frequency such as 13330.
Those frequency values are in radio-frequency limits and produce lower non-ionizing radiation compared to the
microwave frequencies. Two types of antennas have been tested and found that spiral type antenna produces
highly harmonic output and it is not appropriate for such purposes. Meanwhile, toroidal type antennas have
exhibited a better result, that is one of the innovative findings of the present work. Since the frequency is low,
the signal decays tremendously at the receiving circuit. Strictly speaking, 1.2 V is obtained in the receiver circuit
while 140 V is transmitted. Thus high amount of signal is lost in the media.

REFERENCES

[1] Mahariq, I., Berjozkina, S., “Experimental realization of electromigration at high power for copper wires” J. Energy
Syst., 3(4), 158-167, 2019, DOI: 10.30521/jes. 616982
[2] Dursun, B., Kurt, E., Tekerek, M., “A Power circuit design for the poloidal field coils in a torus shaped plasma
system, J. Energy Syst., 3(3), 123-128, 2019, DOI: 10.30521/jes.609667
[3] Feng, W., et al., “Joint 3D Trajectory Design and Time Allocation for UAV-Enabled Wireless Power Transfer
Networks,” in IEEE Trans. on Vehicular Tech., 2020, doi: 10.1109/TVT.2020.2972133
[4] Frah, M., Belyaev, V. (2019). Parameters of electromagnetic pollution from different sources and their hazard impact.
J. Physics: Conference Series. 1309, 012013, doi:10.1088/1742-6596/1309/1/012013.
[5] Syaza, S.K.F., Umar, R., Sabri, N., Kamarudin, M.K.A., Hassan, A., Juahir, H. “Non-ionizing radiation as threat in
daily life,” J. Fund. and Applied Sci., 2018, doi:9.308. 10.4314/jfas.v9i2s.21I.
[6] Mohammed, S.S., Ramasamy, K., Shanmuganantham, T., “Wireless Power Transmission - A Next Generation Power
Transmission System,” Int. J. Computer Appl., 2010, doi:1. 10.5120/274-434.
[7] Inoue, R., Miyagi, D., Tsuda, M., Matsuki, H., "High-Efficiency Transmission of a Wireless Power Transmission
System for Low- Frequency Using REBCO Double-Pancake Coils," in IEEE Trans. Appl. Superc., 27(1), pp. 1-6,
Jan. 2017, Art no. 5400106, doi: 10.1109/TASC.2016.2629263
[8] Shin, S., et al., “Wireless power transfer system for high power application and a method of segmentation,” IEEE
Wireless Power Transfer (WPT), Perugia, 2013, pp. 76-78, 2013.
[9] Minnaert, B., Stevens, N., “Maximizing the power transfer for a mixed inductive and capacitive wireless power
transfer system,” IEEE Wireless Power Transfer Conference (WPTC), Montreal, QC, Canada, pp. 1-4, 2018.

201
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[10] Bechtold, T. , Acevedo, J.A.P., Hohlfeld, D., “Compact model of a wireless power transfer system,”, IEEE Wireless
Power Transfer Conference (WPTC), Taipei, pp. 1-3, 2017
[11] Tan, L., Pan, S., Huang, X., Xu, C., “System optimization for wireless power transfer system with double
transmitters,” IEEE PELS Workshop on Emerging Technologies: Wireless Power Transfer (WoW), Chongqing,
2017, pp. 1-3.
[12] Valone, T., “Geoengineering Tesla's Wireless Power Transmission”, ExtraOrdinary Sci. Tech., pp.31-43, 2017.
[13] Brown, S., Pyke, D., Steenhof, P., “Electric vehicles: The role and importance of standards in an emerging market”,
Energy Policy, 38, pp.3797-3806, 2010, doi:10.1016/j.enpol.2010.02.059.
[14] Zhu, C., Liu, K., Yu, C., Ma, R., Cheng, H., “Simulation and Experimental Analysis on Wireless Energy Transfer on
Magnetic Resonances”, IEEE Vehicle Power and Propulsion Conference, Harbin, China, 3-5 Sep 2008.
[15] Tesla, N., “Apparatus for transmitting electrical energy”, United States Patent Office, US Patent No. 1119732, USA,
1914.
[16] Mahmood, A., Ismail, A., Zaman, Z., Fakhar, H., Najam, Z., Hasan, M.S., Ahmed, S.H., “A Comparative Study of
Wireless Power Transmission Techniques,” J. Basic and Appl. Scien. Res., 4(1), pp.321-326, 2014.
[17] Liu, X., Zhang, F., Hackworth, S.A., Sclabassi, R.J., Sun, M., “Modeling and Simulation of a Thin Film Power
Transfer Cell for Medical Devices and Implants”, IEEE International Symposium on Circuits and Systems (ISCAS),
Taipei, 2009, pp. 3086-3089. doi: 10.1109/ISCAS.2009.5118455
[18] Fernandez, J.M., Borras, J.A., “Contactless battery charger with wireless control link,” US Patent No. 6184651,
USA, 2001.
[19] Scheible, G., Smailus, B., Klaus, M., Garrels, K., Heinemann, L., “System for wirelessly supplying a large number
of actuators of a machine with electrical power”, US Patent No. 6597076, USA, July 2003.
[20] Moffatt, R.A., “Wireless Transfer of Electric Power,” Massachusetts Institute of Technology, 2009.
[21] Diekhans T., Doncker, R.W., “A dual-side controlled inductive power transfer system optimized for large coupling
factor variations,” ECCE2014, Sep., 2014, pp. 652-659.
[22] Nagatsuka, Y., Ehara, N., Kaneko, Y., Abe, S., Yasuda, T., “Compact contactless power transfer system for electric
vehicle,” IPEC2010, Jun. 2010, pp. 807 - 813.
[23] Shijo, T., Ogawa, K., Obayashi, S., “Optimization of thickness and shape of core block in resonator for 7 kW-class
wireless power transfer system for PHEV/EV charging,” ECCE2015, Sep. 2015, pp.3099- 3102.
[24] Kudo, T., Toi, T., Kaneko, Y., Abe, S. “Contactless power transfer system suitable for low voltage and large current
charging for EDLCs,” IPEC-Hiroshima May 2014, pp.1109-1114.
[25] Chigira, M., Nagatsuka, Y., Kaneko, Y., Abe, S., Yasuda, T., Suzuki, A., “Small-size light-weight transformer with
new core structure for contactless electric vehicle power transfer system,” ECCE2011, Sep. 2011, pp.260-266.

202
 IMPLEMENTATION OF 30KHZ SINUSOIDAL CURRENT
GENERATOR BY USING GAN-FET

Liang-Rui Chen
Dept. of Electrical Engineering, National Changhua University of Education, Changhua, Taiwan, ROC,
[email protected], ORCID: 0000-0002-5633-964X

Zi-Yu Lin
Dept. of Electrical Engineering, National Changhua University of Education, Changhua, Taiwan, ROC,
[email protected], ORCID: 0000-0002-4199-2208

Ying-Lin Chen
Dept. of Electrical Engineering, National Changhua University of Education, Changhua, Taiwan, ROC,
[email protected], ORCID: 0000-0002-2161-922X

Chen, L.R., Lin, Z.Y., and Chen, Y.L. Implementation of 30KHz Sinusoidal Current
Cite this paper as: Generator by Using GaN-FET. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,
Turkey

Abstract: In this paper, a 30kHz sinusoidal current generator (SCG) for battery electrochemistry impedance
spectroscopy (EIS) measurement is designed and implemented. A three-phase interleaved
synchronous rectified buck converter topology is used in the proposed 30KHz SCG. In addition,
the enhanced gallium nitride field effect transistors (GaN-FET) are used as switches in the
proposed architecture, which are capable to operate at switching frequency up to 1 MHz and
downsize the circuit volume. A 16-bit microcontroller generates the high resolution sinusoidal
pulse width modulation (SPWM) and it contributes a wide bandwidth between 30Hz and 30kHz
to the proposed SCG. Compared to conventional SCG with MOSFET, the proposed SCG with
GaN-FET has higher sinusoidal current, lower switching loss, lower conduction loss, and higher
efficiency. The three-phase interleaved synchronous rectified buck converter topology can
decrease the inductance and the capacitance, reduce the total harmonic distortion (THD), and
increase the cut-off frequency to reach wide bandwidth under the same inductor current ripple.
Keywords: Gallium nitride(GaN), Three-phase interleaved Converter, Sinusoidal Pulse Width
Modulation(SPWM)
© 2020 Published by ECRES
Nomenclature
EIS Electrochemistry Impedance Spectroscopy
THD Total Harmonic Distortion
GaN-FET Gallium Nitride Field Effect Transistor
SiC-FET Silicon Carbide Field Effect Transistor
SOC State of Charge
SOH State of Health
SPWM Sinusoidal Pulse Width Modulation
SCG Sinusoidal Current Generator

1. INTRODUCTION

Battery energy storage systems have a place in technology nowadays. In order to correctly and effectively charge
the battery, the battery state of charge (SOC) and state of health (SOH) have become important references for
quickly and safely charging the battery [1-3]. Recent years, the development of battery electrochemical
impedance spectroscopy (EIS) is also one of the reference indicators of battery design systems. Without
damaging the battery, the internal state of the electrochemical reaction of the battery can be detected through
EIS and the battery's charging parameters can be confirmed [4-8], as shown in Figure 1. Then we can determine
the charging parameters of the battery. Traditional EIS measurement requires offline measurement. The main
disadvantage is that the battery and the circuit must be separated only for EIS detecting. Therefore, it is necessary

203
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

to develop on-line measurement [9-10]. EIS measurement process is to provide various frequencies sine wave
signals to the battery, and measure the internal impedance of it. Therefore, the development of online EIS
measurement requires the manufacture of sinusoidal current generator (SCG). The enhanced gallium nitride
field effect transistor (GaN-FET) is used as the switch in the proposed architecture to increase the switching
frequency and achieve low THD through a three-phase interleaved synchronous rectification buck converter.

(a) (b)
Figure 1. Battery EIS diagram: (a) Bode diagram and (b) –Nyquist diagram.

2. ADVANTAGES OF WIDE BAND GAP SEMICONDUCTOR COMPONENTS

In order to output high-frequency signals of the proposed architecture, the output bandwidth and the inductor
current slope should be large and wide enough. The inductor value design must be small enough so that the
inductor current slope is large enough. Reducing the inductance will increase the output current ripple.
Therefore, it’s necessary to increase the switching frequency to maintain a small current ripple, which also means
that the switch must seek a faster switching speed. The performance of MOSFET devices at high frequencies
(>500kHz) will be too slow to switching(dv/dt), resulting the switching losses too high to be applied. Wide-
bandgap devices such as gallium nitride or silicon carbide overcome this shortcoming. It can be used in high
power, high frequency and high temperature environment. In addition, converters using wide-bandgap (WBG)
semiconductors can reduce the size and weight of their circuits by more than 70%, while increasing power
density to ensure the feasibility of the SCG proposed in this paper.

3. SYSTEM DESCRIPTION

Figure 2(a) is the system architecture diagram of this paper, including power stage and control stage. The power
stage consists DC power supply, switch, inductor, capacitor and load. The control stage contains MCU and
measurement circuit. The measurement circuit feeds back the current signal to the MCU and compares it with
the sine wave command. After calculation, the SPWM control switch is modulated.

(a) (b)
Figure 2. (a) System architecture diagram, (b) Proposed circuit diagram.

204
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)
Figure 3. (a) Conventional buck topology, (b) Synchronous rectified buck topology

Figure 2(b) is the circuit diagram of the three-phase interleaved synchronous rectification buck converter in this
paper. Using a three-phase interleaved sine wave converter can effectively reduce the current ripple and increase
the output bandwidth. To reduce the circuit loss and recovery time of the diode in the traditional buck converter
structure in Figure 3(a), the proposed architecture changes the diode to a switch, as shown in Figure 3(b).
Choosing GaN-FET as the switch for high switching frequency [11-15]. The following equations are the design
of capacitance and inductance in the buck converter:

𝑉
𝑂
𝐿 = ∆𝑖𝐿 ∙ (1 − 𝐷)𝑇 (1)

𝑇 2 ∙𝑉𝑂 (1−𝐷)
𝐶𝑂 = 8∙∆𝑉𝑂 ∙𝐿
(2)

In Eqs. [1,2], D is defined as the duty cycle of switch S1, and T is defined as the cycle of the switch.
According to Eqs. [1,2]:
1
L ∝ T = 𝑓 (3)

1
C ∝ T = 𝑓 (4)

Therefore, choosing higher switching frequency f can reduce capacitance and inductance. The capacitor and
inductor in the buck converter circuit form an LC low-pass filter, and the cutoff frequency calculation of the
filter is as follows:
1
f = 2𝜋√𝐿𝐶 (5)

In order to obtain a high-frequency sinusoidal signal at the output, the designed capacitance and inductance must
also comply with Eq. [5]. As the inductance decreases, more current ripple will appear. Figure 4. shows the
waveforms of the proposed circuit in this paper. For reducing the current ripple, the PWM phase shifted 120°
per phase, which can reduce higher harmonics and earn lower total harmonic distortion output sine wave.

Figure 4. Key waveforms of the proposed circuit.

205
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. DESIGN EXAMPLE
Table 1. Converter electrical specifications
Item Description Value
Input voltage (Vin) 12V
Output voltage (Vo) 0~4V
Output current (Io) 0~2.5A
Input capacitance (Cin) 20uF
Output inductance (L1, L2, L3) 4.7uH
Output capacitance (C1, C2, C3) 4.7uF
Switching frequency (fsw) 1MHz
Sine wave command frequency (fref) 30~30kHz

Table 1 is the electrical specifications of the proposed converter. Due to the high switching frequency, the
interference caused by parasitic element in the circuit cannot be ignored. Compared with traditional wire
wound inductors, the use of SMD shielded molded inductors greatly reduces the effect of parasitic
capacitance. The self-resonant frequency(SRF) is high and it is suitable for high switching frequency
application. The traditional output filter capacitor is a constant voltage output, and electrolytic capacitors are
usually used to filter out high-frequency components. However, this article designs and outputs high-
frequency sine wave signals, using chip multi-layer ceramic capacitors, small size, high temperature
resistance, so that the converter performs better in the high frequency range. The EPC9203 module with GaN
switch is selected. The high-speed switching feature can support the converter to work at 1MHz. With the half-
bridge gate driver LM5113, the converter performance can be maintained. In order to make the output PWM
have higher resolution, this article uses the 16-bit controller dsPIC33CK256MP508, which has 8 pairs of high-
precision and high-speed PWM modules, which can further reduce the total harmonic distortion of the output
sine wave. Using Powersim(PSIM) and Figure 2(b), the SCG architecture of this paper can be simulated to
confirm the above analysis, as shown in Figure 5.

(a) (b)
Figure 5. PSIM simulation results for the SCG. (a) Vref, Vgs1, IL1-3, Vo, and Io, (b) Zoomed-in waveforms of (a).

5. EXPERIMENTAL RESULTS

To verify the architecture proposed in this paper, a wide-band current can be output, and sine wave commands
at different frequencies can be used to modulate SPWM. The sine wave commands and driver signals are shown
in Figure 6(a).

(a) (b)
Figure 6. Experimental waveforms of (a) SPWM and reference signal, (b) Vds of the GaN-FETs.

206
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

This article selects GaN-FET as the switch, as shown in Figure 6. (b). It can be seen that the upper bridge switches
in the three groups of bucks can operate normally at 1MHz and maintain a phase difference of 120°. Figure 7.
shows the three-phase inductor current waveform. Compared with the non-interleaved architecture, current
stress can be reduced, and high-frequency inductors are used to minimize the influence of the resonant circuit
on the inductor current waveform.

(a) (b)
Figure 7. Experimental waveforms of (a) inductor currents, (b) zoomed-in waveforms of (a).

Figure 8. (a) shows the output waveforms of different frequencies of the non-interleaved circuit architecture.
Although a wide-band sine wave can be modulated, the influence of higher harmonics still exists. Figure 8. (b)
The three-phase interleaved SCG proposed in this paper can clearly see that the noise caused by the switch is
significantly reduced, which has a significant effect on reducing THD.

(a)

(b)
Figure 8. Experimental waveforms of the (a) single-phase, (b) three-phase interleaved SCG.

207
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

6. CONCLUSIONS

This paper proposed a 30kHz sinusoidal current generator (SCG) for battery electrochemistry impedance
spectroscopy (EIS) measurement is designed and implemented. Using GaN-FET as circuit switches, which can
apply at high switching frequency. Due to the high switching frequency, the inductance and capacitance of the
circuit can be reduced, and use SMD packaging to decrease loss and noise and downsize the circuit. The SPWM
to the GaN-FET and the three-phase interleave architecture can obtain a wide bandwidth of 30 Hz to 30 kHz
and low THD.

REFERENCES

[1] J. Jiang, C. Zhang, J. Wen, W. Zhang, S.M. Sharkh. An optimal charging method for li-ion batteries using a fuzzy-
control approach based on polarization properties. IEEE Trans. Veh. Technol. 2013, vol. 62, no. 7, pp. 3000-3009.
< https://doi.org/10.1109/TVT.2013.2252214>
[2] T.T. Vo, X. Chen, W. Shen, A. Kapoor. New charging strategy for lithium-ion batteries based on the integration of
Taguchi method and state of charge estimation. J.Power Sources. 2015, vol. 273, pp. 413-422.
< https://doi.org/10.1016/j.jpowsour.2014.09.108 >
[3] C. H. Lee, M. Y. Chen, S. H. Hsu, J. A. Jiang. Implementation of an SOC-based four-stage constant current charger for
Li-ion batteries. Journal of Energy Storage 2018, vol. 18, pp. 528-537. < https://doi.org/10.1016/j.est.2018.06.010 >
[4] A. Bartlett, J. Marcicki, S. Onori, G. Rizzoni, X. G. Yang, T. Miller. Electrochemical model-based state of charge and
capacity estimation for a composite electrode lithium-ion battery. IEEE Trans. Control Syst. Technol. 2016, vol. 24,
no. 2, pp. 384-399. < https://doi.org/10.1109/TCST.2015.2446947 >
[5] K. Liu, C. Zou, K. Li, T. Wik. Charging pattern optimization for lithium-ion batteries with an electrothermal-aging
model. IEEE Trans. Ind. Informat. 2018, vol. 14, no. 12, pp. 5463-5474. < https://doi.org/10.1109/TII.2018.2866493 >
[6] J. Chen, Q. Ouyang, C. Xu, H. Su. Neural network-based state of charge observer design for lithium-ion batteries. IEEE
Trans. Control Syst. Technol. 2018, vol. 26, no. 1, pp. 313-320. < https://doi.org/10.1109/TCST.2017.2664726 >
[7] Q. Kun Wang, Y. J. He, J. N. Shen, X. S. Hu, Z. F. Ma. State of charge-dependent polynomial equivalent circuit
modeling for electrochemical impedance spectroscopy of lithium-ion batteries. IEEE Trans. Power Electron. 2018, vol.
33, no. 10, pp. 8449-8460. < https://doi.org/10.1109/TPEL.2017.2780184 >
[8] C. Zou, X. Hu, Z. Wei, T. Wik, B. Egardt. Electrochemical estimation and control for lithium-ion battery health-aware
fast charging. IEEE Trans. Ind. Electron. 2018, vol. 65, no. 8, pp. 6635-6645. <
https://doi.org/10.1109/TIE.2017.2772154 >
[9] R. G. Hoffmann, J. E. Slade, J. L. Morrison. Development and test of a real time battery impedance estimation system.
in Proc. IEEE Aerosp. Conf. 2006, pp. 1-8. < https://doi.org/10.1109/AERO.2006.1656091 >
[10] P. A. Lindahl, M. A. Cornachione, S. R. Shaw. A timedomain least squares approach to electrochemical impedance
spectroscopy. IEEE Trans. Instrum. Meas. 2012, vol. 61, no. 12, pp. 3303-3311. <
https://doi.org/10.1109/TIM.2012.2210457 >
[11] B. Ozpineci, L. M. Tolbert, S. K. Islam, M. Chinthavali. Comparison of Wide Bandgap Semiconductors for Power
Applications. In: European Conference on Power Electronics and Applications, 2003.
[12] H. A. Mantooth, M. D. Glover, P. Shepherd. Wide Bandgap Technologies and Their Implications on Miniaturizing
Power Electronic Systems. IEEE Journal of Emerging and Selected Topics in Power Electronics. 2014, vol. 2, no. 3,
pp. 374-385. < https://doi.org/10.1109/JESTPE.2014.2313511 >
[13] A. Knott, T. M. Andersen, P. Kamby, J. A. Pedersen, M. P. Madsen, M. Kovacevic, and M. A. E. Andersen. Evolution
of Very High Frequency Power Supplies. IEEE Journal of Emerging and Selected Topics in Power Electronics. 2014,
vol. 2, no. 3, pp. 386-394. < https://doi.org/10.1109/JESTPE.2013.2294798 >
[14] X. Huang, Z. Liu, Q. Li, F. C. Lee. Evaluation and Application of 600 V GaN HEMT in Cascode Structure. IEEE
Trans. Power Electronics. 2014, vol. 29, no. 5, pp. 2453-2461. < https://doi.org/10.1109/TPEL.2013.2276127 >
[15] W. Huang and J. A. Abu Qahouq. An online battery impedance measurement method using DC-DC power converter
control. IEEE Trans. Ind. Electron. 2014, vol. 61, no. 11, pp. 5987–5995. < https://doi.org/10.1109/TIE.2014.2311389
>

208
 WIND POWER FARMS MODELED BASED ON THE WAKE
EFFECT3

Victorita Radulescu
University Politehnica of Bucharest, Bucharest, Romania, [email protected], ORCID: https://orcid.org/0000-0002-
8707-1914

Victorita Radulescu, Wind power farms modeled based on the wake effect, 8th Eur. Conf. Ren.
Cite this paper as:
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Romania is considered in S-E Europe as having a high wind potential, with a total capacity of
14,000 MW. The present paper illustrates a methodology of wind park realization, based on
optimal turbines distribution. A correlation between the perturbations induced by masts and
nacelles, atmospheric data, and rows number of turbines is realized. To estimate the best structure,
firstly are presented some theoretical aspects referring to the wake effect and their superposition,
by solving the RANS flow equations. The large eddy simulation method, where the computational
domain is structured in the mean stream direction is adopted. Numerical simulation with Ansys
Fluent software to estimate the vortices formed behind the rotor and nacelle was accomplished.
Briefly, are presented some measured data for the wind velocity and its direction and some other
significant air parameters, recorded in the future location, the South part of Moldavia. The
numerical model was tested for several types of turbines with power around 2 MW. Some
conclusions and references are also mentioned at the end of the paper.
Keywords: Atmospheric Measurements, Numerical simulations, Wind turbine, Wind farms
© 2020 Published by ECRES

1. INTRODUCTION
In Romania, in the last decade, the utilization of renewable resources, mainly the wind, became a major interest,
attracting annually more and more investors, due to its high wind potential. Since nowadays the hub of the
turbines is at high levels around 100 m, therefore was realized a new wind map for the registered average values
of wind intensity at 50 m. Due to the present economic and political context, a national objective is to secure
energy for a "wider Europe" competitively, based on the Treaty signed by all EU countries, the Energy
Community Treaty. A new objective is to extend the internal energy markets beyond the boundaries of the EU,
by creating a Pan-European Energy Market [1], [2].
In 2008, Romania began the largest wind power farm project, located on land in Europe, near the Black Sea
region, [3]. It is located in Dobrogea, near the communes Fantanele, Cogealac, and Garden, in the north part of
the Constanta County, about 17 km from the Black Sea, with a total installed capacity of 600 MW. The project
began in October 2008 and, at the end of 2010, a total of 120 turbines were already completed, connected to the
electric distribution network, and have started to operate. The wind park has 240 wind turbines GE-2.5 MW, of
148.3 meters high, and with a rotor diameter of 116 meters. The project was developed in two stages - first at
Fantanele which included 139 wind turbines with a capacity of 347.5 MW and the second in the Cogealac and
Garden, including 101 turbines with a total capacity of 252.5 MW. Through this investment, CEZ Group has
become the largest investor in the Romanian wind energy.
Since 2018 started the development of new capacity in wind turbines equipped with turbines Vestas V100, V112,
GE-2.75-103, and GE 2.5xl. Late last year, a capacity of about 5,000 MW of renewable energy has received
Green Certificates, of which 3105 MW for wind, 1,360 MW photovoltaic, 342 MW hydro, and 142 MW biomass
and biogas. Investments in the renewable energy sector have amounted to about 8 billion euros. On May 14,
2019, the wind turbines in Romania have produced 54 GWh, covering 35% of national consumption. Thus, the
wind potential has allowed our country to be ranked first among the countries with the highest percentage of
wind power in electricity consumption. Our country was followed by Denmark with 22% and 19 GWh, Germany
15% and 221 GWh and Ireland still 15% and 13 GWh. At the European level, 7.5% of the electricity demand
was covered by wind, such that the share of wind in Romania reported to consumption was nearly five times
higher than the European average. Production was higher than the consumption, so Romania in April, May 2019
has exported electricity to the neighboring countries. Each investor is interested to know as many possible details

209
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

about the future location by prior analysis of the wind potential, optimal distribution of turbines in the wind
park, possible energy capable to be produced for each turbine model. Starting January 2018 in Romania was
liberalized the energy market, being listed on the stock exchange. In these conditions, many investors are
interested in developing new wind farms.

2. THEORETICAL ASPECTS OF LARGE EDDY SIMULATION METHOD

The wind turbines that will be implemented have the nacelles at 100-150 m, then the method is used for modeling
a wind power plant equipped with turbines rotors of diameters D = 80- 140 m. Firstly, each rotor is modeled as
a disk of vortices because it is necessary a large domain of turbines with a thrust coefficient of CT=0:75, being
representative for the turbines selected to operate near the Black Sea shore region [4].
A classical approach to study the turbulence is realized by decomposition of the Reynolds efforts into a mean
value and a fluctuating term. The Reynolds Average Navier-Stokes (RANS) equations contain the unknowns
Reynolds stresses, as functions of the turbulence scale, necessary to be modeled, to solve the flow equations.
The main advantage of using the RANS in turbulent modeling consists of the possibility of treating the
turbulence as steady phenomena, even if the flow is unsteady [5].
The RANS models are classified by the number of additional differential equations needed to close the original
set of equations. There are four possibilities:
• Algebraic equation of the mixing length
• An equation model, for example - Spalart Allmaras Model
• Two equations needed to be solved. To solve these equations there are known three models: the k-e - standard
model or the Re-normalized group, the k-w standard model, or the shear stress transport model
• Five equations for a 2D model and seven equations for the 3D model, known as the RSM model
Another approach of turbulence modeling is represented by LES, based on the space filtered Navier-Stokes
equations.
The ground’s roughness has an important role in wind velocity distribution. The vertical distribution of the
horizontal wind component is defined by the logarithmic law [6]:

Uy  1  y 
 ln  (1)
u *   y 0 


Where: U(z)- the average component of the horizontal velocity, y –the vertical direction, u* - the friction velocity,
y0 – the altitude where U(y) is considered almost constant, and k - is the von Karman constant, k = 0,4-0,42. The
usually values for z0 are mentioned in Table 1. The Navier-Stokes equations for the mass conservation (the
continuity equation) and momentum conservation may be written:

u i
0 (2)
x i
 u i u i  p   u i 
   ui     fi    (3)
 t x i  x i xi  x i 

Where: ui - the velocity components, r - density, h - viscosity, fi - the mass forces, and p - pressure. Each velocity
is formed by two components, as an example u x  U  u' than in the equation of the momentum conservation
appear supplementary terms, due to the vortices produced by the turbulent shear stresses.
The wakes behind the wind turbine are characterized by velocity deficits produced by the increased turbulence,
affecting the turbines load and so the wind power farms production. The wakes do not propagate steadily behind
the rotor, but rather tend to oscillate randomly, a phenomenon known as the “wake meandering”. It has important
implications for the turbines operating within the wakes of other turbines.
Recent studies investigating this meandering phenomenon are based on the hypothesis that the wake behaves as
a passive tracer, governed at large-scale by the lateral and horizontal turbulent components of the vortices.
Instead, in experimental modeling, the wake effect is developed in tri-dimensional vortices, with a major effect
on turbines efficiency [7]. A new hypothesis allows modeling the wind park considering the first turbine as
having a roughness effect z0,hi, and the rest of the wind park included in the roughened surface with an increased
momentum flux, with kinetic energy losses, Figure 1-a, b.

210
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The existence of these two logarithmic layers has been confirmed in the Large Eddy Simulations (LES) model.
The results illustrate that there is an intermediate wake region between these two logarithmic layers.
The turbine wake creates a region where the eddy mixing is increased, being modeled by considering an
increased eddy viscosity. This estimation predicts the horizontal distribution of the average velocity in three
layers, the bottom logarithmic layer, the turbine wake layer, and the top logarithmic region.
The wind park may be modeled at large scale as being a fully rough developed flow for the atmospheric air,
modeled with Fluent 6; the mean velocity is:

Uu * 1  Eu * y 

u i2
 ln

   1  C s k s  

(4)

(a) (b)
Figure 1. (a) Subdivision of the selected domain, (b) New hypothesis in modeling the wind farms.

Where E-empirical constant, E = 9.793, u* is determined from Eq. [1], k*S - the turbulent kinetic energy, Cs - a
roughness constant, chosen based on the turbine’s height Cs = 0.5; it must be in the interval [0, 1] and higher
than 0.2. Based on these assumptions where k = 0.4, then Eq. [4] became

U 1  u* y 
  5 ,43
 ln (5)
u*   C s k s 

u*  z 
Uz   ln  z0,lo  z  H (6)
  z0 ,lo 
 

Where the roughness z0, is defined in m, as it is used in Eq. [6]; 10 m is considered the standard reference height
for different types of natural surfaces or buildings. The roughness function takes different forms depending on
the K value. Three regimes are mainly distinguished:
- Aerodynamically smooth K < 2.25
- Transitional 2.25 ≤ K< 90
- Fully rough K ≥ 90, where
k2 z
K  k  z  u*  u 2  u1  (7)
lnz 2 / z 1 

The lower logarithmic region Eq. [6] extends till z = z H - D/2. The wake layer is assumed between this level and
z = zH + D/4, with the velocity distribution, Eq. [7] and the logarithmic region, above the wind turbines region
Eq. [9] which is extending up to internal boundary layer IBL.

u*  z 
Uz   hi ln zH + D/4  z  dABL (8)
  z0 ,hi 
 
u*  z 
Uz   ln  dABL  z  H (9)
  z0 ,lo 
 

211
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The slope of the logarithmic regions above and below the turbines depends on the eddy viscosity, which
increases with the altitude and turbines position in the wind farm.

3. EXPERIMENTAL MEASUREMENTS

The environmental data were monitored for two years, for 16 parameters, with masts of 80 m height. The wind
velocity was registered with anemometers positioned at 70 m, 60 m, and 50 m and the wind directions with wind
vanes at 70 m and 50 m. At 10 m from the ground is placed a data-logger for recording and storing the measured
parameters for about 30 days. In the same box are located the sensors for pressure, temperature, air humidity,
and solar radiation. With a setting selected at 24 h, daily, the registered data send by internet at a local server,
near the future wind park [11]. The database is structured in months and years, for the wind parameters at three
levels from the ground (intensity and direction) and air parameters: temperature, pressure, density, solar
radiation, humidity, dew point, etc. All these parameters affect the wind turbine functioning, implicit its
efficiency. The data recording for this site is made continuously, up to the completion of wind farm execution
and start to deliver electric energy. The column’s structure is mentioned in Table 1. The processed database
contains approximately 4,200,000 measured data. In Table 2 are mentioned the wind intensity as average values,
the maximum recorded at 10 minutes, and the maximum gust recorded at 1 second, for all three levels of
measurements. In Figure 2 are presented the average daily value of the wind intensity at 60 m and the dispersion
of the wind distribution for 3 months November and December 2018 and January 2019.

Table 1. Data Logger configuration

Significance Parameter Settings
C1- SWI C3 Wind velocity High anemometer: 70m, Unity: m/s, Slope, offset: adjusted,
Azimuth: South-East, Static interval: 10 min
C2- SWI C3 Wind velocity High anemometer: 60m, Unity: m/s
Azimuth: South-East, Static interval: 10 min
C3- SWI C3 Wind velocity High anemometer: 50m, Unity: m/s
Azimuth: South-East, Static interval: 10 min
A1- SWI PV1 Wind direction High wind vane: 70m, Unity: 360°, 2500 V, Azimuth: North-West
Static interval: 10 min, Full range: 0-2.5 V, measured: Av., standard deviation
A2- SWI PV1 Wind direction High wind vane: 50m, Unity: 360°, 2500 V, Azimuth: North-West
Static interval: 10 min, Full range: 0 - 2.5 V, measured: Av., standard deviation

U (m /s) Daily Average velocity

30,00

25,00

20,00

15,00

10,00

5,00

0,00
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Day
Sep.2018 Oct.2018
Nov.2018 Dec.2018
Jan.2019 Starting wind turbine

(a) (b)
Figure 2. (a) Average daily value of wind intensity at 60 m, (b) Dispersion of the wind intensity .

Table 2. Recorded wind velocities

High Av. Velocity [m/s] Max. velocity 10’ [m/s] Max velocity 1’’ [m/s]
70m 8.972 17.233 38.457
60m 8.648 15.676 30.3625
50m 8.328 15.194 29.56

212
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. MODELING THE WAKE EFFECT BEHIND TURBINES

If the wind farm consists of wind turbines arranged in several rows, the vertical distribution of the wind velocity
average horizontal components has different values for each row. Numerical modeling was set for turbines GE
2.5 MW, with the rotor diameter D = 116 m, three blades of length L= 56.9 m, hub altitude H h=90 m, and the
turbine height H = 148.3 m.
The mesh grid for an isolated turbine was created from two domains: a spherical area whose radius is about 16
times higher than the radius of the turbine and a field containing the rotational flow induced by the turbine, with
a cylindrical shape, with a height and a radius of the base of 1.6 times bigger than the radius of the turbine. The
computing mesh consists of an average of 8.5 million elements, distributed as follows: 1 million for a fixed grid
associated with the faraway domain or known as boundary conditions, with undisturbed wind speed and 7.5
million elements near the wind turbine. These are distributed as it follows 2.5 million near the turbine, meaning
the tower and nacelle, and the remaining 5 million cells in the necessary layer to analyze the modeling of
development and separation of vortices, the wake effect behind the blades and nacelle.
Vortices separation at the end of the blade (due to the pressure difference between the upper side and lower side)
are transported downstream having a negative influence on the rotor’s performances, creating interferences
between the rows of turbines. A decrease of the air intensity appears in the wind park from one row to another.
For modeling, the wind park was set a fine mesh with 288x192x96 elements which characterizes the wake effect
of the wind turbine.
Figure 3-a illustrates the CFD computations for the vertical distribution of velocity in case of wind turbines
placed in one row and direct wind flow and in Figure 3-b the simulations for the inverse flow, before the turbine
is changing its rotor's position. Due to the produced influence by the nacelle and turbine rotor, the horizontal
velocity will be strongly affected in the case of turbines disposed of many rows.

(a) (b)
Figure 3. Wake effect behind the turbine depending on the wind direction.

It may be observed that a massive circulation and velocity behind the blades and nacelle appear, generating a
decrease of the wind turbine's aerodynamic performance, Figure 4, view modeling from the top of the turbine.

Figure 4. Streamlines before and after the wind turbine.

The obtained values of the vertical and horizontal distribution for the average wind velocity, considering only
the perturbations introduced by the mast are illustrated in Figure 5. It may be observed that it acts as an increased
roughness, from one row to another. Both analyses for the velocities components are made for wind turbines
disposed of one, two, three, four, and five rows. Horizontal lines mark the turbine’s hub position and the limits
of the rotor. Based on the CFD modeling was determined the distribution of the current lines and velocities near
the turbine. If the turbines are located on many rows, the disturbances and induced vortices become higher, due
to the supplementary disturbances from one turbine to another.

213
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

z/D (-) 1 row 4,5

3 2 row s y/D (-) 1 row
3 row s 2 rows
4 row s 4
5 row s 3 rows
2,5 Series6 4 rows
3,5 5 rows
Series6
3 Series7
2
Series8
2,5
1,5
2

1 1,5

1
0,5 x/D (-)
0,5
U(m/s)
0 0
0 2 4 6 8 10 12 0 0,5 1 1,5 2 2,5 3

(a) (b)
Figure 5. Velocity distribution: (a vertical, (b) horizontal.

Based on the wind speed and direction measurements carried out during two years were selected and modeled
5 types of turbines.

5. DISCUSSIONS

Romania with favorable wind resources represents a real interest area for further developments in wind energy.
In 2007 has begun a massive exploration process of wind potential evaluations, by installing more than 140
environmental stations with an average height of 60-80 meters. The measurements confirmed that the Dobrogea
region, Black Sea shore, and south Moldavia represent favorable zones for such wind power farm developments.
A decisive factor in the estimation of the efficiency of the wind farm is represented by the optimal positioning
of the turbines so the interference from one turbine to another be minimal. In the beginning, it was detailed the
mechanism of initiation and downstream propagation of vortices behind the turbine rotor and nacelle. It was
determined the velocity distribution in a vertical and horizontal plane, behind the turbine. The obtained analytical
results are quite similar to those obtained by numerical modeling and experimental measurements, the errors
being below 2%. The selected solver is based on the pressure behind the turbine, by creating a coupling between
the speed and pressure. This allows a better convergence, approximately 12 times faster compared with the
classical algorithms, and ensures approximately two-times fewer memory requirements. For this simulation
model was used 35 GB RAM; the time to achieve the convergence residue on a six-core processor of 3.2 GHz
was more than 68 hours.

ACKNOWLEDGMENT
I thank my colleagues from SC Hidroelectrica SA who funded the measurements and finally, co-funded the
realization of the wind power farm.

REFERENCES
[1] EWEA European Wind Energy Association, Wind in power: European statistics, 2016
[2] Baklanov A., Grisogono G., Atmospheric Boundary Layers Nature, Theory and Applications to Environmental
Modeling and Security, ISBN-978-0-387-74318-9, 2007
[3] Richards, P. J., Hoxey, R. P., Appropriate boundary conditions for computational wind engineering models using the
k-ε turbulence model, Journal of Wind EIA, no 46-47, pgs. 145-153, 2003
[4] Larsen G. C. et al., Dynamic wake meandering modeling, Technical Report Risø-R-1607(EN), Risø National
Laboratory: Roskilde, 2007
[5] Frandsen S., Barthelmie R., Pryor S., et al, Analytical modeling of wind speed deficit in large offshore wind farms,
Wind Energy 9, 39, 2006
[6] Meyers J., and Meneveau C., Optimal turbine spacing in fully developed wind farm boundary layers, Wind Energy 15,
305 (2012), pp. 43-47, 2010
[7]. Søren E., Larsen P, Atmospheric boundary layer over land and sea: focus on the off-shore southern Baltic and North
Sea region, Technical University of Denmark 2013

214
 FREQUENCY AND VOLTAGE CONTROL OF SEIG-WT SYSTEM

Nasir Bilal Abdullah

Northern Technical University, IRAQ, [email protected], ORCID:https://orcid.org/0000-0002-2074-
6841

Nasir, BA. Frequency and Voltage Control of SEIG-WT System. 8th Eur. Conf. Ren. Energy
Cite this paper as :
Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: The three-phase self-excited induction generator (SEIG) plays a basic rule in sources of renewable
energy, such as wind turbines (WT). His main defect is poor regulation of output voltage and
frequency under variable rotor speed and load conditions at stand-alone and isolated area operation
mode. New analytical formulas are used to accurate calculation of minimum and maximum values
of excitation capacitance and generator rotor cut-off and maximum speed. New control strategies
are used to regulate the SEIG voltage and frequency under variable load and turbine speed
conditions when the system operates at stand-alone and isolate mode. The results of the dynamic
model are partially compared with experimental results, and accurate agree are shown.
Keywords: self-excited induction generator, wind turbine, excitation capacitance, voltage and frequency
regulation.
© 2020 Published by ECRES

1. INTRODUCTION

The generator output frequency depends directly on the wind turbine speed. Wind turbines can operate with
fixed or variable speed. Pitch angle control to adjust the variable speed WTs have become a dominating type in
recent years. This will increase the efficiency of the energy captured by the WT and reduce the torque of the
drive train. When the wind speed is below the rated speed, the speed of the generator rotor can be adjusted to
maintain a constant tip speed ratio to obtain the maximum power coefficient [1-2].

The pitch angle controller is important for wind speed higher than the rated speed, and the generator rotor speed
must be kept constant [3-4]. The main aim of the pitch angle controller is to smooth the output power of the
induction generator by controlling the turbine-blade pitch angle [10]. Modern WTs are designed with a pitch
regulator of the rotor blade angle, which has to be able to turn around their axis to face the change in wind
direction. The generator output voltage depends directly on the excitation capacitance. This capacitance can be
controlled using a capacitor bank scheme, static voltage compensator (SVC), and voltage or current source
converter based static compensator (STATCOM) to overcome the poor voltage regulation of the generator [5-
6].

The stand-alone SEIG-WT system in an isolated mode of operation is shown in fig. 1. It consists of four parts:
wind turbine as a prime-mover, SEIG, Excitation capacitor bank, and load. The reactive power supplied to the
load can be obtained by subtracting the reactive power consumed by the generator from the reactive power
generated by the excitation capacitance while the active load power is equal to the generated real power by the
generator.

Figure 1. SEIG - Wind turbine system.

The output voltage and frequency of the SEIG are unidentified in the stand-alone mode of operation and depend
on the machine parameters that need to be measured or computed at a given speed, excitation capacitance, and

215
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

load impedance. The dynamic machine model can be described by differential equations with time-varying
mutual inductances, but such a model tends to be very complex. The 3-phase machine can be represented by an
equivalent two-phase machine with Park's transformation [5-6]. Both stator and rotor variables can be
transformed into a synchronously rotating reference frame that moves with a rotating magnetic field. This model
is very important and preferred for transient analysis and will be used in the present dynamic model of SEIG
analysis by choosing the electrical machine currents as the main variables in the model.

Fig.( 2-a.b) shows the D-Q axes equivalent circuit of SEIG with excitation capacitance, stray load loss resistance,
and inductive load in a synchronously rotating reference frame of the stator and rotor variables.

(a)

(b)

Figure2. (a) D-axis equivalent circuit of SEIG, (b) Q-axis equivalent circuit of SEIG, D-Q axes of the modified equivalent
circuit of SEIG

2-MINIMUM AND MAXIMUM SEIG SPEED AND EXCITATION CAPACITOR

Many calculation methods of minimum and maximum excitation capacitance values to initiate the SEIG-WT
system are proposed in the literature. In these Methods, complicated models are used based on several
approximations to simplify the process of calculations, such as a limited range of rotor speed and resistive load,
neglecting the iron core resistance, the stray load loss resistance, and the effect of magnetizing saturation. Also,
most of the previous works use a numerical iteration method to determine the minimum and maximum excitation
capacitance, but the iteration procedure takes a long time, and maybe a divergence occurs, and this process
cannot be used online. In this section, a new, simple and accurate analytical method depends on the equivalent
circuit parameters of the generator including iron core resistance, stray load loss resistance, and magnetizing
saturation effect is proposed to determine the critical ( minimum and maximum) excitation capacitor and critical
(minimum and maximum) generator rotor speed values. The theoretical results of the proposed formulas are
compared with experimental results, using the parameters of the induction generator that configured in the
laboratory to show the validity of the proposed method. The relation between excitation capacitance and rotor
speed is shown in fig. (3.a). The relation is a hyperbolic shape. When the generator is driven at speed exceed
that of synchronous speed, a minimum capacitance is required to excite the SEIG, but when the SEIG is driven
at low mechanical speed, it needs greater capacitor value for starting up condition.

216
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a)

(b)

Figure 3. (a) Variation of excitation capacitance with rotor speed, (b) Variation of rotor speed with generator terminal
voltage, Variation of excitation capacitance with rotor speed (a) and rotor speed with terminal voltage (b).

In this proposed method, the excitation capacitance calculation based on the resonance between the stator
winding inductance and excitation capacitance at no load. From the equivalent circuit of the machine by selecting
the resonance frequency equals the machine base frequency at no load. The formula for calculation of excitation
capacitance can be derived as:
1
Ce = ω 2 .(L (1)
r m + Lℓs )

In all literature, the authors have used this equation for determining the minimum excitation capacitance which
is not a true assumption, due to the minimum, or maximum excitation capacitance depends on the maximum or
minimum generator rotor speed (𝜔𝑟 ) in electrical rad. / sec. These minimum and maximum excitation
capacitances can be calculated as:
1
𝐶𝑒 (max) = 2 .(𝐿 + 𝐿 ) (2)
𝜔𝑟𝑐 𝑚 ℓ𝑠

1
𝐶𝑒 (min) = 2 .(𝐿 + 𝐿 ) (3)
𝜔𝑟𝑚 𝑚 ℓ𝑠

Where 𝜔𝑟𝑐 = the machine cut-off or minimum rotor speed in (elec. rad./ sec.), as shown in fig.( 3.b).

𝜔𝑟𝑚 is the maximum ( above synchronous speed ) allowable machine speed in electrical. rad./esc.

𝜔𝑟𝑐 = 𝜔𝑠 . (1 − 𝑆𝑚 ) (4)

𝜔𝑟𝑚 = 𝜔𝑠 . (1 + 𝑆𝑚 ) (5)

where 𝜔𝑠 = the machine synchronous-speed or angular-frequency in elec. rad./esc.

𝑆𝑚 = the machine slip at maximum torque or maximum power. The slip is positive when the machine rotates
under synchronous speed, and it is negative when the machine rotates above synchronous speed. The slip at
maximum torque can be calculated from the machine equivalent circuit analysis, taking the effect of iron core
resistance (𝑅𝑖𝑐 ) and stray load loss resistance (𝑅𝑠ℓℓ ) as well as the dynamic saturation into account.

217
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3- BLADE PITCH ANGLE MODEL CONTROL OF WIND TURBINE

The required or demand blade pitch angle (𝛽𝑑 ) can be controlled by introducing the wind speed measured by
the anemometer and the 𝛽 − 𝑉𝑤 relation to the programmable logic controller (PLC). The pitch angle-wind speed
function can be calculated from the wind power coefficient (𝐶𝑝 ), which is a function of 𝛽 and tip speed ratio (𝜆).
Then the PLC will calculate the demand pitch angle (𝛽𝑑 ) and sends this value as a signal to the actuator, as
shown in fig. (4).

Figure 4. Blade pitch-angle control model of the wind turbine.

The pitch actuator (stepper or servo motor) consists of mechanical and hydraulic systems, which are used to turn
the turbine blades along their axis. The actuator model describes the dynamic behavior between the demand
pitch angle received from PLC and executed or measured angle (𝛽𝑚 ) by the actuator. The wind speed can be
measured directly by the anemometer, this method is simple, less unit cost, robustness, better transient
performance, and more accurate than the other control strategies, such as measurement of generator speed by a
tacho-generator or measurement of generator power by a digital wattmeter. The relation between 𝛽 and wind
speed 𝑉𝜔 is shown in fig. (5).This relation is fed to the programmable logic controller (PLC) with the
measurement of wind speed by the anemometer.

Figure 5. Pitch angle - wind speed relation when the wind speed greater than 13 𝑚 ∕ 𝑠 .

4-SEIG VOLTAGE MODEL CONTROL

The generator terminal voltage can be controlled by varying the excitation capacitance. The excitation
capacitance (𝐶𝑒 ) as a function of terminal voltage can be obtained experimentally and using polynomial curve
fitting, as shown in fig.(6).

Figure 6. Variation of excitation capacitance with terminal voltage.

The relation of 𝐶𝑒 − 𝑉𝑝ℎ is used as a simple and accurate control strategy of the terminal voltage. The terminal
voltage can be measured directly at the generator terminals and fed with a required reference voltage to the PLC
as input data.
The PLC compares between the two signals and the error voltage is used to calculate the demand excitation
capacitance (𝐶𝑒𝑑 ) from the 𝐶𝑒 − 𝑉𝑝ℎ curve function.
The PLC will send this new demand excitation capacitance to the actuator. The actuator consists of a group of
relays and switches to execute the required capacitance from the capacitor's bank, as shown in fig. (7).

218
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 7. SEIG terminal voltage control by capacitors bank.

The capacitors bank consists of three strings of capacitors. Each string consists of six capacitors connected in
parallel. One of the string capacitors is fixed as minimum excitation, and the others can be adjusted by the relays
to obtain a constant terminal voltage with a variation of generator load from no load to full load.
For the SEIG used in this project, the minimum excitation capacitance is (30) 𝜇𝐹 calculated from fig. (3.a) and
fixed in the capacitors bank to overcome the VAR load during the no-load operation and the other five capacitors
(each of them has 3𝜇𝐹 are connected with the minimum excitation capacitance in parallel for each string.

5-RESULTS AND DISCUSSION

In the experimental setup, 3-phase, squirrel cage, 220V, Δ- connected induction machine is coupled to a
separately excited DC Motor. The DC motor acts as a prime-mover emulating wind-driven turbine. Test results
and simulation results in Matlab/Simulink of the SEIG-WT system are presented at no load initiation of the self-
excitation. The SEIG is driven at speed 1525 r.p.m (above synchronous speed). Then a connected delta excitation
capacitance bank of 40 𝜇𝐹 was switched across the SEIG stator terminals. Fig.(8-a.b)shows the experimental
and Matlab simulation results of terminal voltage at steady-state conditions. Fig.( 9)shows the simulation result
of the SEIG stator flux linkage trajectory in the D-Q plane. The trajectory is smooth, circular, and builds-up
during the initiating of self-excitation.

(a)

(b)
Figure 8. Experimental result (a) and Matlab simulation result (b) of steady-state generator terminal voltage.

Finally, control strategies are proposed in this paper for voltage and frequency regulation using the PLC as the
main controller. Stepper or servo motor is used at the wind turbine site as an actuator to control the wind turbine
speed by controlling the blade pitch angle to regulate the generator frequency. Relays and switches are used at
the generator side as an actuator to adjust the required excitation capacitance from capacitors bank to regulate
the terminal voltage.

Figure 9. Simulation resul of the trajectory of stator flux linkage in the D-Q plane.

219
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

6-CONCLUSION

New, simple, and accurate analytical formulas are derived for calculation of minimum and maximum excitation
capacitances and rotor speed of SEIG. These formulas are derived from the complete equivalent circuit of the
machine and based on maximum power transfer theory.
A new simple and accurate control strategy is used at the turbine side to regulate the generator frequency by
controlling the pitch angle of the turbine blades via anemometer-PLC and stepper or servo motor as an actuator.
A new simple and accurate control strategy is used at the generator side to regulate the generator output voltage
by variation of excitation capacitors of the capacitor bank via voltage sensor -PLC- and a group of relays and
switches as an actuator.

REFERENCES

[1] V. T. Makvana, R.K. Ahir, D.K. Patel, and J.A. Jadhav, " Study of PID controller based pitch actuator system for variable
speed HAWT using Matlab " , International Journal of Innovative Research in Science, Engineering and Technology,
Vol. 2, Issue 5, pp. 1496-1504, May 2013.
[2] J.M. Ramirez and T.M. Emmanuel, " An electronic load controller for the self-excited induction generator " , IEEE
Transactions on Energy Conversion, Vol. 22, No. 2, pp. 546-548 , June 2007.
[3] K.H. Youssef, M.A. Wahba, H.A. Yousef, and O.A. Sebakhy, " a new method for voltage and frequency control of stand-
alone self-excited induction generator using PWM converter with variable DC link voltage " , 2008 American Control
Conference, Westin Seattle Hotel, Washington, USA, pp. 2486-2491, 11-13June 2008.
[4] H. Gen, D. Xu, B. Wu, and W. Huang, " Direct Voltage control for a stand-alone wind-driven self-excited induction
generator with improved power quality " , IEEE Transactions on Power Electronics, Vol. 26, No. 8, pp. 2358-2363,
August 2011.
[5] A.M. Kassem, " Predictive voltage control of stand-alone wind energy conversion system " , WSEAS Transactions on
Systems and Control, Vol. 7, Issue 3, pp. 97-102, July 2012.
[6] M.L. Elhafyani, S. Zouggar, M. Benkaddour, and A. Aziz, "Behavior an induction generator without and with voltage
regulator" , Smart Grid and Renewable Energy, Vol. 5, pp. 207-219, 2014.

220
 ALGEBRAIC MODELING OF A NEW ELECTROMAGNETIC
MECHANICAL VIBRATION ENERGY HARVESTING

Erol Kurt
Department of Electrical and Electronics Engineering, Technology Faculty, Gazi University,
06500, Beşevler, Ankara/Turkey, [email protected], ORCID: 0000-0002-3615-6926

Aigerim Issimova
Al-Farabi Kazakh National University, Almaty, Kazakhstan, [email protected], ORCID: 0000-0002-
0343-3974

Bekbolat Medetov
Al-Farabi Kazakh National University, Almaty, Kazakhstan, [email protected], ORCID: 0000-0002-5594-8435

Kurt, E., Issimova, A., Medetov, B.. Algebraic modeling of a new electromagnetic mechanical
Cite this paper as: vibration energy harvesting, 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,
Turkey

Abstract: In the current paper, an algebraic analysis of a new mechanical energy harvester has been reported.
The new harvester is considered to work with the mechanical oscillations. It has two magnets and
a housing component between them and the winding itself is wrapped around those magnets. When
an external acceleration occurs on the system, single magnet is considered to move in one direction,
whereas other stays at the rest. The initial analytic study proves that a certain varying magnetic
flux is available between the terminals, thereby a certain electromotive force can be observed to
harvest an electrical power.
Keywords: Electromagnetic, harvester, permanent magnet, flux, electrical power, analytic
© 2020 Published by ECRES

Nomenclature
Mc Coil mass
Ml Mass of lower magnet
Mu Mass of upper magnet
Mp Proof mass
A Vibration amplitude
 Damping coefficient
B0 Tip magnetic flux of permanent magnet
M Magnetization
 Electromotive force
z Displacement of a magnet
 Magnetic flux
R Reluctance
N Number of turn

1. INTRODUCTION

Electrical power is used intensely in everyday life world-widely. The main reason for such an intense usage is
the comfortable settlement of electrical devices in houses, factories and outdoor in addition to their easy
maintenance apart from other mechanical devices. Considering the fossil-fuel consuming systems and fuel-cells,
energy conversion and loss is much easy to understand and apply. Apart from any thermodynamic energy source
system such as fuel fed oven, fossil fuel engines, hydropower motor systems, electricity can be transmitted to
far distances with lower losses compared to mechanical energy sources. Indeed thermodynamic energy
conversion and management has many problems of thermodynamics. Indeed, energy exergy calculations on
those mechanical energy systems yield to complex calculations and lower efficiency systems after all those

221
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

calculations, whereas electricity is easy to handle since it has a certain electrical loss due to Joule heating and
easy conversion mechanisms from electric to magnetic or vice-versa [1,2].
Energy harvesters are mainly considered to provide the energy requirements of low power systems such as
sensors, actuators, wireless nodes, etc. The motivation to work on the harvesters is that these systems do not
require any battery replacement for large time scales, indeed they feed themselves from the natural sources such
as mechanical vibrations, noisy environment, wind, electromagnetic waves, temperature gradient, etc. [3-5].
Following the technical applications world-widely, they find place in many devices such as ships, underwater
pipelines, automobiles, laptops, medical devices, cell phones, etc [6-8]. Since all those devices and un-mentioned
many requires the electrical energy. In this manner, these requirements can be stated as stand-by and operational
powers. Because, many applications have at least 10 times lower stand-by power than its active operation power
while the device is working [9]. Therefore, there exists many configuration of harvesters for their usage
purposes.

In principle, the harvesters can be classified as electromagnetic, electrostatic, thermoelectric and piezoelectric
[4]. Of course, there may be other kind of harvesters depending on the feeding source, however the above
mentioned classes forms the 90 % of all applications. In the present work, we purpose to report a new
electromagnetic harvester system, namely two magnet coil system (TMCS). This work includes only the starting
design and algebraic analysis of this new type of harvester. Further studies will be reported in future
communications.

Section 2 gives a literature summary for electromagnetic harvesters. Section 3 reports the main algebraic study
including the formulation. The concluding remarks are given in the last section.

2. LITERATURE ON ELECTROMAGNERIC HARVESTERS

Piezoelectric, electromagnetic, electrostatic and hybrid energy harvesting systems are encountered in the
literature. Among the classifications above, the electrostatic and electromagnetic harvesters can generate 4
mJ/cm3 energy density [4]. The studies on electromagnetic systems in the literature mainly focus on the
optimized structures. Because the magnetic flux and air gap parameters are important quantities in order to
harvest the maximal power from the system. Besides, N,  and velocity of the moving component such as
magnet or coil are among important parameters to examine in the proposed systems [10]. In the harvesters,
voltage and power are two vital quantities. Power generated by a harvester depends on the electrical load at the
output. From electrical point of view, the maximal power idea defines the exact power point for the maximal
value [11]. It aims to consider the electrical load where the multiplication of current and voltage becomes
maximum; thereby many tests are applied to the devices in order to optimize the system.

The improvements on the rare-earth element makes magnets indispensable components of electromagnetic
harvesters [4,5]. For instance, Amirtharajah and co-worker designed and implemented a harvester in the form
of a cylindrical housing and winding in 1998 [12]. In their geometry, a cylindrical mass was used with a spring
and it was fixed to one of the tips. At the other tip, a permanent magnet (PM) was attached with a mass. This
mass was also a coil to oscillate freely. The moving mass cuts the flux and  was generated in accordance with
Faraday’s Law. Their results have proven that around 400 W was harvested inside a 2-cm-amplitude. El-
Hami’s group considered a cantilever beam system with a C–shaped core [13]. A coupled PM was attached to
two sides of the core. They were oscillated with a beam inside a stable coil. The total volume and the
harvested power were 240 mm3 and 0.53 mW at 322 Hz, respectively.

In Ref. [5], Beeby and his co-workers fabricated a device with discrete permanent magnets (PMs), coils and
other components. The wires were thin about 12 m in diameter and the volume was kept at 150 mm3. That
harvester generated 46 W and 0.428 V at resonance frequency 52 Hz. In other work, Torah et al [14]
obtained 58 W from the same excitation level by using larger PMs for the same configuration. Buren and
Troster [15] worked with a tubular translator with the cylindrical PMs and spacers. That device had a vertical
movement arrangement with parallel springs. The volume and power were 30.4 cm3 and 35 W, respectively,
when it was located on a moving knee. In another work, Yuen and co-workers [16] measured power of 120
W at 70.5 Hz. The proof mass was oscillated in the distance of 250 m. Hadas et al [17], improved a wireless
aircraft and the device generated 3.5 mW at 34.5 Hz within 45 cm3. In a recent study [6], our group has

222
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

designed and implemented a new linear harvester with four small magnets and a movable core-coil system at
the middle of them. The proof mass has been attached to the springs from two ends. The output peak voltage
and power are found as 0.75 V and 14 mW. That yields to a power density of 0.3684 mW/cm 3 with an
efficiency ratio of 36%.

3. ALGEBRAIC ANALYSIS OF TMCS

Fig. 1 sketches the TMCS as an innovative design. Initially, the entire system is easy to implement. Here a pulley
is wrapped via the Copper wires tightly and terminals are directly used as output to an electrical load. Note that
the orientation of magnets is N-S-N-S, thereby a homogeneous magnetic field is produced between the magnets
along z-direction with the air gap d. Note that air gap d is not so large as indicated in Fig. 1. It is exaggerated
here to show the system better. In our case, it is an optimization parameter, which can have a value between 5
mm 10 mm. Note that distance d is stable in TMCS.

Figure 1. The proposed two magnets coil system.

The system indeed can be considered as a linear harvester, however, for excitation in positive and negative z-
direction, only one magnet moves, the other stays at the rest. Strictly speaking, when the pulley and wrapped
winding goes down, only the upper magnet will try to be stay at its actual position according to the Newton’s
First Law (i.e. inertia), but the lower magnet will exactly move down with pulley. Thus the flux variation in time
comes from that reality as an action-reaction mechanism. When the pulley moves up, then the magnets behave
vice versa. Indeed, the move of magnet changes the magnetic flux intensity due to the fact that the flux path is
increased as a result of the acceleration in z-direction. That results with a reluctance (ℜ) increase in the magnetic
circuit, then the magnetic flux is decreased to another value ϕ’. Fig. 2(a,b) shows the operation principle in detail
for both directions.

(a) (b)
Figure 2. Move in (a) –z and (b)+z directions in parallel to Newton’s Principle of Inertia.

223
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

By applying an external shaking to the TMCS, a movement occurs on the proof mass including the coil, wires
and magnets. The proof mass is formulated as follows:

𝑀𝑝 = 𝑀𝑐𝑜𝑖𝑙 + 𝑀𝑢 + 𝑀𝑙 . (1)

Here, one can consider the relation,

𝑀𝑐𝑜𝑖𝑙 ≫ 𝑀𝑢 , 𝑀𝑙 , (2)

since mass of the coil is much larger than the mass of magnets. The velocity and acceleration of the center of
proof mass are determined as follows:

𝑑𝑧
= 𝑣0; (3)
𝑑𝑡

𝑑𝑣0 𝐴
𝑑𝑡
= 𝑀 𝑆𝑖𝑛(𝜔𝑡) (4)
𝑝

Note that Eq. (4) is valid only if the shaking governs the motion compared to friction. Indeed, if the friction
exists, one should also include the damping term. Eq. (4) includes the terms А, Мp, and 𝜔. They are amplitude
of the external force in Newton, the proof mass in kg and the angular frequency of the external vibration in Hz,
respectively.

Now let's look at the movement of each magnet individually. The equation of motion for the upper magnet is as
follows:

𝑑𝑣𝑢 6𝐵0
𝑀𝑢 𝑑𝑡
= −𝛾𝑣𝑢 − 𝐹𝑚𝑢𝑙 + 𝐴 𝑆𝑖𝑛(𝜔𝑡) = −𝛾𝑣𝑢 − 𝜋(𝑧 4 + 𝐴 𝑆𝑖𝑛(𝜔𝑡) (5)
𝑢 +𝑑)

𝑑𝑧𝑢
𝑑𝑡
= 𝑣𝑢 (6)

The equation of motion for the lower magnet:

𝑑𝑣𝑙 0 6𝐵
𝑀𝑙 𝑑𝑡
= −𝛾 𝑣𝑙 + 𝐹𝑚𝑢𝑙 + 𝐴𝑠𝑖𝑛(𝜔𝑡) = −𝛾 𝑣𝑙 + 𝜋(𝑧 +𝑑) 4 + 𝐴 𝑆𝑖𝑛(𝜔𝑡) (7)
𝑙
𝑑𝑧𝑙
𝑑𝑡
= 𝑣𝑙 (8)

where, 𝑀𝑢 , 𝑀𝑙 corresponds to the masses of upper and lower magnets, 𝑣𝑢 , 𝑣𝑙 are velocities of the upper and
lower magnets relative to the coil, 𝛾 is the damping coefficient, 𝐹𝑚𝑢𝑙 is the magnetic force induced by cylindrical
magnets, 𝐴 𝑆𝑖𝑛(𝜔𝑡) is the external oscillating force. In that case, one can write the force of the interaction of
two magnets as follows:

𝜋𝜇0 1 1 2
𝐹𝑚𝑢𝑙 = 𝑀2 𝑅 4 ( 2
+ 2
− ) (9)
4 𝑧 (𝑧+2𝐿) (𝑧+𝐿)2

Note that here z denotes the distance between the magnets whereas L gives the lenght of the magnet. R represents
the diameter of the magnet. Since 𝐿 ≪ 𝑧 sits, Eq.[9] can be simplified as,

3𝜇0 1
𝐹(𝑧) = 𝑀1𝑀2 (10)
2𝜋 𝑧4

224
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3𝜇0 1 3𝜇0 1 4𝐵02 6𝐵02

𝑚1 = 𝑣1 ∙ 𝑀1 = 𝑚12 = = (11)
2𝜋 𝑧4 2𝜋 𝑧4 𝜇02 𝜋𝑧 4

2𝐵02 2 1.22 1
𝑀= = ≅ 107 (12)
𝜇0 4𝜋 10−7 2𝜋

Here the M denotes the mutual magnetization of the magnets. By converting the Faraday’s Law to a velocity-
dependent relation, one arrives at,

𝑑Ф
𝜀 = −𝑁 ( ) |𝑣1 − 𝑣2 | (13)
𝑑𝑧

Eq. 13 gives the electromotive force at the output terminals of windings, where, 𝑁 – number of turns, 𝑣1 , 𝑣2,
𝑑Ф
speeds of the upper and lower magnets, the quantity 𝑑𝑧 is the flux variation by distance inside the coil. Then,
the quantity of flux variation per distance is stated as,

𝑑Ф 𝑑𝑧𝑢 1 𝑑𝑧𝑙 1 1
= −2𝐻𝐿𝜇0 𝐴 [ ∙ 𝑑𝑧 − ∙ 𝑑𝑧 ] (𝑑+|𝑧 2 , (14)
𝑑𝑧 𝑑𝑡 𝑑𝑡 𝑢 −𝑧𝑙 |)
𝑑𝑡 𝑑𝑡

where, 𝐴 is the area perpendicular to the flux direction, 𝜇0 is the vacuum permeability, 𝐿 is thickness of magnets.
Besides, the displacement of magnets are denoted by z1 and z2. Note that Eq. 14 is nothing else than the derivative
of magnetic circuit equation given by,

𝑁𝐼 2𝐻𝐿
𝜙= ℜ
= 𝑑+|𝑧𝑢 −𝑧𝑙 |
(15)
𝜇0 𝐴

Since the reluctance changes by the distance between the magnets, the derivative of this flux relation yields to
Eq. 14.

4. CONCLUSIONS

The algebraic analysis has been performed for a new mechanical energy harvester, namely two magnet coil system TMCS.
The harvester has a basic and specific operation scheme. The magnetic flux path is varied by the move of a magnet as a
natural result of inertia. The proof mass is the total mass of TMCS. This linear harvester has a compact form working with
the acceleration of external mechanically moving system. The new harvester is considered to work with the mechanical
oscillations. Indeed, the velocity also plays an important role to get flux variation in high values. In the future work, the
simulation and experimental studies from the TMCS will be reported.

REFERENCES

[1] İ. Dinçer, M.A. Rosen, Exergy Energy, Environment and Sustainable Development, Elsevier, Oxford, (2007)
[2] R. Terzi, E. Kurt, Improving the efficiency of a nuclear power plant using a thermoelectric cogeneration system, Int.
Journal of Renewable Energy Development, 7(1), 77-84, 2018, DOI:10.14710/ijred.7.1.77-84
[3] E. Kurt, M.M. Kale, S. Akbaba, N. Bizon, Analytical and experimental studies on a new linear energy harvester,
Canadian J. Physics, 96(7), 727-733, 2018, DOI:10.1139/cjp-2017-0708
[4] N. Bizon, N.M. Tabatabaei, F. Blaabjerg, E. Kurt, Energy Harvesting and Energy Efficiency, Springer Verlag London
Ltd., UK (2017)
[5] S. P Beeby, T. O’Donnell, In Energy Harvesting Technologies (Editors: S. Priya, D.J. Inman), Springer, USA, 129
(2009)
[6] B. Mutlu, E. Kurt, N. Bizon, J.M. Lopez Guede, Design and Fabrication of a New Micro-Power Scaled Electromagnetic
Harvester, Journal of Energy Systems 3(2), 51-66, 2019, DOI: 10.30521/jes.554900
[7] F.U. Qureshi, A. Muhtaroglu, K. Tuncay, Sensitivity Analysis for Piezoelectric Energy Harvester and Bluff Body
Design toward Underwater Pipeline Monitoring, Journal of Energy Systems, 1(1), 10-20, 2017, DOI:
10.30521/jes.328600
[8] P.S. Zhang, Design of electromagnetic shock absorbers forenergy harvesting from vehicle suspensions, Dissertation,
Stony Brook University, Stony Brook, NY, USA, 2010.

225
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[9] F. Orfei, R. Mincigrucci, I. Neri, F. Travasso, H. Vocca and L. Gammaitoni, Hybrid autonomous transceivers, 5th
European DSP Education and Research Conference (EDERC), Amsterdam, 173-177, 2012,
DOI:10.1109/EDERC.2012.6532249
[10] E. Kurt, F. Cottone, Y. Uzun, F. Orfei, M. Mattarelli, D. Özhan, Design and implementation of a new contactless triple
piezoelectrics wind energy harvester, Int. J. Hydrogen Energy, 42(28), 17813, 2017
[11] Y. Uzun, E. Kurt, H.H. Kurt, Explorations of displacement and velocity nonlinearities and theireffects to power of a
magnetically-excited piezoelectric pendulum, Sensors and Actuators A 224, 119–130, 2015.
[12] R. Amirtharajah, A.P. Chandrakasan, Self-powered signal processing using vibration-based power generation, IEEE
Journal of Solid-State Circuits, 33(5), 687-695, 1998.
[13] M. El-Hami, P. Glynne-Jones, N.M. White, M. Hill, S. Beeby, E. James, A.D. Brown, J.N. Ross, Design and
fabrication of a new vibration-based electromechanical power generator, Sensors and Actuators A: Physical, 92(1),
335-342, 2001, DOI:10.1016/S0924-4247(01)00569-6
[14] R.N. Torah, P. Glynne-Jones, M.J. Tudor, S.P. Beeby, Energy aware wireless microsystem powered by vibration
energy harvesting, In Pro. 7th Int. Workshop on Micro and Nanotechnology for Power Generation and Energy
Conversion Applications (PowerMEMS 2007), Freiburg, Germany, 323, 28-29 Nov. 2007.
[15] T. von Büren, G. Tröster, Design and optimization of a linear vibration driven electromagnetic micro-power generator,
Sensors and Actuators A: Physical, 135(2), 765-775, 2007.
[16] S.C. Yuen, J.M. Lee, W.J. Li, P.H. Leong, An AA-sized vibration-based microgenerator for wireless sensors, IEEE
Pervasive Computing, 6(1), 64-72, 2007.
[17] Z. Hadas, J. Kurfurst, C. Ondrusek, V. Singule, Artificial intelligence based optimization for vibration energy
harvesting applications, Microsystems Technology, 18, 1003-1014, 2012.

226
 EXPERIMENTAL VALIDATION OF AN ARTIFICIAL NEURAL
NETWORKS (ANN) MPPT CONTROLLER OF AN INSTALLED
SOLAR POWER PLANT AT POLYTECHNIC HIGHER SCHOOL

Amadou Ba
Research team in Energetic System and Efficiency of Physics Department, Alioune Diop University of Bambey,
Senegal, [email protected], ORCID: 0000-0003-4211-6362

Sengahne Mbodji
Research team in Renewable Energies Materials and Laser of Physics Department, Alioune Diop University of
Bambey, Senegal, [email protected], ORCID: 0000-0001-7235-5036

Alphousseyni Ndiaye
Research team in Energetic System and Efficiency of Physics Department, Alioune Diop University of Bambey,
Senegal [email protected], ORCID: 0000-0003-4024-4256

BA, Amadou, Ndiaye Alphousseyni and Mbodji. A Experimental validation of an Artificial

Cite this paper as: Neural Networks (ANN) MPPT controller of an installed solar power plant at Polytechnic
Higher School. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The purpose of this study is to design an Articiel Neuron Network (ANN) Maximum power point
tracking (MPPT) controller to optimise the efficiencies of two solar panels installed at Polytechnic
Higher School and operating in a real conditions. A Multilayer perceptron (MLP) is used as part
of this work and the ANN architecture consists of an input layer of two (02) neurons, a hidden
layer of teen (10) neurons and an output layer of one (01) neuron. The caracteristics of the panels
is used to generate the database required to train the ANN and the panel A is used for the training
and the panel B for the validation of the ANN, respectively. Simulation results under the
environement of matlab/Simulink showed that the MPP contoller is tracking continousually the
three MPPT controllers. The results confirm that ANN ensures exact tracking MPP under different
variation of the meteorologiques conditions. Moreover, the ANN has best performance compared
to the two other optimization techniques like INC and P&O.
Keywords: ANN-INC-P&O-MLP-MPPT
© 2020 Published by ECRES

Nomenclature
𝑃𝑝𝑣𝑟𝑒𝑓 Reference Power of the panel (W)
𝑃𝑜𝑝𝑡 Optimal power of the controller (W)
MAE Mean Absolute Error
MAPE Mean Absolute Percentage Error

RMSE Root Mean Square Error

1. INTRODUCTION

Because of the high installation costs and low energy conversion efficiency of photovoltaic (PV) panels, the
efficiency of PV systems improvement study is considered as a necessity. The nonlinear nature of solar energies
requires to use an intermediate converter implementing a control algorithm operating the PV panels at their
maximum power point (MPP) under different environmental conditions.

In the literature, various control techniques commonly known as (Maximum Power Point Tracking) MPPT
controls [1]have been developped to optimize the efficiency of photovoltaic modules. Among these control
techniques we can mention the perturb and observation (P&O) algorithm [4] which is widely used. They can
disrupt the voltage of the solar panel a small amplitude about its initial value and analyze the power variation.

227
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Following the sudden changes in weather conditions that diverge the control and overcome this drawback found
in the literature of the P&O modified commands [5] [6], a comparative study of modified classic MPPT control
is made in [7] to optimize the conversion efficiency of a photovoltaic system. Simulation results show good
performance of the modified INC control leading to an efficiency of 98.5% compared to the conventional INC.
Other commands based on artificial intelligence have been developed to optimize the conversion efficiency of
PV systems. These commands are fuzzy logic (FL), genetic algorithms, artificial neural networks (ANN), the
Optimization swarms of particles (PSO) and the Adaptive Neuro Fuzzy Inference System (ANFIS) [8] which
the combination of artificial neural network and fuzzy logic. A MPPT control based on fuzzy logic (FL) is
presented to optimize the yield of a photovoltaic system. The main drawbacks of this method is the choice of
the membership function where four optimization algorithms are presented in [9] to optimize the fuzzy
membership (MF) and generate an appropriate duty ratio for the MPPT controller. Finally, to validate the
performance of the optimized FL, it is compared with other techniques such as symmetric fuzzy logic controller
(SLFC) and Perturbation and Observation (P&O) methods; an improvement of the convergence speed is noted
and reduced the oscillations. In [10] a comparative study of the PSO and P&O commands to extract the
maximum power of a PV panel is made. The objective is to overcome the disadvantages of the P&O command
as the convergence speed and the oscillations around the Maximum Power Point (MPP). Simulation results show
the good performance of the meta-heuristic Particle Swarm Optimization (PSO) control compared to the
conventional P&O control. In [11], we note the good performance of ANFIS control compared to a heuristic
INC command to optimize the conversion efficiency of a PV system. The results show that the intelligent control
displays a response time an order of 2.4 s and leads to an efficiency of 99.94% compared to 94.41% obtained
with INC control. The ANN has been widely used to model complex relationships between inputs and outputs
of nonlinear systems and simultaneously contribute to the optimizing of the efficiency of photovoltaic systems.
In [12], a MPPT method based on Artificial Neural Network (ANN) was proposed for continuing the overall of
the MPP in various weather conditions for a stand-alone PV system. The proposed method includes the PV
characteristics scanning procedure with the ANN controller in the shading effect. The simulation results show
that the proposed method is able to find the MPP and ensures rapid convergence towards this point with good
efficiency estimated to 97.64% compared to the results determinated with other commands. A technical hybrid
MPPT approach based on ANN-INC and ANN-P&O algorithm is also proposed in [13]. Compared to
conventional methods, it is noted a higher output power and less oscillation around the MPP with the RNA-INC
control. As a contribution an ANN MPPT controller is proposed in this paper to optimize the conversion
efficiencies of two panels. We also use a real electrical caracteristics of two panels to implemente the ANN
controller. Results are compared to the conventional INC and P&O ones.

2. METHODOLOGY

The experimental set up that allowed us to have meteorological data and electrical characteristics is given in
figure 1. It is installed at Polytechnic Higher School of Cheikh Anta Diop University in Dakar, Senegal. Dakar
is located in the extreme west of Africa, between the 17 ° 10 and 17 ° 32 west longitudes and the 14° 53 and 14
° 35 north latitudes. The temperature mean varies between 17 and 25 ° C from December to April and from 27
to 30 ° C from May to November [14]. The set up consists of two panels whose characteristics are listed in tables
1 and 2.

Table 1. Specific technic of the panel A

𝐼𝑠𝑐 (A) 𝐼𝑜𝑝𝑡 (A) 𝑉𝑜𝑐 (V) 𝑉𝑜𝑝𝑡 (V) P(w)
2.56 2.32 20 14.93 35

Table 2. Specific technic of the panel B

𝐼𝑠𝑐 (A) 𝐼𝑜𝑝𝑡 (A) 𝑉𝑜𝑐 (V) 𝑉𝑜𝑝𝑡 (V) P(w)
2.24 1.93 20.50 15.07 30

Where:

228
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

P (w) is the electric power of solar panel;

Iopt, Vopt are the optimal photocurrent and the phototension, respectively;
Isc is the short-circuit photocurrent;
Voc is the open circuit phototension.

To measure the sunshine and temperature, a pyrometer and a thermocouple are used, respectively. An ammeter
and a voltmeter are used to measure the electrical characteristics of each module and from these values the
corresponding solar cell’s electric power is calculated. We used the set up to determine the electrical
characteristics of each module and both the irradiation and the actual temperature of the site.

Figure 1: Experimental set up

The characteristics of the two modules in different weather conditions are given in figures 2, 3, 4 and 5, below.
Each radiation value (Em) is related to a temperature value (Tm). P-V and I-V characteristics of each solar panel
in different profile of sunshine and temperature are shown in these figures.

3
Cuurent (A)

0
0 5 10 15 20 25
Voltage (V)
Em=729W/m²,Tm=25° C Em=980W/m²,Tm=26° C Em=975W/m²,Tm=27° C
Em=802W/m²,Tm=25° C Em=744W/m²,Tm=26° C Emoy=487W/m²,Tm=27° C
Em=349W/m²,Tm=32° C Em=726W/m²,Tm=34° C Em=329W/m²,Tm=30° C
Figure 2. I-V Characteristic of the solar panel A
40

30
Power (W)

20

10

0
0 5 10 Voltage (V) 15 20 25
Em=729W/m²,Tm=25 ° C Em=980W/m²,Tm=26 ° C Em=975W/m²,Tm=27 ° C
Em=802W/m²,Tm=25 ° C Em=744W/m²,Tm=26 ° C Em=487W/m²,Tm=27 ° C
Em=349W/m²,Tm=32 ° C Em=726W/m²,,Tm=34 ° C Em=329W/m²,Tm=30 ° C

229
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. P-V Characteristic of the solar panel A

2.5

2
Current (A)

1.5

0.5

0
0 5 10 Voltage (V) 15 20 25

Em=761W/m²,Tm=25 ° C Em=980W/m²,Tm=25 ° C Em=977W/m2,Tm=26 ° C

Em=783W/m²,Tm=25 ° C Em=765W/m²,Tm=25 ° C Em=531W/m²,Tm=27 ° C
Em=445W/m²,Tm=32 ° C Em=692W/m²,Tm=32° C Em=348W/m²,Tm=30 ° C
Figure 4. I-V Characteristic of the solar panel B

40

30
Power (W)

20

10

0
0 5 10 Voltage (V) 15 20 25
Em=761W/m²,Tm=25 ° C Em=980W/m²,Tm=25 ° C Em=977W/m²,Tm=26 ° C
Em=783W/m²,Tm=25 ° C Em=765W/m²,Tm=25 ° C Em=531W/m²,Tm=27 ° C
Em=445W/m²,Tm=32 ° C Em=692W/m²,Tm=32 ° C Em=348W/m²,Tm=30 ° C
Figure 5. P-V Caracteristic of the solar panel B

Thess figures show that the power of the photovoltaic is sensitive to the variations of the meteologiques
conditions like the radiation (Em) and the temperature (Tm). The panel give it maximum power only for
particular values of current and voltage that depend on weather conditions. To optimize the efficience of this PV
panel a MPPT controller based on ANN is design. A block diagram of the PV controlled system is developed
under Matlab/ Simulink environment and presented in figure 9. It consists of a PV module, an MPP tracker and
a DC load. The MPP tracker is a device composed of power converter associated with our ANN MPPT controller
unit which uses generally a processor. The converter maintains continuously the PV panel operating at the MPP
in order to extract the maximum power available for each variation of the environmental conditions. We used
the electrical characteristics of the panels A and B, respectively, unlike what exists in the literature for the
learning and the validation of the ANN MPPT controller.

Optimization of the power by ANN

Articial intelligence is considered as the most appropriate control techniques through their coping capacities for
nonlinear systems [15]. It is no longer limited to win championships 'alpha GO'. It is used in the speech
recognition systems and / or in the image, the industries, etc.... Here, in our work we used it to control a converter
to extract the maximum power of a photovoltaic panel. As in articial intelligence technique, we used artificial
neural networks that are highly connected to networks of elementary processors operating in parallel. Each
elementary processor calculates a single output based on the received information.

230
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Modelisation of an artificial neuron

Modeling an artificial neuron is to implement a biological neural network system in an artificial appearance. Each
artificial neuron performs a very simple operation which is actually a weighted sum of its inputs. The model of a neuron
is shown in figure 6. The output S of the neuron is given by Eq. [1].

Figure 6. Model of the neuron

𝑛
𝑆 = 𝑓 (∑ 𝑊𝑖𝑗𝑋𝑖 − 𝑏 ) (1)
𝑖=1

Where
Xi (i =1 to n) are the inputs of the neuron,
Wij is the weight synaptics
With b is the biais of the neuron
Here, in our work we used multilayer perceptrons (MLP) that better fit with the non-linear systems [16]. The neural network
architecture used is given in figure 7. The neurons are connected in multiple layers and the signal goes from the input to
the output layers [17] and it has no interconnection between neurons of the same layer. This conguration is divided into
three layers: an input layer, an intermediate layer called hidden layer and an output layer.

Figure 7. Architecture of a multilayer perceptron

Design of the neuron control

We are going to follow an approach made by the following steps to design the network MLP artificial neurons.

First stage: The number of neurons per layer:

 the input layer: the current and voltage of the module, Ipv and Vpv respectively;
 hidden layers: the number of neurons in these layers is determined by an optimization study;
 the output layer: the panel power 𝑃𝑝𝑣

Second step: Creation of the database:

The objective of this step is to gather a sufficient number of representative data to create a database that will be
used for the learning and the validation of the neural network. In the literature, many researchers use the results
obtained with a conventional control type (INC, P&O, etc.) [13] [18] or the weather conditions (sunshine and
temperature) [19]. In our study we used the electrical characteristics (current-voltage and power) the panels A
and B to create the database that will be used for learning the neural control. The characteristics of the module
A are used for learning and those of module B for validation.

231
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Third step: Choice of the learning algorithm:

Learning is finding optimal synaptic weights to bring the neural network to a desired behavior using random
samples. As training technique we used supervised learning which consists to introduce input pairs (I PV and VPV)
and the desired output power (𝑃𝑝𝑣 ). We chose Adaptive Levenberg-Marquardt algorithm with the nonlinear
systems with a fast response time [20] and does not require a large memory and effective in the treatment of
nonlinear systems. We fixed an error equal to 10-7 to a number of approaches which is equal to 1000.
Step Four: Repeating the database:

This step is repeated by varying the number of hidden layer neuron until it gets a clear end state.

These steps are summarized in the given algorithm presented in figure 8.

Figure 8. Algorithm to design the ANN controller

Figure 9. Diagram of solar PV panel with the MPPT based ANN

232
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. SUMULATION RESULTS AND DISCUSSION

This study aims to optimize the power of two different solar panels at the Polytechnic higher school in Dakar
using an artificial neural networks. To evaluate the performance of our controller we made a comparative study
with two MPPT (INC and P & O). The used criteria to evaluate the performance are the Root Mean Square Error
(RMSE), the Mean Absolute Error (MAE) and the Mean Absolute Percentage Error (MAPE) whom expressions
are presented as Eqs. [2-4] [21] [22]:

𝑛
1 2
𝑅𝑀𝑆𝐸 = √ ∑ (𝑃𝑜𝑝𝑡 − 𝑃𝑝𝑣𝑟𝑒𝑓 ) (2)
𝑛
𝑖=1

𝑛
1 𝑃𝑜𝑝𝑡 − 𝑃𝑝𝑣𝑟𝑒𝑓
𝑀𝐴𝑃𝐸 = 100 ∗ ∑ | | (3)
𝑛 𝑃𝑝𝑣𝑟𝑒𝑓
𝑖=1

𝑛
1
𝑀𝐴𝐸 = ∑ |𝑃𝑜𝑝𝑡 − 𝑃𝑝𝑣𝑟𝑒𝑓 | (4)
𝑛
𝑖=1

With:

n is an entier number ;

𝑃𝑝𝑣𝑟𝑒𝑓 is the reference power of the PV panel;

𝑃𝑜𝑝𝑡 is the optimal power obtained by the MPPT controller.

In figure 10 we present the variation of the irradiation and the temperature versus the time for the solar
panels A and B.

Figure 10. Irradiance and temperature of PV panel A

Evolution of the electric power of the PV panels A and B versus de time is represented in figures 11 and 12,
respectively.

233
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 11. Electric power of PV panel A

Figure 12. Electric power of PV panel B

𝑃𝐴𝑁𝑁 is the power obtained by the ANN controller ;

𝑃𝐼𝑁𝐶 is the power obtained by the InC controller ;

𝑃𝑃&𝑂 is the power obtained by the P&O controller

In figures 11 and 12, we noted that the proposed methods show a good continuation of the maximum power
point for different weather variations. For a variation of sunshine from 700 to 900 W / m², as illustrated in figure
10, the power of the ANN controller is about 34 W and 27 W for solar panels A and B, respectively. All
techniques exhibit very good performance to track the MPP. ANN MPPT controller outperform the others The
MPP is tracked with high rapidity, accuracy and few oscillations. The statistics analysis like the RMSE, the
MAE and the MAPE for each controller are presented in figures 13 and 14 and we plotted the efficiencies of
each controllers in fugure 15 and 16.

Figure 13. Test of performance on panel A of the three MPPT controllers

234
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 14. Test of performance on panel B of the three MPPT controllers

Figure 15. Comparison of efficiencies on PV panel A of the three MPPT controllers

Figure 16. Comparison of efficiencies on panel B of the three MPPT controllers

In order to measure the performance of this controllers, statistical tests were made using the Root Mean Square
Error (RMSE), the Mean Absolute Error (MAE), the Mean Absolute Percentage Error (MAPE) and the solar
panels ’efficiencies. These criterias show the relationship between the electric power of the panel (𝑃𝑝𝑣𝑟𝑒𝑓 ) and
the predicted power of the controller (𝑃𝑜𝑝𝑡 ). A performance tests for each solar panels A and B with the three
considered MPPT controllers is made. These results show that the ANN control has a low RMSE and the MAE
and MAPE give low values compared to the INC command and P & O for both solar panels demonstrating its
stability and accuracy. MAPE helps us to verify if the predicting model results fit the results of the practical
design or not. Usually, smaller values for MAPE indicate a better fitting model. Therefore, the ANN achieves
lower values compared, to P&O and INC controllers figures 13 and 14. The efficiency comparison of ANN,
P&O and INC MPPT algorithm during dynamic variations of solar irradiation and temperature is done for each
panel. It can be observed that in the dynamic variation, ANN attains better efficiency than the P&O and INC
MPPT algorithm.

Figures 15 and 16 show the good performance of the ANN controller compared to the conventional control
which displays an efficiency of about 98% and 99% for the module A and B, respectively. ANN has several

235
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

advantages and it is a command whose modeling does not require the prior knowledge of the mathematical
model of the system contrary to the classical models. It also has a great adaptability especially for very random
atmospheric conditions. ANN, like all MPPT commands, has disadvantages. These are mainly related to learning
where to have a good performance and both a robust control, it requires a large database of data and is added to
that implementation Simulink.

4. CONCLUSION

This study presented an optimization of the electric power of two solar panels using the artificial neural networks
(ANN). We started with the implementation of the ANN controller in matlab using the electrical characteristics
of the first module for the learning and those of the second module for the validation. A comparative study of
the neural control with INC commands and P & O is done for each module. The simulation results showed the
good performance of the MPPT based ANN for each module. In perspective it would be interesting to do the
implementation on a test bench and do a comparative study with the simulation results.

REFERENCES

[1] A. R. Jordehi, «Maximum power point tracking in photovoltaic (PV) systems: A review of different approaches,»
Renewable and Sustainable Energy Reviews, vol. 65, p. 1127–1138, 2016.
[2] M. A. Ramli, S. Twaha et K. I. a. Y. A. Al-Turki, «A review on maximum power point tracking for photovoltaic
systems with and without shading conditions,» Renewable and Sustainable Energy Reviews, vol. 67, p. 144–159,
2017.
[3] S. Motahhir et A. E. H. a. A. E. Ghzizal, «The Most Used MPPT Algorithms: Review and the Suitable Low-cost
Embedded Board for Each Algorithm,» Journal of Cleaner Production, 2019.
[4] N. R. a. B. Rai, «Control of fuzzy logic based PV-battery hybrid system for stand-alone DC applications,» Journal
of Electrical Systems and Information Technology , vol. 5, p. 135–143, 2018.
[5] A. I. Ali, M. A. Sayed et a. E. E. Mohamed, «Modified efficient perturb and observe maximum power point
tracking technique for grid-tied PV system,» Electrical Power and Energy Systems, vol. 99, pp. 192-202, 2018.
[6] R. A. a. A. Jusoh, «An enhanced P&O checking algorithm MPPT for high tracking efficiency of partially shaded
PV module,» Solar Energy, vol. 163, pp. 570-580, 2018.
[7] S. Motahhir, A. Chalh et A. E. G. a. A. Derouich, «Development of a Low-cost PV System using an improved INC
algorithm and a PV panel Proteus model,» Journal of Cleaner Production, 2018.
[8] Z. Liying, M. Liang, L. Zhigang et C. M. a. W. Jianwen, «Implementation and simulation analysis of GMPPT
algorithm under partial shadow condition,» Energy procedia, vol. 158, pp. 418-423, 2019.
[9] S. F. a. S. Hosseini, «Optimization of fuzzy-based MPPT controller via metaheuristic techniques for stand-alone
PV systems,» international journal of hydrogen energy , vol. 44 , pp. 25457-25472, 2019.
[10] S. Meddour, D. Rahem, A. Y. Cherif et W. H. a. L. Hichem, «A novel approach for PV system based on
metaheuristic algorithm The 15th International Symposium on District Heating and Cooli,» Energy Procedia, vol.
162, pp. 57-66, 2019.
[11] E. h. m. Ndiaye, A. Ndiaye, M. A. Tankari et G. Lefebvre, «Adaptive Neuro-Fyzzy inferencesystem application for
the identification of a photovoltaic system and the forecasting of its maximum power point,» 7th international
conference on renewable energy research and applications, pp. 1061-1067, 2018.
[12] LoubnaBouselham, M. Hajji et B. H. a. H. Bouali, «A new MPPT-based ANN for photovoltaic system under partial
shading conditions,» Energy Procedia 924 – 933, vol. 111 , pp. 924-933, 2017 .
[13] «Improved method of maximum power point tracking of photovoltaic (PV) array using hybrid intelligent
controller,» Optik, vol. 168 , pp. 403-415, 2018.
[14] C. FAYE, «Changement climatiques observés sur le littoral sénégalais (Région de Dakar) depuis 1960 : Etude de la
variabilité des tendances sur les températures et la pluviométrie,» Nature & Technology, vol. 20, pp. 65-78, 2019.
[15] C. Vimalarani, N. Kamaraj et B. C. Babu, «Improved method of maximum power point tracking of photovoltaic
(PV) array using hybrid intelligent controller,» Optik, n° %1168, pp. 403-415, 2018.
[16] A. Ndiaye, L. Thiaw et G. Sow, «Application of a new modeling and control for grid connected photovoltaic
sytems based on artificial intelligence,» Journal of Electrical and Electronics Engineering Research, vol. 7, pp. 1-
10, 2014.
[17] L. L. Jiang, R. Srivatsan et D. L. Maskell, «Computational intelligence techniques for maximum power point

236
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

tracking in PV systems: A review,» Renewable and Sustainable Energy Reviews, Vol. 185, pp. 14-45, 2018.
[18] S. Messalti et A. H. a. A. Loukriz, «A new variable step size neural networks MPPT controller: Review, simulation
and hardware implementation,» Renewable and Sustainable Energy Reviews, vol. 68, pp. 221-233, 2017.
[19] M. K. a. E. M. Berkouk, «Artificial intelligence-based maximum power point tracking controllers for Photovoltaic
systems: Comparative study,» Renewable and Sustainable Energy Reviews, vol. 69, pp. 369-386, 2017.
[20] B. H. a. C. Ma, «A Shamanskii-like self-adaptive Levenberg–Marquardt method for nonlinear equations,»
Computers and Mathematics with Applications, vol. 77, pp. 357-373, 2019.
[21] J. H. Yousif, H. A. Kazem et N. N. A. a. I. I. Elhassan, «A comparison study based on artificial neural network for
assessing PV/T solar energy production,» Case Studies in Thermal Engineering 100407, vol. 13, pp. 100-407,
2019.
[22] H. A. K. a. J. H. Yousif, «Comparison of prediction methods of photovoltaic power system production using a
measured dataset,» Energy Conversion and Management, vol. 148 , p. 1070–1081, 2017.

237
 A NOVEL DEGREE-HOUR METHOD FOR RATIONAL DESIGN
LOADING

Andrii Radchenko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0002-8735-9205

Mykola Radchenko
10

Dariusz Mikielevicz
Gdansk University of Technology, Gdansk, Poland, [email protected], ORCID: 0000-0001-8267-7194

Roman Radchenko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0003-2211-3500

Andrii Andreev
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0003-3962-3101

Igor Esin
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, ORCID: 0000-0001-7646-300X

Radchenko, A, Radchenko, M, Mikielevicz, D, Radchenko, R, Andreev, A., Esin, I. A novel

Cite this paper as: Degree-Hour Method for rational design loading. 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: Cooling degree-hour (CDH) calculations are widely used while evaluating the efficiency of
application of ambient air cooling in power generation – engine intake air cooling (EIAC) systems
and in air condition (AC) systems of centralized and combined (outdoor and indoor) types in
climatic conditions of the targeted location. A majority of approaches for designing the EIAC
systems is based on the refrigeration capacity of the chiller selected to provide a maximum total
annual amount of CDH with corresponding maximum annual additional power produced or fuel
consumption reduction due to engine inlet air cooling expected in a particular climate location and
to cover the peak cooling load. Such approaches may lead to considerable oversizing the chillers
designed and the EIAC systems in the whole, including the system for heat removal from the
chillers with cooling towers (chiller condensers). A modified degree-hours calculation method is
developed to obtain more accurate design cooling load to avoid oversizing.
Keywords: ambient air, temperature depression, current thermal load, annual effect
© 2020 Published by ECRES

1. INTRODUCTION

Cooling degree-hour (CDH) calculations are widely used while evaluating the efficiency of application of
ambient air cooling in power generation – combustion engine intake air cooling (EIAC) systems [1, 2], as well
as in air conditioning (AC) systems [3, 4] in climatic conditions of the targeted location [5, 6]. So as the
performance efficiency of cooling systems needs to be analyzed across the full range of operating conditions,
the CDH number is used to determine the total amount ∑CDHs expected in a particular climate region over a
considered period (month, year, season). A commonly used analysis involves calculations of the annual number
∑CDH, while assigning an economic benefit to each CDH number in terms of fuel saved or engine output
increased due to cooling [7, 8]. This calculation gives the total annual reduction of fuel consumption or additional
power that can be produced due to engine inlet air cooling [9, 10] and takes into account the performance

238
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

characteristics of the engine applied.

All typical methods that include summarizing the number of CDH issue from the assumption of a design
capacity, sufficient to provide maximum possible cooling needs for all hours over the full range of yearly
operating conditions [11, 12]. Such approaches may lead to considerable oversizing the chillers designed and
the EIAC systems in the whole, that requires solving the real problem of determining the proper design cooling
load excluding oversizing.

2. LITERATURE REVIEW

Energy analyses of ambient air cooling needs the determination of cooling load profiles [11, 12]. Yearly
(monthly or seasonal) ambient temperature probability distribution is very important input data for energy
analyses and determining a total CDH and a design cooling load. The excellent studies on the input ambient data
when evaluating gas turbine inlet cooling are presented [12, 13]. A sinusoidal function to specify the ambient
temperature probability distribution were proposed [14, 15].
In reality, any estimation method, used in thermal technologies and issuing from annual (monthly, seasonal)
aggregative effect, based on the "hour-by-hour" or other time interval calculation procedure to consider the
actual (varying) site climatic conditions might be considered as such based on cooling/heating degree hour
numbers (C/HDHs). Such methods involve CDH calculations in terms of ambient air temperature depression
(on ambient dry-bulb [13-15], wet-bulb [1, 2] or both of temperatures [12] basis) to save energy due to engine
intake air cooling (EIAC) [16, 17] and operation at rational design cooling loads [18, 19] to match current
cooling demands according to actual climatic conditions and to provide maximum annual effect due to cooling
[20, 21] as well as thermal demand management (TDM) [22, 23] based on different criteria [24, 25] and
combined cooling, heat and power – trigeneration [26, 27], as well as in AC systems for building applications
[28, 29] including modern Variable Refrigerant Flow (VRF) systems [30, 31]. Some studies in thermal demand
management are devoted to development of methods of selecting the rational design cooling loads according to
current loading [26].
Although many scientists consider the (yearly) cumulative cooling degree hours values along with time elapsed
(yearly CDH profile) [12], only a few studies focus on analyzing the behavior of yearly cumulative cooling
profiles in dependence on loading to determine a design cooling load [17, 18, 21].
In power generation applications, input ambient cooling design points, such as those published by ASHRAE,
are frequently used [12]. These involve 0.4-2.0% exceedance values for design dry bulb temperatures, which
are accompanied by mean coincident wet bulb temperatures (DB/MCWB) [12]. In reality the same exceeding
design cooling capacities, rated according to maximum annual CDH numbers, were recommended.
The reason behind this can be found in converse changes in dry-bulb temperatures and relative humidity of
ambient air, that lead to replacing the value of cooling load capacity to cover the ambient temperature peak and
maximum value of thermal load. Proceedings from a deviation of cooling capacity, required to cover the
temperature peak, and the peak of thermal load, we can derive the practical hypothesis, that design cooling
capacities can be reduced compared to their values, corresponding to maximum annular CDH numbers, based
on dry-bulb temperatures, with simultaneous keeping the annular CDH numbers, closed to maximum value, but
keeping a sensible rate of annular CDH number increment to avoid oversizing [17, 18, 21]. Issuing from this the
CDH method was modified to define a design cooling capacity of chiller providing closed to maximum value
of annular CDH. In energetic application the cooling potential needed for cooling duties is to be compared with
its available value due to waste heat recovery, which can be increased by deep utilization of exhaust gas heat in
waist heat recovery boiler as primary source of heat for chillers [32, 33] and its effective conversion into cooling
[34].
The object is to provide an easy to use method of determining the rational design cooling capacity without
overestimation and to enable users to conduct appropriate designing of the overall system for adequate thermal
load.

3. RESEARCH METHODOLOGY

The efficiency of any cooling system performance depends on cooling needs and a duration of its yearly
operation. Therefore, the annual cooling potential ∑CDH is considered as a primary criterion for the choice of
a rational design load. The current degree-hour values CDH are calculated by summing up the differences
between the hourly ambient temperatures ta and a target leaving temperature of chilled air ta2 : Δt = ta – ta2 ,

239
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

multiplied by the corresponding hours lapsed in a year at a outdoor temperature of ta , cooled to a target value
ta2 : CDH = Δt τ, ºС∙h.
The summation of the current hour engine intake air cooling potential for a certain time period (day, month,
year), i.e. CDH = Σ(Δt τ), ºС∙h, gives a total intake air cooling potential (for day, month and year).
The total CDH is arrived at by summing the CDHs derived for the 12 months at a location.
The authors develop a novel method of rational designing based on the annular cooling potential ∑CDH =
Σ(Δt τ), ºС∙h, cumulative characteristic curve dependence on the design specific cooling capacity to choose its
precise rational value, that provides closed to maximum annual cooling potential ∑CDH without oversizing a
chiller. It was done by adding to well known methods [12, 16, 17] some stages aimed to determine a more precise
value of rational design cooling capacity.
In order to generalize the results and simplify calculations for any total cooling capacity Q0 , it is convenient to
present it in relative (specific) value per unit air flow rate (Ga = 1 kg/s) – in the form of specific refrigeration
capacity: q0 = Q0 /Ga , kW/(kg/s), or kJ/kg, where Q0 is the total refrigeration capacity when cooling the air with
the flow rate Ga: Q0 = сa ξ∙Δta ∙Ga , where Δta = ta – ta2 – decrease (depression) of ambient air temperature; ta2 –
target air temperature at the air cooler outlet; ξ – specific heat ratio of total (full) heat, inclooding sensible and
latent heat, to sensible heat of humid air; сma –specific heat of humid (moist) air [kJ/(kg·K)].
There are two addition stages added to calculate an optimal specific cooling capacity q0.opt , providing a
maximum rate of annual cooling potential ∑CDH/q0 and corresponding annual cooling potential ∑CDHopt , as
the second stage of calculation procedure, and determining a precise value of rational specific cooling capacity
q0.rat , providing the annual cooling potential ∑CDHrat closed to its maximum value, but without unproductive
expenses of cooling capacity q0 caused by oversizing chiller without obtaining a sensible increasing the annual
cooling potential ∑CDH, as the third stage. With this a rational specific cooling capacity q0.rat is associated with
second maximum rate of annual cooling potential ∑CDH increment beyond the optimal annual cooling potential
∑CDHopt : (∑CDH – ∑CDHopt )/q0 , where ∑CDH>∑CDHopt .

4. RESULTS

A rational design cooling capacity may be determined from the annual ∑CDH calculated by summing up the
differences between the hourly ambient temperatures and a target cooled air temperature Δta = ta – ta2 , multiplied
by the hours τ lapsed at ambient temperature ta and target air temperatures ta2 = 10 and 15°С (Fig. 1).

ΣCDH, C·h
50000
∑CDH10max
∑CDH10rat Δq0.10
40000
∑CDH10
30000
∑CDH15max
∑CDH15rat
20000
Δq0.15
∑CDH15
10000

0 q0.15rat q0.15max q0.10rat q0.10max

0 5 10 15 20 25 30 35 q0, kW/(kg/s)
40 45
Figure 1. Annual ∑CDH number of cooling ambient air to ta2 = 10 and 15 °С versus design specific cooling capacity q0

As Fig.1 shows, a design specific cooling capacity q0.10 = 33-35 kW/(kg/s) of cooling ambient air from the
current temperature ta to ta2 = 10 °C provides annual cooling potential ∑CDH10 = 44000-45000 °C∙h closed to a
maximum value 46000 °C∙h within sensible high rate of its increment.
While it is evidently seen that the rate of increment of annual CDHs after the value of ∑CDH10 = 44000-45000
°C∙h is negligible, the range of corresponding cooling capacities q0.10>33-35 kW/(kg/s) needed is still wide and

240
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

therefore so will be the range of oversizing. Issuing from this the rational values of cooling capacities are
adequate need to be determined.

It is evident, that it is problematic to select quite precise values of rational specific cooling capacity q0.15rat so as
behaviour of characteristic curve ∑CDH = f(q0) becomes more gentle (Fig. 1). To solve this task the basic method
of determining a design cooling capacity has been improved by adding addition stages aimed for calculation of
optimal design cooling capacity q0.opt, corresponding to a maximum rate of annual ∑CDH number increment:
∑CDH/q0∙(Fig.2), and precise value of rational cooling capacity q0.rat (Fig.3 and 4).

ΣCDH/q0, C·h/kW/(kg/s) ΣCDH, C·h ΣCDH/q0, C·h/kW/(kg/s) ΣCDH, C·h

1200 36000 1600 48000
t2=15 C t2=10 C
1000 30000 1400 42000
ΣCDH/q0.15 ∑CDH10opt
1200 ΣCDH/q0.10 36000
800 24000 1000 30000
∑CDH15opt
600 18000 800 24000
ΣCDH15
400 12000 600 18000
ΣCDH10
400 12000
200 6000 200 6000
0 q0.15opt 0 0 q0.10opt 0
0 5 10 15 20 25 30q0, kW/(kg/s)
35 0 5 10 15 20 25 30 35 40 q45 0, kW/(kg/s)
50
(a) (b)

Figure 2. Relative increment of annual cooling potential ∑CDH/q0 and annual cooling potential ∑CDH of cooling
ambient air to ta2 = 15°C (a) and 10 °C (b) versus design specific cooling capacity q0 : q0.opt and ∑CDHopt – optimal
values

As Fig. 2 shows, a maximum rate of annual cooling potential increment ∑CDH/q0.15 of cooling air to ta2 = 15 °C
takes place at the optimal design cooling capacity q0.15opt=16 kW/(kg/s) and provides covering the annual cooling
potential ∑CDH15opt = 20000 °C∙h significantly less than its maximum value 24000 °C∙h. For cooling air to
ta2 = 10 °C the optimal cooling capacity q0.10opt=27 kW/(kg/s) provides the ∑CDH10opt = 39000 °C∙h.
To achieve closed to maximum annual cooling potential ∑CDH at more precise value of rational cooling
capacity q0.rat without oversizing the chiller, it is proposed to determine the maximum rate of annual cooling
potential increment as (∑CDH– ∑CDHopt )/q0.15 within the range of ∑CDH15 beyond its value ∑CDH15opt =
20000 °C∙h corresponding to optimal design capacity q0.15opt=16 kW/(kg/s) for cooling air to ta2 = 15 °C as well
as beyond ∑CDH10opt = 39000 °C∙h corresponding to q0.10opt=27 kW/(kg/s) for ta2 = 10 °C (Fig. 3 and 4).

ΣCDH/q0,10(ΣCDH-ΣCDHopt)/q0, C·h/kW/(kg/s) ΣCDH/q0,10(ΣCDH-ΣCDHopt)/q0, C·h/kW/(kg/s)

1800 1600
t2=15 C t2=10 C
1600 10(ΣCDH-ΣCDHopt)/q0 1400
1400 1200
1200 ΣCDH/q0
ΣCDH/q0 1000
1000
800
800
600
600 10(ΣCDH-ΣCDHopt)/q0
400 400
200 200
0 q0.15opt q0.15rat 0 q0.10opt q0.10rat
0 5 10 15 20 25 q0,30
kW/(kg/s)
35 0 5 10 15 20 25 30 q0, kW/(kg/s)
35 40 45

(a) (b)
Figure 3. Relative increments of annual cooling potential ∑CDH/q0 and (∑CDH– ∑CDHopt )/q0∙within the range beyond
the optimal value ∑CDHopt (Fig.3) of cooling ambient air to ta2 = 15°C (a) and 10 °C (b) versus design specific cooling
capacity q0: q0.opt– optimal; q0.rat– rational values

241
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

10(ΣCDH-ΣCDHopt)/q0, C·h/kW/(kg/s) ΣCDH, C·h 10(ΣCDH-ΣCDHopt)/q0, C·h/kW/(kg/s) ΣCDH, C·h

1800 36000 1400 49000
t2=15 C t2=10 C ΣCDH10
1600 32000 1200 ∑CDH10rat 42000
10(ΣCDH-ΣCDHopt)/q0.15
1400 28000 10(ΣCDH-ΣCDHopt)/q0.10
1000 35000
1200 ΣCDH15 ∑CDH15rat 24000
1000 20000 800 28000
800 16000 600 21000
600 12000
400 14000
400 8000
200 4000 200 7000
0 q0.15rat 0 0 q0.10rat 0
0 5 10 15 20 25 30 q35 0, kW/(kg/s) 0 5 10 15 20 25 30 35 40 q450, kW/(kg/s)
50
(a) (b)
Figure 4. Annual cooling potential ∑CDH and relative increment of annual cooling potential (∑CDH–
∑CDHopt )/q0∙within the range beyond the optimal value ∑CDHopt (Fig.3) corresponding to cooling ambient air to t a2 =
15°C (a) and 10 °C (b) versus design specific cooling capacity q0: q0.rat and ∑CDHrat – rational values

As Fig. 3 and 4 shows, a maximum rate of annual cooling potential ∑CDH increment (∑CDH–∑CDHopt )/q0.15
for cooling ambient air to ta2 = 15 °C takes place at the rational design cooling capacity q0.15rat=24 kW/(kg/s) and
provides annual cooling potential ∑CDH15rat = 24500 °C∙h that is closed to its maximum value
∑CDH15max =25000 °C∙h and achieved at a refrigeration capacity q0.15rat=24 kW/(kg/s) more than 20 % less than
its oversized value q0.15max = 31 kW/(kg/s). For cooling air to ta2 = 10 °C a maximum of (∑CDH–∑CDHopt )/q0.10
takes place at the rational cooling capacity q0.10rat=37 kW/(kg/s) and provides annual potential ∑CDH10rat =
45500 °C∙h that is closed to its maximum value ∑CDH15max =46000 °C∙h and achieved at a refrigeration capacity
q0.10rat=37 kW/(kg/s) more than 15 % less than its oversized value q0.10max = 44 kW/(kg/s).
Graphs in Fig. 1-5 allow users to match cooling and provide minimum sizing system while maximum rate of
increment of annular cooling effect through operation at optimal cooling capacity q0.10opt (Fig. 2) or to peak
maximum cooling needs without oversizing due to running at a rational cooling capacity q0.rat (Fig. 5).
Thus, we have come to conclusion concerning selection of rational refrigeration capacity values lower than that,
corresponding to maximum annual CDH numbers, i. e. more less than that, published by ASHRAE [12].

5. CONCLUSION

A new method allow to match cooling and provide minimum sizing system while maximum rate of increment
of annular cooling effect through operation at optimal cooling capacity q0.10opt or to peak practically maximum
cooling needs without oversizing system due to running it at a rational value of cooling capacity q0.rat.
The proposed method is quite quantitative to evaluate the effect due to cooling and select the rational design
cooling capacity and estimate a decrease in sizes of chillers.
A modified degree-hour calculation method is aimed to evaluate the effect due to cooling and obtain more
accurate design cooling capacity as important input data for sizing chillers and the overall cooling system. Its
application reduces the efforts, needed to make an economic analysis of cooling systems.
The proposed methodology is quite useful to be extended for application in energy saving technologies of engine
intake air cooling with thermal storage due to using the optimal design cooling capacity q0.opt, providing a
maximum rate of annual cooling potential increment ∑CDH/q0 .

REFERENCES

[1] Chaker, M., Meher-Homji, C.B., Mee, T, Nicolson, A. Inlet Fogging of Gas Turbine Engines- Detailed Climatic
Analysis of Gas Turbine Evaporative Cooling Potential. In: ASME International Gas Turbine and Aeroengine
Conference; 4–7 June 2001: New Orleans, USA, ASME Paper No. 2001-GT-526.
[2] Chaker, M., Meher-Homji, C.B. Inlet fogging of gas turbine engines: climatic analysis of gas turbine evaporative cooling
potential of international locations. In: Proceedings of ASME TURBO EXPO; 3–6 June 2002: Amsterdam, The
Netherlands, GT-2002-30559.
[3] Satman, A, Yalcinkaya, N. Heating and cooling degree-hours for Turkey. Energy 1999, 24 (10), 833–840.
[4] Pegues, J. The benefits of 8760 hour-by-hour building energy analysis. Carrier Software Systems, New York, USA:
HVAC Systems Engineer, 2002.
[5] Duryamaz, A, Kadıoglu, M, Sen, Z. An application of the degree-hours method to estimate the residential heating energy
requirement and fuel consumption in Istanbul. Energy 2000, 25, 1245–1256.

242
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[6] Oktay, Z, Coskun, C, Dincer, I. A new approach for predicting cooling degree-hours and energy requirements in
buildings. Energy 2011, 36 (8), 4855–4863.
[7] Farouk, N., Sheng L., Hayat, Q. Effect of Ambient Temperature on the Performance of Gas Turbines Power Plant.
International Journal of Computer Science 2013, 10, 439–442.
[8] Tiwari A., K., Hasan, M., Muzaffarul, M. I., Effect of ambient temperature on the performance of a combined cycle
power plant. Transactions of the Canadian Society for Mechanical Engineering 2013, 37 (4), 1177–1188.
[9] Zhang, T., Liu, Z., Hao, H., Chang, L.J. Application research of intake-air cooling technologies in gas-steam combined
cycle power plants in China. Power Energy Eng. 2014, 2, 304–311.
[10] Günnur, Şen G., et. al. The effect of ambient temperature on electric power generation in natural gas combined cycle
power plant-A case study. Energy Reports 2018, 4, 682–690.
[11] Komuro T., Ito, E., Sonoda, T., Tomita, Y., Hidaka, K., Shibutani, S. Power Output Augmentation of Gas Turbine
Combined Cycle by Inlet-Air Cooling System of Chiller Type under High Ambient Air Temperature. Mitsubishi Heavy
Industries Technical Review 2010, 47 (4), 33–39.
[12] Forsyth, J.L. Gas turbine inlet air chilling for LNG. In: IGT International Liquefied Natural Gas Conference
Proceedings, 3; 2013, pp. 1763–1778.
[13] Coskun, C., Demiral, D., Ertürk, M., Oktay, Z. Modified Degree-Hour Calculation Method. In Prof. R. Rugescu (Ed.),
Solar Power, InTech, 2012, pp. 55–62.
[14] Coskun, C. A novel approach to degree-hour calculation: Indoor and outdoor reference temperature based degree-hour
calculation. Energy 2010, 35, 2455–2460.
[15] Oktay, Z, Coskun, C, Dincer, I. A new approach for predicting cooling degree-hours and energy requirements in
buildings. Energy 2011, 36 (8), 4855–4863.
[16] Radchenko, R., Radchenko, A., Serbin, S., Kantor, S., Portnoi, B. Gas Turbine Unite Inlet Air Cooling by Using an
Excessive Refrigeration Capacity of Absorption-Ejector Chiller in Booster Air Cooler. In: HTRSE-2018; 2018: E3S
Web of Conferences 70 (03012), 6 p.
[17] Radchenko, A., Bohdal, L., Zongming, Y., Portnoi, B., Tkachenko, V. Rational designing of gas turbine inlet air
cooling system. In: Tonkonogyi V. et al. (eds.) Grabchenko’s International Conference on Advanced Manufacturing
Processes. InterPartner-2019. Lecture Notes in Mechanical Engineering, Springer, Cham; 2020, pp. 591-599.
[18] Radchenko, A., Radchenko, M., Trushliakov, E., Kantor, S., Tkachenko, V. Statistical method to define rational heat
loads on railway air conditioning system for changeable climatic conditions. In: 5th International Conference on
Systems and Informatics. ICSAI 2018; 2018: Jiangsu, Nanjing, China, pp. 1308–1312.
[19] Radchenko, M., Radchenko, R., Tkachenko, V., Kantor, S., Smolyanoy, E. Increasing the Operation Efficiency of
Railway Air Conditioning System on the Base of Its Simulation Along the Route Line. In: Nechyporuk M. et al. (eds)
Integrated Computer Technologies in Mechanical Engineering. Advances in Intelligent Systems and Computing,
Springer, Cham 1113; 2020, pp. 461–467.
[20] Trushliakov, E., Radchenko, M., Bohdal, T., Radchenko, R., Kantor, S. An innovative air conditioning system for
changeable heat loads. In: Tonkonogyi V. et al. (eds.) Grabchenko’s International Conference on Advanced
Manufacturing Processes. InterPartner-2019. Lecture Notes in Mechanical Engineering, Springer, Cham; 2020, pp.
616–625.
[21] Radchenko, A., Radchenko, M., Konovalov, A., Zubarev, A. Increasing electrical power out-put and fuel efficiency
of gas engines in integrated energy system by absorption chiller scavenge air cooling on the base of monitoring data
treatment. In: HTRSE-2018; 2018: E3S Web of Conferences70 (03011), 6 p.
[22] Canova, A., Cavallero, C., Freschi, F., Giaccone, L., Repetto, M., Tartaglia, M. Optimal energy management. IEEE
Industry Applications Magazine 2009, 15, 62–65
[23] Suamir, I. N., Tassou, S. A. Performance evaluation of integrated trigeneration and CO2 refrigeration systems. Applied
Thermal Engineering 2013, 50 (2), 1487–1495.
[24] Rocha, M.S., Andreos, R., Simões-Moreira, J.R. Performance tests of two small trigeneration pilot plants. Applied
Thermal Engineering 2012, 41, 84–91.
[25] Freschi, F., Giaccone, L., Lazzeroni, P., Repetto, M. Economic and environmental analysis of a trigeneration system
for food-industry: A case study. Applied Energy 2013, 107, 157–172.
[26] Cardona, E., Piacentino, A. A methodology for sizing a trigeneration plant in mediterranean areas. Applied Thermal
Engineering 2003, 23 (13), 1665–1680.
[27] Rodriguez-Aumente, P.A., Rodriguez-Hidalgo, M. C., Nogueira, J. I., Lecuona, A., Venegas, M. C. District heating
and cooling for business buildings in Madrid. Applied Thermal Engineering 2013, 50 (2), 1496–1503.
[28] Marques, R. P., Hacon, D, Tessarollo, A., Parise, J.A.R. Thermodynamic analysis of trigeneration systems taking into
account refrigeration, heating and electricity load demands. Energy and Buildings 2010, 42 (12), 2323–2330
[29] Ortiga, J. Bruno, J.C., Coronas, A. Operational optimization of a complex trigeneration system connected to a district
heating and cooling network. Applied Thermal Engineering 2013, 50 (2), 1536–1542.
[30] Im, P., Malhotra, M., Munk, J.D., Lee, J. Cooling Season Full and Part Load Performance Evaluation of Variable
Refrigerant Flow (VRF) System Using an Occupancy Simulated Research Building. In Proceedings of the 16th
International Refrigeration and Air Conditioning Conference at Purdue; July 11–14 2016: West Lafayette, USA.
[31] Lee, J. H., Yoon, H. J., Im, P., Song, Y.–H. Verification of Energy Reduction Effect Through Control Optimization of

243
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Supply Air Temperature in VRF-OAP System. Energies 2018, 11 (1).

[32] Kornienko, V., Radchenko, R., Stachel, A., Andreev A., Pyrysunko, M. Correlations for pollution on condensing
surfaces of exhaust gas boilers with water-fuel emulsion combustion. In: Tonkonogyi V. et al. (eds.) Grabchenko’s
International Conference on Advanced Manufacturing Processes. InterPartner-2019. Lecture Notes in Mechanical
Engineering, Springer, Cham; 2020, pp. 530-539.
[33] Radchenko, R., Kornienko, V., Pyrysunko, M., Bogdanov, M., Andreev, A. Enhancing the Efficiency of Marine Diesel
Engine by Deep Waste Heat Recovery on the Base of Its Simulation Along the Route Line. In: Nechyporuk M. et al.
(eds) Integrated Computer Technologies in Mechanical Engineering. Advances in Intelligent Systems and Computing,
Springer, Cham 1113; 2020, pp. 337-350.
[34] Radchenko, M., Radchenko, R., Ostapenko, O., Zubarev, A., Hrych, A. Enhancing the utilization of gas engine module
exhaust heat by two-stage chillers for combined electricity, heat and refrigeration. In: 5th International Conference on
Systems and Informatics, ICSAI 2018; 2018: Jiangsu, Nanjing, China, pp. 227–231.
[35] Konovalov, D., Kobalava H. Efficiency Analysis of Gas Turbine Plant Cycles with Water Injection by the
Aerothermopressor. In: Ivanov V. et al. (eds) Advances in Design, Simulation and Manufacturing II. DSMIE 2019.
Lecture Notes in Mechanical Engineering, Springer, Cham. 20; 2020: pp. 581-591.

244
 THE RECOVERY OF HISTORIC CENTRES: A MULTIPLE
STRATEGY FOR THE ENERGY SAVINGS

Losco Giuseppe
University of Camerino, Italy, [email protected], ORCID: 0000-0002-3260-236X

Pierleoni Andrea
University of Camerino, Italy, [email protected], ORCID: 0000-0002-7902-0068

Roncaccia Elisa
University of Camerino, Italy, [email protected], ORCID: 0000-0002-4982-5850

Losco, G, Pierleoni, A, Roncaccia, E. The recovery of historic centres: a multiple strategy for
Cite this paper as:
the energy savings. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Sustainable criteria and comfort objectives are nowadays the basis of any design process that seeks
to rise the liveability of an urban context and the opportunity of use of its built heritage. This study
starts by considering the same criteria and objectives for the reconstruction phases of those small
historical cities severely hit by the earthquake: case study is Arquata del Tronto, located in central
Italy and affected by the seismic activity starting in 2016. The study we propose illustrates a
methodology for achieving the energy efficiency from a multilevel perspective, identifying three
different dimensions -the territory, the urban context and the building-. For each dimension we
move from a cognitive framework to a phase of planning, providing a series of operative
indications. The strategy is not to consider the possible energy solutions independently from each
other, but to think about an integration of solutions, by combining the retrofit interventions on
buildings with the use of various type of renewable energies. In this regard, the understanding of
the territorial resources becomes necessary to suggest reconstructive solutions consistent with the
characteristics of the place, while the knowledge about the features of the buildings allows to
carefully evaluate the possible impact of the interventions, so to identify the most appropriate
energy strategies in respect of the historical value.
The study therefore outlines how to supply all the energy demand of the urban center with only
retrofit intervention and renewable sources, so to minimise the need of fossil energy. It provides a
comparison between the investment cost required to perform all the interventions planned and the
total energy savings over the years, so to calculate the amount of time necessary to the full return
of investments.
In conclusion, the project provides a multi-scale guideline text usable for supporting the
reconstruction process and, in general, the recovery of the built heritage, orienting the design
activities towards the sustainability requirements and the energy efficiency.
Keywords: Recovery, guidelines, integrated strategies, energy efficiency
© 2020 Published by ECRES

1. INTRODUCTION

The energy efficiency of buildings has undergone an important evolution from a regulatory point of view, which
was also matched by a satisfactory response from all the private players who invested in energy savings – thanks
to political-economic strategies such as the 65% ecobonus. The achievement of the objectives of the European
Action Plan for Energy Efficiency of 2011 (known as "Europe 20-20-20") is still far away though and by 2030
further targets are expected to be pressing with the scenarios developed at EU level, such as the Clean Energy
Package. The most important challenge that the Italian building sector will have to set in order to get closer to
these objectives is to intervene on the existing public and private building heritage. In fact, around 60% of the
built fabric of the country is made up of buildings characterised by the lowest energy efficiency class (G) and
the 88% of Italian buildings are prior to 1991, when the first framework law came into force regarding the
rational use of energy, energy saving and the development of renewable sources. That is why it is so important
to consider the issue of energy improvement in the minor historic centres that widely characterise the national

245
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

territory and on which the freedom of design is constrained by the need to preserve the original technical and
architectural features. The current study arises from dramatic circumstances such us the earthquakes that struck
Central Italy from 24 August 2016, which become the occasion to try out an operational mode to promote a
reconstruction oriented to sustainability and energy efficiency criteria. An historical centre strongly affected by
the earthquake is Arquata del Tronto, whose municipality commissioned the present research to the School of
Architecture and Design of Ascoli Piceno. The aim was to provide public administrations and designers involved
in the reconstruction phase with guidelines to suggest technological solutions for the energy improvement from
the territorial scale to the single building.
The actual shape of an historic centre, characterised by several construction techniques and an irregular
distribution of closed and open spaces, is the result of a series of successive transformations over the centuries.
It is therefore the understanding of these transformations that allows to identify the historical values and the
identity of the built areas, providing real indications on the elements needed to promote an active conservation
of the urban-fabric. At the same time, in order to improve liveability and environmental comfort conditions, it
is necessary to study the climatic context in which the historic center is located. This research starts from a
climatic study of the territory of Arquata del Tronto and then goes on to provide the operational suggestions in
order to support the project activities, both on an urban and building scale.

1. CLIMATE ANALYSIS

The bioclimatic analysis of the Arquata del Tronto area started with the assessment of the parameters of
sunshine, radiation, ventilation, humidity and consistency of precipitation, carried out in both summer and winter
scenarios.
Combining the values obtained, it is possible to identify the critical environmental zones and the areas which
are potentially predisposed to the introduction of renewable energy systems, through which we can maximise
the comfort perceived by the inhabitants and minimise the use of fossil energy.
The case of Arquata del Tronto, whose urban context is notably widespread and fragmented within an
altimetrically articulated territory, has required to adopt a cognitive path divided into successive levels of in-
depth analysis. The research was therefore structured on different phases: the territorial climate, the urban
climate and the climatic analysis on the scale of a single building.

1.1. Territorial climate analysis

Arquata del Tronto is the last municipality in the province of Ascoli Piceno, on the border between Abruzzo and
Lazio. It extends for 92 km square in a mountainous area and is the only municipality in Europe to belong to
two national parks: the Monti Sibillini National Park and the Gran Sasso and Monti della Laga National Park.
The territory is crossed by the Tronto river which runs along the flat area where the narrow and steep valleys of
the northern and southern slopes converge. According to the Decree of the President of the Republic n. 412 of
26 August 1993, which establishes a subdivision of the Italian territory into climatic zones, the municipality of
Arquata del Tronto falls within the climatic zone E.

Figure 1. View of the land of Arquata del Tronto with its 14 hamlets
1.FORCA CANAPINE – 2.CAPODACQUA – 3.TUFO – 4.PESCARA DEL TRONTO – 5. VEZZANO – 6.SPELONGA – 7.FAETE 8.TRISUNGO – 9.ARQUATA CAPOLUOGO – 10.BORGO
– 11.CAMARTINA – 12.PIEDILAMA – 13.PRETARE – 14.COLLE

The territorial location data (latitude, longitude and height above the sea level) were entered into the Meteonorm
7 software, using the climatic data reported in the UNI10349 standard. It was possible to obtain the files about

246
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the climatic data that characterise the 14 different locations in the municipal area of Arquata del Tronto
(temperature, humidity, wind speed and direction, solar radiation). Using an additional software, Weather Tool,
the outcomes have been reprocessed and it has been possible to obtain graphs and to conduct comparative
evaluations which are useful for a more comprehensive understanding of the case study.
It was found, for example, that directionality and frequency of the wind are common to almost all the localities,
with some differences which are more evident as the speed increases – since Spelonga, Colle and Forca Canapine
differ from the other fractions. According to the parameters of temperature and humidity, Trisungo and Tufo,
situated in the valley, differ from the centres at higher altitudes, like Colle and Forca Canapine. Focusing on
rainy or snowy perturbations, we observed their distribution over time: it was found that the villages in the valley
and those south of the Tronto river receive more precipitations, while those of medium height north of the river
are less rainy.
This first general climatic analysis allowed to organise the localities of Arquata del Tronto in homogeneous
climatic zones, grouping them according to similar climatic and geomorphological conditions.

Figure 2. Homogeneous climatic zones grouping the 14 hamlets of Arquata del Tronto according with climatic and
geomorphological similarities.

The territory is in fact characterised by multiple exposures and environmental situations, since there is a short
distance between the valley floor and the mountain ridges, passing from an altitude of 600 m (Trisungo) to 1500
m (Forca Canapine). Five distinct climate zones were outlined (Fig.2):
- Zone 1. Villages on the valley floor and on the first hills (Trisungo, Vezzano, Pescara del Tronto, Tufo, Arquata e
Capodacqua)
- Zone 2. Piedmont villages north of the Tronto river (Pretare, Piedilama, Camartina e Borgo)
- Zone 3. Piedmont villages south of the Tronto river (Faete e Spelonga)
- Zone 4. Mountain villages north of the Tronto river (Forca Canapine)
- Zone 5. Mountain villages south of the Tronto river (Colle)

1.2. Urban climate analysis

After identifying the climatic zones, we assumed some fractions as representative case studies for the three
central homogeneous zones of the territory: Trisungo for the zone 1, Arquata Capoluogo for the zone 2 and
Spelonga for the zone 3.
Under the assumption that the buildings will be reconstructed preserving the original volumes, the climatic study
was deepened, aiming to obtain indications strictly related to the relationship between soil and buildings. A fluid

247
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

dynamics analysis was then carried out using the ENVI-met software4 (the results are described in sub-chapter
3.1) while the climatic data relating to solar geometries and psychrometric diagrams were elaborated.
On the three localities assumed as case studies, we simulated the distribution of the shadows and the lighting of
the building fronts. Since they are located in three distinct areas, the fractions differ in exposure. In the mountain
villages, a mountain or a woodland area can be a barrier to the winds but also an obstacle for a correct daytime
lighting. The 3D solar path was then simulated through the Ecotect software. It allows to provide a study of
daylight during standardised periods of the day and year: morning, noon and afternoon on the days of summer
solstice (21 June), winter solstice (21 December) and equinox (21 March). Once the simulations were completed,
critical situations have been noticed: persistent shadow in winter and continuous sunlight in summer. After
evaluating the degree of illumination of the fronts in the different locations it was possible to quantify the average
duration with which the fronts of the buildings are hit by the sun's rays.
According to the calculated sunshine we can consider four types of building fronts:
- Fronts with less than 1 hour of sunshine
- Fronts with 1÷4 hours of sunshine
- Fronts with 4÷8 hours of sunshine
- Fronts with more than 8 hours of sunshine

Observing the maps obtained, we note that in the village of Arquata, located on the hill, the north-facing parts
present a critical situation, because they are significantly shady, especially in winter (Fig.3).

For Arquata del Tronto we wanted to carry out a study on the pre-earthquake situation in order to understand
the critical points of the urban-fabric currently destroyed and to provide useful suggestions for a possible
reconstruction. In fact, some fronts could be lowered to allow the illumination of the secondary streets, in order
to improve internal comfort both from a luminous and thermal point of view. In general, considering the
reconstruction or reconfiguration of the existing villages, a study of this type is useful to reorganise the
distribution of interior spaces by dedicating more illuminated areas to day use destinations.
The study of the sunshine follows the reading of the psychrometric chart. Normally the psychrometric diagram
is used during the design phase of a building but the same can also be used for the assessment of the thermo-
hygrometric comfort of an existing building.

Figure 3. Map regarding the sun exposure of the building fronts. Arquata Capoluogo. 31 December edifici. Arquata
Capoluogo – 31 december

It starts with the construction of the psychrometric chart with reference to the climatic characteristics of the
place. Everything within it could in a first instance determine a specific standard comfort area (pink line) and
observe the graph offering us a rate of reasoning on intervention strategies that contribute to widening the initial
comfort area. As can be seen from Fig.4, the best and most effective intervention strategies that prolong comfort
in both winter and summer are of four types:

4
ENVI-met is a three-dimensional, non-hydrostatic and microclimatic model, able to provide simulations with a spatial
resolution of 0.5÷10 meters and a temporal resolution of 10 seconds. The software is based on the fundamental laws of
fluid dynamics and thermodynamics.

248
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

- Passive solar heating, interventions that provide for the heating of the premises through passive solar
systems such as solar greenhouses, large windows, or natural air heating systems;
- Thermal mass effects, interventions in which we tend to use building materials that store the heat
produced by the sun rays that hit the building and slowly release it over time;
- Exposed mass and night purge ventilation, the most used strategy for the summer period with which it’s
possible to cool in a building using the insertion of opposing openings of the casing, so as to move from
a cooler area to a warmer one. Both for day and night time.
- Natural ventilation, it can be used for passive cooling systems, called dissipative, which requires
considerations with respect to the characteristics of the local winds and the vertical distribution of the
communicating internal environments.

Figure 4: Example of psychrometric chart. Arquata Capoluogo – Annual scenario

In case of restructuring operations, it is possible to reduce the energy consumption for heating or cooling by
extending the internal comfort area without focusing on plants, but using passive technologies (solar
greenhouses, solar air collectors, exploitation of the thermal mass and optimal arrangement of the openings so
as to guarantee natural internal ventilation). The analysis of microclimatic aspects on an urban scale was
performed using the ENVI-met software, capable of simulating surface-plant-air interactions in an urban
environment. ENVI-met is particularly useful in assessing the impact of new urban-design interventions,
because it provides numerical forecasts on the environmental impact of new structures or green areas. It also
simulates the flow of wind between buildings, heat exchange processes and steam at ground and wall level,
atmospheric turbulence, some vegetation parameters and bioclimatology. The parameters calculated by the
software are:

1. Atmosphere: wind, temperature, steam, turbulence, pollution; soil - temperature, water flow, water
concentrations;
2. Vegetation: leaf temperature, heat exchange, steam exchange, water transport, water interception;
3. Surfaces: ground level flows, wall and roof flows, heat transfer through the walls;
4. Biometeorology: PMV (Predicted Mean Vote) value.

Figure 5. Example of graphic representation for critical areas: PMV (Predicted Mean Vote) and PPD (Predicted
Percentage of Dissatisfied) parameters. Arquata Capoluogo – Summer scenario

249
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In particular, the mapping of the PMV and PPD (Predicted Percentage of Dissatisfied) parameters allows to
identify the most critical areas of the urban context. The maps show, as an example, the criticalities outlined
through the fluid-dynamic analysis conducted on the Arquata Capoluogo case study (Fig.5).

1.3. Climate analysis of the building

This last level of in-depth analysis in the thermal analysis of the building can be achieved both instrumentally
(thermography, thermoflowmetry - Fig.6), and analytically, through the knowledge of the construction
characteristics that characterize the building envelope and the systems. In particular, to have a reference picture
of the energy-environmental performance of the building, it’s necessary to consider further factors such as
energy consumption, thermal insulation, attenuation and phase shift, air permeability, passive solar thermal
loads.

Fig. 6: Example of a thermographic survey

The data obtained from the analysis of the energy-environmental behavior are to be considered together with
those deriving from the environmental analysis, in order to choose which type of intervention to operate. In
addition to the analysis of the energy-environmental behavior of the building, it is essential to proceed with the
analysis of the "building-plant system", through which the specific technical-construction and technological
characteristics of the building envelope and of the energy supply systems are identified. This will make it
possible to evaluate the possibility of maintaining its presence in the new project solution, to evaluate their
integrability, or to declare its age and therefore its ineffectiveness with respect to new energy performances.

2. WORKING SUGGESTIONS
If the current legislative framework requires an unconditional adjustment to the regulatory obligations,
undoubtedly the pursuit of energy efficiency objectives for the architectural heritage needs to be regulated by a
careful evaluation of the possible impact that the interventions would entail on historical buildings: any type of
intervention, if not calibrated through conservative criteria, could lead to a transformation and to the distortion
of the buildings and the areas where they are located.
The present research therefore intends to introduce, in the reconstruction process of the analysed centre, a
technological and energy sustainability method based on evaluative and critical steps. In this way it is possible
to place the interventions that can be implemented in relation to the pre-existing urban-fabric and the landscape.
The “improvement” logic so adopted does not lead to a mere adaptation of regulatory standards and minimum
requirements, but is focused on the assessment of the landscape compatibility, in respect of a territory that bears
witness to the millennial balance of man and nature.
In the following paragraphs some strategies used for the energy efficiency improvement of historical centers
will be showed. The research aims to reduce energy consumptions in the historic center and integrate renewable
energy sources (active and passive systems for energy production), by direct actions on individual buildings or
by urban-design applications on collective open spaces.

250
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2.1. Urban-Design for the common spaces

Resuming the assessments described in sub-chapter 2.2 and after a cognitive stage, which allowed the
identification of the areas with the greatest critical environmental comfort, we arrived at a proposal phase, in
which the same ENVI-met software was used to prove the effectiveness of the hypothesised interventions as the
introduction of new vegetation, the use of appropriate materials (low albedo and emissivity values) for exterior
pavings or the realization of fountains, green roof and little pools of water. ENVI-met provides quantitative
forecasts on the impact of newly designed green areas or structures, supporting the urban-design process in order
to achieve the following objectives: reducing the summer temperatures by making outdoor spaces more
accessible, protecting buildings from direct sun radiation, avoiding excessive heating of the interior and
improving comfort even in the winter season. The graphic representations regarding Arquata Capoluogo are
reported below, showing the comparison made between the energy simulations before and after interventions
(Fig. 7, 8).

Figure 7. Example of comparative evaluation for the PPD value: before and after the interventions. Arquata Capoluogo –
Summer scenario
To solve problems related to temperature and humidity values of the summer scenario, it is advisable to introduce
specific deciduous vegetational plants, which lower the air temperature and shield the solar radiation.
Considering the PMV values, that are not optimal during the summer seasons, the third frame shows also the
positive effect of introducing pools and fountains, as well as using materials suitable for paving open spaces
(stone, gravel and porous materials). The effectiveness of the interventions is clear observing the yellow and the
green colours that appear in the final scenario and that correspond to a lower percentage of dissatisfied.

Figure 8. Example of comparative evaluation for the PPD value: before and after the interventions. Arquata Capoluogo –
Winter scenario

To improve temperature and ventilation values of the winter scenario, it is advisable to introduce high evergreen
trees that act as a barrier to the strong and cold winter currents coming from the north/north-west. As proof of
the correctness of the actions planned, the third frame of the winter scenario shows that the yellow and the green
colours have been reduced, in favour of a lower percentage of dissatisfied.

2.2. Use of renewable energies to cover urban energy needs

The objective of this part of research is to reduce the consumption of fossil energy through the use of renewable
energies. In this regard, the understanding of the territorial resources becomes necessary to suggest

251
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

reconstructive solutions consistent with the characteristics of the place. Considering the village of Arquata del
Tronto as a case study, we imagine to cover all the energy needs with renewable energy sources only. We assume
to install photovoltaic and thermal solar panels according to the features of the study area and to use the territory's
own resources. We refer to the electricity supplied by small hydroelectric generators fed by watercourses and
the zero-mile thermal energy generated by the combustion of biomass collected in the territory. A significant
amount of energy is added as a result of a possible geothermal system. A time of three years is considered to
perform all the measures suggested.
The verification of the total coverage of the energy needs through the renewable sources is carried out starting
from the buildings’ square meters and the average needs of the users, using pre-dimensioning calculation charts.
The steps of the energy calculations are described in figure 9. This table regards the only hamlet of Arquata
Capoluogo but same data are collected for all the other localities of Arquata del Tronto.
For each type of renewable sources, the figure shows the corresponding percentage of coverage of the total
amount and the value of residual need that has to be covered through fossil energy. It also outlines the reduction
of the energy demand after the interventions planned.

Figure 9. Energy calculations – Arquata Capoluogo

As further deepening, we chose Arquata Capoluogo to carry out a cost estimates of the interventions required
for the energy savings. Details of calculations are showed in figure 10 and they regard only the investment costs
necessary to perform the interventions planned.

Figure 10. Estimate of investment costs – Arquata Capoluogo

Figure 11a and figure 11b show the same kind of data: while figure 11a is referred only to the hamlet of Arquata
Capoluogo, figure 11b provides the total values for the whole territory of Arquata del Tronto, calculated as sum
of the partial values about the fourteen hamlets. In the tables the costs associated to the existing conditions are

252
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

compared to the ones regarding the new scenarios (during and after the project). The tables also show the total
cost of the energy actually required (divided into electrical and thermal energy) [a] and then the cost of the share
of energy that is still required after interventions [b]. The difference between the two values is the amount of
energy saving [c], in terms of both €/year and GWh/year.

Figure 11a. Total energy saving – Arquata Capoluogo

Figure 11b. Total energy saving – Arquata del Tronto

To evaluate the gains achieved with the project it’s possible to look at the table 4. The period considered covers
10 years because the Italian legislation (D. Lgs.192/2005) requires the return of investments within this amount
of time.
Following a description of the calculations given in figure 12:

Row [0]: It shows the cost of the primary energy that would be required every year in the existing condition.

Row [1]: It shows the investments divided in equal parts over the three years during which the project is
performed.

Row [2]: It shows the decrease of the energy demand, hypothesizing that in the first year there is still no energy
savings and that the full capacity is reached when the project ends (only residual need remains for the following
years).

253
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Row [1+2]: It shows the total cost, considered as the sum of the investment for the execution of the interventions
and the cost of the decreasing energy. Red values outline that during the first three years the total cost is major
then the one that would be required if the project was not carried out.

Row [0-(1+2): It shows the difference between the cost in the existing condition and the total cost during the
project. As evident from the green cells of the table, starting from the fourth year a considerable gain is obtained.

Figure 12. Payback time – Arquata Capoluogo

In conclusion the values outline that over ten years the gains are capable of repaying all the initial loss, so what
remains in the future is exclusively the annual energy saving.

2.3. Energy improvement of individual buildings

In the case of historical buildings, which characterise the urban settlement of Arquata del Tronto, improving the
energy saving does not mean reaching the legislative parameters established for the new constructions. Attention
however must be paid to the technical-design solutions in order to identify those that make a more conscious
and rational use of energy resources.
Looking at a conscious use of energy-environmental resources, it is possible to list a series of actions according
to the indications collected in the guidelines of the Italian Ministry of Cultural Heritage.
These have been systematically reorganised, following three aspects:
1) the elements of the building on which the intervention is applied (opaque and transparent surfaces);
2) the type of action carried out;
3) the impact that the intervention produces on the historic building.

Figure 13a. List of the eligible actions on historical buildings and their description in terms of aim and impact

254
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 13b. List of the eligible actions on historical buildings and their description in terms of aim and impact

3. CONCLUSION

The research conducted verifies the methodological, operational and technological complexity of the theme, but
also its centrality for the identification of the strategies for a sustainable and energetically oriented enhancement
of the existing historical building heritage.
Starting from the territory and the knowledge of the historical values, it is necessary that the intervention projects
safeguard the identity and the global functioning of the building and its context. Therefore, knowing the territory
and understanding the complex historical and architectural stratification in which the buildings are located, it’s
vital to promote an increasingly efficient energy management.
The study carried out on Arquata allowed us to outline a hierarchy of operational suggestions, which go from
the territorial scale to the individual buildings, by proceeding through successive levels of in-depth analysis.
These suggestions provide cues for an appropriate recovery process in respect of environmental issues.

ACKNOWLEDGMENT
We thank the Municipality of Arquata that has commissioned this research and Silvia Gialluca for her valuable
cooperation.

REFERENCES
[1] AA.VV., L'efficienza energetica nei regolamenti edilizi. Linee guida, Provincia di Milano, 2006.
[2] AA.VV., Linee d’indirizzo per il miglioramento dell’efficienza energetica nel patrimonio culturale. Roma: Mibac, 2015.
[3] Calzolari, M. Prestazione energetica delle architetture storiche: sfide e soluzioni: analisi dei metodi di calcolo per la
definizione del comportamento energetico. Milano: Franco Angeli, 2016.
[4] Cetica, P.A. L’Architettura dei Muri Intelligenti. Esperienze di climatizzazione sostenibile nell’Ottocento. Firenze:
Angelo Pontecorvi Editore, 2004.
[5] Corti, L, Rigano, E. Edifici storici ad alta efficienza energetica, Padova: Overview editore, 2013.
[6] Crawley, D. Hand, Kummert, M. Griffith, B. Contrasting the capabilities of building energy performance simulation
programs. Building and Envirorment 2008; 43: 661-673.
[7] Davoli, P. Il recupero energetico ambientale del costruito. Santarcangelo di Romagna: Maggioli Editore, 2010.
[8] De Santoli, L. Efficienza energetica negli edifici storici, Milano: Edizioni Delfino, 2014.
[9] Di Ruocco, G. Strategie innovative non invasive per l’efficientamento energetico degli edifici storici. Fisciano: Cues
Edizioni, 2014.
[10] D’Olimpio, D. Il retrofitting energetico e bioclimatico nella riqualificazione edilizia. Roma: Legislazione Tecnica,
2017.
[11] Franco, G. Innovazione e sostenibilità in un paesaggio culturale. Techne 2010; 5: 129-134.
[12] Franco, G. Paesaggi ed energia un equilibrio delicato. Monfalcone: Edicom Edizioni, 2015.
[13] Lucchi, E. Pracchi, V. Efficienza energetica e patrimonio costruito: la sfida del miglioramento delle prestazioni
nell’edilizia storica, Santarcangelo di Romagna: Maggioli Editore, 2013.
[14] Marino, V. Pagani, R. Un’infrastruttura geotermica per un borgo storico in Toscana: riflessioni sulla sostenibilità
delle soluzioni di riqualificazione energetica. Tecne 2011; 11: 87-96.
[15] Ryhl-Svendsen, M. Padfied, T. Smith, V.A. De Santis, F. The indoor climate in historic buildings without mechanical
ventilation system, EU 5th Framework project Microclimate Indoor Monitoring in Cultural Heritage Preservation
(MIMIC), 2004
[16] Rolli, G. L. Salvare i centri storici minori: proposte per un atlante urbanistico dei centri d'Abruzzo, Firenze: Alinea
Editrice, 2008.
[17] Rotilio, M. Linee guida per il recupero dei centri minori in contesti insediativi di valore ambientale e storico. Due casi
di studio: il borgo di Navelli e quello di Caporciano (Aq). In Trento, A. L’Attività di ricerca nel dottorato: verso un
sapere tecnico condiviso. Roma: Aton Immagine e Comunicazione, 2011.
[18] Torresi, F. Manuale del recupero della Provincia di Ascoli Piceno, Acquaviva: Fast Edit, 2004.

255
 FEATURES OF MOBILE PLATFORMS' EQUIPMENT SECOND
ORDER DIGITAL FREQUENCY-DEPENDENT COMPONENTS
PHASE-FREQUENCY RESPONSE AT TUNING

Hanna Ukhina
Odesa National Polytechnic University, Odesa, Ukraine, [email protected], ORCID: 0000-0003-3797-1460

Ivan Afanasyev
Odesa National Polytechnic University, Odesa, Ukraine, [email protected],

ValeriiSytnikov
Odesa National Polytechnic University, Odesa, Ukraine, [email protected], ORCID: 0000-0003-3229-5096

Oleg Streltsov
Odesa National Polytechnic University, Odesa, Ukraine, [email protected], ORCID: 0000-0002-4691-5703

Pavel Stupen
Odesa National Polytechnic University, Odesa, Ukraine, [email protected], ORCID: 0000-0003-1952-6144

Ukhina H, Afanasyev I, Sytnikov V, Streltsov O, Stupen P. Features of Mobile Platforms'

Cite this paper as: Equipment Second Order Digital Frequency-Dependent Components Phase-Frequency
Response at Tuning. 8th Eur. Conf. Ren. Energy Sys.24-25 August 2020, Istanbul

Abstract: This study analyzes the influence of digital filter transfer function coefficients on filter responses'
properties. Equations used to find the transfer function denominator coefficients for Chebyshev
digital filter-1 are obtained. This study results revealed that the PFR depends on the level of
pulsation meanwhile never depending on the cut-off frequency. Carried out are the experimental
study with subsequent simulation that allowed us to obtain a number of values for this dependence.
However, it is quite difficult to apply such a dependence on platforms with limited computational
and energy resources, so the obtained dependence has been approximated for the convenience of
its operational use.
Keywords: Industry 4.0, transfer function coefficient, ripple index, Chebyshev filter-1, Butterworth filter.

Nomenclature
IoT Internet of Things
IIoT Industrial Internet of Things
PFR Phase frequency response
AFR Amplitude-frequency response
FDC Frequency-dependent component
LPF Low pass filter
© 2020 Published by ECRES

1. INTRODUCTION

The modern state-of-art in the "Computer engineering" special field development does fully correspond to the
Industry 4.0. concept. Requirements applied to the Internet technologies and their components, the computer
systems' components, as well as to software, do imply and stipulate the theoretical and practical developments'
continued improvement in accordance with this concept.
The modern production processes' progress trend is moving towards a complete automation with all processes
real-time management, taking into account their changing external conditions. In this case, the Internet
technologies and their components must ensure reliable communication between the production process
components, and between the personnel engaged. This implies a necessity to improve further the development

256
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

of components capable for their responses comprehensive rearrangement, self-tuning and self-training in
accordance with the said concept [1].
The Internet of things (IoT) technology uses not only information exchange between people, but also the data
exchange between all sorts of "things", i.e. machines, devices, and sensors. On the one hand, "things" equipped
with sensors can exchange data and process data without human engaging to that process, and on the other hand,
a human can actively participate in the process, for example, when it comes to "smart home" or "smart enterprise
" concepts.
The Industrial Internet of Things, (IIoT) embodies a kind of Internet of Things (IoT) properly said. This direction
allows us to implement a completely automated production. Today, the technological equipment and facilities
components are equipped with multifunctional sensors, actuators and controllers. The collected data upon
processing are sent to the enterprise's relevant services, that allows the personnel quickly making respective
operational well balanced and grounded decisions, on completely informed basis. But the maximum task here
is to achieve such a level of the enterprise's automation when in all production areas where it is possible, the
machinery and automatic tools are operated without the personnel's intrusion. The role of personnel in this case
is reduced to monitoring the machines operation rendering an intrusive response only when emergency situations
[1-3].
Such tasks are also encountered in robotics on various mobile platforms, as well as in critical application systems
[4-7]. Here there arises the problem of building components for special-purpose computer systems, information-
measuring or data-management-and-control systems that could comprehensively rearrange their response
characteristics using software and hardware methods, with respect to the noise-signal environment, with the aim
of such system efficiency improving.
Usually, the sensors-produced signal processing path includes frequency-dependent components whose
response characteristics are capable for rearrangement. In general, such systems embody a complex architectural
combination of hardware and software means implemented using a variety of connections [7-11].
When processing signals with phase distortion compensation, it is important to know what happens to the
computer system component's phase-frequency response in the bandwidth or at the bandpass cutoff frequency
boundary; in addition, such phase data are often used in further processing.
In this case, necessary is to analyze the processing path component' transfer function coefficients influence on
the phase frequency response (PFR). A similar problem is encountered when developing adaptive and tunable
devices that use, for example, standard links of second-order control systems, as well as the filters specific with
the same mathematical description [12].

2. ANALYSIS OF THE DIGITAL FILTER TRANSFER FUNCTION COEFFICIENTS INFLUENCE

ON ITS RESPONSE PROPERTIES

Here we shall consider the widely used typical digital filters as a component of signal processing and filtering
path. Well known is that for ease of tuning the high-order components are built on the basis of first and second
orders frequency-dependent components (FDC) [7]. Therefore, the analysis of the digital filter transfer function
coefficients' influence on the digital filter responses properties is carried out using the second-order transfer
function, which has the form

𝑎0 + 𝑎1 𝑧 −1 + 𝑎2 𝑧 −2
𝐻(𝑧) = (1)
1 + 𝑏1 𝑧 −1 + 𝑏2 𝑧 −2

where 𝑎0 , 𝑎1 , 𝑎2 , 𝑏1 , 𝑏2 are the real coefficients of the numerator and denominator.

When substituting 𝑧 −1 = 𝑒 −𝑗𝜔̀ or using Euler's formula 𝑧 −1 = 𝑐𝑜𝑠𝜔̀ − 𝑗𝑠𝑖𝑛𝜔̀ , where 𝜔̀ is the normalized
𝑓
2𝜋
𝑓𝑑
angular frequency, 𝜔̀ = , 𝜔̀ ∈ [0 ÷ 𝜋], 𝑓, 𝑓𝑑 are the linear frequency and the sampling frequency,
𝜋
respectively, we get a complex transfer coefficient, and based on this coefficient, the amplitude-frequency
response (AFR) and phase-frequency response (PFR): at 𝑎0 = 𝑎2 [13]

257
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑎 2 𝑎
2 + (𝑎1 ) + 4 (𝑎1 ) 𝑐𝑜𝑠𝜔̀ + 2𝑐𝑜𝑠2𝜔̀
𝐻 (𝜔̀ ) = √𝑎02 0 0 (2)
1 + 𝑏12 + 𝑏22 − 2𝑏1 (1 + 𝑏2 )𝑐𝑜𝑠𝜔̀ + 2𝑏2 𝑐𝑜𝑠2𝜔̀

(𝑏2 − 1)𝑠𝑖𝑛𝜔̀
𝜑(𝜔̀ ) = 𝑎𝑟𝑐𝑡𝑔 (3)
(𝑏2 + 1)𝑐𝑜𝑠𝜔̀ + 𝑏1

It should be noted that similar equations are used to describe the typical second order links at automatic control
systems. Therefore, we consider the influence of the transfer function coefficients onto the frequency response
of second order low-pass frequencies (LPF) Butterworth and Chebyshev-1 digital filters, as the most used
frequency-dependent component used at computer systems and control systems [14-15].
In these digital filters, the numerator coefficients, as a rule, include dependence 𝑎1 = 2𝑎0, which makes it
possible to get the amplitude frequency response square from the formula (2) upon certain transformations
carried out [16-17]

4𝑎02 (1 + 𝑐𝑜𝑠𝜔̀ )2
𝐻2 (𝜔̀ ) (4)
(1 − 𝑏2 )2 + (𝑏1 + 2𝑐𝑜𝑠2𝜔̀ )(𝑏1 + 2𝑏2 𝑐𝑜𝑠𝜔̀ )

From equation (3) for the Butterworth filter and with respect to PFR type (Fig.1), it can be specified that
−𝜋
𝜑(𝜔̀ 𝑐 ) = 2 . Then the PFR denominator will appear like

(𝑏2 + 1)𝑐𝑜𝑠𝜔̀ + 𝑏1 = 0

We define 𝑏1 as

𝑏1 = −(𝑏2 + 1)𝑐𝑜𝑠𝜔̀ (5)

Figure 1. Frequency and phase response of the second order Butterworth filter

For the Chebyshev filter-1, this solution is not possible, because here the PFR has different phase values at the
cutoff frequency level𝑐 set by the ripple value in the bandwidth𝜀 . It should be noted that in most mathematical
packages, the ripple level is set in RP decibels further recalculated to 𝑐 as follows

1
𝑐=
√100,1𝑅𝑃
258
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

For the Butterworth digital filter, the RP value is constantly -3 dB (𝑐 = 0.707), and for other filters it has
different values, but the value 𝑐is in the range 0 < 𝑐 < 1, refer to Fig. 2 and 3.

Figure 2. Frequency response and frequency response of the second order Chebyshev filter-1 at RP= -3 dB

Figure 3. Frequency response and frequency response of the second order Chebyshev filter -1, at RP= - 7dB

Studies of PFR have shown that one of its main features is not a dependence on the cutoff frequency, but the
dependence on the ripple level. This phenomenon is manifested differently in different filters.
After all transformations based on equations (2) and (3), the coefficients 𝑏1 and 𝑏2 for the Chebyshev filter -1
will have the form

1 − 𝑏2
𝑏1 = − [(1 + 𝑏2 )𝑐𝑜𝑠𝜔̀ 𝑐 + 𝑠𝑖𝑛𝜔̀ 𝑐 ]
𝛷

[2√1 + 𝛷2 + (1 + 𝑐𝑜𝑠𝜔̀ 𝑐 )] − 𝛷𝑠𝑖𝑛𝜔̀ 𝑐

𝑏2 =
[2√1 + 𝛷2 + (1 + 𝑐𝑜𝑠𝜔̀ 𝑐 )] + 𝛷𝑠𝑖𝑛𝜔̀ 𝑐

where 𝛷 = 𝑡𝑔(𝜑) from formula (3).

As evidently seen, to calculate the coefficients 𝑏1 and 𝑏2 it is necessary to know the cutoff frequency PFR values
at different ripple level values in the bandwidth.
It follows from Figures 2 and 3 that the phase values at the cutoff frequency depend only on the ripple level in
the bandwidth, Fig.4 [12].

259
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 4. Dependence of Chebyshev's LPF-1 FDC PFR at cutoff frequency 𝜔̀ = 0.5 on the ripple level

To be noted is that if the ripple level is constant and the cutoff frequency varies, this value does never change.
Since it was not possible to obtain an analytical expression for the PFR depending on the ripple level, an
experimental study with subsequent simulation have been performed, that allowed us to obtain a number of
values for this dependence. However, using such a dependence, for example, on mobile platforms with limited
computing and energy resources is quite difficult, so there arose a problem of such dependence approximating
for the needs of operational use.
One of the first approximations gave us the dependence of the kind

𝜑 = −0.537𝑙𝑛 (𝑅𝑃 ) − 1.4355

At that the root-square-mean deviation of such approximation is 2%.

Although of this solution simplicity it was not entirely satisfactory when applied on a mobile platform.
Therefore, the new approximation is elaborated based on the hyperbolic tangent, Fig. 5.

𝜑 = −𝜋[𝑡ℎ (0.0792𝑅𝑃 + 0.4521)] (6)

In this form, it is not possible to implement the hyperbolic tangent. So we decompose the hyperbolic tangent
into a series [18]:

1 2 17 7 62 9
𝜑 = −𝜋𝑡ℎ𝑥 = −𝜋 [𝑥 − 𝑥 3 + 𝑥 5 − 𝑥 + 𝑥 ]
3 15 315 2835
Converting this relation to the form

𝜑 = −𝜋𝑡ℎ𝑥 = −𝜋𝑥[1 − 0.333(1 − 0.4𝑥 2 )(1 + 0.162𝑥 4 )𝑥 2 ]

where 𝑥 = 0.0792𝑅𝑃 + 0.4521 from the ratio (6).

𝜋
This expression only works if the |𝑥| ≤ 2 value. In our case, the RP value should never exceed 14, Fig. 6.

260
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 5. Tabulated values (1) and approximating function (2) graphs.

Figure 6. Tabulated values (1) and approximating function (2) graphs.

As we can see from the graph represented in Fig. 6, a good approximation goes up to RP=7 with a standard error
of 0,68%.
With RP broader changes, for example, reaching up to 40. we need switching to a different representation. Let
we express the hyperbolic tangent using the hyperbolic sine and cosine:

𝑠ℎ𝑥
𝑡ℎ𝑥 =
𝑐ℎ𝑥

Now we decompose these components in a series [18]:

𝑥3 𝑥5 𝑥7 𝑥9 𝑥2 𝑥4 𝑥6 𝑥8
𝑠ℎ𝑥 = 𝑥 + 3!
+ 5!
+ 7!
+ 9!
, 𝑐ℎ𝑥 = 1 + 2!
+ 4!
+ 6!
+ 8!

261
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Well known is that when implemented, the series are decomposed up to rows consisting of series 5 members.
This is due to the fact that microcontrollers will not be able to cope with factorial calculation for larger numbers.
We convert these formulas reaching the form:

𝑠ℎ𝑥 = 𝑥[(1 + 0.1666𝑥 2 ) + 0.008𝑥 4 (1 + 0.0241𝑥 2 )];

𝑐ℎ𝑥 = (1 + 0.5𝑥 2 ) + 0.0417𝑥 4 (1 + 0.0335𝑥 2 ).

Dividing these relations, we obtain a hyperbolic tangent for the value𝑥 = 0.0792𝑅𝑃 + 0.4521 . Multiplying the
result by– 𝜋, we bring it closer to the tabulated values.

𝑥[(1 + 0.1666𝑥 2 ) + 0.008𝑥 4 (1 + 0.0241𝑥 2 )]

𝜑 = −𝜋𝑡ℎ𝑥 = −𝜋 [ ]
(1 + 0.5𝑥 2 ) + 0.0417𝑥 4 (1 + 0.0335𝑥 2 )

It is worth noting that this expression works at RP up to 40 with a standard error of 0,06%. The resulting
transformation is shown in Figure 7.

Figure 7. Tabulated values (1) and approximating function (2) graphs.

If we consider the Chebyshev filter 1 phase value as the limit value equal to -π, then the expansion into series
allows using it only up to the bandwidth ripple level value equal to 35.

4. CONCLUSION

The obtained approximations of the Phase-frequency response dependence on the ripple level, as well as
equations derived for finding the digital filter transfer function denominator coefficients do efficiently contribute
to these components more operative use on platforms with limited computing and energy resources.

REFERENCES

[1] The Industry 4.0 Standards Landscape from a Semantic Integration PerspectiveConference Paper (PDF Available) ·
September 2017 with 4,699 Reads DOI: 10.1109/ETFA.2017.8247584 Conference: Conference: 2017 IEEE 22nd
International Conference on Emerging Technologies and Factory Automation (ETFA), At Limassol, Cyprus, available:
https://www.researchgate.net/publication/318208930_The_Industry_40_Standards_Landscape_from_a_Semantic_Int
egration_Perspective
[2] Industry 4.0: an overview. Conference Paper (PDF Available) · July 2018 with7,147 Reads, available at:
https://www.researchgate.net/publication/326352993_Industry_40_an_overview
[3] Industry 4.0, available: https://www.cognex.com/ru-ru/what-is/industry-4-0-machine-vision/development.

262
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[4] Semenov, S., Voloshyn, D., Ahmed, A.N. “Mathematical model of the implementation process of flight task of
unmanned aerial vehicle in the conditions of external impact”, International Journal of Advanced Trends in Computer
Science and Engineering, vol. 8(1), pp. 7-13, 2019.
[5] Zhuravska, I., Musyienko, M., Tohoiev, O. “Development the heat leak detection method for hidden thermal objects by
means the information- measuring computer system”. CEUR WorkshopProceedings, 2353, pp. 350-364, 2019.
[6] Musiyenko, M.P., Denysov, O.O., Zhuravska, I.M., Burlachenko, I.S. “Development of double median filter for optical
navigation problems”. Proceedings of the 2016 IEEE 1st International Conference on Data Stream Mining and
Processing, DSMP 2016, 7583535, pp. 177-181, 2016.
[7] Ukhina, H., Sytnikov, V., Streltsov, O., Stupen, P., Yakovlev, D. “Transfer Function Coefficients Influence on the
Processing Path Bandpass Frequency-Dependent Components' Amplitude-Frequency Characteristics Properties at the
NPP TP ACS”. Conference Proceedings of 2019 10th International Conference on Dependable Systems, Services and
Technologies, DESSERT 2019, June 2019, Номерстатьи 8770050, Pages 193-19610th International Conference on
Dependable Systems, Services and Technologies, DESSERT 2019; Leeds; United Kingdom; 5 June 2019 до 7 June
2019, DOI: 10.1109/DESSERT.2019.8770050
[8] Eliseenko V.V., “Information and control systems of NPPs produced by SNPO “Impulse”. State, prospects”, Jadernaja
i radiatsionnaja bezopasnost, No 4(60), 2013, 61-64 pp. (in Russian)
[9] NishaHaridas, Elizabeth Elias, “Efficient variable bandwidth filters for digital hearing aid using Farrow structure”,
Journal of Advanced Research, vol. 7(2), 2016, pp. 255–262.
[10] SumedhDhabu, Abhishek Ambede, Smitha K. G., Sumit Darak, A. P. Vinod, “Variable Cutoff Frequency FIR Filters:
A Survey”, 2018, available: https://arxiv.org/ftp/arxiv/papers/1804/1804.02891.pdf
[11] ShunsukeKoshita, Masahide Abe and Masayuki Kawamata, “Recent Advances in Variable Digital Filters”, 2018,
available: https://www.intechopen.com/books/digital-systems/recent-advances-in-variable-digital-filters
[12] Ukhina H.V., Afanasyev I.V., Sytnikov V.S. The phase-frequency characteristic study of the second order digital
frequency-dependent components of mobile platforms // «Modern Information Tecnology 2019» (May, 23-24, 2019,
Odesa). – Odesa: Ecology, 2019, 19-20pp.
[13] Dzhigan, V.A. Adaptivnajafiljtratsija signalov: teorija i praktika [Adaptive signals filtering: theory and algorithms],
M.: Tekhnosfera, 528 p., 2013.
[14] Ukhina, H., Sytnikov, V., Streltsov, O., Stupen, P., Yakovlev, D. “Stability evaluation based on the sustainability
triangle application for transfer functions above 2nd order”. CSAE 2019: Proceedings of the 3rd International
Conference on Computer Science and Application EngineeringOctober 2019, Article No.: 6, pp 1–
5https://doi.org/10.1145/3331453.3361305
[15] Ukhina, H., Sytnikov, V., Streltsov, O., Stupen, P., Yakovlev, D., “Specialized computer systems digital bandpass
frequency-dependent components rearrangement”. Proceedings of the 2019 10th IEEE International Conference on
Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, IDAACS 2019, DOI:
10.1109/IDAACS.2019.8924368
[16] Zhengyou, He. Wavelet Analysis and Transient Signal Processing Applications for Power Systems, Chine Electric
Power Press, 252 p.,2016.
[17] SumbalZahoor, Shahzad Naseem, “Design and implementation of an efficient FIR digital filter”, Electrical &
electronic engineering | Research article, vol. 4, i. 1, 2017, pp. 34-46.
[18] Korn G., Korn T. Mathematical handbook for scietists and engineers – McGraw Hill Book Company, N.York, 1974 –
832 p.

263
 HOW CAN ROMANIA IMPROVE THE POTENTIAL OF
SUSTAINABLE ENERGY DEVELOPMENT USING FOREST
BIOMASS
Laurențiu Ciornei
Center for Study and Research for AgroForestry Biodiversity “Acad. David Davidescu”, Romanian Academy, Bucharest,
Romania, [email protected], ORCID: 0000-0002-9381-5051.

Paula Munteanu
School of Advanced Studies of the Romanian Academy (SCOSAAR), Romanian Academy, Bucharest, Romania,
[email protected], ORCID: 0000-0003-0366-8762

Ciornei, L., Munteanu, P. Romania's potential for sustainable energy development. 8th Eur.
Cite this paper as:
Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: More than 66% of the gross final energy consumption in Romania is based on non-renewable
resources. The rest of 34% originates from slowly renewable resources and other unconventionals.
Decreasing the share of non-renewable energy sources is a major objective that can be achieved
through rationally exploiting renewable resources and adopting new technologies.
Forest is a slowly renewable resource that requires efficient methods of exploitation to ensure long-
term energy sustainability. For these reasons, the biomass* resulting from forestry and primary
processing of uncontaminated wood can successfully replace the relevant primary energy
resources.
Cogeneration plants could contribute to the rational use of primary energy resources, being some
of most efficient solutions, because they generate both, electricity and heat.
For these reasons, in our study, we started from biomass that Romania can annually produce. Based
on capacities of modern cogeneration plants, the result of econometric analysis shows that using
cogeneration plants and biomass could contribute to the annual saving of a forest area between
5.200 – 6.000 ha and to the generation of electricity (340-400 MWe/h) and thermic energy (1160-
1350 MWt/h) which could heat between 205.350-234.950 households, annually.
In addition, as an alternative of biomass usage, the results confirm that the quantity of coal from
the final energy consumption could be reduced by approximately 25%.
Keywords: sustainable energy development, renewable resources, biomass,
© 2020 Published by ECRES

Nomenclature
*Biomass In our study the biomass refers only to that quantity of wood that results after the primary
exploitation of trees in forests (twigs, foliage, bark, nails and roots).
INS Romanian National Institute of Statistics
WCEF World Circular Economy Forum

1. INTRODUCTION

Through this study, our aim is to contribute to the scientific research on the potential of sustainable energy
development that Romania has, in terms of renewable resources. For these reasons, we focused our attention on
the usage of the remaining biomass after the exploitation of the wood as primary energy source for cogeneration
plants that produce, simultaneously, both, heat and electricity.

264
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In this study, we used an example of good practice that Romania already has in terms of biomass-based
cogeneration. Starting from the quantity of wood harvested in Romania and from the capacities that the
cogeneration plants have, we used the results and realized forecasts on the wider use of biomass, considering
the natural potential that Romania has, from this point of view, on the time horizon 2020-2050. We wanted to
illustrate that the intelligent use of the "waste" left after the exploitation of the raw wood, contributes to the
significant decrease of both, renewable and non-renewable resources, with a major impact on achieving the
goals of sustainable development, environmental protection and conservation of natural ecosystems and
biodiversity for future generations.

1. THEORETICAL AND EMPIRICAL DATA

a) The structure of the primary energy resources consumption

In order to reach the objectives, in this study we started from the current situation of the use of primary energy
resources, in Romania. As can be seen in Figure 1, over 66% of the final gross energy consumption in Romania
has based on non-renewable, exhaustible resources. The rest of 34% of the consumption originates from slowly
renewable resources and other sources of alternative and unconventional energy.

Figure 1. Structure of final energy consumption_2018

According to estimates, in Romania, the degree of assurance of lignite resources is limited to the next 28 years,
and to the pit coal ones, to the next 104 years. Moreover, Romania imports about 36% of the primary energy
resource needs, mainly oil, coal and natural gas (Figure 2).

Figure 2. Source of primary energy resources as of 31.12.2018

265
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

As can be seen in Figure 3, one third of Romania's energy consumption refers to households, the consumption
has based on the use of raw wood, about 40% of the total.

Figure 3. Structure of energy consumption by categories of consumers and resources_2018

According to the INS address no.14347 dated on 13.05.2018, in Romania a number of 7.469 million households
had identified, and 45.72% of them use firewood as a heating resource (Table 1). A percentage of 14.22% of
households use firewood for food preparation.

Table 1. Wood consumption of households_2018

Households Firewood Firewood
heated with consumption by consumption Firewood
Household firewood HH (thousands tons by HH consumption by
HH coming (thousands) - (thousands) - of oil equivalent)- (thousands HH (thousands
from: INS INS Eurostat tons) m3)

Urban 4,207.00 642.00 568.59 2,338.08 3,896.79

Rural 3,262.00 2,773.00 2,455.92 10,098.88 16,831.47

Total 7,469.00 3,415.00 3,024.52 12,436.96 20,728.27

According to Eurostat data, the wood consumption of households is 3,024 thousand tons of oil equivalent,
meaning 20,408.95 thousand m3 of wood, annually.

Looking at the European level (Figure 4), Romania does not occupy a leading place with regard to the
consumption registered by the household. However, the weight of wood consumption mentioned above must be
take into account.

Figure 4. Energy consumption of households in Europe_2018 (Thousand tons of oil equivalent)

266
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The energy consumption of households recorded the evolution shown in Figure 5, being between 7,375.18 and
8,102.01 thousand tons of oil equivalent, the minimum reached in 2015, and the maximum in 2010.

Figure 5. Evolution of energy consumption of households in Romania (thousand tons of oil equivalent)

b). Energy strategies from a sustainable development perspective:

According to the definition contained in the WCEF Report, also called the Bruntland Report, "sustainable
development is development that meets the needs of the present, without compromising the opportunities of
future generations to meet their own needs."
Given the imminent depletion of non-renewable resources, given the limited time horizon of exploitation,
decreasing the share of these energy sources is a major goal of sustainable development that could only be
achieve through strategies of rational exploitation of renewable resources, adoption of new technologies and
finding alternative energy resources.
In this regard, we also considered the energy strategy of Romania for the time horizon 2018-2030 and the
projections targeting 2030-2050 interval that refer to the growth of the energy sector under conditions of
sustainability. The development of the energy sector is regarded as part of the development process of Romania
as a whole. Moreover, two of the strategic objectives have aimed at Romania's contribution to regional energy
development, by increasing the country's energy contribution on regional and European markets (through better
capitalization of national primary energy resources) and becoming Romania as a regional energy security
provider.

Figure 6. Projections regarding energy production from primary resources

c). Wood mass that could be generate in a year:

Regarding the annual quantity of wood mass that is generated in Romania, during one year, in this study we
considered the statistical data included in the National Forest Inventory 2013-2018. According to the inventory,
“the total volume of forests is estimated at over 2,354.79 million m3 […] the main forest species being beech
(37%), spruce (32%) and oak (13%), the total growth of forests is estimated at 58.6 million m3/year.” According
to INS, the volume of wood harvested at national level is currently 19.46 million m3. The value of the biomass

267
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(result of the wood exploitation) is around 30%, but it can be use between 5%-15%, the rest being necessary for
the forest ecosystem.
Currently, Romania does not efficiently use the remains that result from the primary exploitation of trees (the
top, the cracks, the foliage, the bark, even the cuttings with partial roots), as the more advanced countries in the
field (for example, the Nordic countries). Romania is limiting on biomass harvesting of only about 5%.
Biomass is the abbreviation used for the biological mass derived from the photosynthesis of chlorophyll that can
come from any organic substance, living or dead, resulting after the forestry exploitation and the primary
processing of the uncontaminated wood.

d). New methods and technologies for reducing consumption of primary energy resources - Cogeneration:

Nowadays, cogeneration is one of the most economical solutions available on the market. A cogeneration plant
produces electricity and heat at the same time. The residual heat produced by the generation of electricity has
efficiently used in heating systems and those for the preparation of household water. In this way, primary energy
has used efficiently. Energy efficiency and overall efficiency increase, in many cases, even over 100%.
By cogeneration, the wasted energy has reduced by almost half. As an example, annual US energy losses in the
form of waste heat, within power plants, are greater than Japan's total energy consumption.
Therefore, a cogeneration plant can save 50% primary energy, or more, compared to a conventional heating
solution without electricity generation. This also reduces CO2 emissions.
The most efficient cogeneration plant in Europe is located in Romania, in the city of Râmnicu-Vâlcea. The
energy efficiency of over 90% of this cogeneration plant determined the construction in 2015 of a second power
station of this type. The Cogeneration II facility, that has considered being the world's leading technology,
produces more electricity and heat than consumes the entire city of Râmnicu-Vâlcea, plus several other rural
localities bordering the municipality. As result of the two investments, Chimcomplex (the largest chemical
compound in Romania) reduced its energy costs by 42%. In terms of environmental impact, due to the annual
savings of natural gas generated by the project, CO2 emissions have reduced by over 17,800 tons per year.
If the above-mentioned cogeneration plant uses resources other than biomass, the one located in the village of
Telciu use exclusively the biomass resulting from forestry exploitation, as well as biomass of agricultural origin.
Village of Telciu is located in county Bistrița-Năsăud, a heavily forested county, with rich forest resources,
benefiting from about 18,000 ha of forest. In the first phase, the project aims the replacement of a number of
several old-generation thermic power stations that serve public-owned buildings, also ensure the consumption
of electricity, including street lighting. The technology chosen for the production of electricity in cogeneration
refers to the Rankine Organic Cycle, which will replace the cycle of energy production in conventional
cogeneration with steam boiler and water heat.
Village of Telciu has an area of 28,530 hectares (870 ha urban and 27.660 ha, outside the built-up area). The
annual logging is about 60,000 cubic meters. The biomass that has used by the cogeneration station, by applying
the minimum percentage of 5% to the exploited wood mass, is about 3,000 m3. According to the authorities, all
the objectives set out in the first implementation phase consume as much as the average consumption of more
than 300 households in the same locality. We have to mention that in the village of Telciu there are 5,798
inhabitants, 2,073 households, 1,807 dwellings, 7 kindergartens, 4 schools and a high school. The project has to
be extend to the total number of households.
The resulting thermal energy is about 1.7 MWh and electricity 0.5 MWh that is deliverable in the National
Energy System, thus compensating the consumption invoiced to the town hall's office. Fuel consumption is
approximately 1,550 t/year (for operation of 2,000 hours per year) and refers exclusively to biomass. Thus, the
biomass is valued higher by higher energy conversion efficiency and by the possibility of using depreciated
biomass.

2. METHODOLOGY

The analysis started from the quantity of wood mass that Romania can harvest annually. Subsequently, we
determined the potential amount of additional biomass that can be annually generate at national level (using the
unexploited difference in double percentage, of 10%) and that can be used as a primary energy resource, instead
of raw wood. Through the econometric analysis performed in Eviews, we predicted the area of forest that would
remain unexploited annually, as result of the efficient consumption of primary resources based on biomass.

268
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 7. The output of econometric analysis (Eviews) - the area of forest that can be annually save

Figure 8. Projections targeting the evolution of the forest area that can be save annually

Starting from the quantity of timber harvested between 2000-2018 (considering that the statistics for 2019 have
not yet been publish) and the surface of the forest that would have been saved through the biomass usage to its
full potential, we made forecasts for the period 2019-2050 on the forest area that would remain standing for
future generations. Moreover, considering the technical parameters of the cogeneration plant based on the
Rankine Organic Cycle, we have determined the annual amount of energy that can be generate from this primary
resource, at national level.

Table 3. 2020-2050 time interval projections

Additional 10% of biomass will generate, on annually basis: UM Min-Max
Heat MWt/h 1160-1350
Values
Electricity MWe/h 340-400
Number of household receiving energy number 205350-234950
Forest saved ha 5200-6000
Alternatives for reducing coal consumption Thousand 430-500
Estimation of total surface of forest that could be saved, for the period 2020- ha
tonnes of 174.198
2050 oil
Thus, using only biomass as a renewable energy source, the results show that a forest area of 5,200 - 6,000 ha
equivalent
could be annually save. Moreover, it could generate about 340-400 MWe/h(TOE) electricity and 1,160-1,350 MWt/h
thermic energy that could serve a number of households between 205,350 and 234,950 that are located in the
mountain area, without access/with limited access to the gas and electricity network. Currently, we have to
underline that these households produce heat with the raw wood, based exclusively on log consumption.

269
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

As noted in Table 3, we made forecasts regarding the reduction of coal consumption, as an alternative of biomass
usage. In other words, if the biomass resulting from the forests logging would be exploit to its full capacity and
used to replace the consumption of non-renewable resources, in particular the consumption of coal, we would
have an annually decrease of 430-500 thousand tons oil equivalent. This quantity, compared to the current final
consumption of coal (1,861.22 thousand tons oil equivalent) means a reduction of approximately 25% of the
coal consumption.
We have considered this alternative of using biomass because the caloric power of coal (6 - 15 MJ/kg) is similar
to the caloric power of wood (9 - 16 MJ/kg). Moreover, most of the big cities from Romania receive thermic
energy produced by heating plants based on the consumption of solid fuel - coal. As shown in Figure 2, part of
the coal consumption is imported.

3. RESULTS

The results indicate that the increasing the biomass exploitation from 5% to 15% (the full potential that Romania
has) leads to the increase of the surfaces with forests that remain standing. Thus, between 5,200 - 6,000 hectares
of forest can be yearly save, which in a time horizon of about 30 years (2020-2050) would increase the forest
areas with 174,198 hectares.
Taking into account the total amount of biomass annually available in Romania and the capacity that the
cogeneration plants have in producing both, electric and thermal energy, at least 340-400 MWe/h of electricity
and 1,160-1,350 MWt/h thermic energy would be generate.
Referring to the domestic consumers located in the mountain area only, the cogeneration plants based on biomass
can serve a number of households between 205,350 and 234,950.
Looking over the National Energy Strategy projections, we concluded that a percentage of 25% reduction in
final coal consumption could be achieve, if the efficient use of biomass would replace the consumption of non-
renewable resources.
Moreover, regarding the perspective of air pollution, the cogeneration plants parameters are below the legal
limits allowed. If we take into account the significant decrease in consumption of raw wood, this would lead to
an increase in the storage capacity of the main pollutants. Looking from a biodiversity perspective, the fact that
forests remain standing means the perpetuation of species for future generations and induction of perpetual
positive externalities.

CONCLUSIONS

The results of our study indicate that biomass (what remains after the primary exploitation of trees), the resource
that Romania enjoys at present, if efficiently used by the modern technologies of electricity production in
cogeneration, would result in a reduction of the consumption of renewable resources, forests remaining alive,
able to perpetuate species and protect the entire ecosystem.
We have to mention that the Romanian state encourages the population from the mountain area to implement
projects regarding the energy management of the mountain area and the exploitation of the energy potential of
the renewable resources. The investments in the energy sector in the mountain area benefit from funds allocated
from the Romanian state budget (Law 197/2018 - Mountain Law) and European funds, too.
Considering all these elements, we appreciate that our study comes to meet the sustainable development
objectives regarding the potential of renewable energy that Romania has.
Unfortunately, even if Romania has a good potential in this regard, logging is being excessively done,
jeopardizing future energy security, threatening biodiversity and promoting increased pollution by reducing CO 2
storage capacity. Romania has been accuse of violating EU legislation, approving harvesting of tens thousands
hectares of secular and primary forests from protected areas, being - and for this reason - in a state of
infringement.
For these reasons, our study will continue with the projections we wish to undertake regarding afforestation and
their ability to produce future biomass.

REFERENCES

[1] https://newstrategycenter.ro
[2] http://roifn.ro/site/rezultate-ifn-2/
[3] http://statistici.insse.ro:8077/tempo-online/

270
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[4] http://rosilva.ro
[5] Badea, A., Necula, H., Renewable energy sources, Bucharest, Agir Publishing House, 2014.
[6] Băhnăreanu, C., Romania's energy security in a European context, Bucharest, Ed. “Carol I”
Universit y, 2010
[7] Eurostat, Energy production and imports http://ec.europa.eu/eurostat/statistics-explained/index.php/
[8] Georgescu, G., Reforma economică şi dezvoltarea durabilă. Editura Economică, Bucureşti, 1995.
[9] Jula, D., Jula, N.M., Econometrie, Editura Mustang, București, 2019
[10] Georgescu, G., Reforma economică şi dezvoltarea durabilă. Editura Economică, Bucureşti, 1995.
[11] Rey, R., Ionașcu Gh., Strategia de dezvoltare durabilă a zonei montane. Editura Academiei Române (2008).
[12] The Government of Romania, Ministry of Waters and Forests, 2018-2027 National Forest Strategy
13] Zaman, Gh., Geamănu, M., Economic efficiency under sustainable development conditions, Bucharest, Publishing
House Romania Foundation of Tomorrow, 2014.
[14] Vasile, Valentina, Zaman, Gh. Dezvoltarea durabilă. Editura Expert, Bucureşti, 2015.
[15] World Commission on Environment and Development, Our Common Future, 1987.

APPENDIXES
Primary energy resources_sources_31.12.2018 (thousands tons of oil
equivalent)

Total
Primary energy Internal Initial Thousand
Import
resources Production balance tons of oil
equivalent

Solid fossil fuels 4016 459 393 4868

Peat and peat
products 454 454
Natural gas 8562 1220 1305 11087
Oil 3491 8263 731 12485
Petroleum
products 3290 3290
Firewood
(including
biomass) 3443 57 52 3552
Hydroelectric,
nuclear-electric
and imported
electricity 5044 253 5297
Energy from
unconventional
sources 61 61
Others 362 172 19 553
Total 24979 14168 2500 41647
Final energy consumption_2018 (thousands tons of oil equivalent)
Primary resources Quantity
Solid fossil fuels 1.861,22
Peat and peat products 14,86
Natural gas 5.845,88
Oil and petroleum products 7.876,26
Wood and biofuels 3.700,10
Energy 3.918,01
Non-renewable waste 170,90
Total 23.387,24

271
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Final energy consumption_2018_HH (thousands tons of oil

equivalent)
thousands tons of oil
Primary resources equivalent

Natural gas 2.513,81

Primary solid biofuels 3.024,52
Electricity 1.098,87
Heat 781,78
Solid fossil fuels 33,63
Liquefied petroleum gases 277,26
Other kerosene 19,83
Total 7.749,69

Harvested wood and additional biomass 2000-2018_(Source:INS)

Biomass
Forest additional
(th ha) 10%
(th ha)
2000 14.285,00 1.428,50
2001 13.410,00 1.341,00
2002 16.383,00 1.638,30
2003 16.692,00 1.669,20
2004 17.082,00 1.708,20
2005 15.671,00 1.567,10
2006 15.684,00 1.568,40
2007 17.238,00 1.723,80
2008 16.705,00 1.670,50
2009 16.520,00 1.652,00
2010 16.992,00 1.699,20
2011 18.705,00 1.870,50
2012 19.081,00 1.908,10
2013 19.282,00 1.928,20
2014 17.889,00 1.788,90
2015 18.133,00 1.813,30
2016 17.198,00 1.719,80
2017 18.316,00 1.831,60
2018 19.462,00 1.946,20

272
 ARTIFICIAL NEURAL NETWORK DESIGN FOR AN AXIAL
FLUX PERMANENT MAGNET GENERATOR

Adem Tekerek
Gazi University, Information Technology Department, Ankara, Turkey, [email protected], ORCID: 0000-0002-
0880-7955

Erol Kurt
Gazi University, Department of Electrical and Electronics Engineering, Technology Faculty, Ankara, Turkey,
[email protected], ORCID: 0000-0002-3615-6926

Tekerek, A., Kurt, E., Artificial Neural Network Design For An Axial Flux Permanent
Cite this paper as:
Magnet Generator. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In the present study, an artificial neural network (ANN) design has been proposed to analyze and
estimate the output quantities of a new manufactured generator, namely axial flux permanent
magnet generator (AFPMG). The machine has been designed with a maximum power of 3 kW for
the household electric power generation. As one of the innovative points, this machine has a
variable air gap between the stator and the rotors, thereby the maximal power scale can be
determined easily by adjusting the air gap between 2 mm and 7 mm. It means that the maximal
power has values between 3 kW and 1.5 kW to, respectively. ANN approach for such a machine
is important in the sense that the estimation of output power, voltage and current values under
various speeds and electrical loads is vital for the optimum operation of the machine.
Keywords: Artificial neural network, generator, axial flux, air gap, power
© 2020 Published by ECRES

1. INTRODUCTION

Permanent magnet generators (PMGs) are easy to construct and reliable with a sinusoidal output waveform if
an optimized magnetic flux path is designed and implemented [1]. According to literature, there are two types
of permanent magnet generators, namely axial flux (AF) and radial flux (RF) [2-4]. AF machines have certain
superiorities over the RF ones [5]. RF machines have hard design and construction processes, in addition they
have worse ratio of power-to-weight compared to the AF machines. Thereby, especially for vertical turbine
systems, AF machines are more preferable solutions. There exist various types of AF machines in the literature:
They can be both core-free and cored structure [6,7]. While the cored machines have high flux densities, core-
free ones have lower cogging torque values [8,9]. The core-free AF machines are generally utilized in low scale
applications [10], however, they suffer from the low power output compared to the cored ones. When the
literature on AF generators is examined, different AF topologies such as funnel, toroidal, single side, double
side, multi-side can be seen as main machine types [11-15].

The output power of any AF machine depends on air gap, windings mass and outer diameter of the
machine[16,17], indeed the outer diameter cannot be extended more because of the mechanical restrictions in
the usage area. In the literature, Ishikawa et al [11] designed a machine with axial and radial flux morphology
and performed some analyzes on flux and reluctance structures. The engineers do not deal with only the flux or
reluctance; in addition, optimization of machine at high speeds is also vital. It has been known that high speed
operation causes warming in the windings, back-irons and cores. In such scenario, Li et al [18,19] studied the
temperature dependence of the machine components. They have proven that the temperature increases from 50
°C to 150 °C at high speeds, that results certain problems in magnetic components due to the decreasing
magnetization. In addition to efficient cooling, other mechanical aspects such as acoustic noises and vibrations
are important quantities to be lowered for an optimized machined [20]. In order to get maximum efficiency,
decrease of loses in electrical, magnetic and mechanical manner is also important. In this manner, Vansompel
et al [21] studied the core geometry, mass, and lamination type and recorded their effects to machine efficiency.
Air gap variation decreased the core loses within the rate 8% according to their findings [22]. AF machines have

273
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

different power densities varying from 10 to 100 kW/m3 [23-25]. One of our different design, a single phase 12
poles AF permanent magnet generator had power density of 28.5 kW/m3 [2]. Wind turbine blades usually rotates
at low speeds (i.e. about 400 rpm or less). Therefore, high number of pole pairs are generally used in the market
in order to get 50 Hz frequency from the output [26]. Especially, in the cases of developing countries, the rural
populations require small and cheap electrical power solutions. They can be either off-grid solutions, or on-grid
applications for wind energy promoting regions.

The present AF machine has a variable power density due to adjustable air gap, whereas the maximum power
density is measured as 111.7 kW / m3 at 2 mm air gap according to experimental results [27]. That is also a
good value compared to the RF machines in the literature. Besides, the RF machines have high total harmonic
distortion (THD) values, whereas the present design and prototype generate outputs with 4% maximum THD.
Apart from our earlier designs, the dimensions of the new machine are larger and more energetic. The reason
for that is to adjust thinner air gap with 2 mm, to use a new magnet positioning strategy and to increase the
electromagnetic active volume for windings and cores. The combination of those above-mentioned points
presents a better power up to 3 kW. The present machine also operates well under different electrical load
conditions. To sum up the points above, the present strategy to form a good machine is to design a machine with
high power density, stable, having easy-maintenance, low cogging torque and good cooling units, however,
exploration of all those parameters in laboratory condition is hard job, and it can be simplified by ANN analyses
for many system parameters mentioned above. One of the aims of the present study is to prove this idea. This
paper is organized as follows: In Section 2, a literature review on the ANN applications of electric machines
will be presented. In Section 3, the methodology is described. Next section gives the main results of ANN
analysis and discuss the findings. Finally, the paper ends with the concluding remarks.

2. ANN APPLICATIONS IN ELECTRICAL MACHINES

There exist many studies on the applications of ANN to the electric machines. For instance, Ding and co-workers
have developed an expert knowledge regarding the conventional time-frequency analysis diagnosis methods
[28]. In their work, authors have proposed an intelligent fault diagnosis method for a rotating electrical machine
based on Generative Adversarial Network under small sample size conditions. According to their findings, their
approach has achieved higher accuracy and low volatility diagnosis results over other diagnosis methods. The
proposed method could stably distinguish fault modes under different working conditions in an adaptive way,
even under though few training samples are available [28]. In Ref. [29], the authors have proposed an ANN
based method for islanding detection of distributed synchronous generators. The proposed method takes
advantage of ANN as pattern classifiers. It is capable of identifying the islanding condition based on samples of
the voltage waveform measured at the distributed generator terminals only, which is an important advantage
over other ANN-based ant islanding methods [29]. In a different work, Bouzid and G. Champenois have
proposed an automatic and intelligent stator fault diagnosis system for the permanent magnet synchronous
generators. The proposed system is based on the use of a feedforward multilayer perception ANN performed by
the back-propagation training algorithm. Two fault indicators are the phase angle of the negative sequence
voltage and the frequency of the machine voltages. The accurate results provided by the ANN prove the
reliability of the proposed system to automatically diagnose different stator faults under variable speed, variable
fault current, and variable load conditions overcoming different disturbances, such as the noisy signals and the
harmonics in the machine [30]. In Ref. [31], the authors have proposed a method to prevent limitations of ANN
for distribution system monitoring. The objective of their study is to overcome the limitations of existing
approaches. Authors do that by presenting an ANN-based scheme, advancing the state-of-the-art in several ways:
Their scheme copes with a very low number of measurements, far less than traditionally required by the state-
of-the-art weighted least squares state estimation (WLS SE). The contribution of their work consists of a method
to generate useful training data by using a scenario generator and a number of hyper parameters that define the
ANN architecture [31]. In another study [32], a new method is proposed to determine the real-time transient
stability status and identification of the coherent generator groups by predicting the rotor angle values following
a large disturbance through radial basis function neural network. In the frame of their work, the first six cycles
of synchronously sampled post-fault data measurements from Phasor Measurement Units (PMU) consisting of
rotor angles and voltages of generators are taken as the input to the neural network to predict the future state of
the system [32].

274
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In 2017, Baptista et al have proposed a specific tool, to assist the hardware implementation of an artificial neural
network, i.e. Automatic General Purpose Neural Hardware Generator. Because an artificial neural network
trained using only the data of solar radiation presents a good solution to predict, in real time for the produced
power values of a photovoltaic system. Even though the neural network can run on a computer, it is expensive
to have a control room with a computer for small photovoltaic installations. The proposed tool allows for an
automatic configuration system enabling the user to configure the ANN, releasing the user from the details of
the physical implementation [33]. In another study [33], the authors presented intelligent controllers as a
maximum power point tracking system for a switched reluctance generator driven by a variable speed wind
turbine to attain the maximum power. The intelligent controller systems are ANN controller and fuzzy logic
(FL) controller. The wind plant is connected to the grid through a DC–AC inverter system and two step up power
transformers [34]. In a different work, the authors have proposed flow phenomena through the axial fan and the
rotor dynamic performance analysis of a permanent magnet generator. The velocities for different blade angles
have been analyzed, numerically. As a result of their simulations, the optimum flat angle of the blade is
determined after having the highest velocity value from the outlet of the simulation. In addition, the rotor fan
power is obtained from the pressure differences between the inlet and outlet. According to the results, the highest
velocity has been predicted as 1.86 m/s and power has been calculated as 0.48 Wat 65 degrees of blade [35]. In
Ref. [36], a new strategy is reported for the control of voltage and frequency of a self-excited induction generator.
An external excitation circuit, including a power switched inductance or capacitor, is used to compensate for the
reactive demand. The conventional dynamic modeling of the system is enhanced by using an ANN to model the
induction generator. The proposed ANN model is used to obtain the inverse model of the self-excited induction
generator and trained by using the output voltage, the load and the wind turbine speed as an input vector and the
required capacitor bank as the target output.

For a steam generator, the pulsed eddy current data are obtained by using a single driver with an array of eight
pick-up coils as a different application area of ANN [37]. In-service inspection of complex systems such as
nuclear steam generator tubes and their surrounding support structures, overlapping degradation modes are
considered. Voltage data are processed by a modified principal component analysis to reduce data
dimensionality, and modified principal component analysis scores are taken as input into an ANN that
simultaneously targeted four parameters associated with support structure hole size, tube off-centering in two
dimensions, and fret depth. The estimates of hole ID and tube position are improved when fret depth is used as
an input [37]. In Ref. [38], authors presented a new strategy for estimating the states (rotor flux and speed) and
the load torque to implement a multivariable controller for a sensorless three-phase squirrel-cage induction
machine. The multivariable control is carried out using input-output feedback law and its objective is to track
profiles of the rotational speed and the rotor flux amplitude. In their study, the state estimation considerably
improves the performance of the rotor flux based model reference adaptive system in the variable speed region
of operation. The proposed technique uses Kalman filter as a rotor flux observer and an ANN adaption
mechanism to estimate the rotor speed. The state estimation requires only the measurements of the stator voltages
and currents [38]. In another work, the authors propose an ANN algorithm for a permanent magnet synchronous
generator (PMSG) connected to a wind turbine to generate the maximum power by using the integral sliding
mode control technique and ANN. The real power produced by the generator is found dependent on the internal
characteristic of the wind turbine such as speed and power coefficient [39].

3. AXIAL FLUX PERMANENT MAGNET GENERATOR (AFPMG)

The experimental setup is founded as in Fig. 1. Indeed, there are two basic aims to perform the experimental
work: Initially, testing the design parameters found in the theory, second, making a feed-back to tune the
parameters for the use in real world. As seen in Fig. 1, the machine locates at the right hand-side in front on the
specific experimental table. It coupled to an induction motor via a coupling apparatus in order to explore the
characteristics. This induction motor is driven by a controller, which adjusts the speed of the motor. The main
measurement devices sit on the measurement table. From left to right, beside controller, a Yokogawa DLM4038
oscilloscope, a Fluke 434 Series II power analyzer and Delorenzo DL1017R, DL1017L, DL1017C electrical
loads with resistive, inductive and capacitive type exist, respectively. Note that while the power analyzer
measures and records the phase voltage, current and power in a synchronized manner, the oscilloscope measures
and records the waveforms at the same time.

275
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The machine has the characteristic features given in Table 1. According to table, the most attractive point is that
we have a chance to see the characteristics of the machine for different air gap values. That is indeed the most
important superiority of the axial machines. Strictly speaking one can easily vary the air gap of the proposed
machine by adjusting two screws. Thus, we have determined the features of the proposed machine for various
air gaps, easily. That feature also gives one to change the rated power of the machine just by changing the air
gap, easily.
In the manufacturing process, we have produced a 1.3 mm air gap machine, however, it caused a high cogging
torque, and it is not useful at low wind speeds. Therefore, 2 mm and 3 mm values are tested for the air gap to
have a chance to use in low speeds such as 30 rpm.

Figure 1. The experimental setup with the AFPMG.

Table 1 gives the specifications of the machine. According to the table, the machine has characteristic geometric
features and parameters such as core structure, back iron structure, rotor and stator geometry. Therefore, an
ANN model can be good tool to identify the relation among all these parameters.

Table 1. Parameters of the AFPMG

Components Features
Inner radius of Rotor 1 (mm) 130
Outer radius of Rotor 1 (mm) 160
Inner radius of Rotor 2 (mm) 75
Outer radius of Rotor 2 (mm) 105
Inner radius of Stator Side 1 (mm) 110
Outer radius of Stator Side 1 (mm) 180
Inner radius of Stator Side 2 (mm) 68
Outer radius of Stator Side 2 (mm) 112
Phase No. 3
Resistance per phase (Ω) 8.8
Number of coils per phase (4 large & 4 small) 8
Winding Turns (big / small coils) 500 / 250
Wire diameter (mm) 1.10
Magnet Type NdFeB
Magnet Shape Circular
Magnet diameter (mm) 30
Magnet thickness (mm) 10
Magnet number 32
Core M19
Back-iron M530-50A
Core type Axial laminated with 40 pieces
Air gap (mm) 2, 3
Filling material Aluminum (Al7000)
Core coefficients (W/m3)
kh 164.20
ke 1.30
ka 1.72

Two representative oscilloscope images are shown in Fig. 2(a,b). These are the generated voltage waveforms
and received at no-load case. Note that it fits to the theoretical expectations in terms of waveform, which is ideal

276
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

sinusoidal. Note that while Fig. 2(a) was measured from the machine configuration having 3 mm air gap, Fig.
2(b) was taken under the condition of 7 mm air gap. It is obvious that there is 120-degree phase difference among
the phases.

(a)

(b)
Figure 2. The oscilloscope patterns of voltage waveforms with no-load condition at 1400 rpm under the air gap
configurations in the cases of (a) 3 mm and (b) 7 mm.

Fig. 3 shows the measurements of the rms values of output voltage per phase for different speed scenarios from
100 rpm to 1300 rpm under no-load conditions. The generator generates more voltage as parallel to the theory
when the air gap decreases. Strictly speaking, the rms phase voltage is measured as 355.2 V at 1400 rpm under
7 mm and 437.6 V under 3 rpm and 517.3 rpm under 2 mm air gap conditions. That gives us a structural factor
that a factor of 0.42 decrease in air gap results 1.23 increase in voltage. Fig. 3 also shows a clear linear increase
by speed as the same with the theoretical explanation of Faraday’s Law. Note that the slopes of the voltage
curves decrease substantially with air gap increase as a structural characteristics of the proposed machine.

Figure 3. The experimental voltage plots depending on the rotor speed for the no-load condition of the generator for 2
mm, 3 mm and 7 mm air gap.

277
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. MODEL

In this section, ANN method study is briefly explained, and steps of obtaining data, cleaning the unnecessary
and noise data and normalizing the data process is described. Then, the data is divided as training and test data,
and the creation of the ANN model is explained.

Artificial Neural Networks and Data

ANN is a machine learning algorithm that is modeled according to the biological neuron cell structure and has
a self-learning feature. ANN can be used in many different applications such as pattern recognition, medicine,
signal processing, prediction and especially system modeling. ANN is formed by connecting neurons to each
other. ANN learns by making inferences from the data and thus new information is predictable and used in the
solution of nonlinear problems. The data is divided into two sets as training and test sets. The aim of the training
is to minimize the error level by adjusting the weights in the artificial neural network. The training process
continues until the intended output is achieved. The performance of the training process is obtained by testing
the data not used in training in the neural network. In Fig. 4, a simple ANN network scheme presented.

Bias
bk
Wk1
Activation
Function

Wk2 Vk ᵩ(Vk) yk

Wkn

Weights
Figure 4. A simple ANN network

The ANN network given in Fig. 4 can be defined mathematically as follows:

𝑛
𝑢𝑘 = ∑ 𝑤𝑘𝑗 𝑥𝑗
𝑗=1

𝑣𝑘 = 𝑢𝑘 + 𝑏𝑘 (1)
𝑦𝑘 = 𝜑(𝑢𝑘 + 𝑏𝑘 )

𝑦𝑘 = 𝜑(𝑣𝑘 )

ANN is a combination of neurons with connections. It is trained using the training data and can therefore make
new inferences and predictions. ANN is used to solve non-linear problems. To train the ANN network, the data
is divided into training and test datasets. The aim of the training process is to reduce or minimize the error level
by adjusting the weights in the neural network. This process continues until the intended output is obtained. The
performance level of the training process is obtained by testing the data not used during training in the neural
network. In train of the neural network, the feed forward back propagation architecture is effective. The feed
forward neural network moves in one direction from the input layer of the model to the output layer.

The training of the network is carried out by arranging the weights of the connections between neurons in each
layer. This process is done by the error function presented in Eq. 2 (dj: targeted result, oj: actual result).

1 𝑝 𝑝 2
𝐸 𝑝 = 2 ∑𝑗 (𝑑𝑗 − 𝑜𝑗 ) (2)

In Eq. 3, the error function is given. This equation is used to rearrange the weights,

278
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝜕𝐸𝑝
∆𝑝 𝑤𝑗𝑖 = −𝜂. (𝜕𝑤 ) (3)
𝑗𝑖

In Eq. 3, for the η constant (learning rate) any value can be assigned. Eq. 4 is used to rearrange the weights:

𝑤𝑖𝑗 (𝑡 + 1) ≅ 𝑤𝑖𝑗 (𝑡) + 𝜂. 𝛿𝑗 . 𝑖𝑖 (4)

In Eq. (4) wij(t) is weight, i result value of node or error term of 𝛿𝑗 with j node. Output layer node error is 𝛿𝑗 is
calculated by,

𝛿𝑗 ≅ 𝑜𝑗 . (1 − 𝑜𝑗 ). (𝑑𝑗 − 𝑜𝑗 ) (5)

Here, j is a hidden node and the error term is calculated by,

𝛿𝑗 ≅ 𝑜𝑗 . (1 − 𝑜𝑗 ). ∑𝑘 𝛿𝑘 . 𝑤𝑗𝑘 (6)

The weight changes can be revised by adding any α moment term,

𝑤𝑖𝑗 (𝑡 + 1) ≅ 𝑤𝑖𝑗 (𝑡) + 𝜂 (𝑑𝑗 − 𝑜𝑗 ) 𝑖𝑖 + 𝛼 (𝑤𝑖𝑗 (𝑡) + 𝑤𝑖𝑗 (𝑡 − 1)) (7)

The feed forward back propagation neural network architecture is effective in the training of the network. The
feed forward neural network moves from the input layer of the model to the output layer in one direction. The
feedforward ANN model consists of the input layer first, then one or two hidden layers and finally the output
layer. The data used in the study were collected using an electric generator. The dataset consists of a total of 392
lines data. Dataset consist of five feature, such as air gap (in mm), ohmic load (in ohm), generator speed (in
rpm), rms phase voltage (in volt), rms phase current (in amper) and total output power (in watt). In Table 2, the
data structure is presented.

Table 2. Data structure obtained from electric generator.

Air Gap Generator RMS Phase RMS Phase Total Output
(mm) Speed (RPM) Voltage (V) Current (A) Power (W)
2 100 66.40 0.060 11.95
2 100 67.10 0.085 17.11
2 600 217.00 0.482 313.78
2 500 185.00 0.609 338.00
3 600 151.00 1.19 538.606
3 800 191.00 1.26 720.020
3 400 132.00 0.61 239.780
3 1300 328.00 1.08 1058.203
3 1000 289.00 0.65 563.063
3 800 259.00 0.33 259.668
3 1000 313.00 0.30 278.585

ANN Model

Before ANN model was created, the features which are selected as ANN model inputs were determined. Ohmic
load, generator speed (RMS), and air gap, phase current (A), which are the experimental results of the electric
generator, are considered for prediction of phase voltage and output power.

(a) Normalizing of data: The data are normalized in order to prevent unnecessary repetition. The data is
rearranged to the 0-1 range and increase the performance of the ANN model. In the normalization process, the
min-max method was used. Here, Vr represents the actual value of the input, Xmin represents the minimum
input value and Xmax represents the maximum input value.

279
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑋𝑟 −𝑋𝑚𝑖𝑛
𝑋𝑛 = 𝑋 (8)
𝑚𝑎𝑥 −𝑋𝑚𝑖𝑛

(b)The training and testing process: After the normalization process, train and test operations are carried out
using the MATLAB program. Dataset is divided in two group, 70% of data is selected for training and 30% of
is selected for testing randomly. Many attempts have been made in different ways to create the proposed model.
As a result of the experiments, the best model (high R and low error rate) is selected. The ANN model structure
is given in Fig. 5. A feedforward multilayer network with an input layer, one hidden layers and an output layer
are considered.

Input Layer (3) Hidden Layer (1:10) Output Layer (2)

X0 W0,0
Speed (RPM)
W0,0
Power (W)
W0,9
W9,0
X1 W1,0
Air Gap (AG)
W0,1
W1,9 Voltage (V)

W2,0 W9,1
X2
Current (A)

W2,9

Bias

Figure 5. Suggested ANN structure

According to the test results, the model has high performance and low error rate. Input layer neurons do not have
transfer functions, such a network structure is often called a three-layer network. In this design, Speed (RMS),
Current (A), and air gap are the inputs, while the corresponding outputs are the Voltage (V), Power (W). The
input and output layers have a number of neurons equal to the respective number of variables. There are 10
neurons in the hidden layer with hyperbolic tangent activation function, which is also used for the output neuron.
This particular network, 3-10-2 plus bias nodes with the relative activation functions is selected.

5. ANN RESULTS AND DISCUSSION

The best model is created as a result of the experiments after the training regression ratio: 0.99988, test regression
ratio: 0.9999, validation regression ratio: 0.99987 and mean square error rate is 1.2509x10-05, and a multi-
layered structure regression and low error rate. According to the experimental results, the ANN model structure
(3-10-2) has a very low performance while it is a one hidden layer with 10 neurons.

280
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 6. Training, validation and test regression

The obtained values were tested with test data to measure the memorization status and performance level of the
model. train-test regression values and error graph of the determined model are given in Fig. 6. As a result of
the experiments, the lowest error rate was determined in 409 iterations. In Fig.7, training, validation and test
performance error rate graph is presented. The error results are very well diverging by epochs.

Figure 7. Performance error rate graph of the model

A comparative representation of some of the real and predictive values obtained from the model is given in Fig.
8(a,b). It is seen that the prediction results obtained with the model using real project data overlap during the
experiment period are largely overlapping with real results and the error rates are very low (close to zero).

281
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a)

(b)
Figure 8. Comparison of experimental and predicted values of (a) output voltage and (b) output power in the case of 3
mm air gap.

6. CONCLUSIONS

An artificial neural network (ANN) design has been performed in order to predict the outputs of an axial flux
permanent magnet generator (AFPMG). Following the training of the real experimental data sets, the new ANN
model can predict the output voltage and power with a very well accuracy. Indeed, the error becomes too low
within the value of 0.1 % for many cases. Note that such a result can also divert the air gap, whether it is 2 mm
or 3 mm, since voltage and power change too much with it. By using this ANN model, one can predict the output
power and voltage for a given air gap, phase current and rotor speed with a perfect accuracy.

ACKNOWLEDGMENT
This research has been supported by The Scientific and Technological Research Council of Turkey (TUBITAK)
under grant No. MAG-315M483. This product has two patents under Turkish Patent Institute with No. TR 2015
04164 B and TR 2013 13062 B and cannot be manufactured without the permission of authors.

REFERENCES

[1] Guannan D, Haifeng W, Hui G, Guobiao G. Direct drive permanent magnet wind generator design and
electromagnetic field finite element analysis. IEEE Tran Appl Supercond 2010;20(3):1883-7.
[2] Kurt E, Gor H, Demirtas‚ M. Theoretical and experimental analyses of a single phase permanent magnet generator
(PMG) with multiple cores having axial and radial directed fluxes. Energy Convers Manag 2014;7:163-72.
[3] Wallace RR, Lipo TA, Moran LA, Tapiai JA. Design and construction of a permanent magnet axial flux
synchronous generator. Milwaukee: Electric Machines and Drives Conference IEEE; 1997. MA1/4.1-MA1/4.3.
[4] Cavagnino, A, Lazzari, M, Profumo, F, Tenconi, A. A comparison between the axial flux and the radial flux
structures for PM synchronous motors. IEEE Transactions on Industry Applications, 2002; 38(6): 1517-1524, DOI:
10.1109/TIA.2002.805572.
[5] I. Bouloukza, M. Mordjaoui, E. Kurt, G. Bal and C. Ökmen, “Electromagnetic design of new radial flux permanent
magnet motor,” Journal of Energy Systems, vol. 2(1), pp. 13-27, 2018.

282
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[6] H.J. Shin, J.Y. Choi, Y.S. Park, M.M. Koo, S.M. Jang and H. Han, “Electromagnetic vibration analysis and
measurements of double-sided axial-flux permanent magnet generator with slotless stator,” IEEE Transactions on
Magnetics, vol. 50(11), pp. 1-4, 2014.
[7] S. Javadi, and M. Mirsalim, “A coreless axial-flux permanent-magnet generator for automotive applications,” IEEE
Transactions on Magnetics, vol. 44(12), pp. 4591-4598, 2008.
[8] Praglowska, NR, Borkowski, D, Wegiel, T. Model of coreless axial flux permanent magnet generator. In:
International Symposium on Electrical Machines (SME); 18-21 June 2017: IEEE, Naleczow, Poland, DOI:
10.1109/ISEM.2017.7993568.
[9] Wu, D, Yang, J, Chan, T.T, Lai, C.S, et al. Axial-flux permanent-magnet synchronous generator with coreless
armature and non-integral coil–pole ratio. IET Renewable Power Generation, 2019; 13(2): 242-252, DOI:
10.1049/iet-rpg.2018.5311.
[10] Daghigh, A, Javadi, H, Javadi, A. Improved analytical modeling of permanent magnet leakage flux in design of the
coreless axial flux permanent magnet generator. Canadian Journal of Electrical and Computer Engineering, 2017;
40(1): 3-11, DOI: 10.1109/CJECE.
[11] Ishikawa, T, Amada, S, Segawa, K, Kurita N. Proposal of a radial- and axial-flux permanent-magnet synchronous
generator. IEEE Transactions on Magnetics, 2017; 53(6): 1-4, DOI: 10.1109/TMAG.2017.2665440.
[12] W,. Zhao, W, Lipo, T.A, Kwon, B.I. Comparative study on novel dual stator radial flux and axial flux permanent
magnet motors with ferrite magnets for traction application. IEEE Trans. Magn; 2014; 50(11), DOI:
10.1109/TMAG.2014.2329506.
[13] Gör, H, Kurt, E. Effects of back iron components on efficiency and generated power for new wind energy
Generators. Electric Power Components and Systems, 2018; 46(10):1103–1120,
DOI:10.1080/15325008.2018.148801.
[14] Shin, H.J, Choi, J.Y, Park, Y.S, Koo, M.M, Jang, S.M, Han, H. Electromagnetic vibration analysis and
measurements of double-sided axial-flux permanent magnet generator with slotless stator. IEEE Transactions on
Magnetics, 2014; 50(11): 1-4.
[15] Javadi, S, Mirsalim, M. A coreless axial-flux permanent-magnet generator for automotive applications. IEEE
Transactions on Magnetics, 2008; 44(12): 4591-4598, DOI: 10.1109/TMAG.2008.2004333.
[16] Sezenoğlu, C, Balıkçı, A. Design of axial flux permanent magnet generator for generator driven electromagnetic
launcher. In: 9th International Conference on Electrical and Electronics Engineering (ELECO); 26-28 Nov. 2015:
IEEE, pp. 595-598, Bursa, Turkey.
[17] Fei, W.Z, P.C.K, Luk. Design and performance analysis of a high-speed air-cored axial-flux permanent-magnet
generator with circular magnets and coils. In: IEEE International Electric Machines and Drives Conference; 3-6
May 2009: IEEE, pp. 1617 - 1624.
[18] Li J, Choi DW, Cho YH. Development of a natural cooled axial flux permanent magnet generator for wind turbine.
Hangzhou: Industrial Electronics (ISIE), 2012 IEEE International Symposium; 2012. p. 635-40.
[19] Bumby JR, Martin R. Axial flux permanent magnet generator for engine integration. In: Published at the 12. Int.
Stirling Engine Conf., Durham; 2005.
[20] Chan CC. Axial-field electrical machines-design and applications. IEEE Trans Energy Conv 1987;2(2):294-300.
[21] Vansompel H, Sergeant P, Dupre L. Optimized design considering the mass influence of an axial flux permanent-
magnet synchronous generator with concentrated pole windings. Magn IEEE Trans 2010;46(12):4101-7.
[22] Vansompel H, Sergeant P, Dupre L, Van den Bossche A. Axialflux PM machines with variable air gap. Ind
Electron IEEE Trans 2014;61(2):730-7.
[23] Bumby JR, Martin R. Axial-flux permanent-magnet air-cored generator for small-scale wind turbines. Proc IEE-
Electr Power Appl 2006;152(5):63-73.
[24] Qamaruzzaman A, Pradikta P, Dahono A. Analytical prediction of inductances of slotless axial-flux permanent
magnet synchronous generator using Quasi-3D method. Int J Electr Eng Inf 2009;1(2):115-25.
[25] Gholomian SA, Yousefi A. Power density comparison for three phase non-slotted double-sided AFPM motors.
Aust J Basic Sci 2010;4(12):5497-955.
[26] Wu, W. Xie, D., Lu, Y., Zhao, Z., Yu, S., Oscillation mode analysis considering the interaction between a DFIG-
based wind turbine and the grid, IOP Conference Series: Earth and Environmental Science, 52 (2017), 012007,
doi:10.1088/1755-1315/52/1/012007.
[27] Gör, H, Kurt, E. Preliminary studies of a new permanent magnet generator PMG with the axial and radial flux
morphology. International Journal of Hydrogen Energy, 2016; 41(17): 7005–7018,
DOI:10.1016/j.ijhydene.2015.12.195.
[28] Y. Ding, L. Ma, J. Ma, C. Wang and C. Lu, (2019), A Generative Adversarial Network-Based Intelligent Fault

283
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Diagnosis Method for Rotating Machinery Under Small Sample Size Conditions, IEEE Access, vol. 7, pp:149736-
149749.
[29] V. L. Merlin et al., (2016), A new artificial neural network based method for islanding detection of distributed
generators, International Journal of Electrical Power & Energy Systems, vol. 75, pp:139-151.
[30] M. B. K. Bouzid and G. Champenois, (2017), An Efficient Simplified Physical Faulty Model of a Permanent
Magnet Synchronous Generator Dedicated to Stator Fault Diagnosis Part II: Automatic Stator Fault
Diagnosis, IEEE Transactions on Industry Applications, vol. 53, no. 3, pp: 2762-2771.
[31] Jan-Hendrik Menke, Nils Bornhorst, Martin Braun, (2019), Distribution system monitoring for smart power grids
with distributed generation using artificial neural networks, International Journal of Electrical Power & Energy
Systems, vol. 113, pp:472-480.
[32] 32.S. A. Siddiqui, K. Verma, K. R. Niazi and M. Fozdar, (2018), Real-Time Monitoring of Post-Fault Scenario for
Determining Generator Coherency and Transient Stability Through ANN, IEEE Transactions on Industry
Applications, vol. 54, no. 1, pp: 685-692.
[33] Baptista, D., Abreu, S., Travieso-Gonzalez, C., & Morgado-Dias, F. (2017), Hardware implementation of an
artificial neural network model to predict the energy production of a photovoltaic system, Microprocessors and
Microsystems, Vol. 49, pp:77-86.
[34] Rahmanian, E., Akbari, H., & Sheisi, G. H. (2017), Maximum power point tracking in grid connected wind plant by
using intelligent controller and switched reluctance generator. IEEE transactions on sustainable energy, Vol. 8,
no.3, pp:1313-1320.
[35] İlbaş, M., Demirci, M., & Kurt, E. (2019). Modeling and experimental validation of flow phenomena for optimum
rotor blades of a new type permanent magnet generator. SN Applied Sciences, Vol.1, No. 12, pp:1544.
[36] Zidani, Y., Zouggar, S., & Elbacha, A. (2018), Steady-state analysis and voltage control of the self-excited
induction generator using artificial neural network and an active filter, Iranian Journal of Science and Technology,
Transactions of Electrical Engineering, 42(1), 41-48.
[37] Buck, J. A., Underhill, P. R., Morelli, J. E., & Krause, T. W. (2016), Simultaneous multiparameter measurement in
pulsed eddy current steam generator data using artificial neural networks. IEEE Transactions on Instrumentation
and Measurement, Vol. 65 No. 3, pp:672-679.
[38] Merabet, A., Tanvir, AA., Beddek, K., (2016), Speed control of sensorless induction generator by artificial neural
network in wind energy conversion system, IET Renewable Power Generation, Vol. 10, No. 10, pp:1597-1606.
[39] Agarwal, N. K., Sadhu, P. K., & Chakraborty, S. (2019), MPPT Based PMSG Wind Turbine System Using Sliding
Model Control (SMC) and Artificial Neural Network (ANN) Based Regression Analysis. IETE Journal of
Research, pp:1-9.

284
 NUMERICAL APPROXIMATION OF THE HYDROLOGICAL
TIME OF CONCENTRATION

Barrón-Fernández, J.R.
Department of Building, University of Extremadura, Cáceres, Spain,[email protected], ORCID: 0000-0001-8374-
2330

Calvo-Jurado C.
Department of Mathematics, University of Extremadura, Cáceres, Spain, [email protected], ORCID: 0000-0001-9842-
081X

Barrón-Fernández, J.R., Calvo-Jurado, C. Numerical approximation of the hydrological

Cite this paper as:
concentration time. 8th Eur. Conf. Ren. Energy Sys.24-25 August 2020, Istanbul, Turkey

Abstract: The time of concentration, i.e. the time it takes for a single "drop of water" to move
superficially from the most distant point of the watershed to the exit point, it is a
fundamental parameter of the hydrological analysis. Even more recently due to growing
concern for the environment and the responsible use of resources so that allow
optimizing the currently available water reservoirs. A number of studies have been
conducted to propose empirical formulas to calculate the time of concentration. One of
the best known is the Temez formula based on time series data collected in accounts in
Spain with areas of less than 3000 km2 . This expression uses the main channel length as
a parameter that in many works, for small slopes is approximated by the distance
between the geographic coordinates between the starting and ending points, leading for
larger catchments and slopes to approaches with a large error. In this work, using a
proper discretization of the curve, by using polynomial interpolation methods, we
improve the calculation of the length of the main channel and therefore, we provide a
more reliable method for calculating the time of concentration using the Temez
expression. We illustrate the proposed scheme with different numerical examples
comparing the results with those provided by other methods.
Keywords: Concentration term, main channel length, Temez formula, interpolation methods
© 2020 Published by ECRES

1. INTRODUCTION

In hydrological analyses, data and parameters are needed to quantify the variables studied by different methods
and models. An important and difficult parameter to study due to its variability is the concentration time, in the
following denoted as 𝑇𝑐 . According to some authors, the time of concentration can be defined as the time it takes
for a single drop of water to travel from the most distant point in the basins to the point of exit [1]. 𝑇𝑐 functionality
is defined as the design of rainwater systems, hydraulic structures using rational methods and operations in
hydraulic infrastructures [8]. There are a variety of empirical equations to calculate this parameter, some of
which are:

Bransby-Williams’s Model [10]

𝑇𝑐 = 0,2426 𝐿𝑐 𝐴−0,1 𝑆𝑐 −0,2 (1)

Where L denotes the length of main watercourse channel (km); Sc the average main watercourse slope (mm-0,1)
and A refers to the catchment area (km2).
Multiple regression derived from observations of floods in rural catchments in the region of India, for areas
smaller than 129,50 km2.

285
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Giandotti’s Model [3]

1,5 𝐿𝑐
𝑇𝑐 = 4√𝐴 + (2)
0,8√𝐻
where H is difference between the average elevation of the whole catchment area and the outlet section.
This expression developed through the study of twelve large catchment areas in northern and central Italy, for
areas between 170 and 70.000 km2.

Kirpich’s Model [7]

𝑇𝑐 = 0,0663 𝐿𝑐 0,77 𝑆𝑐 −0,385 (3)

It is deduced from multiple regression of rainfall time series in small agricultural basins in Tennessee and
Pennsylvania, for areas between 0,004 and 0,453 km2, and average slope between 3% and 10%.

Johnstone and Cross’s Model [6]

𝑇𝑐 = 0,0543 𝐿𝑐 0,5 𝑆𝑐 −0,5 (4)

This formula was developed from the study of nineteen sub-basins, located in the area of Scioto and Sandysky
(Ohio). The area of the basin must be between 65 to 4206 km2.

NRCS-SCS Method [11]

1,000
𝑇𝑐 = 0,057 ( − 9) 𝐿𝐶 0,8 𝑆𝐶 −0,5 (5)
𝐶𝑁
Where CN denotes the number of the run-off curve. This expression was developed by Soil Conservation Service
(SCS) which later became Natural Resources Conservation Service (NRCS). It is based on was based on data
collected from homogeneous agricultural catchments with areas of up to 8 km2.

Chow’s Model [1]

𝑇𝑐 = 0,1602 𝐿𝑐 0,64 𝑆𝑐 −0,32 (6)
This equation was deduced for twenty rural catchments in the U.S. whose area varies from 0,01 to 18,5 km2, and
average slope from 0,0051 to 0,09 (mm-1).

Temez’s Formula [9]

𝑇𝑐 = 0,3 𝐿𝑐 0,76 𝑆𝑐 −0,19 (7)

This formula was born from time series data collected in basins in Spain with areas less than 3.000 km2.

Ecs. [1-7] have as a common parameter, the average slope of the main channel Sc or the points of highest and
lowest elevation of the basin area H. In all of them, it is calculated in the same way for small basins as for large
basins [11, 1, 6, 7, 3, 10]. In small basins the approximate slope with the geographic coordinates between the
starting point and the end point may be acceptable, however for large basins the slope is very changeable along
the riverbed, and therefore, these methods and in particular Temez’s formula produce large errors in the
calculations of large basins. Young rivers are located in mountainous riverbeds with large slopes and a v-shaped
cross section and are irregular. Mature rivers have low slopes. Old rivers have very low slopes and no rapids or
falls [4]. For this reason, the calculations developed to obtain the average slope are not precise. For this reasons,
new reliable method for calculating Sc must be deduced. In this sense, the main objective of this study is to
propose another method of calculating mean slope with a greater adjustment than the current one and illustrate
its implementation by applying it to a real case in a province of Cáceres, Spain.

2. MATERIAL AND METHODOLOGY

In order to determine the parameter Sc and therefore the concentration time, different methods are proposed in
the literature. In [2], authors estimate Sc as follows
𝐻
𝑆𝑐 = 𝐿 (8)

286
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure1. Approximation of the average slope parameter proposed in [11].

where L’ has been approximated by the main channel length. Indeed they assume that the slope is enough small
to consider L’=L cos ~L which does not have to happen especially in large and irregular basins. For this reason,
in this work we propose a more accurate method based on the approximation of L by using different interpolation
methods. For this, in a first step, we need a discrete approximation of the terrain main channel orography using
proper software.

2.1 GvSIG software

The gvSIG software (v2.5.0.2930) is a geographic information system (GIS) based on free software, a project
promoted by the Valencian Community (Spain). Firstly, the cartography was downloaded from the spatial data
infrastructure (SDI), which was: National Plan of Aerial Orthophotography (PNOA) and Digital Terrain Model
with a 25 meters mesh pitch (MDT25), of the study area. The layers were loaded in gvSIG and several
geoprocesses were performed in the MDT25 layer. The first geoprocessing done to the layer was "depressions
removal", followed by "flow accumulation", "drainage network". At this point, the PNOA layer was visualized
to obtain the control point through a script called "coordinate capture". In the MDT25 layer, the geoprocess
"catchment through the control point" was executed. The PNOA layer was reloaded and with the mentioned
script a new starting coordinate of the main river was obtained. In the MDT25 layer, the "profile according to
flow line" geoprocess was executed to obtain a table with the coordinates of the points (x,y), the elevation (z),
the horizontal distance between the points (dh) and the real distance between the points (dr).

2.2 Numerical modeling of the channel

GVSIG has provides a set of n+1 geographical Cartesian coordinates D={(xi , yi , zi)}ni=0 where zi represent the
altitude or height in meters from the ground of a place located at the position (xi , yi ), i=0,…n. Considering the
orthogonal projection on the XZ plane and the following table of values,

Table 1. Channel riverbed profile values

X 𝑥0 𝑥1 … 𝑥𝑛
Z 𝑧0 𝑧1 … 𝑧𝑛

the path of the main channel can be approximated by a polynomial P of degree n. Thus, define the mean slope
as the mean of the channel slopes in each subinterval

∑𝑛 ′
1 𝑃 (𝑐𝑖 ) 𝑥𝑖 + 𝑥𝑖+1
𝑆𝑐 = 𝑛
where 𝑐𝑖 = 2
, i=0,….n. (8)

3. RESULTS

To illustrate the scheme proposed in section 2, using GvSIG software we obtain 79 cartesian coordinates that
describes the "Regato Celadilla" (Figure 2) river basin in Cáceres, Spain (see Figure 4 (a)).

287
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. (a) Hypsometric map of "Regato Celadilla" watershed (Cáceres, Spain). (b )Map of the "Regato Celadilla"
watershed (Cáceres, Spain), contour lines with river drainage network. (c) Map of the "Regato Celadilla" river basin
(Cáceres, Spain), points for calculating the average slope of the main riverbed.

Using the Lagrange polynomial interpolation we construct a 78 degree polynomial that as we can see in Figure
3 does nor provides good approximation of the riverbed in the at the extreme points of the interval

𝑃 = 418.831 − 2.699784037494087 × 1016 𝑥 + 5.642748677181165 × 1015 𝑥 2

− 5.408136737963023 × 1014 𝑥 3 + 3.194622101925719 × 1013 𝑥 4 − ⋯
+ 8.665166556017677 × 10−199 𝑥 76 − 2.047792108557352 × 10−203 𝑥 77
+ 2.376944085604347 × 10−208 𝑥 78

Figure3. Lagrange interpolation polynomial approximates the 79 data obtained with GVSIG.

As consequence, it is necessary provides better approximations of the set of data. In this sense we construct P
by using polynomial interpolation to pieces, of first, second and third degree (𝑆 𝑘 , 𝑘 = 1,2 ,3). We will denote
𝑆 𝑘 by the set of functions of k-differentiable that are polynomials to pieces of degree k in each of the partition
intervals:

𝑆 𝑘 (𝐷) = {𝑠𝑖 (𝑥) (𝑘 − 1) − 𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑡𝑖𝑎𝑏𝑙𝑒 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛: |𝑠𝑖 |[𝑥 ∈ 𝑃𝑘 }

𝑖,,𝑥𝑖+1 ] (9)

288
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

By imposing the differentiability conditions (9) at the ends of the intervals, 78 polynomials are obtained.

S3 = {s1= 418.831 − 0.0027525639608865877x 2 + 0.00007160785024784927x 3 ,

s2= 431.6293852364462, s3= − 1.5248274705853295x +
0.05780452383042911x 2 − 0.0007300480643843417x 3 ,… … , s78= −
(10)
22670.92377768599 + 30.401744113190002x − 0.013359675974951895x 2 +
0.000001956363805974555x 3 }

𝑆 2 = {𝑠1= 418.831 − 0.0009494782916457017𝑥 2 , 𝑠2= 435.53322304832756

− 1.32662613568924𝑥
+ 0.025393375872159698𝑥 2 , 𝑠3= 415.36083122632107
+ 0.1641466405740309𝑥 (11)
− 0.0021492612387162377𝑥 2 , … , 𝑠78= 35243.59637185137
− 30.35137681332042𝑥 + 0.006606066297746681𝑥 2 }

We can see that the approximation P obtained using cubic and one degree splines is the most accurate (see
Figures 4 and 5).

Figure 4. (a) Projection on the xy plane of the z coordinate (height of the point whose geographic coordinates are given by
(x, y)). The sample consists on 79 points. (b) Closed polygonal connecting the 79 points of the sample. (c) Quadratic splines
that interpolate the sample. (d) Cubic splines that approximate the points. Note that the curve described by the cubic splines
provides better approximations to the data.

Applying expression (8) to the approximations obtained in (10) and (11) it follows the following approximation
of the average slope 𝑆𝑐 parameter

Table 2. Values of 𝑆𝑐 using different methods

Splines-order 1 −0.015357939198492711 Splines-order 3 −0.01568477762546392
Splines-order 2 −0.015554835854883335 Temez method −0.015289636168351805

289
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 5. (a) Plot of a subset of the sample (29 points) where positive slopes can be observed in the interval [𝑥21, 𝑥22]. (b)
Approximation through the polygonal curve that joins them. (c) Approach using quadratic splines.(d) Interpolation using
cubic splines. Tangent lines have been drawn at the midpoints of the subintervals [𝑥7 , 𝑥8 ], [𝑥11, 𝑥12] and [𝑥21, 𝑥22] in
magenta (c) and black (d). (e) Note that cubic splines offer better approximations of the data and therefore of the slopes
(positives and negatives).

4. CONCLUSIONS

In this paper, a method to calculate the average slope has been proposed. Using a geographic information system
a sample of discrete information of the channel profile values of the river bed is generated. Using piecewise
cubic, quadratic and linear interpolation the riverbed is approximated. Using the lines tangent slopes to the
splines at the midpoint of each of the subintervals, we obtain a more precise value for the mean slope 𝑆𝑐 than
those commonly proposed in the literature. This allows to determine a more accurate estimation for the time of
concentration that constitutes a fundamental parameter of the hydrological analysis.

ACKNOWLEDGMENT
This work has been partially supported through Ministerio de Economía y Competitividad [MTM2017-83583-
P] of Spain and from the Junta de Extremadura through Research Group Grants [FQM-022].

REFERENCES
[1] Chow VT, Maidement DR, Mays LW. Applied hydrology. Singapore: McGraw-Hill, 1988.
[2] Departamento de Ingeniería Industrial e Ingeniería Civil. Caudal de Referencia. Método racional modificado de Temez.
Universidad de Cádiz.
[3] Giandotti M, Previsione dell epiene e dell emagre dei corsid’acqua, 107. Memorie e studi idrografici Servizio
Idrografico Italiano, Roma, Italia, 1934.
[4] Gracia-Sánchez J, Maza-Álvarez JA, Morfología de Ríos. Manual de Ingeniería de Ríos. México: Instituto de Ingeniería
de la UNAM, 1997. Chap. 11.
[5] Grimaldi S, Petroselli A, Tauro F, Porfiri M, Time of concentration: A paradox in modern hydrology. Hydrol. Sci.
J.2012; 57 (2): 217-228. https://doi.org/10.1080/02626667.2011.644244
[6] Johnstone D, Cross WP., Elements of applied hydrology. New York: Ronald Press Company, 1949.
[7] Kirpich ZP, Time of concentration of small agricultural watersheds. Civ. Eng1940; 10 (6): 362-368.
[8] Perdikaris J, Gharabaghi B, Rudra R., Evaluation of the sim- plified dynamic wave, diffusion wave and the full dynamic
wave flood routing models. Earth Sci. Res.2018; 7 (2): 14https://doi.org/10.5539/esr .v7n2p14
[9] Temez JR. Cálculo hidrometeorológico de caudales máximos en pequeñas cuencas naturales. Dirección General de
Carreteras, Ministerio de Obras Públicas y Urbanismo, España, 1978.
[10] Williams GB, Flood discharges and the dimensions of spillways in India. Engineering 1940; 134 (9): 321-322.
[11] Woodward time of concentration. In: C. C. Hoeft, A. Humpal, and G. Cerrelli, editors. Hydrology, national engineering
handbook. Washington, DC: USDA, 2010. Chap. 15.

290
 INFLUENCE OF TRAFFIC AND ROAD SURFACE PROPERTIES
ON ELASTIC BEHAVIOR OF LAYERED PAVEMENTS

Roldán-Oliden P.
Department of Mathematics, University of Extremadura, Cáceres, Spain, [email protected], ORCID: 0000-0002-5845-
9469

Calvo-Jurado C.
Department of Mathematics, University of Extremadura, Cáceres, Spain, [email protected], ORCID: 0000-0001-9842-
081X

Roldán-Oliden, P., Calvo-Jurado, C. Influence of Traffic and road surface properties on

Cite this paper as: elastic behavior of layered pavements.8th Eur. Conf. Ren. Energy Sys.24-25 August 2020,
Istanbul, Turkey

Abstract: In this paper we present the overall elastic properties of different layered road pavements. Using
some micromechanical theory, and specifically the Voigt/Reuss and Hashin-Strikman bounds, we
determine all the elastic moduli that characterize the stiffness of the composite medium. Following
the Spanish Highway Administration classification of the pavement sections, we will present in
detail the effective elastic behavior of any of the cases covered by this document from their
composition and the type of traffic. Although numerous experimental and laboratory works have
been developed to study the mechanisms that govern different types of pavements, there are few
works where they are explicitly determined. That is why the main novelty of this work consists in
presenting analytical methods capable of quantifying the performance of the pavement based on
the materials that compose it and the traffic loads to which it is subjected. We illustrate the
implementation of scheme in some scenarios by showing the dependence of the different elastic
modulus on the volume fraction of the different materials that made the layered composite
pavement.
Keywords: Pavements, layered materials, elastic properties, traffic loads, composite materials
© 2020 Published by ECRES

1. INTRODUCTION

Traffic is one of the most important variables in road design. The structural section of pavement is basically
conditioned by the average daily intensity of heavy vehicles, since those are the ones whom transmit the most
axle load and shorten the life of the infrastructure.
The high cost of construction and maintenance of road networks leads to numerous studies that evaluate the
resistance of different pavement sections [9], numerical studies where new reinforcing materials are introduced
[11] or research evaluating sections of recycled pavements [15].
In the last decades, many authors have focused on the development of finite element models (FEM) that
faithfully represent the properties of pavements [2], [4]. However, the application of these models is very
expensive due to the complex structure of these materials.
The main objective of all those studies is to provide to the designer with standards that make easier the
relationship between life, traffic and layer thickness.
The Spanish Government (BOE, 2003) suggests a catalogue of pavements relating to heavy traffic intensities
and permissible levels of damage for each type of soil. Moreover, for each previous combination proposes
different sections based in the different materials that compose it.
However, as far as the authors are aware no specifications about the elastic properties of the proposed composites
are available. For this reason, the aim of this work will provide explicit predictions of the anisotropic elastic
performance of the pavement as function of the elastic properties of the mixture components and the traffic
category.

291
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2 STATEMENT OF THE PROBLEM

The catalog of firm sections of the Spanish Highway Administration [3] publishes the relationships in each type
of structural section, between the intensities of heavy traffic and the types of esplanade. By means of a series of
tables such as those indicated in Figure 1, this document lists the construction solutions according to the case as
well as the minimum thickness of the layers.

Figure 1. List of pavement sections for heavy traffic categories T3 and T4, depending on the esplanade category.
MB: bituminous mixture, HF: rich concrete, SC: Cement stabilized soil, ZA: artificial aggregate (minimum thickness in
cm)

Each section is designated by three or four numbers. The first (if it is three digits) or the first two (if it is four
digits) indicate the category of heavy traffic, from T00 to T42. The penultimate of the figures represents the type
of esplanade, while the last one refers to the type of pavement that is denoted as follows:

1. Bituminous mixtures on granular layer.

2. Bituminous mixtures on cement stabilized soil.
3. Bituminous mixtures on gravel cement built on cement stabilized soil.
4. Concrete pavement

Despite this, although the construction standards try to reduce the effects of traffic or environmental loads, the
specific conditions or mechanisms that govern the behavior of pavements are, at least, unclear and in what refers
to the standard [1] have remained practically without quantification. That is why in the following section we
intend to predict the elastic behavior of the materials described in Table 1 through the characterization of all the
elastic constants that are involved in Hooke's Law, i.e.:

𝟏 𝜹𝒖 𝜹𝒖𝒋 (1)
(x)=C(x)e(x), 𝐞𝒊𝒋 (x)=𝟐 (𝜹𝒙𝒊 + 𝜹𝒙 )
𝒋 𝒊

satisfying the equations of elastostatics

- div (x)=f in 𝛺 C(x)e(x)(x)=𝒖

̌ on ∂𝛺 (2)

In Eq. [2] div is the divergence operator ∂𝜴 the boundary of the heterogeneous domain, the outer unit normal
to∂𝜴 , e is the linear strain tensor, u is the displacement field and f is the force field. In Eq. [1]  denotes the
Cauchy stress tensor and C=C(x) (we work in Cartesian coordinates x=(𝒙𝟏,𝒙𝟐,𝒙𝟑)) that in our case, for a
transversely isotropic material is related to the engineer constants as follows

292
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝐶11 𝐶12 𝐶13 0 0 0

𝐶12 𝐶11 𝐶13 0 0 0
𝐶13 𝐶13 𝐶33 0 0 0
𝑪= 0 0 0 𝐶44 0 0
0 0 0 0 𝐶44 0 (3)
𝐶11 − 𝐶12
( 0 0 0 0 0
2 )

that in terms of the engineering constants (E: Young modulus, v= Poisson modulus) can be expressed as
1−𝜈23 𝜈32 𝜈21 −𝜈31 𝜈23 𝜈31 −𝜈21 𝜈32
𝐸1 𝐸1 𝐸1 0 0 0
Δ Δ Δ
𝜈21 −𝜈31 𝜈23 1−𝜈23 𝜈32 𝜈31 −𝜈21 𝜈32
𝐸1 𝐸1 𝐸1 0 0 0
Δ Δ Δ
𝜈31 −𝜈21 𝜈32 𝜈31 −𝜈21 𝜈32 1−𝜈12 𝜈21
𝐸1 𝐸1 𝐸3 0 0 0
Δ Δ Δ
C= 𝐸1
0 0 0 0 0 (4)
2(1+𝜈31 )
𝐸1
0 0 0 0 0
2(1+𝜈31 )
1−𝜈23 𝜈32 𝜈 −𝜈 𝜈
𝐸1𝐻 −𝐸1𝐻 21 31 23
Δ Δ
( 0 0 0 0 0 )
2

Where Δ = 1 − 𝜈31 𝜈21 − 𝜈23 𝜈32 𝜈31 𝜈13 − 2𝜈21 𝜈32 𝜈13 and 13 = 𝐶44 , 12 = 𝐶66 are the in-plane and anti-plane shear
modulus respectively.

2. METODOLOGY

In order to determine the effective properties of the layered pavement we will use some micromechanical
methods.
The main reason is the complex geometric structure of these kind of materials, usually depending of small
parameters  (Figure 2) describing the size of the heterogeneities or the periodicity of its structure. It makes that
the application of the usual numerical methods for the resolution of the partial differential equations (EDPs) that
model them, it turns out impossible to implement. From the numerical point on view it is necessary to carry out
some type of average for its study. Micromechanical techniques can be useful when looking for global
information, that is, at macroscopic level, in a medium heterogeneous from which microscopic information is
obtained. Key results in this scientific field are due to Voigt [13] and Reuss [12] that together with other authors
([5], [6], [7], [14]) and since the 1960s, they have studied the properties of the effective composite materials,
especially fiber reinforced and laminated. Voigt-Reuss bounds estimate the isotension and isodeformation states
of the composite as upper and lower dimensions of the elastic properties of the same. Specifically, for composites
made of two different materials (Figure 1), in the following denoted by phases denoted by C1 and C2 the effective
properties C∗ of the composite can be determine by

CR ≤ C∗ ≤ CV (5)
where
C R = (1 (C1 )−1 + 2 (C 2 )−1 )^(−1)C V = 1 C1 + 2 C 2 (6)

Motivated by knowing the behavior of fiber-reinforced materials, in the early 1960s, numerous works were
carried out. In particular, Hashin and Shtrikman (HS) [6] through a variational principle for elastostatics
presented bounds on the effective bulk and shear moduli that are currently known as Hashin–Shtrikman bounds.
For example, their expressions for bulk modulus [7] are given by

k L ≤ C∗ ≤ k U (7)

2 1
k L = 𝑘1 + 1 31 , k U = 𝑘2 + 1 32
(8)
+ +
𝑘2 −𝑘1 3𝑘1 +41 𝑘1 −𝑘2 3𝑘2 +42

293
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

These estimates depend on the microstructure of the problem and therefore provide finer dimensions than the
Voigt-Reuss bounds that only depend only on the phase volume fraction (and phase material properties).

Figure 2. Heterogeneous layered composite with small size of the inhomogeneities

In general, (HS) bounds that appear in the literature are almost always merely "stated" and it is often unclear
how to construct them when the material is not of isotropic medium. In this work, using the shorthand notation
and basis tensor defined in [14] we will deduce the effective properties for the two-phase layered composites
described in Figure 1. Assuming that the inclusion 𝜴 has an ellipsoidal shape, these expressions predict the
behavior inside the composite through the socalled concentration tensor Ain the following way

e|𝜴 = 𝐀e, 𝐀 = (𝐈 + 𝐏(C m − C  ))−𝟏 (9)

where I is the fourth order identity tensor and P is the Hill Tensor (also known as P-tensor).

For layered materials, from Backus’ expressions [1] the P-tensor is given by
e|𝜴 = 𝐀e, 𝐀 = (𝐈 + 𝐏(C m − C  ))−𝟏 (10)

k −l2 1 (11)
𝐏 = (0,0,0, 2
, 2
, 0, ),
−l + kn n(−l + kn) 2p
Where the parameters are k, l, n, p are defined as in Hill [8].

3. NUMERICAL RESULTS

In this section, using proper numerical software we illustrate the above schemes showing the elastic properties
of composite pavements described in Figure 1. To do it we will use the properties described in Table 1
empirically determine by several authors [10]

Table 1. Elastic constants of materials for pavements ([10])

Material Young Module E (GPA) Poisson Coefficient ν
Bituminous mixtures 6 0.35
Asphalt concrete 1.2 0.25
Cement stabilized soil 1.5 0.25
Granular layer 0.46 0.35

First, in Figure 3 using the expressions Eqs. [9,10] we consider the case of transversely isotropic behavior of
bituminous pavements (host phase) reinforced with artificial aggregate for different volume fractions of the
inclusion phase (aggregate). Effective anti-plane 13 and plane 12 shear moduli are presented showing the
improvement of
Hashin-Strikman bounds over Voigt-Reuss ones. Analogously in Figure 4 bounds on bulk modulus for different
esplanades are represented. The compressibility modulus for the anisotropic laminate are plotted depending on
the volume fraction inclusion of artificial aggregate and cement stabilized soil.

294
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Hashin-Strikman (black solid lines) bounds of the effective shear13 , 12 modulus of a transversely isotropic
pavement, made of a bituminous mixture reinforced with artificial aggregate with the volume fraction indicated in Figure
1 (𝑇3111 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙). Voigt-Reuss bounds are also plotted (red dashed lines) notingtheimprovementoftheHashin-
StrikmanboundsoverVoigt-Reussones.

Figure 4. Hashin-Strikman (black solid lines) bounds of the bulk modulus of a transversely isotropic pavement, made of
a bituminous mixture reinforced with artificial aggregate (on the left, 𝑇3111 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙) and cement stabilized soil (on the
right, 𝑇3112 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙) with the volume fraction indicated. Voigt-Reuss bounds are also plotted (red dashed lines). Note
that Hashin-Strikman bounds (solid lines) are tighter than Voigt-Reuss ones (dashed lines).

For simplicity we will present the elastic modulus of one pavement of each type of the classification.

Table 2. HS and Voigt-Reuss bounds on the bulk modulus of pavement sections for different heavy traffic categories
BULK Reuss bound Voigt bound HS lower bound HS upper bound
MODULUS
( GPa)
T3111 1.1573333333333333 0.33110556869180846 0.46730435637747614 0.606110085295573
T3121 1.3761975308641978 0.37105601735896787 0.5458884306522523 0.7138774415116693
T3131 1.2968592592592594 0.3555067120246861 0.5160312966108286 0.6735979705187106
T3212 1.1330074074074075 0.5264848762861927 0.6406396923533942 0.6839809833983175
T3222 1.0407407407407407 0.5061370958427021 0.6050448942804901 0.6424530377247044
T3232 1.2201481481481482 0.5473311106506145 0.6761350287193165 0.7252347193512455
T4114 0.2735802469135802 0.2642891506769284 0.2859116015670019 0.2864751053512516
T4122 0.9111111111111111 0.29533325307973196 0.39121230406456453 0.4955381321012999
T4132 1.0205432098765432 0.3102296542456776 0.42362127710823216 0.5434875467438034
T4214 0.2717234567901235 0.2626380266039726 0.28372732559803515 0.28438773977678217
T4222 0.8869629629629627 0.47562533401963875 0.54971219972463 0.5775791582094717
T4232 0.24572839506172842 0.24099372505066424 0.251246299503611 0.25228013240392044

295
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 3. Effective axial and transverse Young’s modulus HS bounds on the bulk modulus of pavement sections for
different heavy traffic categories
BULK Effective Young’s Effective Young’s HS lower bound HS upper bound
MODULUS modulus (𝐸12 ∗
) ∗
modulus (𝐸13 ) (shear modulus 12 ) (shear modulus 12 )
( GPa)
T3111 0.9020712977909856 0.9458528605461283 0.41170015926429887 0.7587108499365659
T3121 1.0700287890143534 1.0716134989985673 0.4941536380645635 0.9112116115571373
T3131 1.0062637257603282 1.0229354349635293 0.46267316900172045 0.8552526814428191
T3212 1.7606165209983542 2.153023783830408 0.9219746920124119 1.0034265832261922
T3222 1.674700583719813 2.0749610422242757 0.8771697604788169 0.9503410851606532
T3232 1.8445564595609092 2.231943160299745 0.9655788016583235 1.0542081629951727
T4114 0.6213051571300484 0.8434774413530317 0.29766913177506404 0.31287197385335597
T4122 0.7393330710597408 0.830889259564751 0.33328787135060745 0.5939254963530239
T4132 0.8086216669712566 0.879095593505824, 0.36649246012993364 0.6663044367055164
T4214 0.6111692419319678 0.8308905387342802 0.29171449606549776 0.30673806898789635
T4222 1.5377261785531957 1.955930783509008 0.8054460120158035 0.8633904753428512
T4232 0.47838306172925976 0.6822942654116554 0.21638303122320313 0.22477477595445045

4. CONCLUSIONS
Explicit estimations for elastic properties of heterogeneous pavements are presented. The approximations are
based on the implementation of the Hashin-Shtrikman bounds using transversely isotropic tensors for layered
materials. Following the Spanish Highway Administration classification of the pavement sections, the proposed
scheme easily allows the prediction of pavement performance as function of the elastic properties of the mixture
components and the traffic category.

ACKNOWLEDGMENT
This work has been partially supported through Ministerio de Economía y Competitividad [MTM2017-83583-
P] of Spain and from the Junta de Extremadura through Research Group Grants [FQM-022].

REFERENCES
[1] Backus, G. Long-wave elastic anisotropy produced by horizontal layering. J. Geophys. Res 1962, 67:4427–4440.
[2] Beskou, ND; Tsinopoulos, SV, Theodorakopoulos, DD. Dynamic elastic analysis of 3-D flexible pavements under
moving vehicles: A unified FEM treatment. Soil Dynamics and Earthquake Engineering 2016, 82: 63-72.
[3] BOE. ORDEN FOM/3460/2003, de 28 de noviembre, por la que se aprueba la norma 6.1-ICSecciones de firme de la
Instrucción de Carreteras.
[4] Cho, YH; Mccullough, BF, Weissmann, J. Considerations on finite-element method application in pavement structural
analysis. Transportation Research Record 1996, 1539: 96-101.
[5] Eshelby JD. The determination of the elastic field of an ellipsoidal inclusion. Proc. R. Soc. Lond. A. 1957, 241:376–
396
[6] Hashin Z, Shtrikman S. On some variational principles in anisotropic and non-homogeneous elasticity. J Mech Phys
Solids 1962, 10:335–342.
[7] Hashin Z, Shtrikman S. A variational approach to the theory of the elastic behaviour of multiphase materials. J. Mech.
Phys. Solids.1963, 11:127–140.
[8] Hill R. Theory of mechanical properties of fibre-strengthened materials: I. Elastic behaviour. J Mech. Phys. Solids 1964,
12:199–212.
[9] Kraemer, C, Albelda, R. Evaluación técnico-económica de las secciones de firme de la Norma 6.1-IC. En VI Congreso
Nacional de Firmes: Normalización e Innovación, Asociación Española de la Carretera, Madrid. 2004.
[10] Peng, Y, Yunlong, HE. Structural characteristics of cement-stabilized soil bases with 3D finite element method.
Frontiers of Architecture and Civil Engineering in China, 2009, 3: (4), pp. 428.
[11] Perkins, SW. Constitutive modeling of geosynthetics. Geotextiles and Geomembranes, 2000, 8:(5) 273-292.
[12] Reuss, A. Calculation of the flow limits of mixed crystals on the basis of the plasticity of mono-crystals. Z. Angew.
Math. Mech. 1929, 9:49–58.
[13] Voigt W. Ueber die Beziehung zwischeden beiden Elasticitats constanten. Annalen der Physik. 1889, 38: 573-587.
[14] Parnell, WJ, Calvo-Jurado, C. On the computation of the Hashin–Shtrikman bounds for transversely isotropic two-
phase linear elastic fibre-reinforced composites, J. Eng. Math. 2015, 95:295–323.
[15] Zhao, Sheng, et al. Comparative evaluation of warm mix asphalt containing high percentages of reclaimed asphalt
pavement. Construction and Building Materials. 2013, 44: 92-100.

296
 SUPER TWISTING ALGORITHM- BASED DISCRETE-TIME
CONTROL OF INDUCTION MOTOR

Farhi Salah Eddine

Laboratory of Electrical Engineering and Automatic, LGEA, Larbi Ben M’hidi University, Oum El Bouaghi, Algeria
[email protected], ORCID: 0000-0002-0465-1527

Sakri Djamel
Laboratory of Electrical Engineering and Automatic, LGEA, Larbi Ben M’hidi University, Oum El Bouaghi, Algeria
[email protected], ORCID: 0000-0002-4694-810X

Goléa Noureddine
Laboratory of Electrical Engineering and Automatic, LGEA, Larbi Ben M’hidi University, Oum El Bouaghi, Algeria
[email protected], ORCID: 0000-0002-1414-3075

Farhi, SE, Sakri, D, Goléa, N. Super Twisting Algorithm- Based Discrete-Time Control of
Cite this paper as:
Induction Motor. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: This paper presents a discrete-time modeling and control of three-phase induction motors (IM).
IM is modeled by a variational integrator formulation in stationary reference frame and the control
objective is to ensure a decoupling between flux and speed.
Standard sliding-mode control (SMC) is a style of variable structure control (VSC) that is robust
to parameters variation and different system uncertainties. However, higher-order sliding mode
(HOSM) is one of the robust controls by SMC, for his ability to overcome chattering phenomenon
which appear in all traditional control based on sliding mode approach.
In this paper, a discrete-time super-twisting like algorithm (DSTA) is used to control of the IM
with an integral action. Moreover, two observers are presented in discrete-time domain, the first
one for non- measurable rotor fluxes designed on DSTA while the second observer is applied to
estimate the external load torque. The proof of the stability of algorithms is also offered. Results
of digital simulation are performed out to show that the discrete-time HOSM provides a high-
performance dynamic and robustness against speed and load torque variations.
Keywords: induction motors (IM),higher-order sliding mode (HOSM),discrete-time super-twisting algorithm
(DSTA),observer
© 2020 Published by ECRES

Nomenclature
Vs , k Stator voltages
 r ,k Rotor flux vector
is , k Stator current vector
k Electrical rotor speed
Ls , Lr , M Stator, rotor and mutual inductance
Rs , Rr Stator and rotor resistance
Ts Sampling time
k Sampling instant
p Number of pole pairs
J Moment of inertia
F Friction coefficient
Tm , k , TL , k Motor torque and load torque

1. INTRODUCTION

Three-phase induction motor (IM) is one of the most desired in modern industries, due to its reliability, simple
mechanical design, ease of maintenance and more important low cost compared to the DC motor [1].
However, control of IM is a difficult task since its behavior is nonlinear. In the literature, numerous non-linear
control approaches have been utilized to develop their performance [2], e.g. SMC which is one of the most well-

297
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

known methodologies in the control of electrical motors due to its precise robustness in instances where the
uncertainty and disturbances existed in the system [3], [4].
Indeed, the emergence of digital technology has seen a significant revolution in the applications of advanced
control systems based on the computer [5], which is especially true with developed software like digital signal
processors (DSP) or microprocessors, hardware solutions like field-programmable gate arrays (FPGA) [6].
Typically, these devices are used for executing control algorithms. In digital computer-based control, most
algorithms are set up in continuous-time, and then discretized for a digital processor using the symplectic Euler,
forward Euler, or Tustin approximations methods [7]. Often Explicit Euler method is used for approximation of
the discrete-time control to the continuous-time [8]. So that, higher response and better robustness can be
obtained by designing directly the control system in the discrete-time [7].
Control of IM by digital control did not get much interest since the discrete model of the IM is difficult to reach,
regardless of sampling rates [9]. Variational integrator formulation overcame these problems and provided a
great discrete model better than the model developed by the explicit Euler method with lower sampling rates,
this is an important fact for IM control algorithms [10].
One of the first discrete-time integral SMC applications for speed control of IM was presented in [11], and one
of the new algorithms is introduced in [12] known as discrete-time super-twisting-like algorithm (DSTA). This
name was used to take into account the likenesses with the Euler discretization applicable to the continuous STA
version. DSTA algorithm has a great property when it is employed as a controller or an observer despite the
existence of disturbances and parametric uncertainties in the system [13]. Also, this algorithm is characterized
by having a relatively low switching frequency.
Therefore, this paper is structured as follows: in the first section, the discrete-time model for IM in (α.β) axes
developed by Rivera in [10] is presented. In the second section, the control of IM using DSTA is developed. In
the last section, two observers are designed in the discrete-time domain to improve the control technique, the
first one for rotor fluxes while the second ensure the estimation of external load torque.

2. DISCRETE-TIME INDUCTION MOTOR MODEL

The discrete-time model for IM in the stationary reference frame (α.β) presented by Rivera in [10] can be
written in the form :
r , k 1   d e r , k  M  d Ts e is , k
Ts
is , k 1  yd is , k   d  d r , k   d  d2 e (r , k  M  Ts is , k )  Vs , k (1)
d
F T
k 1  k  Ts  k  d Ts isT, k I1r , k  s TL , k
J J
The electromagnetic torque of motor is given as follows:

Tm, k  d JisT, k I1r , k (2)

T T T
Where r , k  r , k r  , k  , is , k  is , k is , k  ,Vs , k  Vs , k Vs ,k 
I1 is a skew-symmetric matrix, and e is:

cos(Ts pk )  sin(Ts pk )  0 1

e    , I1   
sin(Ts pk ) cos(Ts pk )  1 0 

The IM coefficients are:

1 R 3 M M RT M2 M2
d   d ,  d  ,  r ,   p , d  , yd  1  s s ,  d     d2 Ts ,   Ls 
1  Ts Lr 2 JLr Lr  d d Lr Lr

The study and the comparison between this model and the model derived with explicit Euler method can be
found in [10]. In the next section, we'll design the control with the discrete SMC to demonstrate finite-time
convergence for controlled variables. The parameters of IM are shown in Appendix.

298
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. DISCRETE-TIME SLIDING MODE CONTROL

SMC is a particular type of VSC, the main idea of this control is that once the sliding surface was reached, the
system is immune to disturbances and parametric variations [2], [14]. For an IM, the general goal is to control
the rotor flux and the speed of the motor [15]. We will design a powerful controller treated in [16] which depends
on a discrete-time version of the Super-Twisting developed in [12].
The output tracking error is:

z1, k  yk  yr , k (3)

Where z1, k   z11, k z12, k  , yk  k m, k  and m,k  rT,kr ,k .

T T

The controller will be designed to force yk to the reference y r , k , by taking one step ahead, the dynamics were
obtained:

z1, k 1  f1, k  DGk is , k (4)

Where f1,k  f k  d k

 F   Ts 
 k  Ts k  r , k 1    TL , k    T 0  r  , k r , k 
fk  dk  D d s G 
 
J J
 2     r , k r  , k 
 d m, k  m, k 1  ( M  d Ts ) im, k 
2 0

With   2M  d2Ts and im,k  isT, k is, k is the stator current square modulus. Now, the following assumptions will
be used in the control design [16]:

TL , k 1  TL , k (5)

Due to a sinusoidal distribution of the stator voltages:

im, k 1  im, k (6)

Returning to Eq. [4], the desired dynamics can be assigned imposing:

z1, k 1  f1, k  DGk is , k  k1 z1, k (7)

Where
 k11 0 
k1   k11 , k12  1
0 k12 

From Eq. [7], the stator reference current is introduced as:

is , r , k  Gk1 D 1   f k  k1 z1, k  (8)

Note that matrix D is not singular and Gk1   m, k , we can always choose the initial conditions r , k (0) to
guarantee Gk1 exists. From Eq. [8], an integral action is added to cancel the effect of any disturbance on the
follow-up of the reference:

is , r , k  Gk1 D 1   f k  k1 z1, k  ki zi , k 
(9)
zi , k 1  zi , k  Ts z1, k

T  ki1 0 
Where zi , k   zi 1, k zi 2,k  , ki    , ki1 , ki 2 are constant gains.
0 ki 2 

299
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In order to control of induction motor, we use DSTA with the reference and measured current, we define the
current tracking error:

z2, k   z21, k z22, k   is , k  is , r , k

T
(10)

The control law for the system based on discrete-time Super-Twisting like algorithm is given by:

Vs , k  M1 0 sign( z2, k )  z3, k

(11)
z3, k 1  z3, k  Ts  M 2 sign( z2, k )  q0 z3, k 

Where:
 1

 z21, k 0 
2
T
0   1  , sign( z2, k )   sign( z21, k ) sign( z22, k ) 
 0 z22, k 2 
 

By substituting Eq. [10] and Eq. [9] in Eq. [7], the closed loop of transformed system Eq. [7] is:

z1, k 1  k1 z1, k  ki zi , k  DGk z2, k  d k

(12)
zi , k 1  zi , k  Ts z1, k

The stability analysis of the DSTA control can be found in [16] or in [12].

4. ROTOR FLUX ESTIMATION

Generally, the direct measurement of rotor flux is not realized in practice due to its difficulty [17].The state
equation model of the motor is an alternative way to estimate this value, through utilization of an easily
measurable quantities such as voltage and current [18]. In order to estimate the rotor flux, we suggest observer
proposed in [16] which is designed using DSTA. The observer model equations (  ,  coordinate) is:

ˆr , k 1   d e ˆr , k  M  d Ts e is , k  kv vk
T (13)
iˆs , k 1  yd iˆs , k   d  d ˆr , k   d  d2 e (ˆr , k  M  Ts is , k )  s Vs , k  vk
d

ˆr , k , iˆs , k are the rotor flux and stator currents estimated values, vk is the injected signal to the observer.
The tracking errors are defined as:

r , k  r , k  ˆr , k
(14)
is , k  is , k  iˆs , k

By taking one step ahead, dynamics of the tracking errors are:

r , k 1   d e r , k  kv vk
(15)
is , k 1  yd is , k   d  d r , k   d  d2 e r , k  vk

The switching function (sliding function) is chosen as is , k . The DSTA to select vk is defined by:

vk  M 3 1sign( is , k )  v1, k
(16)
v1, k 1  v1, k  Ts  M 4 sign( is , k )  q1v1, k 

Where:

300
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

 1

 is , k 0 
2
T
1   1  , sign( is , k )   sign( is , k ) sign( is  , k ) 
 0 is , k 2 
 

The coefficients M 3 , M 4 , q1 are chosen according to requirements of the sliding mode system as in [12]. When
the sliding mode occurs by Super-Twisting algorithm, the equivalent injected signals are determined from
is , k 1  0 :

veq, k  d dr ,k  dd2 e r ,k  1 (17)

The ideal sliding mode dynamics 1  0 .By replacing Eq. [17] in Eq. [15]:

r ,k 1  c1e r ,k  c2r ,k (18)

Where:
c1   d (1   d  d kv )
c2   d  d kv

To ensure global stability, the Lyapunov function is proposed as:

Vk  rT,kr ,k (19)

The increment of such function is:

Vk  rT, k 1r , k 1  rT, k r , k

(20)
 (c12  c2 2  1  2c1c2 cos  )rT, k r , k

Vk is negative if the following condition is satisfied:

1 1 1  d
  kv  (21)
d d 1   d

5. LOAD TORQUE OBSERVER

The load torque can be taken as an external disturbance or as a system uncertainty case when it is unidentified
or is very variable over time [19].This may influence the efficacy of the control that must be robust under this
uncertainty. Since the flux can be estimated, speed rotor and stator current are available by measures, therefore
the load torque can be estimated using the mechanical equation of the motor model [20]. Moreover, we are
interested to estimate load torque using an observer designed in [8]. The observer model is described as follows:

ˆ ˆ F ˆ   T iT I ˆ  Ts Tˆ  l (  
ˆ )
k 1   k  Ts  k d s s,k 1 r ,k L,k 1 k k
J J (22)
TˆL , k 1  TˆL , k  l2 (k  
ˆ )
k

ˆ , Tˆ are the rotor speed and load torque estimated variables. The estimation errors can be expressed as:
 k L,k

ˆ
k 1  k   k
(23)
T  T T ˆ
L, k L,k L,k

From Eq. [22] and Eq. [23] the dynamics of the observation errors are:

301
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

F T
k 1  k  Ts  k  d Ts isT, k I1r , k  s TL , k  l1 k
J J (24)
TL , k 1  TL , k  l2 k

In matrix form:

xk 1  Gxk  H r , k (25)

Where:

 F Ts 
1  Ts l1   d Ts isT, k I1 
xk  k TL , k  G   J H  
T


J
 
 0 
 l2 1 

Since r , k 1  0 through Eq. [21], system Eq. [25] can be reduced as:

xk 1  Gxk (26)

The characteristic polynomial is given by:

F F T
det( zI - G )  z 2  (l1  Ts  2) z  (1  Ts l1  l2 s ) (27)
J J J

Where I is 2 × 2 identity matrix.

In order to determine values for l1 and l2 , the coefficients of the polynomial Eq.[27] are compared with a desired
characteristic polynomial. If the desired poles are defined by:

D( z )  z 2  ( p1  p2 ) z  p1 p2 (28)

With poles p1 and p2 both inside the unit circle in the complex z–plane. By comparison, the solution allows
calculating the coefficients:

F
l1  p1  p2  2  Ts
J
(29)
J
l2   ( p1 p2  p1  p2  1)
Ts

The block diagram of DSTA control applied to IM drive is shown as in figure 1:

Figure 1. Diagram of the control scheme.

302
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

6. SIMULATION AND DISCUSSION

To validate the control technique based on DSTA, simulations have been carried out by means of Matlab-
Simulink software. The parameters of induction motor and DSTA control are given in Appendix. Simulation
results are shown through various figures:

Figure 2. Speed response with sampling time variation.

In this subsection, the initial task is to study the behavior of the DSTA controller with different sampling time
values, and without application of load torque. It is mention that for the flux observer, the initial condition is set
different from zero ˆr , k (0)  0 to avoid the singularities in Eq. [09]. The gains values of DSTA controller are
selected and presented in Appendix.
Figure (2) displays the output tracking of the rotor speed with an increase in sampling time value. For Ts =10μs
figure 2 (a) shows a good tracking of the response. Even in the case in increased the sampling time Ts =50μs the
sampled dynamics of the IM are not affected by this augmentation and maintains its response.
In figure 2 (c) when Ts =100μs we can observe a few oscillations in response (zoomed interval). These
oscillations disappear in a transitional regime and response still performs well. In figure 2 (d) when Ts =150μs ,
the response is affected but tracks its reference properly.
It is visible that the measured speed for the 4 sampling time values has a fast convergence and reaches its rated
reference value (equal to 148.7 rad/s) within a short time approximately t=0.12s which is illustrated by the zoom.
For more accuracy, these results can be evaluated through presenting the mean value of the error between the
reference and measured speed in each case of sampling time variation. As shown in table 1, it can be noticed
that the mean error value at Ts =10μs is remarkably smaller equal 0.002116rad / s , after the increases in the
sampling time, this value slowly increasing to reach 0.08943rad / s for Ts =100μs . In the case when Ts =150μs ,
we get great error value but the output tracking maintains its precision.

Table 1. The mean value of speed error in steady-state

Sampling time Mean error  rad / s 
 
 s
10 0.002116
50 0.06285
100 0.08943
150 0.2158

This sampled dynamic model can be given good results and better accuracy for sampling periods above 500 μs
without losing its accuracy, this is what distinguishes the variational integrator formulation from other methods.

303
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Speed response. Figure 4. Estimated rotor flux.

In order to show the stator current and estimated rotor flux responses, the used sampling time is set at 10 μs:
Figure (3) presents the stator currents response with a small sampling period. As usual in electrical drives at the
starting, both currents present an initial peak with few oscillations during the transient period around 0.12s,
which are admissible values. At steady –state, stator currents stabilize at an approximate value of 4A which
represents the no-load current.
In figure (4), the estimated rotor flux is illustrated with sampling time relatively small. The reference rotor flux
is fixed at its rated value of r , m, k  1Wb . As we can notice, the two components are perfectly in quadrature with
identical amplitudes and without a chattering phenomenon. We can see that the observer estimates the discrete
rotor fluxes with a good waveform, especially in the permanent regime (starting from t=0.1s), the zoom shows
this accurate estimate.
In this a second subsection, the control scheme is tested through several profiles of speed (-148.7/+50/+148.7
rad/s) and for three values of load torque (0/10/5 Nm), the sampling time is kept at 10 μs:
Figure 5. (a) shows the evolution of measured rotor speed compared to a reference speed profile. It is seen that
after a transitory time about 0.12s, the measured speed follows the reference perfectly and has good tracking. At
time t=0.6 s when the rated load torque is applied or halved at t=0.8s, the speed shows a small attenuation, which
disappears thereafter to regain its reference value. The zoom illustrate that the chattering is negligible in this
response.

(a) (b)
Figure 5. (a)Speed tracking performance.(b) Estimated load torque and motor torque tracking performance.

304
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 6. Rotor flux observer under speed variation and load torque application.

Figure 5. (b) shows the estimation of load torque and motor torque for several values of an external disturbance
load. At t = 0.6 s, the nominal load torque is applied and again decreases to half at t = 0.8 s. The curve obtained
shows the increases and decreases of the electromagnetic torque according to these changes, however, as we can
see clearly that the phenomenon of chattering appears in this case but it can be eliminated by the setting of the
sampling time. On the other hand, the load torque estimation using DSTA observer is accurately occurred and
not affected by load variations. Zoom shows the tracking performance of this observer.
Except for the transient regime, the estimations provided by the rotor flux observer are not affected by the speed
profile or the load torque application whether at = 0.6s or at t=0.8s, this is very visible in Figure 6.
Through these results, it is noticed that the DSTA controls the IM as if they are in a continuous system. It is
worth mentioning that the control of IM based on STA version as a continuous-time system, it can be found in
[21].
Note that in the same context, others studies have shown that in SMC, discrete implementation of the controller
is a source of chattering phenomena, also this phenomena are related to the sampling period [22],[23].

7. CONCLUSION

In this paper, the discrete-time model in the stationary reference frame for IM has been presented. This model
has been gotten by variational integrator formulation. Among the benefits of this technique is that it approaches
the sampled dynamics of IM better than the model obtained with the classical Euler method. In addition, a
designed controller based on DSTA allows checking the performance of the motor discrete model.
In order to eliminate the utilization of sensors, two observers in the discrete-time domain have been introduced,
the first for rotor flux using DSTA and the second for load torque. Simulation results show that the designed
model improves the dynamic of IM and the control provides good performance and high robustness under
variations of speed and external load torque.

REFERENCES

[1] Oliveira,CMR , Aguiar, ML , Monteiro, JRBA , Pereira,WCA , Paula,GT,Almeida,TEP . Vector Control of Induction
Motor Using an Integral Sliding Mode Controller with Anti-windup.Journal of Control, Automation and Electrical
Systems 2016; vol. 27, no 2, p. 169‑178. Doi: 10.1007/s40313-016-0228-4.
[2] Quintero-Manriquez, E , Sanchez, EN , Felix, RA . Induction motor torque control via discrete-time sliding mode.
World Automation Congress (WAC) 2016; p. 1‑5. Doi: 10.1109/WAC.2016.7582984.
[3] Milosavljević, Č , Peruničić-Draženović, B , Veselić,B , Petronijević,M . High-performance discrete-time chattering-
free sliding mode-based speed control of induction motor. Electrical Engineering 2017; vol. 99, no 2, p. 583‑593. Doi:
10.1007/s00202-016-0386-1.
[4] Zhang, L , Zhang, H , Obeid, H , Laghrouche,S . Time-varying state observer based twisting control of linear induction

305
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

motor considering dynamic end effects with unknown load torque. ISA Transactions 2019; vol. 93, p. 290‑301. Doi:
10.1016/j.isatra.2019.03.008.
[5] Castillo-Toledo, B , Di Gennaro, S , Loukianov, AG , Rivera, J . Discrete time sliding mode control with application to
induction motors. Automatica 2008; vol. 44, no 12, p. 3036‑3045. Doi: 10.1016/j.automatica.2008.05.009.
[6] Monmasson , E , Cirstea, MN . FPGA Design Methodology for Industrial Control Systems—A Review. IEEE
Transactions on Industrial Electronics 2007; vol. 54, no 4, p. 1824‑1842. Doi: 10.1109/TIE.2007.898281.
[7] Awan, HAA , Tuovinen, T , Saarakkala, SE , Hinkkanen,M . Discrete-Time Observer Design for Sensorless
Synchronous Motor Drives. IEEE Transactions on Industry Applications 2016;vol. 52, no 5, p. 3968‑3979. Doi:
10.1109/TIA.2016.2572105.
[8] Dueñas, IE , Rivera, J , Di Gennaro, S , Ortega,S.Discrete–time Field Oriented Control for Induction Motors. IFAC-
PapersOnLine 2017; vol. 50, no 1, p. 15760‑15765, juill. Doi: 10.1016/j.ifacol.2017.08.2310.
[9] Chihi,A , Ben Azza, H , Jemli, M , Sellami,A. Nonlinear Discrete-Time Integral Sliding Mode Control of an Induction
Motor: Real-Time Implementation . Studies in Informatics and Control 2017; vol. 26, no 1. Doi:
10.24846/v26i1y201703.
[10] Rivera Domínguez,J. Discrete-Time Modeling and Control of Induction Motors by Means of Variational Integrators
and Sliding Modes—Part I: Mathematical Modeling. IEEE Transactions on Industrial Electronics 2015; vol. 62, no 9,
p. 5393‑5401. Doi: 10.1109/TIE.2015.2414391.
[11] Chern, T.-L , Liu, C.-S , Jong, C.-F , Yan,G.-M. Discrete integral variable structure model following control for
induction motor drivers. IEE Proceedings - Electric Power Applications 1996; vol. 143, no 6, p. 467‑474. Doi:
10.1049/ip-epa:19960545.
[12] Salgado, I , Kamal, S , Chairez, I , Bandyopadhyay, B , Fridman, L . Super-twisting-like algorithm in discrete time
nonlinear systems. In: The 2011 International Conference on Advanced Mechatronic Systems;August 2011:IEEE
Xplorepp. 497‑502.
[13] Salgado, I , Chairez, I , Bandyopadhyay, B , Fridman, L , Camacho,O . Discrete-time non-linear state observer based
on a super twisting-like algorithm. IET Control Theory Applications 2014; vol. 8, no 10, p. 803‑812. Doi: 10.1049/iet-
cta.2013.0568.
[14] Oliveira, CMR , Aguiar, ML , Monteiro, JBA , Pereira, WC , Paula, GT , Santana, MP . Vector control of induction
motor using a sliding mode controller with chattering reduction. In:COBEP/SPEC 2015. IEEE 13th Brazilian Power
Electronics Conference and 1st Southern Power Electronics Conference; 2015:IEEE Xplore, pp. 1‑6. Doi:
10.1109/COBEP.2015.7420071.
[15] Comanescu, M . Discrete Time Sliding Mode Controller and Observer Scheme for Decoupled Current Control of the
Induction Motor Drive. In:PECI2019. IEEE Power and Energy Conference at Illinois; 2019, pp. 1‑5. Doi:
10.1109/PECI.2019.8698783.
[16] Domínguez, JR . Discrete-Time Modeling and Control of Induction Motors by Means of Variational Integrators and
Sliding Modes—Part II: Control Design. IEEE Transactions on Industrial Electronics 2015; vol. 62, no 10, p.
6183‑6193. Doi: 10.1109/TIE.2015.2431632.
[17] Utrata, G , Rolek, J , Kaplon, A .The Novel Rotor Flux Estimation Scheme Based on the Induction Motor
MathematicalModel Including Rotor Deep-Bar Effect . Energies 2019; vol. 12, no 14, p. 2676. Doi:
10.3390/en12142676.
[18] Benlaloui, I , Khamari, D , Drid,S , Chrifi-Alaoui,L , Hamzaoui,M . Flux observers design for induction motor using
linearparameter varying observer and sliding mode observer. In: 2016 17th International Conference on Sciences and
Techniques ofAutomatic Control and ComputerEngineering (STA); 2016, p. 646‑651. Doi:
10.1109/STA.2016.7952012.
[19] Barambones, O , Alkorta, P .Position Control of the Induction Motor Using an Adaptive Sliding-Mode Controller and
Observers. IEEE Transactions on Industrial Electronics 2014; vol. 61, no 12, p. 6556‑6565. Doi:
10.1109/TIE.2014.2316239.
[20] Benderradji, H , Benamor, A , Chrifi-Alaoui, L , Bussy, P, Makouf,A.Second order sliding mode induction motor
control with a new Lyapunov approach. In: International Multi-Conference on Systems, Signals Devices; 2012:IEEE
Xplore, p. 1‑6. Doi: 10.1109/SSD.2012.6197961.
[21] Rashed, M , Goh, KB , Dunnigan, MW , MacConnell, PFA , Stronach, AF , Williams, BW . Sensorless second-
order sliding-mode speed control of a voltage-fed induction-motor drive using nonlinear state feedback . IEE
Proceedings - Electric Power Applications 2005; vol. 152, no 5, p. 1127‑1136. Doi: 10.1049/ip-epa:20050042.
[22]Svečko, R , Gleich, D , Sarjaš, A . The Effective Chattering Suppression Technique with Adaptive Super-Twisted
Sliding Mode Controller Based on the Quasi-Barrier Function; An Experimentation Setup . Applied Sciences 2020;
vol. 10, no 2, p. 595. Doi: 10.3390/app10020595.
[23]Zhang, J , Liu, Y , Gao, S , Han, C . Control Technology of Ground-Based Laser Communication Servo Turntable via
a Novel Digital Sliding Mode Controller . Applied Sciences 2019; vol. 9, no 19, p. 4051. Doi: 10.3390/app9194051.

306
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

APPENDIX
Table 2. Parameters of induction motor and control
Parameters of IM
Pn  1500 W n  1420 rpm F  0.001136 N.m.s/rad
Rr  3.805 Ω Rs  4.85 Ω Ls  0.274 H
M  0.258 H p2 J  0.031 kg .m 2
Parameters of DSMC
k11  0.0099 k12  0.0699 ki1  -0.1
ki 2  -0.1 q0  q1  10 kv  10 6
p1  0.8 p2  0.9 M1  0.9 / Ts
M 2  0.065 / Ts M 3  10  / T
5
M 4  4, 5 *105  / Ts

307
 RATIONAL LOADS OF TURBINE INLET AIR BSORPTION-
EJECTOR COOLING SYSTEMS

Mykola Radchenko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0002-1596-6508

Andrii Radchenko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0002-8735-9205

Serhiy Kantor
PJSC "Zavod "Ekvator", Mykolayiv, Ukraine, [email protected], ORCID: 0000-0001-5050-5937

Bohdan Portnoi
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-
0002-3142-2148

Anatolii Zubarev
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID:
0000-0001-7868-4519

Ivan Kalinichenko
Admiral Makarov National University of Shipbuilding, Mykolayiv, Ukraine, [email protected], ORCID: 0000-0001-
6765-6168

Radchenko, M., Radchenko, A., Kantor, S., Portnoi, B., Zubarev A., Kalinichenko, I. Rational
Cite this paper as: loads of turbine inlet air cooling systems. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
2020, Istanbul, Turkey

Abstract: An increase in gas turbine efficiency is possible by inlet air cooling in chillers converting a heat
of exhaust gas to refrigeration. Decrease in turbine intake air temperature and fuel consumption
accordingly is affected by ambient air temperature ta and a target chilled air temperature ta2, which
depends on the temperature of a coolant, i.e. on the type of chiller. In absorption lithium-bromide
chillers (ACh) an intake air can be cooled practically to 15 ºС (chilled water temperature t w ≈ 7
°С) with high coefficients of performance COP = 0.7…0.8, whereas in the most simple and cheap
refrigerant ejector chillers (ECh) – to more lower temperatures 5-10 °С (boiling refrigerant
temperature t0 = 1–5 °С), but with low COP = 0.2…0.3. The chillers of combined type with using
a high efficient ACh as a high-temperature stage of ambient intake air cooling to 15 °C and with
less efficient ECh as a low-temperature stage for further cooling air lower than 15 °C are
considered. A novel simple method of calculating a design cooling load to achieve practically
maximum annual fuel saving and to avoid oversizing is proposed.
Keywords: design cooling capacity, current thermal load, turbine, annual fuel saving.
© 2020 Published by ECRES

1. INTRODUCTION

Thermodynamic efficiency of gas turbines (GT) reduces with increasing intake ambient air temperature ta [1, 2].
Thus, power output drops by 0.5–0.9% and specific fuel consumption increases by 0.35–0.7 g/(kW∙h) for every
1°C rise in intake air temperature [3, 4]. An increase in GT efficiency at high air temperatures is possible by
turbine inlet air cooling (TIAC) in waste heat recovery chillers using exhaust gas heat [5, 6]. Decrease in turbine
intake air temperature Δt = ta – ta2 and fuel consumption accordingly depends on ambient air temperature ta and
a target chilled air temperature ta2, which depends on the temperature of coolant (refrigerant or cooled water

308
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

from chillers), i.e. on the type of chiller. So, in absorption lithium-bromide chillers (ACh) an intake air can be
cooled practically to t a2 = 15 ºС (temperature of chilled water tw ≈ 7 °С) with high coefficients of performance
COP = 0.7–0.8 [7, 8], whereas in the simplest and cheap ejector chillers (ECh) using high volatile fluids
(refrigerants) [9, 10] – to lower temperatures ta2 = 5-10 °С (temperature of boiling refrigerant t 0 = 1–5 °С), but
with low COP = 0.2–0.3 [11, 12]. Therefore it is quite reasonable to apply the chillers of combined AECh type
with using a high efficient ACh as a high-temperature stage of inlet air cooling from ambient air temperatures ta
to ta2 = 15 °C and with less efficient ECh as a low-temperature stage for further cooling air lower than 15 °C
[13, 14]. A design refrigeration capacity of the chiller should be selected to cover the current cooling needs and
provide a closed to maximum annual effect avoiding oversizing [15, 16].

2. LITERATURE REVIEW

A large number of publications is aimed for enhancing the efficiency of exhaust gas heat conversion to
refrigeration for turbine inlet air cooling (TIAC) [17, 18]. To increase the efficiency of cooling air in actual
climatic conditions a two-stage deep cooling in combined AECh [19, 20] that convert the exhaust heat has been
proposed as a novel approach in combustion engine inlet air cooling. Deep utilization of exhaust gas heat in
waste heat recovery boiler [21, 22] as primary source of heat for chillers [23] provide increasing the available
heat to be converted to refrigeration. Air cooling in AECh provides about 1.5 times greater annual reduction in
fuel consumption compared to cooling in ACh [19, 24].
For estimation of GT intake air cooling potential a simple parameter calculated as air temperature depression
Δt = ta – ta2 multiplied by the corresponding hours τ lapsed in terms of cooling degree hours (CDH) is used:
CDH = Δt·τ, °С∙h [25, 26] and their annual values ΣCDH = Σ(Δt·τ), °С∙h [25, 26]. Yearly ambient temperature
distribution is important input data for energy analyses and a design cooling load [27, 28]. A sinusoidal function
for the ambient temperature probability distribution was proposed [29–31].
A lot of researches deal with rational sizing the chilling systems [32, 33] including cooling towers [34, 35]. A
majority of approaches for designing the TIAC systems is based on the refrigeration capacity to cover the peak
cooling load [27], that can lead to considerable oversizing as shown in [15, 16].
The goal of research is to increase the fuel efficiency of GT by rational designing the TIAC system of combined
absorption-ejector type with ACh as a high-temperature and ECh as a low-temperature cooling stages.

3. RESEARCH METHODOLOGY

The installed (design) refrigeration capacity Q0 of TIAC system, on the one hand, has to cover GT intake air
cooling demand during as long time of GT operation year around as it is possible, providing the greatest annual
fuel saving. On the other hand, a design refrigeration capacity Q0 should not be overestimated that the most part
of year a TIAC system would operate at a high load level closed to a design value. Otherwise the low efficiency
of TIAC system operation will take place, and on the contrary, if the underestimated Q0, – insufficient GT intake
air cooling at high ambient temperatures.
The annual fuel saving ∑B is calculated by summarizing current fuel reduction B on a step (hour) by step (hour)
basis: ∑B = ∑[(Δt·τ) (Δbe /Δt )∙Pe]. With this a climatic characteristic of GT as dependence of specific fuel
consumption be and power output Pe on intake air temperature are used [1–4, 19, 24]. As example we use the
value of decrease Δbe in specific fuel consumption for every 1°C drop in GT intake air temperature Δbe /Δt =
0.35 and GT of 10000 kW for easy calculation of fuel saving data ∑B for other GT power output.
A total refrigeration capacity Q0 , when cooling air with flow rate Ga: Q0 = сa ξ∙Δta ∙Ga , where Δta = ta – ta2 –
decrease (depression) of ambient air temperature; ta2 – target chilled air temperature at the air cooler outlet; ξ –
specific heat ratio; сa –specific heat of humid air [kJ/(kg·K)].
The authors develop an original method of TIAC system rational designing based on the annular fuel saving ∑B
cumulative characteristics versus design cooling capacity Q0 as ∑B= f(Q0 ), modified according to sizing TIAC
intention. To calculate an optimal design cooling capacity Q0.opt , providing a maximum rate ∑B/Q0 of annual
fuel saving increment and corresponding annual fuel saving ∑Bopt , as the second stage of calculation procedure,
and to determine a precise value of rational design cooling capacity Q0.rat , providing the annual fuel saving ∑Brat
closed to its maximum value, but without unproductive wasting cooling capacity Q0, caused by oversizing chiller
without obtaining a noticeable increasing the annual fuel saving ∑B, as the third stage. With this a rational
cooling capacity Q0.rat is associated with second maximum rate of annual fuel saving ∑B increment beyond the
optimal annual fuel saving ∑Bopt : [∑B–∑Bopt ]/Q0 , where ∑B>∑Bopt .

309
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. RESULTS OF INVESTIGATION

It is reasonable to cool intake air from current ambient air temperature tamb to ta2 = 15 С in ACh with high
coefficient of performance COP=0.7−0.8 but deep air cooling from ta2 = 15 °С down to ta2 = 7 and 10 °С has to
be performed in ECh with low coefficient of performance COP = 0.2−0.3, i.e. to apply combined two-stage
absorption-ejector chiller (AECh) (Fig. 1).
Ambient air
ACHT
Cooling Condenser
Tower Pump
Generator Cooling
Tower
Absorption Pump Exhaust Boiler
Chiler Exhaust Condencer
Gas Ejector
Chiler
Evaporator Combustion Chamber Exp. ACLT
Valve
Absorber Ejector
Gas Turbine
Compressor Pump
Turbine
Generator

Figure 1. A scheme of two-stage absorption-ejector GT intake air cooling system: ACHT – air cooler high-temperature
stage; ACLT – air cooler low-temperature stage; Exp.Valve – expansion valve;

A rational design cooling capacity may be determined om the basic annual fuel saving ∑B cumulative
characteristics versus design cooling capacity Q0 (Fig.2).

1st t2=7,10,15 C
ΣB, t
1 ∑B7max ∑B7rat
700
t2=7 C
0,9 t2=10 C ∑B7 600
0,8 t2=15 C
∑B10max ∑B10rat
0,7 500
0,6 400
0,5 ∑B10
∑B15max ∑B15rat 300
0,4
∑B15
0,3 ΔQ0.7 200
0,2 ΔQ0.15 ΔQ0.10
0,1 100
15max 10max 7max
0 Q0.15rat Q0.10ratQ0.7rat Q0.7max 0
0 500 1000 1500 2000 Q25000, kW
Figure 2. Annual fuel saving ∑B due to cooling ambient air at the inlet of GT to temperatures ta2 = 7, 10 and 15 °С
versus a design cooling capacity Q0

As Fig.2 shows, a design cooling capacity Q0.10 = 1300–1500 kW provides cooling ambient air from the current
temperature ta to ta2 = 10 °C with annual fuel saving ∑B10 ≈ 500 ton that is closed to a maximum value.
While it is evidently seen that the rate of increment of annual fuel saving ∑B is negligible, the range of
corresponding cooling capacities Q0 needed is still wide and therefore so will be the range of oversizing. So, the
proper values of cooling capacities Q0 need to be determined for appropriately sized TIC system.

It is evident, that with increasing a target temperature of cooled air ta2 to 15 °C it is more problematic to select
quite precise values of rational cooling capacity Q0.15rat so as behaviour of characteristic curve ∑B = f(Q0)
becomes more gentle (Fig. 2). To solve this task the original method of determining a rational design cooling
capacity has been improved by adding addition stages aimed for calculation of optimal design cooling capacity
Q0.opt, corresponding to a maximum rate of annual fuel saving increment ∑B/Q0∙(Fig.3), and more precise value
of rational specific cooling capacity Q0.rat (Fig. 4).

310
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

ΣB/Q0,10(ΣB-ΣBopt)/Q0, t/kW 2st t2=7,10,15 C ΣB, t

0,45 ΣB/Q0.7 ∑B7
700
t2=7 C
0,4 t2=10 C ΣB/Q ∑B7opt 600
0.10
0,35 t2=15 C
∑B10
∑B10opt 500
0,3 ΣB/Q0.15
0,25 400
0,2 ∑B15 300
∑B15opt
0,15
200
0,1
0,05 100
0 opt15 opt10 opt7 0
0 500 1000 1500 2000 Q2500
0, kW
Figure 3. Relative increment of annual fuel saving ∑B/Q0 and annual fuel saving ∑B due to cooling ambient air to target
temperatures ta2 = 7, 10 and 15 °C versus a design cooling capacity Q0

As Fig. 3 shows, a maximum rate of annual fuel saving ∑B/Q0.15 due to cooling ambient air to the temperature
ta2 = 15 °C takes place at the optimal design cooling capacity Q0.15opt=16 kW/(kg/s) and provides covering annual
fuel saving ∑B15opt = 20000 °C∙h considerably less than its maximum value 24000 °C∙h.
To achieve closed to maximum annual fuel saving ∑B at more precise value of rational cooling capacity Q0.rat
without oversizing the chiller it is proposed to determine the maximum rate of annual fuel saving increment
∑B/Q0.15∙within the range beyond its value ∑B15opt = 20000 °C∙h corresponding to optimal design cooling
capacity Q0.15opt=16 kW/(kg/s), as (∑B– ∑Bopt )/Q0.15∙(Fig. 4 and 5).

ΣB/Q0,10(ΣB-ΣBopt)/Q0, t/kW 3st t2=7,10,15 C

0,5 t2=7 C 10(ΣB-ΣBopt)/Q0.15 ΣB/Q0.7
0,45 t2=10 C 10(ΣB-ΣBopt)/Q0.7
0,4 t2=15 C
0,35 ΣB/Q0.10
0,3 10(ΣB-ΣBopt)/Q0.10
ΣB/Q0.15
0,25
0,2
0,15
0,1
0,05
0 opt15 rat15 opt10 opt7 rat10 rat7
0 500 1000 1500 2000 Q02500
, kW
Figure 4. Relative increments of annual fuel saving ∑B/Q0 and (∑B– ∑Bopt )/Q0∙within the range beyond the optimal value
∑Bopt (Fig.3) due to cooling ambient air to target temperatures ta2 = 7, 10 and 15 °C versus capacities Q0 needed

As Fig. 4 and 5 shows, a maximum rate of annual fuel saving ∑B increment (∑B–∑Bopt )/Q0.15 within the range
beyond its value ∑Bopt,15 = 20000 °C∙h (in Fig.3) corresponding to optimal design cooling capacity Q0.15opt=16
kW due to cooling ambient air to the temperature ta2 = 15 °C takes place at the rational design cooling capacity
Q0.rat=24 kW/(kg/s) and provides quite precise annual fuel saving ∑Brat = 24500 °C∙h that is very closed to its
maximum value 25000 °C∙h, but at a design refrigeration capacity Q0.rat= 24 kW/(kg/s) much less than it
oversized value Q0 = 31 kW/(kg/s).

311
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

ΣB/Q0,10(ΣB-ΣBopt)/Q0, t/kW 4st t2= 7,10,15 C ΣB, t

0,5 ∑B7 700
t =7 C 10(ΣB-ΣBopt)/Q0 ∑B7rat
0,45 t 2=10 C 600
2
0,4 t =15 C
2 ∑B10 ∑B10rat 500
0,35
0,3 400
0,25 ∑B15 ∑B15rat 300
0,2
0,15 200
0,1
0,05 100
0 Q0.15rat Q0.10rat Q0.7rat 0
0 500 1000 1500 2000 Q ,
2500
0 kW
Figure 5. Annual fuel saving ∑B and relative increment of annual fuel saving (∑B– ∑Bopt )/Q0∙within the range beyond
the optimal value ∑Bopt (Fig.3) to find rational value Q0.rat versus cooling capacities Q0 needed for cooling ambient air to
target temperatures ta2 = 7, 10 and 15 °C
Graphs such as this allow users to match cooling to provide minimum sizing system while maximum rate of
annular cooling effect or to peak maximum demand needs without oversizing system.
Thus, a developed new method enables to determine a rational design cooling capacity that provides nearly
maximum annual fuel saving ∑B according to actual climatic conditions and avoid oversizing a chiller.

5. CONCLUSION

An improved simple method to calculate an optimal refrigeration capacity Q0.opt , providing a maximum rate of
annual fuel saving increment ∑B/Q0 , and a precise value of rational design refrigeration capacity Q0.rat ,
providing closed to maximum annual fuel saving ∑Brat without oversizing chiller is developed. With this a
rational refrigeration capacity Q0.rat is associated with second maximum rate of annual fuel saving increment just
beyond the optimal annual fuel saving ∑Bopt : [∑B–∑Bopt ]/Q0 , where ∑B>∑Bopt . The application of proposed
method is demonstrated by calculating the optimal Q0.opt and rational Q0.rat design refrigeration capacities for two
stage cooling air at the inlet of GT in combined AECh with high efficient ACh as a high temperature cooling
stage and less efficient ECh as a low temperature stage. It is shown, that for temperate climatic conditions the
annual fuel saving due to intake air cooling by AECh practically twice surpasses its value gained due to intake
air absorption cooling. The method to calculate an optimal specific refrigeration capacity Q0.opt , providing a
maximum rate of annual fuel saving increment is a very effective energy saving method of maintenance of
refrigeration capacity, in particular, through accumulation of excessive refrigeration capacity at lowered current
thermal loads to cover peak loads. The proposed methodology is quite useful to be extended for bringing to
cooling, heating and trigeneration plants.

REFERENCES

[1] Tiwari, A.K., Hasan, M., Muzaffarul, M.I. Effect of ambient temperature on the performance of a combined cycle power
plant. Transactions of the Canadian Society for Mechanical Engineering 2013, 37 (4), 1177–1188.
[2] Zhang, T., Liu, Z., Hao, H., Chang, L.J. Application research of intake-air cooling technologies in gas-steam combined
cycle power plants in China. Power Energy Eng. 2014, 2, 304–311.
[3] Günnur, Şen G. et. al. The effect of ambient temperature on electric power generation in natural gas combined cycle
power plant-A case study. Energy Reports 2018, 4, 682–690.
[4] Erdem, H.H., Sevilgen, S.H. Case study: Effect of ambient temperature on the electricity production and fuel
consumption of a simple cycle gas turbine in Turkey. Applied Thermal Engineering 2006, 26, 320–326.
[5] Al-Ibrahim, A.M., Varnhamm, A. A review of inlet air-cooling technologies for enhancing the performance of
combustion turbines in Saudi Arabia. Applied Thermal Engineering 2010, 30, 1879–1888.
[6] Chacartegui, R., Jiménez-Espadafor, F., Sánchez, D., Sánchez, T. Analysis of combustion turbine inlet air cooling
systems applied to an operating cogeneration power plant. Energy Convers. and Managem. 2008, 49, 2130–2141.
[7] Popli, S., Rodgers, P., Eveloy, V. Gas turbine efficiency enhancement using waste heat powered absorption chillers in
the oil and gas industry. Applied Thermal Engineering 2013, 50, 918–931.
[8] Zhang, T., Liu, Z., Hao, H., Chang, L.J. Application research of intake-air cooling technologies in gas-steam combined
cycle power plants in China. Power Energy Eng. 2014, 2, 304–311.

312
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[9] Butrymowicz, D., et. al. Investigations of prototype ejection refrigeration system driven by low grade heat. In: HTRSE-
2018; 2018: E3S Web of Conferences70 (03002), 7 p.
[10] Konovalov, D., Kobalava H. Efficiency Analysis of Gas Turbine Plant Cycles with Water Injection by the
Aerothermopressor. In: Ivanov V. et al. (eds). Advances in Design, Simulation and Manufacturing II. DSMIE 2019.
Lecture Notes in Mechanical Engineering, Springer, Cham. 20, 2020, pp. 581-591.
[11] Smierciew, K., Gagan, J., Butrymowicz, D., Karwacki, J. Experimental investigations of solar driven ejector air-
conditioning system. Energy and Buildings 2014, 80, 260–267.
[12] Elbel, S., Lawrence, N. Review of recent developments in advanced ejector technology. International Journal of
Refrigeration 2016, 62, 1–18.
[13] Radchenko, A., Mikielewicz, D., Forduy, S., Radchenko, M., Zubarev, A. Monitoring the Fuel Efficiency of Gas
Engine in Integrated Energy System. In: Nechyporuk M. et al. (eds). Integrated Computer Technologies in Mechanical
Engineering. Advances in Intelligent Systems and Computing, Springer, Cham 1113, 2020, pp. 363-370.
[14] Radchenko, M., Radchenko, R., Ostapenko, O., Zubarev, A., Hrych, A. Enhancing the utilization of gas engine module
exhaust heat by two-stage chillers for combined electricity, heat and refrigeration. In: 5th International Conference on
Systems and Informatics, ICSAI 2018; 2018: Jiangsu, Nanjing, China, pp. 227–231.
[15] Trushliakov E., Radchenko, M., Radchenko, A., Kantor, S., Zongming, Y. Statistical approach to improve the
efficiency of air conditioning system performance in changeable climatic conditions. In: 5th International Conference
on Systems and Informatics, ICSAI 2018; 2018: Jiangsu, Nanjing, China, pp. 1303–1307.
[16] Trushliakov, E., Radchenko, M., Bohdal, T., Radchenko, R., Kantor, S. An innovative air conditioning system for
changeable heat loads. In: Tonkonogyi V. et al. (eds.) Advanced Manufacturing Processes. InterPartner-2019. Lecture
Notes in Mechanical Engineering, Springer, Cham, 2020, pp. 616–625.
[17] Popli, S., Rodgers, P., Eveloy, V. Trigeneration scheme for energy efficiency enhancement in a natural gas processing
plant through turbine exhaust gas waste heat utilization. Applied Energy 2012, 93, 623–636.
[18] Ghaebi, Н., Karimkashi, Sh., Saidi, M.H. Integration of an absorption chiller in a total CHP site for utilizing its cooling
production potential based on R-curveconcept. International journal of refrigeration 2012, 35, 384–1392.
[19] Radchenko, R., Radchenko, A., Serbin, S., Kantor, S., Portnoi, B. Gas Turbine Unite Inlet Air Cooling by Using an
Excessive Refrigeration Capacity of Absorption-Ejector Chiller in Booster Air Cooler. In: HTRSE-2018; 2018: E3S
Web of Conferences 70 (03012), 6 p.
[20] Forduy, S., Radchenko, A., Kuczynski, W., Zubarev, A., Konovalov, D. Enhancing the Gas Engines Fuel Efficiency
in integrated energy system by chilling cyclic air. In: Tonkonogyi V. et al. (eds.) Advanced Manufacturing Processes.
InterPartner-2019. Lecture Notes in Mechanical Engineering, Springer, Cham, 2020, pp. 500-509.
[21] Kornienko, V., Radchenko, R., Stachel, A., Andreev A., Pyrysunko, M. Correlations for pollution on condensing
surfaces of exhaust gas boilers with water-fuel emulsion combustion. In: Tonkonogyi V. et al. (eds.) Advanced
Manufacturing Processes. InterPartner-2019. Lecture Notes in Mechanical Engineering, Springer, Cham, 2020, pp.
530-539.
[22] Radchenko, R., Kornienko, V., Pyrysunko, M., Bogdanov, M., Andreev, A. Enhancing the Efficiency of Marine Diesel
Engine by Deep Waste Heat Recovery on the Base of Its Simulation Along the Route Line. In: Nechyporuk M. et al.
(eds) Integrated Computer Technologies in Mechanical Engineering. Advances in Intelligent Systems and Computing,
Springer, Cham 1113, 2020, pp. 337-350.
[23] Radchenko, A., Radchenko, M., Konovalov, A., Zubarev, A. Increasing electrical power out-put and fuel efficiency
of gas engines in integrated energy system by absorption chiller scavenge air cooling on the base of monitoring data
treatment. In: HTRSE-2018; 2018: E3S Web of Conferences70 (03011), 6 p.
[24] Radchenko, A., Bohdal, L., Zongming, Y., Portnoi, B., Tkachenko, V. Rational designing of gas turbine inlet air
cooling system. In: Tonkonogyi V. et al. (eds.) Advanced Manufacturing Processes. InterPartner-2019. Lecture Notes
in Mechanical Engineering, Springer, Cham, 2020, pp. 591-599.
[23] Chaker, M., Meher-Homji, C.B., Mee, T, Nicolson, A. Inlet Fogging of Gas Turbine Engines- Detailed Climatic
Analysis of Gas Turbine Evaporative Cooling Potential. In: ASME International Gas Turbine and Aeroengine
Conference; 4–7 June 2001: New Orleans, USA, ASME Paper No. 2001-GT-526.
[24] Chaker, M., Meher-Homji, C.B. Inlet fogging of gas turbine engines: climatic analysis of gas turbine evaporative
cooling potential of international locations. In: Proceedings of ASME TURBO EXPO; 3–6 June 2002: Amsterdam,
The Netherlands, GT-2002-30559.
[25] Satman, A, Yalcinkaya, N. Heating and cooling degree-hours for Turkey. Energy 1999, 24 (10), 833–840
[26] Pegues, J. The benefits of 8760 hour-by-hour building energy analysis. Carrier Software Systems, New York, USA:
HVAC Systems Engineer, 2002.
[27] Forsyth, J.L. Gas turbine inlet air chilling for LNG. In: IGT International Liquefied Natural Gas Conference
Proceedings 3; 2013, pp. 1763–1778.
[28] Komuro, T. et. al. Power Output Augmentation of Gas Turbine Combined Cycle by Inlet-Air Cooling System of Chiller
Type under High Ambient Air Temperature. Mitsubishi Heavy Indust. Technical Rev. 2010, 47(4), 33–39.
[29] Coskun, C., Demiral, D., Ertürk, M., Oktay, Z. Modified Degree-Hour Calculation Method. In Prof. R. Rugescu (Ed.),
Solar Power, InTech, 2012, pp. 55–62.
[30] Coskun, C. A novel approach to degree-hour calculation: Indoor and outdoor reference temperature based degree-hour

313
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

calculation. Energy 2010, 35, 2455–2460.

[31] Oktay, Z, Coskun, C, Dincer, I. A new approach for predicting cooling degree-hours and energy requirements in
buildings. Energy 2011, 36 (8), 4855–4863.
[32] Carvalho, M, Lozano, MA, Serra, LM. Multicriteria synthesis of trigeneration systems considering economic and
environmental aspects. Applied Energy 2012, 91, 245–254.
[33] Kavvadias, KC, Tosios, AP, Maroulis, ZB. Design of a combined heating, cooling and power system: sizing, operation
strategy selection and parametric analysis. Energy Convers Manage 2010, 51, 833–845.
[34] Mulyandasari, V. Cooling tower selection and sizing (engineering design guideline). KLM Technology Group.
Practical Engineering Guidelines for Processing Plant Solutions, 2011, 52 p.
[35] N. Liu, Z.Li. The Feasibility on the Case that the Air Conditioning Condensate Water is used as the Make-up Water
of Cooling Tower. Procedia Engineering 2017, 205, 3557–3562.

314
 INTERRUPTION-LESS CHARGING OF PLUG-IN ELECTRIC
VEHICLES USING THE GRID-CONNECTED PHOTOVOLTAIC
SYSTEM

Abdul Rauf Bhatti

Department of Electrical Engineering, Government College University Faisalabad, Faisalabad, Pakistan,
[email protected], ORCID: 0000-0001-9609-4563

Arslan Dawood Butt*

Department of Electrical Engineering, Government College University Faisalabad, Faisalabad, Pakistan,
[email protected], ORCID: 0000-0002-4704-2553

Yawar Ali Sheikh

Department of Electrical Engineering, Government College University Faisalabad, Faisalabad, Pakistan,
[email protected], ORCID: 0000-0002-1458-3898

Kashif Nisar Paracha

Department of Electrical Engineering, Government College University Faisalabad, Faisalabad, Pakistan,
[email protected], ORCID: 0000-0001-6379-7385

Bhatti, AR, Butt, AD, Sheikh, YA, Paracha, KN. Interruption-less Charging of Plug-In
Cite this paper as: Electric Vehicles Using the Grid-Connected Photovoltaic System. 8th Eur. Conf. Ren. Energy
Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The widespread application of Plug-in Electric Vehicles (PEVs) can reduce the problems
associated with greenhouse gas emissions of internal combustion engines. However, during the
(daytime) busy hours, charging of PEV inflicts an added encumbrance on the utility grid. This
paper proposes a solution to this vehicle charging problem by introducing a combination of Energy
Storage Unit (ESU) and Photovoltaic (PV) panel charging along with the grid. The proposed
solution is efficient in providing charging for PEV without any interruption in an office parking
scenario. The important benefit of proposed system is the rapid fulfillment of charging demand of
all PEV in parallel to lessening the added grid burden. An algorithm is designed according to
schemes that are rule based. The efficacy of the proposed technique is evaluated under different
charging situations using Matlab simulations. The simulation results prove that PEV charging
using proposed system is considerably lower the added burden from the grid. This work is
envisioned to offer stimulating research prospects in the area of vehicle charging using grid-
connected renewable sources.
Keywords: Interruption-less charging, plug-in EVs, PV system, energy storage unit
© 2020 Published by ECRES

1. INTRODUCTION

The rise of greenhouse gasses and their imminent environmental impacts have accelerated the research on
Renewable Energy (RE) sources [1, 2]. It has been established that one of the main sources of the pollution is
the greenhouse gasses and CO2 emissions from the conventional Internal Combustion Engine (ICE) vehicles.
The growing interest of researchers in Electric Vehicle (EV) is a measure to reduce the dependency on ICE
vehicles. The main link between EVs and the electric grid is the charging of the batteries. Due to the easiness
and availability of unrestrained energy from the grid, direct charging of EVs from the grid is extensively adopted.
The charger only requires a simple power outlet that provides the EV users with easy access to the charging
point [3]. Conversely, charging directly from the grid generates an added load for the electric grid. This is
predominantly critical during the peak (day time) hours when there is a very high demand of electrical power.
The consequences are distribution losses, degradation in the power quality and voltage deviation due to the
congestion of the grid [4, 5]. However, in countries where climate is hot and sun shines a larger part of the day,
roofing parking lots with PV plates provides protection from sun and rain in addition to the renewable energy

315
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

which facilitates the vision of “charging while parking” [6, 7] and ideal setup for office environment can be
achieved. Another impetus to incorporate PV is the sharp decline in the price of modules; the cost for
photovoltaic generators has dropped by more than 50% between 2010 and 2014 [8]. Another preferable option
is to combine PV array and Energy Storage Unit (ESU) to deal with fluctuated power [9, 10, 11]. However, the
hybridization of the sources (PV, ESU, and grid) increases the complexity of the system. Moreover, optimizing
the performance of the hybrid system in the presence of intermittent PV power and volatile grid electricity prices
is a difficult task. Furthermore, the system must meet the random EV demand while reducing the grid burden.

Inevitably, to effectively utilize the energy sources, an efficient and intelligent scheme for the proper energy
management is required. The purpose is to manage the interactions among the system components, as well as to
deal with the various uncertainties, namely the random EV power demand, PV intermittency and variation in
the grid energy prices [12, 13]. The scheme can be embedded into a controller for real-time execution; thus, the
charging station can be automatically operated without the presence of human operators [14].

2. MODELING OF PROPOSED SYSTEM

To simulate the proposed system, it is inevitable to model a PV-grid charging system as shown in Figure 1. For
this purpose, the components of charging system are modeled first.

Figure 1. PV to EV system model

Power Demand of EV

The mathematical model used in this work to simulate power demand of EV (EV_Dmd) is as described in [15].
Using this model, the power requirement of any individual PEV connected to the grid can be calculated at time
t using Eq. [1].
𝑃𝑒𝑣,𝑡 = 𝑃𝑒𝑣,𝑟𝑒𝑞 × 𝑆𝑡 × 𝑤𝑡 × 𝑐𝑡 (1)

Here, in time step ∆t, Pev,req shows the charging power measurements which are required by the PEV’s battery
to upsurge between the state of charge (SOC0) and maximum (SOCmax) conditions.

To check SOC, Variable St is used as a control signal. Whereas office working days and working hours are
shown by wt and ct respectively. For a range of connected EVs, the summation of Pev,t is for the total EV_Dmd
is shown in Eq. [2].
𝑁
𝑛
∑ 𝑃𝑒𝑣,𝑡 , (𝑡 = 9,10, . . . , 18)
𝐸𝑉_𝐷𝑚𝑑𝑡 = { (2)
𝑛=1
0, 𝑒𝑙𝑠𝑒

316
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Since the rule-based energy management scheme (REMS) is intended to be used for daytime charging, The
office working hours are assumed to be between 9:00 am to 6:00 pm. For the same office hours, EV_Dmd is
produced, to fulfil the need during the mentioned hours. [16].

The PV output power is determined by using the single diode model [17, 18]. The output PV current Ipv can be
attained in Eq. [3], by assuming available values of Temperature (T) and solar irradiance (G).
𝑉𝑝𝑣 + 𝐼𝑝𝑣 𝑅𝑆
𝐺 ( ) 𝑉𝑝𝑣 + 𝐼𝑝𝑣 𝑅𝑆
𝐼𝑝𝑣 = [(𝐼𝑆𝐶𝑆𝑇𝐶 + 𝑘𝑖 (𝑇 − 𝑇𝑆𝑇𝐶 )) ] − 𝐼0 [𝑒 𝑉𝑇 − 1] − ( ) (3)
𝐺𝑆𝑇𝐶 𝑅𝑃

Where Io is diode saturation current; ISC_STC is the short circuit current at STC (GSTC= 1000 W/m2, TSTC= 298 K);
ki is the short circuit current coefficient; VT = akT/q is the thermal voltage; q is electron charge (1.602×10-19 C);
a is the ideality factor (1 <a <2) [19]; k is Boltzmann constant (1.381×10-23 J/K). The RS shows the series whereas
RP shows the shunt resistances respectively. Newton-Raphson iteration method for its rapid convergence, is used
to solve Eq. [3] because of its transcendental nature, by varying the output voltage (Vpv) from zero to maximum
open circuit voltage [20]. The (Vpv) can be concluded as its final value equal to the voltage at Maximum Power
Point (MPP). By assuming that there is no partial shading and the inverter is functioning at MPP, the single
module output power can be formulated by Eq. [4]

𝑃𝑝𝑣 = 𝑚𝑎𝑥( 𝐼𝑝𝑣 × 𝑉𝑝𝑣 ) (4)

where Vpv is the PV module output voltage. At this stage, we can calculate the entire power of the PV array
(PV_Pwr) by using Eq. [5].

𝑃𝑉_𝑃𝑤𝑟 = 𝑃𝑝𝑣 × 𝑁𝑝𝑣 (5)

where Npv is the number of modules used in PV array. In this work, the KD325GX-LFB polycrystalline modules
are used [21]. The data values of for irradiance (G) and temperature (T) are collected from the National
Renewable Energy Laboratory (NREL) located in Southern California (GPS: 33.966674° N, 18.42282° W) [12,
22, 23].

The Energy Storage Unit (ESU) Model

For the approximate calculations of SOC of the ESU, the model is taken from [24].
𝑃𝐵 (𝑡)
𝑆𝑂𝐶 (𝑡) = 𝑆𝑂𝐶 (𝑡 − 1). (1 − 𝛿𝑏𝑎𝑡 (𝑡)) + ( ) . 𝜂𝑏𝑎𝑡 . 𝛥𝑡 (6)
𝑉𝑏𝑢𝑠

In Eq. [6], δbat(t) is the self-discharge rate, calculated on hourly basis. It is considered to be zero for practical
purposes. The bat denotes the efficiency of charging and discharging of the battery, which in both cases, is
supposed to be 100%. The voltage of dc common bus is shown by Vbus; in this scenario, Vbus= 500 V; PB(t) shows
the charging or discharging power. The value of PB(t) varies with respect to the available and required ESU
Power. The required ESU power (Req_ESU_Pwr) is the minimum amount of power (in kW) that increases the
SOC from initial to the upper limit (SOCU) in time step ∆t and is described in Eq. [7]. In time step ∆t, if we
want to vary the SOC from lower to upper limit (SOCU), we need the required ESU Power (Req_ESU_Pwr) as
the minimum value of power (in kW).

(𝑆𝑂𝐶𝑈 − 𝑆𝑂𝐶(𝑡)) × 𝐶𝑏𝑎𝑡 × 𝑁𝑏𝑎𝑡 (7)

𝑅𝑒𝑞_𝐸𝑆𝑈_𝑃𝑤𝑟(𝑡) =
𝛥𝑡
Similarly, ESU can provide continuous power for time step ∆t, before it reaches from its SOC to its lower limit
(SOCL), this continuous power of ESU is the maximum amount of power (in kW); shown as available ESU
power (Avl_ESU_Pwr). It is characterized as

(𝑆𝑂𝐶(𝑡) − 𝑆𝑂𝐶𝐿) × 𝐶𝑏𝑎𝑡 × 𝑁𝑏𝑎𝑡 (8)

𝐴𝑣𝑙_𝐸𝑆𝑈_𝑃𝑤𝑟(𝑡) =
𝛥𝑡

317
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In Eq. [7] and Eq. [8], Nbat represents the total batteries available in the ESU, whereas Cbat is considered as
apparent value of the capacity of a single battery. To get the maximum longevity from the batteries, the battery
charging and discharging states should be restricted as per the mentioned SOCL and SOCU limits. Which are
set as 10% and 100%, respectively [25].

3. MAIN FLOW OF RULE-BASED ENERGY MANAGEMENT SCHEME (REMS)

The projected REMS is developed on rule-based strategies. The rules may be taken as two cases of ‘if-then’
description, in which the ‘if’ cases are connected to the different situations and ‘then’ cases implements the
operational modes. The overall operation of the charging station is divided into four main scenarios. The main
scenarios are defined based on PV_Pwr and the EV_Dmd. However, some additional scenarios appear under the
mains which depend on other parameters, namely GE_Pr and SOC of the ESU. The operation under each
scenario is accomplished using single or multiple operating modes to achieve the required objectives of REMS.
The main four scenarios are defined as;

 Overload, when EV_Dmd > PV_Pwr.

 Underload, when EV_Dmd is present but EV_Dmd is ≤ PV_Pwr.
 No-load, when PV_Pwr is available but the EV_Dmd = 0.
 Idle, when both EV_Dmd=0, and PV_Pwr =0.
By using the outputs of already developed sub-models, the PV-grid charging station is simulated using REMS.
The objective of the REMS based control is to charge the EVs of office employees without interruption at
specified constant price (lower than Avg_GE_Pr). Since the REMS is exclusively intended to aid the office
employees without concerns of generating revenue from the charging station, the proposed scheme operates the
charging station on ‘break even’. The ‘break even’ strategy will facilitate the PEV user with lowest charging
price as possible and also care for the charging station of bearing financial distress. For this purpose, it is crucial
to find the optimum size of charging station under which resilient operation of REMS could be realized without
experiencing financial losses to the station. The main flow of PV-grid operation under the control of REMS is
given in the form of the flowchart as shown in Figure 2 (see Appendix).

In this flowchart the four main operating scenarios are shown with diamond shapes and the possible operating
modes under each scenario are mentioned in the boxes. To simulate the operation of charging station, the REMS
algorithm requires eight inputs: i) PV_Pwr (kW), ii) SOC of ESU (kW), iii) EV_Dmd (kW), iv) GE_Pr
(cents/kWh), v) PV_Pr (cents/kWh), vi) ESU_Pr (cents/kWh), vii) List of holidays and viii) date (dd/mm/yyyy).
One year long database consisting the hourly values of PV_Pwr, EV_Dmd and GE_Pr are generated with the
help of developed sub-models of the PV-grid system [26]. The database is loaded within the REMS algorithm
to use for the simulation of the charging station. Although, the SOC is represented in percentage but it is
differently defined here in terms of power to hold the similarity of plotting parameters (that is PV_Pwr, EV_Dmd,
SOC). Thus, it is represented as 𝑆𝑂𝐶 = % 𝑆𝑂𝐶(𝑡) × 𝐶𝑏𝑎𝑡 × 𝑁𝑏𝑎𝑡 which can be defined as, the maximum amount
of power (in kW) which ESU can deliver continuously for one hour. Initially, the SOC of ESU is considered to
be 30% which will be changing after every hour of operational job under the given particular circumstances
[27]. Similarly, PV_Pwr, GE_Pr, and EV_Dmd are changing with the new value in every hour of simulation.

To specify the simulation period, the REMS algorithm asks the user to enter a couple of dates at the beginning
of the simulation. This duration can be varied from one day up to one year for this research. According the
provided input duration of the user, the database is loaded and the information is extracted from the database as
per the input period. The office working days are confirmed by comparing the user specified dates with the
preloaded list of holidays. Moreover, on office working days, the office working hours are considered to be from
9:00 am to 6:00 pm [16, 28, 29].

318
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Uninterrupted charging Uninterrupted charging

using PV only using PV and ESU

Selling surplus PV
energy to grid

Figure 3. The results of REMS at PL on 11th June

After every hour of the operation, the prices of energy selling and purchasing are calculated based on the
predefined rules. The difference of the prices of energy selling and purchasing gives the amount of hourly profit
in cents. At the end of the simulation, these hourly profits are added and converted into USD to get the total
profit (J). The J is considered as a fitness function to ascertain the number of PV modules (Npv) and ESU batteries
(Nbat) in the charging station at which the system can be made functional under the resistive control of REMS
without bearing financial distress.

4. RESULTS

Figure 3 shows the results of charging for a specific day (11/06/2013) in summer. It is to be noted that the PV
produces more power compared to the winter day. As the days in summer are long, the power started to be
generated by PV before the office working hours and carry on to be produced even after the closing office hours.
This is depicted in the plot (a). The additional PV_Pwr lessens the grid reliance by restricting the triggering of
Gd2EV as illustrated by the plot (b). Additionally, the excess PV_Pwr compensates the system financial losses
by selling the surplus energy to the grid using mode PV2Gd, shown by the plot (c) with label “Selling surplus
energy to grid”.

During low irradiance, for example, during the hours 18 to 19, the ESU is activated to meet the EV_Dmd using
mode ESU2EV which is labeled as “Uninterrupted charging using PV and ESU”. Although the ESU loses its
energy during ESU2EV mode, it is regained during hour 19 to 20 via mode PV2ESU when PV_Pwr exceeds
EV_Dmd as shown in plot (c). Similar to the winter day operation, the results for summer show the resiliency of
REMS for constant price charging throughout the day. This can be interpreted in plot (d). Consequently, the
robustness of REMS can also be seen in plot (c) for the provision of uninterrupted charging, against the different
parameter conditions of PV_Pwr, EV_Dmd and electricity prices.

5. CONCLUSION

In this paper, a novel technique is presented for the charging of PEVs during the day by making use of the
charging station based on PV-grid combination. The proposed technique REMS, which constitute schemes that
are rule based, helps the charging station to provide continual charging. The scheme has been tested on the
typical system. The initial results show that charging PEVs by making use of the PV-grid combination along
with the proposed strategy is more financially beneficial as compared to the traditional basic grid alone charging.
In addition, the proposed strategy helps lessening the additional PEV charging encumbrance on the utility grid.
It is anticipated that the proposed technique will offer interesting and encouraging research avenues for PEV
charging by exploiting renewable energy sources.

319
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] Ahmed, A, Al-Amin, AQ, Ambrose, AF, Saidur, R.. Hydrogen fuel and transport system: A sustainable and
environmental future. International Journal of Hydrogen Energy 2016, vol. 41, pp. 1369-1380.
[2] Hossain, M.. Solar energy integration into advanced building design for meeting energy demand and environment
problem. International Journal of Energy Research 2016, vol. 40, pp. 1293-1300.
[3] Yilmaz, M, Krein, PT.. Review of battery charger topologies, charging power levels, and infrastructure for plug-in
electric and hybrid vehicles. IEEE Transactions on Power Electronics 2013, vol. 28, pp. 2151-2169.
[4] Sundström, O, Binding, C.. Flexible charging optimization for electric vehicles considering distribution grid constraints.
IEEE Transactions on Smart Grid 2012, vol. 3, pp. 26-37.
[5] Akhtar, Z, Opatovsky, M, Chaudhuri, B, Hui, SYR.. Comparison of point-of-load versus mid-feeder compensation in
LV distribution networks with high penetration of solar photovoltaic generation and electric vehicle charging stations.
IET Smart Grid 2019, vol. 2, pp. 283-92.
[6] Van Roy, J, Leemput, N, Geth, F, Salenbien, R, Buscher, J, Driesen, J.. Apartment Building Electricity System Impact
of Operational Electric Vehicle Charging Strategies. IEEE Transactions on Sustainable Energy 2013, vol. 5, pp. 264
- 272.
[7] Esfandyari, A, Norton, B, Conlon, M, McCormack, SJ.. Performance of a campus photovoltaic electric vehicle charging
station in a temperate climate. Solar Energy 2019, vol. 177, pp. 762-71.
[8] Tripathi, L, Mishra, A, Dubey, AK, Tripathi, C, Baredar, P.. Renewable energy: An overview on its contribution in
current energy scenario of India. Renewable and Sustainable Energy Reviews 2016, vol. 60, pp. 226-233.
[9] Zakeri, B, Syri, S.. Electrical energy storage systems: A comparative life cycle cost analysis. Renewable and Sustainable
Energy Reviews 2015, vol. 42, pp. 569-596.
[10] Bhatti, AR, Salam, Z, Aziz, MJBA, Yee, KP.. A critical review of electric vehicle charging using solar photovoltaic.
International Journal of Energy Research 2016, vol. 40, pp. 439-461.
[11] Bhatti, AR, Salam, Z, Sultana, B, Rasheed, N, Awan, AB, Sultana, U, et al.. Optimized sizing of photovoltaic grid‐
connected electric vehicle charging system using particle swarm optimization. International Journal of Energy
Research 2019, vol. 43, pp. 500-22.
[12] García-Triviño, P, Torreglosa, JP, Fernández-Ramírez, LM, Jurado, F.. Control and operation of power sources in a
medium-voltage direct-current microgrid for an electric vehicle fast charging station with a photovoltaic and a battery
energy storage system. Energy 2016, vol. 115, pp. 38-48.
[13] Ramadhani, UH, Shepero, M, Munkhammar, J, Widén, J, Etherden, N.. Review of probabilistic load flow approaches
for power distribution systems with photovoltaic generation and electric vehicle charging. International Journal of
Electrical Power & Energy Systems 2020, vol.120, p. 106003.
[14] Liu, N, Chen, Q, Liu, J, Lu, X, Li, P, Lei, J, et al.. A Heuristic Operation Strategy for Commercial Building Microgrids
Containing EVs and PV System. IEEE Transactions on Industrial Electronics 2015, vol. 62, pp. 2560-2570.
[15] Shao, S, Pipattanasomporn, M, Rahman, S.. Development of physical-based demand response-enabled residential load
models. IEEE Transactions on Power Systems 2013, vol. 28, pp. 607-614.
[16] Zhang, W, Ge, W, Huang, M, Jiang, J.. Optimal Day-Time Charging Strategies for Electric Vehicles considering
Photovoltaic Power System and Distribution Grid Constraints. Mathematical Problems in Engineering 2015, vol.
2015, pp. 1-9.
[17] Ahmed, J, Salam, Z.. An improved perturb and observe (P&O) maximum power point tracking (MPPT) algorithm for
higher efficiency. Applied Energy 2015, vol. 150, pp. 97-108.
[18] Ahmed, J, Salam, Z.. A Maximum Power Point Tracking (MPPT) for PV system using Cuckoo Search with partial
shading capability. Applied Energy 2014, vol. 119, pp. 118-130.
[19] Ishaque, K, Salam, Z.. An improved modeling method to determine the model parameters of photovoltaic (PV)
modules using differential evolution (DE). Solar Energy 2011, vol. 85, pp. 2349-2359.
[20] Ishaque, K, Salam, Z, Taheri, H.. Accurate MATLAB simulink PV system simulator based on a two-diode model.
Journal of Power Electronics 2011, vol. 11, pp. 179-187.
[21] ([accessed 15.04.2020]). Multicrystal KD325GX-LFB model. Available from: http://www.kyocerasolar.com
[22] Andreas, A, Stoffel, T.. ([accessed 14.04.2020]). NREL Solar Radiation Research Laboratory (SRRL). Available:
http://dx.doi.org/10.5439/1052221
[23] Hsu, RC, Liu, CT, Chen, WY, Hsieh, HI, Wang, HL.. A Reinforcement Learning-Based Maximum Power Point
Tracking Method for Photovoltaic Array. International Journal of Photoenergy 2015, vol. 501, pp. 1-12.
[24] Belfkira, R, Zhang, L, Barakat, G.. Optimal sizing study of hybrid wind/PV/diesel power generation unit. Solar Energy
2011, vol. 85, pp. 100-110.
[25] Li, X, Hui, D, Lai, X.. Battery Energy Storage Station (BESS)-Based Smoothing Control of Photovoltaic (PV) and
Wind Power Generation Fluctuations. IEEE Transactions on Sustainable Energy 2013, vol. 4, pp. 464-473.
[26] Changsong, C, Shanxu, D.. Optimal Integration of Plug-In Hybrid Electric Vehicles in Microgrids. IEEE Transactions
on Industrial Informatics 2014, vol. 10, pp. 1917-1926.
[27] Maleki, A, Ameri, M, Keynia, F.. Scrutiny of multifarious particle swarm optimization for finding the optimal size of
a PV/wind/battery hybrid system. Renewable Energy 2015, vol. 80, pp. 552-563.

320
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[28] Mohamed, A, Salehi, V, Tan, M, Mohammed, O.. Real-Time Energy Management Algorithm for Plug-In Hybrid
Electric Vehicle Charging Parks Involving Sustainable Energy. IEEE Transactions on Sustainable Energy 2014, vol.
5, pp. 577-586.
[29] Locment, F, Sechilariu, M.. Modeling and Simulation of DC Microgrids for Electric Vehicle Charging Stations.
Energies 2015, vol. 8, pp. 4335-4356.

Appendix

Start

REMS Inputs:
Load data: PV_Pwr (kW), SOC of ESU (kW), EV_Dmd
(kW), GE_Pr (cents/kWh), PV_Pr (cents/kWh),
ESU_Pr (cents/kWh) and the list of holidays
Specify simulation period: Date (dd/mm/yyyy)

Update SOC of ESU, PV_Pwr, GE_Pr, EV_Dmd

Chrg_Pr = Yes Office No Chrg_Pr = 0

PV_Pr 1.0 working
(cents/kWh)
(cents/kWh) hour

End End End

No
Underload scenario No
EV_Dmd > 0 AND Idle scenario
EV_Dmd <=PV_Pwr PV_Pwr = 0 AND
No EV_Dmd = 0
Overload scenario No load scenario
EV_Dmd > PV_Pwr EV_Dmd = 0 AND Yes
Yes
No PV_Pwr > 0
End
Yes Yes
Operating Operating Operating Operating
Modes Modes Modes Modes
PV2EV, PV2EV,
ESU2EV, PV2ESU, PV2ESU,
Gd2EV, PV2Gd, PV2Gd,
Gd2ESU Gd2ESU Gd2ESU Gd2ESU

Purchasing price = (PV energy x LCOE of PV) + (ESU energy x LCOE of ESU) +
(Grid energy x GE_Pr)
Selling price = (Energy sold to EV x Chrg_Pr) + (PV energy sold to grid x
(GE_Pr 0.1)) + (PV energy sold to ESU x PV_Pr )
Hourly Profit (in cents) = (Selling price – Purchasing price)

No Last hour of given

period reached

Yes
Resiliency of REMS:
Total Profit (in USD) = J = J is used as fitness
Σ(Hourly Profit)/100 function in PSO to
find Npv and Nbat

End

Figure 2. Simplified flowchart for REMS to charge PEVs.

321
 HOW DO STUDENTS’ COMPETITIONS LIKE SOLAR
DECATHLON INCREASE RENEWABLE ENERGY AWARENESS
AND WHO DOES BENEFIT?

Liudmila Cazacova
American University of Ras Al Khaimah, Ras Al Khaimah, United Arab Emirates, [email protected],
ORCID: 0000-0003-2363-2361

Eka Sediadi
American University of Ras Al Khaimah, Ras Al Khaimah, United Arab Emirates, [email protected],
ORCID: 0000-0003-0691-0999

Cazacova, L. Sediadi, E. How do students’ competitions like Solar Decathlon increase

Cite this paper as: renewable energy awareness and who does benefit? 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: This article aims to demonstrate how Solar Decathlon and other similar competitions disseminate
knowledge in renewable energy and particularly in solar energy by bringing together students,
academicians, industry and community for designing and building sustainable houses that operate
with solar energy. Through providing examples of competitions - Solar Decathlon Middle East
2018 and 2020 and Oman Eco House Design competition, presenting students learning outcomes
and the benefits for stakeholders, this article shows that such competitions increase public
awareness in renewable energy and contribute the “Mind change” change of the community.
Keywords: Renewable energy awareness and education, Solar energy, SDME, OEHDC
© 2020 Published by ECRES
Nomenclature
SD Solar Decathlon
SDME Solar Decathlon Middle East
OEHDC Oman Eco-House Design Competition
CO2 Carbon Dioxide
CH4 Methane
AURAK American University of Ras Al Khaimah
UD University of Dubai
DU Dhofar University

1. INTRODUCTION

Currently, offering clean, green, renewable and yet affordable energy for residents and businesses it’s
prerogative for every country. Providing energy from renewable sources means improving the environment as
it minimizes resources’ depletion and carbon footprint. The replacement of conventional resources of energy by
renewable ones also offers new economic prospects for countries and, therefore, new workplaces for young
skilled in renewable energy technologies graduates [1].
The importance of employing renewable energy and raising public awareness in this field is emphasized by
Broman and Kandral [2], who also listed the reasons why raising public awareness is so important to us: a) our
planet is lonely in the vast universe and once or if destroyed moving its inhabitants to another livable planet
won’t be possible as Venus is too hot, and Mars is too cold; b) climate change or so called global warming is
due to abundant emissions of greenhouse gasses like CO2, CH4; c) one of the major source of emissions that
contribute to global warming is combustion of fossil fuels for energy generation; and d) the consumption of
electricity should be reduced worldwide and the energy should be replaced by clean one [3].
For improving community awareness in renewable energy, Broman, Kandpal and Ott [2] suggest the
introduction of notion “Mind change”, which is similar, to “Climate change” (that monitors the changes in
climate), but stresses on changing the minds of humans in aspects of transition from fossil to renewable sources
of energy via educating the community. In their opinion community’s “Mind change” should start with

322
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

elementary/school education by introducing basic knowledge in renewable energy to the students [4] and should
continue with higher education that will train students in related field and supply well trained human resources
for the renewable energy industry [1]. During this process is essential to ensure the relevance of the knowledge
and skills imparted through the renewable energy education programs and course and their connection with the
needs of industry [2]. Here, great role plays hand-on experience or any other educational approach that engage
students into activities and suggest knowledge and skills through experiences as defined by Budiyanto and
Widiastuti [4]. This can be implemented through practical lectures (labs), trainings, projects, competitions, etc.
[4]. Community’s continuous education that starts from elementary school will lead to the formation of a new
generation, competent in renewable energy technologies or in other words “Mind changed”.
In this process of “Mind change” the role of completions such as Solar Decathlon (SD), Solar Decathlon Middle
East (SDME) and Oman Eco House Design Competition (OEHDC) organized among universities is substantial
as they provide students with hand-one experience in renewable energy. During the competition students design,
build and operate sustainable houses that utilize solar energy. The competition engages faculties, researchers,
industry, authorities and volunteers from local communities that work hand by hand with students for achieving
the best self-sustainable solar operated house [5], [6], [7].
The aim of this paper is to describe the experiences of the universities – American University of Ras Al Khaimah
(AURAK) and Dhofar University (DU) – in SDME 2018, SDME 2020 and OEHD competitions. The paper
demonstrates how via above listed completions students obtain knowledge and skills in renewable energy and
how these competitions influence community’s transformation, or its “Mind change”. The authors of this paper
as team members of the above listed competitions express their point of view toward the competitions, students
learning outcomes, their achievements and who benefited from the competitions.

2. SOLAR DECATHLON AND OTHER SIMILAR COMPETITIONS OVERVIEW

This chapter provides brief description of the competitions - Solar Decathlon, Solar Decathlon Middle East and
Oman Eco house Design Competition, and pots out competitions aims.

Solar Decathlon (SD)

SD is a competition organized by the United States Department of Energy among universities all around the
word that takes place every two years and in which students are designing and building a small-scale
demountable residence that operates with solar energy. The residence must provide adequate living conditions
and is assessed according to ten contests – architecture, market potential, engineering, communication,
innovation, water, energy, appliances, health and comfort and home life. The aim of the competition is to
demonstrate the viability of solar powered residences, encourage the application of renewable sources of energy
in residential sector, and increase community’s awareness in solar energy [5], [8].

Solar Decathlon Middle East (SDME)

SDME is a result of an agreement signed between Dubai Supreme Council of Energy, Dubai Electricity and
Water Authority, and the United States Department of Energy that aims to integrate unique and regional
characteristics into the design of solar powered houses. This competition intends to transform Dubai into a global
hub for green economy and renewable energy and strengthen its status as a place that fosters innovations,
creativity and sustainability for ensuring clean, healthy and safe environments for future generations.
The first edition of competition took place in 2018 and the houses were assembled in Mohammed bin Rashid Al
Maktoum Solar Park in Dubai for testing and exhibition. The second ongoing edition, SDME 2020, ends in
November when the houses will be tested and open for the public.
The competition challenges teams formed of students, faculties and industries’ partners to design and built
houses that are not only sustainable and solar powered but are also responsive to the local culture and climate.
The houses are assessed according to the following ten criteria – architecture, energy management, comfort
conditions, sustainable transportation, communication, engineering and construction, energy efficiency, house
functioning, sustainability and innovation [9].

Oman Eco House Design Competition (EHDC)

OEHDC sponsored by The Research Council of Sultanate of Oman and organized between the high education
institutions of the country aimed to develop research and innovative thinking skills of Oman’s youth and foster
awareness of renewable energy, sustainable social and economic development. The core of the competition

323
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

structure and rules derived from the SD and were adapted to the local climate, social and economic condition of
the country.
The competition was open to all high education institution of the country and challenged teams to design, build
and operate eco-friendly houses that are cost-effective, energy-efficient, and responsive to local climate and
culture and operate with solar energy.
Unlike SDME, for OEHDC the teams built five eco-friendly solar powered houses that are permanent structures
on the universities’ campuses. The assessment of the houses was according to the following ten contests:
Conceptual Design & Design Development, Architecture, Engineering, Sustainability, Communications,
Comfort Zone, Appliances/Lighting/Electronics and Energy Balance. The winner of the competition had to score
the biggest amount of points and build a house that blend consumer appeal and design excellence with optimal
energy production and maximum efficiency [10], [11].

3. DISCUSSIONS

This chapter discusses how these competitions influence institutions’ teaching approaches and the engagement
of students’ from AURAK and DU in SDME 2018, SDME 2020 and OEHDC. Students learning outcomes of
the competitions are pointed out as well as the stakeholders’ benefits.
Learning renewable energy and sustainability in architecture and engineering using conventional theoretical
approach only is not enough. There are aspects that students should learn by doing. Practical session, application
of measuring tools and simulation software help to fill in the gaps. The experiences of the authors show that
students are more active when they learn that helps achieving the learning outcomes of the course. Due to the
SDME competitions AURAK ‘s students were offered this opportunity of learning by doing. For the benefit of
the project, several courses in architectural and engineering departments were integrated with the project and
became “Green” oriented. These courses provided students with knowledge and skills they needed competitions.
Students studied the relationship of build environment with the eco-systems through a holistic approach and
with emphasis on energy efficiency, renewable energy and appropriate use of resources, including materials,
water, and land. They also gained understanding of the constraints of healthy ecology places on the growth of
human civilization; the ways buildings are historically related to their environment; how design can minimize
the impact of buildings and urban areas on the larger environment and skills in energy simulation software to
evaluate design options; measurement of local weather conditions using simple portable equipment;
measurement of indoor temperature and relative humidity using HOBO thermometer available in architectural
laboratory; sun path simulations. These knowledge and skills students applied in their architectural and
engineering design of the houses for the competitions with an aim of achieving better design solution by house’s
orientation and improved envelope that minimizes heat gain infiltration and, therefore, the cooling loads.
Following are the description of the competitions and students’ projects.

The HUB team, American University of Ras Al Khaimah, United Arab Emirates (SDME 2020)
SDME 2020 is an ongoing competition that started in 2019 and in which participate twenty-two teams from
eleven countries. The teams are still working on improving the systems design and the competition’s testing
contest will take place in November 2020. For this contest the teams will assemble their houses, operate them
and prepare them for testing. The competition ends with an exhibition of houses and open for public visits.
THE HUB team consists of thirty-three students, faculties from different colleges of the two universities –
AURAK and UD, and partners from local industry. THE HUB house, where the past experiences in Middle East
residence’s building accumulated by generations meet the most advanced technologies, building materials and
construction techniques, aims to offer a modern sustainable house that satisfies present generation’s desires for
better living conditions and operates with solar energy. Team’s previous experience in SDME 2018 helps to
achieve better architectural and engineering design solutions. While working enthusiastically on house’s smart
systems design students are developing innovative ideas.
As team members, students have opportunity to learn about solar energy implication in residential design,
sustainability in architecture and engineering. Besides that, during common with other teams’ meetings and
workshops, students learn from each other’s experiences and cultures, share ideas and make new friends.

324
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Al Bait Al Kamel team, American University of Ras Al Khaimah, United Arab Emirates and SoLLite-
Salalah,
Dhofar University, Oman (SDME 2018)
SDME 2018 competition in which participated fifteen teams from eleven countries started in 2016 and ended in
November 2018 with testing period and an exhibition of solar powered sustainable houses in Mohammed bin
Rashid Al Maktoum Solar Park in Dubai. This competition brought together over six hundred students and
academics from fifty-four nationalities from twenty-eight universities.

a) b)
Figure 1. (a) Al Bait Al Camel house model, (b) Al Bait Al Camel team.
Al Bait Al Kamel team from AURAK was one of the fifteen teams that finished the race and assembled their
house for the competition contest. Composed of forty-three students and twenty-one academics from
architectural, civil, mechanical, electrical, computer engineering and mass communication departments, the
team was supported by twenty-three sponsors from local industry.
Al Bait Al Kamel solar powered sustainable house design concept, based of customized modules tailored to
different function and house owner’s needs, suggest an innovative idea for modular construction of solar
powered residences. Made of modules located around a courtyard, the house can grow (horizontally) with the
expansion of the family. This “growing system” creates spaces customized for every family.
SoLLite-Salalah team from DU, Oman, led by one of the authors was one of the twenty-two teams that initially
entered the competition SDME 2018. Unfortunately, due to lack of sponsoring partners the team withdrawn
from the competition just before the assembly period. The team was composed of fifty-six students and
academics from architectural engineering, interior architecture, graphic design, civil, electrical, mechanical and
computer engineering, communication and English language departments of the university. Several graduates
of the university that had experience with OEHDC joined the team as volunteers.
Aiming to design and build a light structure that is easy to assemble and provides mobility if needed under
stressed circumstances (natural disasters, refugees, high tourist season, etc.) that operates with solar energy only,
the team worked excitedly to achieve an unique design for their SoLLite house. Driven by its slogan “go beyond
your own walls” students suggested a design concept based on grid that provides the owner of the house to be a
designer and change the layout and spaces according to the family’s desires by simple moving the walls along
the rails.
Due to the participation in these competitions students from AURAK and DU were able to obtain knowledge in
solar energy, sustainable materials and eco housing, engineering design and smart house’s systems. They
obtained also skills in design, drawing production, construction materials making and communication. With an
aim of attracting sponsor and increase community awareness, students visited numerous companies in Oman
and UAE, conducted several exhibitions.

325
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. (a) SoLLite-Salalah house model, (b) Students team, (c) Students testing invented bricks for walls
construction, (d) Exhibition by SoLLite-Salalah.

Oman Eco House Design Competition (EHDC)

For this competition five teams were selected: German University of Technology (Muscat), Sultan Qaboos
University (Muscat), Nizwa University (Nizwa), Higher College of Technology (Muscat) and Dhofar University
(Salalah) and, therefore, five eco-friendly houses were built: three in different districts in Muscat, one in Nizwa
and one in Salalah. The competition lasted three and half years from 2011 to 2015.
Dhofar University’s team as one of the competitors designed and built the Dhofari Eco House based on
traditional local architecture that incorporates modernized two wind catchers and provides contemporary
comfort conditions to its inhabitants. The house operates with solar energy, is a zero-net house (generates as
much energy as needed to its systems operation) and accessible on university campus for students and faculties
for study and for public visits. Dhofari Eco House’s multidisciplinary team was composed of students, faculties
and from different colleges and departments of the university, supported by partners from local industry and
governing authorities and led by one of the authors of the article. During three and half years of the competition
over one hundred students were engaged in different stages of the project. Several courses in architectural, civil,
electrical and mechanical engineering were integrated with the competition. Students enthusiastically worked
on assigned tasks and, therefore, were able to acquire knowledge and skills in housing sustainable design, solar
energy, design, construction drawings and specifications, bidding, procurement, construction supervision, house
operation, etc. Due to the phased evaluation of the project, its juried and testing contests, exhibition and house’s
visits students improved their communication skills. As they were well trained in diverse areas in architecture,
engineering and renewable energy, their employability increased and they were able to find good jobs.

(a) (b)
Figure 1. (a) Dhofari Eco House and its unique features, (b) Dhofari Eco House students are happy with the result of
their work and their achievements.

SDME 2018, SDME 2020 and EHDC outcomes

The experiences of AURAK and DU in SDME 2018 (Al Bait Al Camel and SoLLite-Salalah) and 2002 (THE
HUB) and OENDC (Dhofari Eco House) shows that all the stakeholders benefit from competitions – students,
academicians, universities, organizers, industry, governing authorities and communities. Therefore, the
outcomes and benefits were categorized by stakeholders and presented in Table 1, 2, 3, 4 and 5.
Table 1 is summing up students learning outcomes and shows students’ knowledge and skills in renewable
energy obtained due to the participation in competitions. Besides the field of renewable energy students obtained
knowledge and skills in sustainable housing design - architectural design, engineering systems design,

326
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

installation and operation, construction drawings, specifications, project scheduling and cost estimation,
construction supervision, project management, teamwork and communication due to the required multiple
deliverables, juried and testing contest of the competitions and through workshops, presentations, exhibitions,
and public visits to the house.
Students acquired experiences that they are not able to get in the class as they gained their experience through
the execution of each phase of the project and competition starting with design, through construction, assembly,
juried and testing contest and finishing with exhibition. Furthermore, due to their collaboration with the local
industry, students had opportunity to learn and got exposure, which all together with obtained skills during the
competition increased their possible employability upon graduation.

Table1. Competitions outcomes - Students learning outcomes – renewable energy

Category Outcome Notes
Knowledge Renewable energy and particularly solar Study on PV panels systems for choosing the most
energy suitable; Study of alternative sources of energy
Sustainable design Study on energy efficient systems
Smart building's systems Study on smart building automation and controls for
minimizing electricity budget
Local and international market, products Students learned the availability and types of Solar
and materials – PV systems Panels
Electrical cars Study on electrical cars that are run during the testing
contest
Skills PV systems design Power budget estimation and PV system and number of
panels – competition requirements
PV systems installation Assembly of the house and PV panels on site for
competition
PV systems operation Operation during the testing contest
Electrical car operation Electrical car driven by students during testing contest
was charged with electricity generated by PV panels

As the research for competitions and projects were led by academicians that worked hand by hand with students
and industry partners with an aim of finding the best solution for minimizing the energy budget for the house
and selecting the best solar powered system, this resulted with articles publications and collaborative research
with industrial partners as shown in Table 2.

Table 2. Competitions outcomes – awareness in renewable energy and benefits for academicians.
Category Outcome Notes
Research Publications in journals, conference proceedings, Academics conduct research in renewable
books, book chapters energy, sustainable housing and engineering
systems. aiming to have the best project.
Collaborative research projects with industrial Contribution to the industry movement
partners with an aim of finding the best solar power toward application of renewable energy and
system for the house, and best architectural and sustainable housing
engineering solutions.
Availability of the house on campus for research Academicians have an opportunity to
continue research after the competition on
house solar system improvement

Due to the universities teams’ participation in competition, the entire family of the universities (students,
academicians, technical and administrative staff) learned about solar energy application in residential sector, as
after the competition the houses are available on universities campuses, and assessable for universities’ students,
academic and administrative staff visits as well as for public visits (schools, companies, etc). Courses are
integrated into projects and new elective courses in renewable energy offered which leads to increased popularity
of the exciting programs. Table 3 given below shows how renewable energy awareness increases among
universities and how do universities benefit from competitions.

327
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 3. Competitions outcomes – awareness in renewable energy and benefits for academicians.
Category Outcome Notes
Exposure University’s reputation and visibility by supporting the International competitions with teams
team, competition and the application of renewable energy from different countries, extensive media
in housing coverage
Graduates Improved quality of graduates Students gained knowledge and skills in
renewable energy
Enrolment Increase in number of students Due to the exposure and alumni
experienced in renewable energy

Partners from local and international industry supporting students’ teams have also opportunity to contribute to
the renewable energy employment in housing and to suggest new ways of their products application. Together
with academicians and students they come with innovative ideas reducing energy consumption in buildings and,
therefore, with new products. Table 4 demonstrates the benefits of partner companies due to their participation
in competitions.

Table 4. Competitions outcomes – awareness in renewable energy and benefits for industries.
Category Outcome Notes
Reputation Increased reputation of the company as Increased visibility of the company and its
supporter of energy efficient housing and products due to their employment in students’
renewable energy application house.
Products New products and their promotion; sales Students are promoting sponsoring teams and their
increase products – competition requirement
Staff Employment of staff qualified in renewable Ability to employ university graduates qualified in
energy renewable energy and collaborate with
academicians.

Due to the multiple stakeholders of the competitions – students, academicians, universities, industries - and to
the publicity of the competitions, the community awareness in renewable energy increases. Local and global
community benefits by having qualified specialists in renewable energy, local industry’s products and educated
in the field population (Table 5). This “Mind change” of the community opens for countries’ governing
authorities opportunities for substituting conventional energy for building sector maintenance for renewable
ones and also for creating sustainable communities.

Table 5. Competitions outcomes – awareness in renewable energy and benefits for the community.
Category Outcome Notes
Mind change Educated in renewable energy Opportunity to replace the conventional energy to the
community renewable once
Opportunity to create sustainable communities.

4. CONCLUSION

Current situation shows that the process of merging renewable energy with education it’s in its beginning phase.
Even though, institutions are embedding technical courses in the field of renewable energy into the
undergraduate curriculum of programs offered, this won’t be enough to satisfy the fast-growing renewable
energy industry and its demand for qualified specialists in the field, nor will this be enough for increasing
community’s awareness in renewable energy [1].
Moreover, as indicated in International Renewable Energy Agency report [12], building sector is one of the
biggest consumer of electricity and, therefore, building codes are aiming for near-zero or even energy positive
buildings in nearest future. Meaning, more and more buildings will be operated by renewable energy and
households will rely mostly on solar energy. This will require not only technically educated specialists but also
an educated community for operating those households [13].
Multiple editions of Solar Decathlon demonstrate a new enhanced approach in education in which a multi-
disciplinary, project-led teaching method is successfully implemented in existing degrees of educational
institutions involved in competitions [8]. Such competitions motivate multi-disciplinary teams (students,

328
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

faculties, industrial partners), targeting to win the competition, for research and innovations in field of solar
energy and sustainable housing as the best outcome for the competing team is to see their house displayed and
open for public visits [9].
The outcomes of the competitions SDME 2018 and 2020 and OEHDC presented in this paper shows that such
competitions provide students with a unique opportunity in their educational experience [8]. This hands-on
approach in education through competitions facilitates institutions to graduate better qualified specialists in
renewable energy that are ready to meet architectural and engineering challenges of twenty-first century. With
their experience in renewable energy, competition, teamwork they are prepared for designing and building near-
zero or even energy positive buildings and infrastructures [8].
The outcomes also show that such competitions among students promote the application of renewable energy,
sustainable living and contribute to the creation of an educated community [9]. More ever, the article shows that
due to the multiple stakeholders of the competitions - universities, industrial partners, governing authorities,
countries and community, the awareness in renewable energy increases. Such competitions contribute
significantly to the “Mind change” of the country’s and global community via educating students, institutions,
industries and citizens.

ACKNOWLEDGMENT

The authors would like to express their gratitude to American University of Ras al Khaimah and Dhofar
University for the support during the competitions as well as to the organizers of the Competitions – Dubai
Water and Electricity Authority and United States Department of Energy and the Research Council of Oman.

REFERENCES

[1] Bajpai, S, Kidwai, NR. Renewable Energy Education in India. Comparative Professional Pedagogy 2017, 7[4], 103-
113; DOI: 10.1515/rpp-2017-0057.
[2] Broman L, Kandpal TC, Ott A. On the Importance of Education When Implementing Renewable Energy. In: ISES
Conference Proceedings; EuroSun, 7 June 2018: International Energy Society. DOI: 10.18086/eurosun2018.0706.
Available at: http://proceedings.ises.org.
[3] Broman, L, Kandpal TC. PURE - Public Understanding of Renewable Energy. In: World Renewable Energy Congress
2011, 8-13 May 2011: Policy Issues, pp. 2478-2484.
[4] Budiyanto, C, Widiastuti, I. Hands-on Learning on Renewable Energy – A Proposed Approach for Technology
Dissemination. In: AIP Conference Proceedings 1977, 060017 (2018); https://doi.org/10.1063/1.5043029.
[5] Herrera-Limones, R, León-Rodríguez, AL, López-Escamilla A. Solar Decathlon Latin America and Caribbean: comfort
and the balance between passive and active design. Sustainability 2019, 11, 3498; DOI:
http://doi.org/10.3390/su11133498.
[6] Eastment, M, Hayter, S, Nahan, R, Stafford, B, Warner, C, National Renewable Energy Laboratory. Solar Decathlon
2002: The Event in Review. U.S. Department of Energy. Energy Efficiency and Renewable Energy.
[7] Arhipova, A. Academy-Industry Cooperation: Case Solar Decathlon Competition. MSc. Aalto University, 2010.
[8] Bailer, R, Tao, Y. An interdisciplinary pedagogical teaching approach for engineering, in conjunction with architecture
and construction with solar decathlon project. American Society for Engineering Education 2007, 227-240.
[9] Cazacova, M, Farhan, MB. Innovative building material for the house construction for Solar Decathlon Middle East
2018 competition. MATEC Web of Conferences 2018, 249, 01009. DOI:
https://dio.org/10.1051/mateconf/201824901009.
[10] The Research Councils of Oman. Available at :https://www.trc.gov.om/trcweb/topics/projects.
[11] Oman EcoHouse Design Competition. Available at: https://ecohouse.trc.gov.om/ecohouse/.
[12] Dhifari Eco House. Project Manual. Available at: https://ecohouse.trc.gov.om/ecohouse/.
[13] IRENA. A Road Map to 20150. Global Energy Transformation. International Renewable Energy Report, 2018.
Available at: www.irena.org.
[14] Benchikh, O. UNESCO’s Global Renewable Energy Education and Training Programme. UNESCO’s GREET
Programme, Science Forum, 2004.

329
 MAPPING THE SOLAR ROOF POTENTIAL OF THE NORTH
CAMPUS BUILDINGS AT THE UNIVERSITY OF ALBERTA

Joseph Nofech
University of Alberta, Edmonton, Canada, [email protected], ORCID: 0000-0002-9553-953X

Nima Narjabadifam
University of Alberta, Edmonton, Canada, [email protected], ORCID: 0000-0001-9549-0484

Hadia Awad
University of Alberta, Edmonton, Canada, [email protected], ORCID: 0000-0002-6452-3973

Mustafa Gül
University of Alberta, Edmonton, Canada, [email protected], ORCID: 0000-0002-7750-0906

Nofech, J., Narjabadifam, N., Awad, H., Gül, M. Mapping the solar roof potential of the North
Cite this paper as: Campus buildings at the University of Alberta. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: Optimization of solar photovoltaic (PV) systems requires individual rooftops to be identified, and
performing this on a large scale is especially reliable given a topographical elevation map of the location.
Using 3D models from Google Maps as input, we reconstruct elevation maps of ten buildings at the
University of Alberta (with the possibility for expansion to a city-wide scale in the future) located in
Edmonton, Canada (53.5461° N, 113.4938° W), and present a preliminary rooftop identification algorithm
that circumvents the half-metre-scale imperfections in these 3D models. We also carry out a preliminary
cost/energy study on the identified rooftops to find tilt and profile angles in ways that minimize the
installation cost per kilowatt-hour and maximize the energy generation. Furthermore, rooftop electricity
generation potential and solar PV system installation costs are estimated based on these optimum tilt and
profile angles. Preliminary comparisons between all ten buildings show that in cold-climate regions (high
latitude regions), fixing the tilt and profile angles to 25.5 degrees and 48.12 degrees, respectively,
maximizes the annual energy generation. In the case of minimizing the installation cost per kWh, it is not
possible to generalize the results because the cost of installation varies from region to region.
Keywords: photovoltaic system optimization, roof detection, flat roofs, image processing, high latitude
© 2020 Published by ECRES

1. INTRODUCTION

The energy and environmental crisis in recent years has led countries around the world to take a different
approach to energy issues, including the replacement of fossil fuels with renewable sources of energy, to reduce
and save energy consumption, to control supply and demand, and to reduce their carbon footprints. To this end,
all potential locations in which a photovoltaic system could be placed (e.g. rooftops, walls, etc.) in urban and
rural areas should be exploited to optimally utilize renewable energy. This includes the detection of residential,
commercial, and office rooftops for optimum installation and production of photovoltaic (PV) systems. In order
to extract the rooftop information required for optimizing solar PV installations, a variety of different methods
exist and are frequently used in this field of research. Roof detection is possible through extensive image
processing of satellite photographs of cities [1], and an increasingly common method is to obtain a digital surface
model using light detection and ranging (LiDAR; the interested reader can refer to studies by Mansouri
Kouhestani et al. [2] and Jakubiec and Reinhart [3] for examples), often in combination with geographic
information system (GIS) data. Google in particular produces full 3D reconstructions of locations using
photogrammetric analysis of numerous photographs obtained via airplane, for use in their Google Maps and
Google Earth applications [4].

330
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In this study, we present the preliminary results of our ongoing work on mapping the solar roof potential of the
North Campus buildings at the University of Alberta, with the possibility of being expanded to a city-wide scale
in the future. The overall purpose of this project is to procedurally identify usable rooftop space and to develop
a software tool that optimizes solar PV placement and energy generation on these rooftops, thus determining
their solar potential. Rooftop information is extracted from 3D models, and then the electricity generation
capacity of each identified rooftop is investigated. Section 2 details the current rooftop recognition process,
along with the methodology for the solar generation/cost analysis. Section 3 includes the current results and a
discussion of the process, and the work completed thus far is summarized in Section 4.
This work is expected to have a positive impact on the environment by procedurally optimizing solar PV
installations, paving a path for economically-viable renewable energy. We end this introduction by noting that
this project is currently a work in progress, and these preliminary results should not be considered final at this
stage.

2. METHODOLOGY

Building Models
Information about each building's rooftops is obtained using Google's 3D models of the buildings as they appear
in Google Maps [5]. Because these models are used primarily for aesthetic and interactivity purposes in the
Google Maps interface, Google has not provided a method of parsing or otherwise downloading the models. We
instead obtain the models by using RenderDoc [6] to capture "snapshots" of these models directly from the GPU
as they are being rendered. After capturing, the RenderDoc output is converted into the more common .STL
format by opening it in Blender [7] with the use of a custom add-on [8] and exporting the mesh alone without
any texture information. Finally, based on the assumption that solar panels will not be placed on vertical surfaces
(or on surfaces that have a surface directly above), the 3D model of each building is simplified into a 2D "height
map". As an example, Figure 1 shows the textured model of the Centennial Centre for Interdisciplinary Science
(a University of Alberta building, hereafter CCIS) in the Blender interface (left) and its corresponding height
map (right).
While the 3D models appear to have arbitrary scaling, a conversion factor between model units and real-world
units can be found by selecting two points on the model, and finding the corresponding real-world horizontal
distance between these points using a tool in Google Maps. For the time being, this is performed manually for
each building; however, we intend to automate this process in the future.
A benefit of obtaining rooftop information via this method is that Google Maps provides relatively accurate, up-
to-date models for every building in most populated cities, meaning that this method can ideally be applied to
nearly any building in any city without needing to request building plans (and without needing to accommodate
for differences in drawing format). This means that the process can, in principle, be applied to projects of any
scale, provided that all of the necessary Google Maps models exist.

Figure 1. CCIS Google Maps model with textures (left), and height map in metres (right). Imagery, Map Data © 2019
Google.

However, a significant downside is the accuracy of these models. A closer inspection of each building will reveal
that the photogrammetry process generates models with a resolution sufficient for the aesthetic purposes of
Google Maps/Earth, but due to the sheer size of the triangles (that is, the triangular "faces" comprising the

331
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

model), imperfections on the scale of up to several tens of centimetres are rampant. This causes even flat rooftops
to have a distinctly "lumpy" appearance, and simple parapet structures are often represented by a gradual
increase in height (if at all) rather than the actual vertical bearing of a parapet at the end of a roof.
In addition, issues can also occur during rendering that may cause difficulties in the analysis. Triangles will
occasionally fail to render for each building (punching distinct triangular "holes" into rooftops), and on occasion,
entire areas can fail to render behind large objects due to occlusion culling (an optimization feature that causes
the GPU to not render obstructed objects for the sake of performance). These issues are currently dealt with on
a case-by-case basis for each building. This overall process has not yet been automated—models are currently
obtained, fixed, and converted to real-world units manually, requiring up to several minutes of effort per
building.
With this in mind, we present the current state of our work as performed on the following ten University of
Alberta buildings: the Administration Building (ADMIN), Biological Sciences Building (BioSci), Cameron
Library (CAM), Centennial Centre for Interdisciplinary Science (CCIS), Donadeo Innovation Centre for
Engineering (DICE), Engineering Teaching and Learning Complex (ETLC), General Services Building (GSB),
National Institute for Nanotechnology (NINT), Natural Resources Engineering Facility (NREF), and Student
Union Building (SUB).

Roof Identification
Rooftops are identified directly from a building's height map. An individual roof in this case is defined as a
region that is isolated from the rest of the map by a sharp discontinuity in height, at least 3.5 m above the ground
height (which is approximated for the full height map by identifying and fitting to low-lying regions), and able
to fit at least one solar panel with a 1.8 m walkway around it (following current regulations in the California
Solar Photovoltaic Installation Guideline [9]). This process separates the height map into distinct isolated regions
that we will refer to as "blobs" [10], with the end goal being that each blob represents a distinct face of a rooftop
with its own height and slope. For example, a hip rooftop would be separated into four blobs representing its
four faces.
At this point in the algorithm, due to limitations in the models, some distinct rooftops may be part of a single
blob due to the edge between them not being correctly identified (e.g. if the height difference between them is
too small). To address this, each blob is checked for regions where the height does not increase significantly,
and if multiple such regions are found and significantly differ in height, then the region connecting them is
removed.
In addition, due to the inaccuracies inherent in Google's building models, many rooftop blobs feature sudden,
inaccurate "spikes" in height. These regions are detected and flagged as "unreliable". They are not removed from
the blob for the purpose of retaining shape, but are omitted from subsequent calculations involving height or
slope.

Roof Classification
At this point, individual rooftops have for the most part been isolated and are ready to be analyzed in closer
detail. Curved or otherwise irregularly-shaped rooftops are not considered due to physical difficulties in
installation. As such, rooftops will be classified into one of three categories: "flat" (i.e. rooftops that have a
roughly constant height, with or without parapets), "slanted" (i.e. rooftops that have a fairly constant slope), or
"irregular" (i.e. regions that do not have a constant slope, such as curved roofs or treetops, and are not considered
for PV installation).
Each individual rooftop is first checked for flatness by comparing the percentile values to percent ranks of its
height map. If most of the rooftop has a relatively constant height, then the rooftop is classified as "flat", and the
mean of the height map gives the final height of a flat rooftop. If the comparison between height percentile
values and percent ranks reveals multiple distinct flat regions, then parapet height (if any) is approximated by
the difference between the final measured height and the highest "flat region" of the blob. Otherwise, parapet
height is simply approximated as the difference between final height and the 90th percentile of the height map.
Any rooftop that is not flat is then checked for slantedness by fitting a plane to it. If the standard deviation of
the residuals between the roof's heightmap and the plane fit is below 0.09 m, then the rooftop is classified as
"slanted", and the slope magnitude, slope direction, and (peak) roof height are given by the plane fit.
Any rooftop that is neither flat nor slanted is then treated as "irregular". Here, we note that rooftop blob
separation has not considered sharp changes in slope (e.g. the highest "ridge" of a gabled rooftop), due to the
irregularities inherent in Google Earth's building models. As a result, many roof types that are composed of

332
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

connected, distinctly-sloped regions (e.g. gabled, hip, and butterfly rooftops) are incorrectly considered as
single, irregular blobs at this stage. To check for this, the "ridges" that separate rooftop faces are identified by
finding local maxima and minima running along various directions, and removing the ridges only if they split a
blob into multiple distinctly-sloped blobs (as many removed ridges would otherwise be false positives caused
by imperfections in Google's building models).
Once the irregular rooftops of a building have been split into their component blobs, these component blobs are
fed into the classification code and identified as "flat", "slanted", or "irregular". This recursive process is run
several times; afterwards, any rooftop that is still classified as "irregular" will simply stay that way. With all of
the rooftop blobs properly classified, the usable rooftops can now be analyzed for solar PV system cost and
energy gains.

Solar Calculations
A case study is being conducted for ten rooftops on the University of Alberta campus located in the city of
Edmonton, Canada (53.5461° N, 113.4938° W), which is in a cold-climate zone. When optimizing solar PV
placement, the tilt angles and spacing must be accounted for such that panels are not excessively shaded by
panels in front of them; however, it is sometimes inevitable. The shade of each panel on the rooftop and the
fraction of the area of the panel in row r+1 shaded by the panel in row r could be found based on the work
authored by Duffie and Beckman [12]. Bear in mind that the shade area of each panel on the rooftop includes
the projected area of the panel itself on the rooftop too. To find the number of panels that could be installed on
each harvestable area or blob, as mentioned before, the effective harvestable area is divided by the shade area
of each panel on the rooftop. The effective harvestable area (Areaeff,blob ) of the rooftop is calculated using Eq.
[1]:

𝐴𝑟𝑒𝑎𝑒𝑓𝑓,𝑏𝑙𝑜𝑏 = 𝜎(𝐴𝑟𝑒𝑎𝑎𝑐𝑡𝑢𝑎𝑙,𝑏𝑙𝑜𝑏 − 𝐴𝑟𝑒𝑎𝑜𝑓𝑓𝑠𝑒𝑡 ) (1)

where Areaactual,blob is the actual or total area of each blob, Areaoffset the offset walkway area around each blob,
and 𝜎 the coefficient for considering the walkway for maintenance.
In order to calculate energy generation of system, we need to introduce a parameter called the monthly
generation yield (Ymonth) defined as:
(2)
𝐺𝑃𝑉,𝑠𝑦𝑠
𝑌𝑚𝑜𝑛𝑡ℎ =
𝑛×𝑃

where GPV,sys is the monthly energy generation of solar PV system (kWh), P the panel capacity (kW), and n the
number of installed panels. Ymonth values are extracted from PVWatts®. The values for monthly generation yield
for different tilt angles of PV panels are represented in a study authored by Awad [13], and are obtained based
on the latitude of the city being studied and the cloudiness of the weather.
Consequently, using Eq. [2] and the monthly yield value, one can calculate the monthly energy generation of
the solar PV system (GPV, sys) according to Eq. [3]:
𝐺𝑃𝑉,𝑠𝑦𝑠 = 𝑌𝑚𝑜𝑛𝑡ℎ 𝑃𝑛(1 − 𝐿𝑠𝑛𝑜𝑤 )𝛼 (3)

where Lsnow is the snow coverage loss factor, and 𝛼 is the fraction of the panels that is not shaded. Lsnow values
for different tilt angles could be obtained from reference [13].
Finally, the installation cost (Costinstall, $) for the solar PV system can be estimated as shown in Eq. [4]:
(4)
𝐶𝑜𝑠𝑡𝑖𝑛𝑠𝑡𝑎𝑙𝑙 = 𝐶𝑤 𝑛𝑃

where Cw is the cost per installed watt, n is the number of installed panels, and P is the panel capacity. Cw's value
is between 2.77 and 3.02 dollars per watt for Alberta, Canada. As the size of the solar PV system increases, the
cost of production per watt approaches the lower boundary of the range [14]. Since the size of installed PV

333
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

systems considered in this study are in megawatts, we choose 2.77 dollars per watt for installation cost
estimation. All costs are in Canadian dollars.

3. RESULTS AND DISCUSSIONS

Rooftop Recognition
After fully identifying/classifying the rooftops for the ten buildings, we note overall that while the behaviour of
the algorithm with respect to wide, simple rooftops is largely consistent with our expectations, there are a few
noteworthy issues that present themselves (caused by both inaccuracies in the Google Maps models, and
limitations in the current algorithm itself). For instance, some buildings have small objects on them (e.g. narrow
pipes on top of GSB) that do not appear in Google Maps' 3D models at all, and must be accounted for with
another method. Some particularly small slanted rooftops are mistakenly classified as "flat". Certain blobs with
relatively complex shapes can also fail to separate into their component rooftops. In addition, the current code
that splits irregular rooftops based on local minima and maxima does not support gambrel roofs, or roofs where
"edges" do not manifest as a peak or trough in height; these roofs are still treated as connected, "irregular"
objects. Issues such as these will be worked through in later iterations of the code.
In addition, rooftops are currently "classified" in the sense that they are each given measurements for their height,
their parapet height (if any), their slope, their direction (if any), and their surface area. A crucial goal of this
research is to establish a method of optimizing solar panel installation within a given arbitrary rooftop shape;
however, as will be mentioned later, this part of the project is still in its infancy and will be further developed in
the future.

Minimizing the Cost per kWh Energy Generation

MATLAB® is used to find the optimum values for tilt and profile angles. Cost per kWh varies with tilt angles
of PV panels and profile angles as shown in Figure 2. When the tilt angle of panels and profile angle are fixed
to 50 degrees and 13.02 degrees (minimum cost per kWh conditions), respectively, the minimum cost per kWh
for solar PV systems is achieved. For the minimum cost per kWh conditions, Figure 3 presents the energy
generation per harvestable and gross area for a 300 W CanadianSolar PV panel with dimensions of 1.650 m ×
0.992 m on the buildings under study at the University of Alberta.
The gross rooftop area for each building is obtained by summing the area of all types of rooftops including flat,
slanted, and irregular. On the other hand, the harvestable area is equal to the sum of flat and slanted areas.
As shown in Figure 3, when the tilt and profile angles are set to 50 degrees and 13.02 degrees, respectively, the
NREF building produces the most electricity per gross and harvestable area annually, which are 71.1 and 84.4
kWh/m2, respectively.
Table 1 shows the annual energy generation in MWh and installation cost of PV systems in thousand dollars for
different buildings when the tilt and profile angles are fixed to 50 degrees and 13.02 degrees, respectively. The
NREF and the ADMIN buildings generate the most and the least electricity annually, 649.0 MWh and 56.2
MWh, respectively. Accordingly, the highest and lowest installation costs are 1,352.0 thousand dollars and 117.2
thousand dollars, respectively.

Figure 2. Cost per kWh variance as a function of tilt and profile angles.

334
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Energy generation per gross and harvestable area for the buildings under study for the minimum cost per kWh
conditions.

Table 1. Electricity generation and installation cost for the buildings under study for the minimum cost per kWh conditions
ADMIN BioSci CAM CCIS DICE ETLC GSB NINT NREF SUB

Annual Generation (MWh) 56.2 159.6 440.0 468.7 516.2 296.0 136.0 225.8 649.0 409.3

Installation Cost ($000s) 117.2 332.4 916.6 976.4 1,075.3 616.6 283.4 470.4 1,352.0 852.6

Maximizing the Energy Generation

The other way to calculate the annual electricity generation is to find the maximum energy generation. Figure 4
shows the variation of non-dimensional energy generation with respect to the tilt and profile angles.

Figure 4. Non-dimensional energy generation versus tilt and profile angles.

The non-dimensional energy generation is obtained by dividing the generation by its maximum value. The
maximum energy generation of the solar PV system happens when the tilt and profile angles are 25.5 degrees
and 48.12 degrees, respectively (maximum generation conditions). For the maximum generation conditions,
Figure 5 depicts the energy generation per harvestable and gross area for the aforementioned PV panel for the
buildings under study at the University of Alberta. For the maximum generation conditions, SUB produces the
most electricity per gross and harvestable area annually, which are 116.4 kWh/m 2 and 119.2 kWh/m2,
respectively. Table 2 shows the annual energy generation in MWh and the installation cost of PV systems in
thousands of dollars for the buildings under study for the maximum generation conditions.

335
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 5. Energy generation per gross and harvestable area for the buildings under study for the maximum energy
generation conditions.

Table 2. Electricity generation and installation cost for different buildings for the maximum generation conditions
ADMI BioSc CAM CCIS DICE ETLC GSB NINT NREF SUB
Annual Generation N
141.1 i
386.8 755.4 868.3 1,179. 729.8 336. 507.4 757.9 1,007.
(MWh)
Installation Cost ($000s) 366.5 1,004. 1,962. 2,255. 3
3,063. 1,895. 9
875. 1,318. 1,968. 2
2,616.
7 0 3 1 5 0 0 6 0
The DICE and ADMIN buildings generate the most and the least electricity annually, 1,179.3 MWh and 141.1
MWh, respectively. Accordingly, the highest and lowest installation costs are $3,063,100 and $366,500,
respectively.
Comparing the results shown in Table 2, it can be concluded that the generation rate of buildings is higher when
the maximum generation conditions are fixed rather than using the conditions of the minimum cost per kWh.
However, installation costs will also be proportionally higher.

4. SUMMARY AND CONCLUSION

In this study, we aim to automatically identify usable rooftop regions and optimize photovoltaic (PV)
installations on these regions using 3D building models as input. This is performed across a variety of buildings,
starting with ten buildings at the University of Alberta with plans to eventually move to a city-wide application.
With 3D models of these buildings scraped from Google Maps, and employing a series of statistical methods
and simple morphological processing methods, we present rudimentary detection of flat and slanted rooftops,
showing largely consistent (although somewhat buggy) results across these ten buildings.
Following this, the possible rooftop energy generation, energy generation per gross and harvestable area, and
installation costs of PV systems for different buildings were evaluated using the identified rooftops. To find the
best solar PV system, two different parameters were investigated: cost per kWh energy generation and electricity
generation. The results showed that fixing the profile and tilt angles to 13.02 degrees and 50 degrees,
respectively, minimizes the cost per kWh, and when profile and tilt angles are 48.12 degrees and 25.5 degrees,
respectively, the electricity generation is maximized. It is worth mentioning that tilt and profile angles obtained
for maximizing generation could be generalized for high latitude regions with cold-climate conditions; however,
those attained for cost minimization could not be generalized since the installation cost per watt varies from
region to region. Regarding minimum cost per kWh conditions, the NREF and BioSci buildings produce the
highest and the lowest energy, respectively, per harvestable and gross area. The highest annual electricity
generation and installation costs belong to the NREF building with 649.0 MWh and $1,352,000, respectively.
However, using the conditions of maximizing energy generation, the SUB and BioSci buildings have the highest
and the least electricity generation potential per gross and harvestable area, respectively. The DICE building has
the potential to generate 1,179.3 MWh, which costs $3,063,100.
Overall, the approximately half-metre-scale imperfections in Google Maps' building models cause rooftop
classification to become deceptively difficult. Flat rooftops often exhibit a "lumpy" appearance in the 3D model,
and sudden variations in slope of the triangles that comprise a 3D rooftop model cause simple edge detection
methods to occasionally fail. However, despite the difficulties presented, we succeeded in classifying many of
the rooftops captured so far, with larger rooftops resulting in the most consistently ideal algorithm behaviour.
In the future, in addition to improving the current roof-identification algorithm to reduce inaccuracies, we also
intend to implement functionality such that a rooftop of an arbitrarily complicated shape can have an optimal

336
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

solar PV configuration generated automatically, without the need for applying a simplifying assumption as to a
rooftop region's shape. This will act as the crucial connecting step between identifying/classifying rooftops and
making practical use of that information. We will also analyze the year-round shading, and account for its effect
on the solar potential across each rooftop. Also, the optimization of solar PV systems could be conducted for all
of the University of Alberta’s buildings. Risk assessment and financial analysis would be the next steps in the
future.

ACKNOWLEDGMENT
We express our deepest gratitude to members of the University of Alberta's Energy Management and Sustainable
Operations (EMSO) project, including Michael Versteege, Shannon Leblanc, and Jason Yuzyk, for their insight
and guidance in this study.

REFERENCES

[1] Saeedi, P., Zwick, H.. Automatic Building Detection in Aerial and Satellite Images. IEEE International Conference on
Control, Automation, Robotics and Vision 2008, 623 - 629. DOI: https://doi.org/10.1109/ICARCV.2008.4795590
[2] Mansouri Kouhestani, F., Byrne, J., Johnson, D. et al. Evaluating solar energy technical and economic potential on
rooftops in an urban setting: the city of Lethbridge, Canada. International Journal of Energy and Environmental
Engineering 2019, 10, 13–32. DOI: https://doi.org/10.1007/s40095-018-0289-1
[3] Jakubiec, J., Reinhart, C.. A method for predicting city-wide electric production from photovoltaic panels based on
LiDAR and GIS data combined with hourly DAYSIM simulations. Solar Energy 2013, 93. 127-43. DOI:
https://doi.org/10.1016/j.solener.2013.03.022.
[4] Dennis, N.. Google Earth's Incredible 3D Imagery, Explained [Video file]. 2017, April 18. Retrieved from
https://www.youtube.com/watch?v=suo_aUTUpps.
[5] Google Earth 7.3 (2019). University of Alberta 5331'36.52"N, 11331'30.98"W. 3D map, Buildings data layer, viewed
December 2019.
[6] Karlsson, B.. RenderDoc version 1.1. 2019. Retrieved from https://renderdoc.org/.
[7] Blender Foundation. Blender: A 3D modelling and rendering package version 2.81. Stichting, Amsterdam, 2019.
Retrieved from http://www.blender.org.
[8] Michel, E.. Maps Models Importer. 2019. GitHub repository, https://github.com/eliemichel/MapsModelsImporter
[9] Awad, H., et al.. Solar photovoltaic optimization for commercial flat rooftops in cold regions. In: 2016 IEEE Conference
on Technologies for Sustainability (SusTech); 9-11 October 2016, IEEE, Phoenix, AZ: pp.39-46.
[10] Solomon, C., Breckon, T.. Fundamentals of Digital Image Processing: A Practical Approach with Examples in Matlab.
UNITED KINGDOM: Wiley-Blackwell, 2012.
[11] Hutcheon, N. B., Handegord, G. O. P.. Building science for a cold climate. TORONTO: Wiley, 1983.
[12] Duffie, J. A., and Beckman, W. A. Solar engineering of thermal processes. Hoboken, New Jersey, USA: John Wiley
& Sons, Inc, 2013.
[13] Awad, H.. "Integrating Solar PV Systems into Residential Buildings in Cold-climate Regions: The Impact of Energy-
efficient Homes on Shaping the Future Smart Grid". PhD, University of Alberta, Edmonton, Canada (2018).
[14] https://energyhub.org/cost-solar-power-canada/ (accessed February 21, 2020)

337
 INXGA1-XP AND INXAL1-XP HETEROSTRUCTURES FOR
CASCADE SOLAR CELLS INVESTIGATION

Аndrey Lebedev
JSC Scientific-production plant “Kvant”, Moscow, Russia, National university of science and technology “MISiS”
Moscow, Russia, [email protected],ORCID: 0000-0101-7911-8495

Anna Naumova
JSC Scientific-production plant “Kvant”, Moscow, Russia, [email protected], ORCID: 0000-0198-2919-8497

Boris Zhalnin
JSC Scientific-production plant “Kvant”, Moscow, Russia, [email protected], ORCID: 0000-0129-8219-8455

Nina Vagapovа
JSC Scientific-production plant “Kvant”, Moscow, Russia,
National university of science and technology “MISiS” Moscow, Russia, [email protected], ORCID: 0000-0199-
7111-8015

Еvgeny Slyschenko
National university of science and technology “MISiS” Moscow, Russia, [email protected], ORCID: 0000-0871-
3111-5995

Sergey Sharov
JSC Scientific-production plant “Kvant”, Moscow, Russia, [email protected], ORCID: 0000-0187-8219-8429

Ekaterina Obrucheva
JSC Scientific-production plant “Kvant”, Moscow, Russia, [email protected], ORCID: 0000-0112-8219-8414

Sergey Didenko
National university of science and technology “MISiS” Moscow, Russia, [email protected], ORCID: 0000-0820-
3100-5823

Anatoliy Izotov
National university of science and technology “MISiS” Moscow, Russia, [email protected], ORCID: 0000-0810-
3100-5802

Andrey Lebedev, Sergey Didenko. InXGa1-XP and InnAl1-XP heterostructures for cascade solar
Cite this paper as:
cells investigation. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: This work is focused on InGaP and InAlP epitaxial layers growth. Such heterostructures are the
main materials not only for contemporary triple-cascade solar cell, but also in other promising
photovoltaic devices. The optimal technological parameters for the InGaP and InAlP epitaxial
growth are determined. The growth results of single-crystal InxGa1-xP epitaxial layers with
different Indium and Gallium atoms concentration on a Ge substrate, as well as In xAl1-xP layers
with different Aluminum and Indium atoms concentration on Ge and GaAs substrates are
discussed. The rocking curves obtained using high-resolution two-crystal X-ray diffractometry
were investigated. By a high-resolution two-crystal X-ray diffractometry the lattice parameter and
the ratio of indium / gallium / aluminum atoms in the layers were calculated. The dependence of
the solid phase composition versus the gas phase composition was detected based on the lattice
parameter data. As a criterion for grown structure quality investigation, diffraction X-ray peaks
broadening can be used.
It is detected that for the In xGa1-xP solid solution in the range X = 45–53 % and for InxAl1-xP in the
range X = 46-51 %, a high perfection of the single-crystal structure and a slight diffraction X-ray
peaks broadening is observed.
Keywords: Heterostructure, solar cell, X-ray diffractometry
© 2020 Published by ECRES

338
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION

Solar battery (SB) converts sunlight directly into electricity with high efficiency, creating almost constant power
according to the increase in requirements for spacecraft systems, there is a need to create spacecraft with high
operational, energy characteristics and an extended life (over 15 years). The most promising solar cells (SC) for
modern satellites operated in the spacecraft are cascade solar cells based on AIIIBV materials. A modern three-
cascade solar cell is a complex planar device. Semiconductor structure for a solar cell consists of three dozen
InGaP / InGaAs / Ge layer cascades. For the semiconductor structures growth MOCVD is used. However, due
to the relatively large area of solar cells, which is not characteristic for other semiconductor devices, about 30
cm2 in order to achieve the high characteristics, it is particular important to meet the requirements for the
properties uniformity in each structure layer [1-3]. To identify possible defects basis, it is necessary to determine
exactly the dependence between the structure properties and the MOCVD multiple technological of epitaxial
growth parameters.
Among the factors that determine the crystal structure perfection and the achievement of high homogeneity
composition on the entire wafer surface, the lattice parameters correspondence between the structure layers has
high importance. Therefore, for the epitaxial layers growth with a AxB1-xC given composition with a precise
coincidence of the lattice parameter in the technological growth process, it is necessary to have a dependence of
the solid phase composition versus the third group elements ratio A / (A + B) elements in the vapour phase flow
for specific growth conditions. This work is for the InxGa1-xP and InxA1-xP materials investigation, which are the
main ones not only in the structure for the modern three-stage SC, but also in other perspective photoelectric
devices.

2. EXPERIMENT SETUP

For a detailed investigation epitaxial layers samples were grown: InxGa1-xP (60 pcs) with various indium and
gallium atoms concentration (X = 38-53 %), doped by silicon / tellurium on a germanium substrate; InxAl1-xP (8
pcs) with a different indium and aluminum atoms concentration (X = 46-51 %) on the germanium and gallium
arsenide substrates. All samples were grown by the MOCVD using an Veeco reactor (vertical type).
To obtain an epitaxial layer with specify properties, such as the crystal lattice parameter, the forbidden band, it
is necessary to control the chemical composition of the growing layers, determined by the gas mixture
composition and its distribution in the reactor. In turn, for the applied constructions of alkyl sources (bubblers),
the mole flow rate (the source stream concentration) at the bubbler outlet can be controlled by changing the
hydrogen flow rate, the pressure in the bubbler and the bubbler temperature [3-4]. Increasing the pressure in the
bubbler (Pb) reduces the molar velocity, the increase in temperature (equivalent to an increase in P v), and also
the flow of hydrogen - increases the molar velocity. Such parameters dependence can be described by the
equation (1):
S  Pv
gas
VTMGa,TMIn  , (1)
(Pb  Pv )  V
where S - the substance flow through the bubbler, cm3 / min, Pb - pressure in the bubbler, Pa, Pv - partial pressure
of organometallic compounds, Pa, V - one mole volume of an ideal gas, cm3 / mol.

By the hydrogen fluxes change at the entrance into the TMGa and TMIn bubboders at the determined V / III
value, a necessary spread of the vapour phase composition for the experimental samples growth is obtained.
The technological parameters for each layer used in the calculations, as well as schematic samples structures
scheme with an InxGa1-xP epitaxial layer (Fig. 1a) and with the InxAl1-xP layer (Fig. 1b), were obtained using
RCP Analysis prescriptions software for the analysis and comparison.

a) b)
Figure 1. Sample structures scheme: а – with InxGa1-хP epitaxial layer, b - with InxAl1-хP epitaxial layer

339
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

To control the epitaxial growth parameters, the internal and the external monitoring systems are used [3-5]. One
of the external system elements is X-ray diffractometry (XRD), implemented on a two-crystal Vector
diffractometer. It is used to determine the lattice parameter of individual layers by processing the rocking curves.
The maximum contribution to the intensity is given by the thickest layer of the structure, in our case the Ge
substrate. The broadening and some other separately standing peaks shows layers presence that are mismatched
by lattice parameters. When a peak with a lower intensity is located to the left of the main one, the lattice
parameter of the layer increases according to the substrate, with a location to the right of the base - decreases.
To calculate the In / (In + Ga) ratio of the InxGa1-xP epitaxial layer in the solid phase, it is necessary to measure
the distance between peaks equal to d2θ. Using the relationship between the lattice period and the interplanar
distance for a cubic cell described by expression (2), as well as by the Wolf-Bragg formula, it can be determined
the lattice parameter. Then according to Vegard law (3), it is possible to calculate the epitaxial layer chemical
composition:
d2
a 
2
 ( H 2  K 2  L2 ), (2)
n

where а – lattice constant, nm, d – interplaner distance between reflecting planes, nm, n – diffraction reflection
coefficient, H, K, L – interference coefficient.
a InxGa1 x P  x  a InP  (1  x)  aGaP , (3)
where X – In content in solid solution.

3. RESULT AND DISCUSSION

The obtained experimental data corresponds with liner dependence according to the Vegard law, but there are
some deviations near the Ge lattice parameter. Since in this case it is difficult to separate the substrate diffraction
peaks and the epitaxial layer, an error arises in determining the distance between them. For samples with a strong
deviation of the layer lattice parameter, the rocking curves are characterized by a large peak broadening value -
about 800 arc seconds (Fig. 2). For a classical rocking curve of a defect-free structure, the distance between the
peaks is of the order about 200 - 400 angular units and the broadening is less than 200 (Fig. 2b).

a) b)

c)
Figure 2. Dependence of the InxGa1-хP epitaxial layers lattice parameter: a, b - rocking curves for different In / (In
+ Ga) ratios in the solid phase (a - sharp peak, b - an two spaced peaks overlap), c - the dependence of the epitaxial
layers InxGa1-хP lattice parameter vs the In / (In + Ga) ratio in the vapour phase

340
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The InxGa1-xP lattice parameter on the In / (In + Ga) ratio in the vapour phase dependence (Fig. 2c) based on the
XRD results, data processing of the samples rocking curves, and also data of the solid solution composition from
the vapour phase. As it is known, the InxGa1-xP lattice parameter increases with increasing indium percentage.
The lattice parameter mismatch is 0.1 % - 0.26 %. In the In compositions range 45 - 53 %, the single-crystal
structure is characterized by high perfection. At an In and Ga ratio - 1:1, the epitaxial layer is characterized by
the best properties with the closest lattice parameter value to the Ge lattice parameter. The deviation from the
composition for these samples is in the range − 0.5 - 3% [5-8].
Investigations to determine the solid solution composition homogeneity over the entire surface of the sample
were carried out. Due to the reactor central symmetry, the individual oscillation curves are measured along the
sample radius (coinciding with the growth chamber radius). The results show that the lattice parameters
mismatch does not exceed 0.1 % (the values scatter does not exceed 6 %), which makes it reasonable to perform
measurements at the central point of the sample area. In a similar way, samples with In xAl1-xP epitaxial layers
were investigated. To evaluate the substrate influence on the InxAl1-xP epitaxial layers lattice parameter, samples
on a germanium and gallium arsenide substrate were investigated. For values X = 46 - 53 %, there is a good
agreement between structures lattice parameters and a linear dependence according to the Vegard law (Fig. 3).

Figure 3. Dependence of the InxAl1-хP epitaxial layers lattice parameter versus the In / (Al + In) ratio in the vapour phase:
squares - InAlP / GaAs structure samples; circle - InAlP / InGaP / Ge structure samples; triangle - structure InAlP / GaAs
/ Ge samples; the dotted line - calculated according to Vegard law on the basis of table data (GaAs substrate); the straight
line - by Vegard law with correction (Ge substrate); dash dotted line - table value of the Ge lattice parameter; dotted line
- the table the GaAs lattice parameter value

For InxAl1-xP / ... / Ge a decrease by 0.005 Å of the linear dependence according to values calculated by Vegard
law, which is valid for InxAl1-xP / GaAs, was detected. To determine the AxB1-xC solid solution composition it
can be used the method for photoluminescence (PL) measuring (spectra vs the width dependence of the material
forbidden band specified by the solid solution composition). The obtained spectral maps (Fig. 4a) show the PL
distribution of integrated intensity, the wavelength peak, the intensity peak and the width value at a half the
spectrum height on the epitaxial structure (Fig. 4b) [2-3].

a) b)
Figure 4. Spectral map and single PL spectrum for one sample: a - the spectral map, b - the unit peak of the PL spectrum

341
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Multistage processing, data maps visualization in comparison with the reactor growth chamber elements location
(Fig. 5a) allows us to give an properties uniformity estimation and to reconstruct the conditions for epitaxial
growth (Fig. 5b).

a) b)
Figure 5. Stages of processing PL maps: a - combined maps visualization of several plates for a single technological
process in comparison with the arrangement of elements the reactor growth chamber, b - comparison of the samples
linear profiles of several processes (the square points - the profile for the case with two spaced peaks, triangular -
for the case with one sharp peak and one wide peak, phomboid-shaped - for the case of two closely spaced peaks)

The forbidden band value in the processing of PL spectra was calculated by the expression (4):
hc
Eg  , (4)

where Еg – photon energy, eV, λ – photon wavelength (PL characterization), nm, h – Planck constant, eV/s,
с –velocity of light, m/s.

The epitaxial layer composition (for Ga xInx-1P at X <0.63) was determined by (5):
E g  1.34  0.69  x  0.48  x 2 (5)

The results obtained show a uniform distribution of the PL spectra peaks over the sample entire surface. The PL
spectra maxima deviation is 2.8 %.
This method is easier to use, the equipment makes it possible to obtain maps for the analysis of the properties
uniformity, however, when comparing the In / (In + Ga) ratio in the solid phase, measured by two methods for
samples doped with Si and Te (Fig. 6), it was detected that the composition data obtained by the PL measurement
differ from the analogous values determined by the XRD method. Due to the dopant influence on the PL
spectrum, these data should be considered as estimative.

Figure 6. In / (In + Ga) ratio dependence of the InxGa1-хP epitaxial layer in the solid phase, obtained by the XRD
and PL on the ratio in the vapour phase methods: dots - the results obtained by the XRD method; triangular points
- the results obtained with the PL method, the straight line - constructed on the table data basis according to the
Vegard law.

4. EXPERIMENTAL FINDINGS

The following important findings were obtained during the work:

342
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

- it was found that for the InxGa1-xP solid solution in the range 45 - 53 %, for InxAl1-xP in the range 46 - 51
% there is a single-crystal structure high perfection, diffraction X-ray peaks insignificant broadening (less than
200 angular seconds);
- it was shown that, in addition to the mismatch of the diffraction maxima, the diffraction peak broadening
of the structure can be used as a criterion for evaluating the grown structure quality.
It was detected that the composition of the solid solution determined by the method of the photoluminescence
effect measuring should be regarded as an estimative.
The results obtained make it possible to proceed to the next stage of work on the determination of optimal
technological parameters for the perfect InxAl1-xP and InxGa1-xP epitaxial layers growth in the cascade high-
efficiency SC for space applications, namely the required electrical characteristics achievement by precision
doping.

REFERENCES

[1] Breiland, W, Coltrin, M, Creighton, J, Hou, H, Moffat, H, Tsao, Y. Organometallic vapor phase epitaxy. Materials
Science and Engineering: R: Reports 1999; 24: 241-274 <https://doi.org/10.1016/S0927-796X(98)00018-7>
[2] Lebedev, A, Naumova, A, Smirnov, A. MOCVD heterostructure growth growth. Electronic and electromechanical
systems and devices 2018; 2:12-46 <https://doi.org/10.1015/S0927-796X(2018)00011-7>
[3] Fedorchenko, I, Ril, A, Marenkin, S, Rabinovich, O, Legotin, S, Didenko S, Skupiñski, P, Kilanski L, Dobrowolski, W.
Phase diagram of the ZnSiAs2–MnAs system. Journal of Crystal Growth 2017; 468C: 683-687
<https://doi.org/10.1016/j.jcrysgro.2016.10.029>
[4] Rabinovich, O. AIIIBV heterostructure simulation and investigation. Journal of Alloys and Compounds 2014; 586: S258-
S261 <https://doi.org/10.1016/j.jallcom.2013.03.214>
[5] Fedorchenko I, Kushkov, A, Gaev D, Rabinovich, O, Marenkin, S, Didenko, S, Legotin, S, Orlova, M, Krasnov, A.
Growth method for AIIIBV and AIVBVI heterostructures Journal of crystal growth 2018: 483: 245–250
<https://doi.org/10.1016/j.jcrysgro.2017.12.013>
[6] Rabinovich, O, Savchuk, A, Didenko, S, Orlova, M, Marenkin, S, Ril, A, Podgornaya, S. AlGaN optimization for
photodetectors. Optical and Quantum Electronics 2019; 51: 681-690 <https://doi.org/10.1007/s11082-019-1791-2>
[7] Orlova, M, Abdullaev, O, Mezhenny, M, Chelny, A, Savchuk, A, Rabinovich, O, Didenko, S, Osipov, Yu, Kourova, N,
Akhmerov, Yu, Marenkin, S. Influence of Growth Parameters on a-Plane InGaN/GaN Heterostructures on r-Sapphire.
Phys. Status Solidi B 2019: 10: 18003711- 180037117 <https://doi.org/10.1002/pssb.201800371>
[8] Abdullaev, O, Mezhenny, M, Chelny, A, Savchuk, A, Ahmerov, Yu, Rabinovich, O, Didenko, S, Osipov, Yu, Orlova,
M, Aluyev, A. MOCVD growth GaN on sapphire. IOP Conf. Series: Materials Science and Engineering 2019; 617:
0120151-0120156 <https://doi.org/10.1088/1742-6596/1410/1/012022>

343
 BOOSTING THE ELECTRIC OUTPUT OF A CANTILEVER
PIEZOELECTRIC HARVESTER BY TIP CURVATURE
BLOCKING ELEMENTS

Claudia Irina Borzea

Romanian Research and Development Institute for Gas Turbines COMOTI, Bucharest, Romania,
[email protected], ORCID: 0000-0001-9578-1222

Constantin Daniel Comeagă

University Politehnica of Bucharest, Bucharest, Romania, [email protected], ORCID: 0000-0001-6175-
4832

Adrian Săvescu
Romanian Research and Development Institute for Gas Turbines COMOTI, Bucharest, Romania,
[email protected], ORCID: 0000-0003-0559-1172

Borzea, C I, Comeagă, C D, Săvescu, A. Boosting the Electric Output of a Cantilever

Cite this paper as: Piezoelectric Harvester by Tip Curvature Blocking Elements. 8th Eur. Conf. Ren. Energy Sys.
24-25 August 2020, Istanbul, Turkey

Abstract: The paper studies the influence of tip curvature on the electric response of a cantilevered
piezoelectric harvester, with four active PZT-5H wafers. When structures with large enough width
to height ratios are subjected to vibrations, a slight curving of the free tip occurs on a perpendicular
plane to the charge generation one. Dynamic finite element analyses were conducted for the
piezoelectric transducer, showing a voltage increase when the tip curvature on yOz plane is
blocked between two bars, leaving free only the deflection on xOz plane. Damping loss factor was
calculated analytically and introduced in COMSOL Multiphysics. The quality factor is also
determined experimentally. The comparison was realised between cantilever tip constrained with
bars and with an equivalent concentrated proof mass, respectively. Even if the elements used in
the two cases have the same mass of 4.0 grams, an output voltage of 8.2 V/g was obtained by
blocking tip curvature. Due to inducing higher stress into the structure, voltage increases with 7.8%
comparing with concentrated tip mass case rendering 7.6 V/g. The simulations are validated by
experimental tests. Frequency response was defined between base vibration (electromagnetic
shaker table and monitoring accelerometer) and electrical output voltage, measured with a signal
analyser.
Keywords: Energy Harvesting, Piezoelectric transducer, Vibrations, Finite Element Method
© 2020 Published by ECRES

Nomenclature
CAD Computer Aided Design
FEM Finite Element Method
FFT Fast Fourier Transform
FR4 Flame Retardant Type 4 glass-reinforced epoxy
PZT-5H Soft Lead Zirconate Titanate
RMS Root Mean Square
E Young’s modulus (equivalent one for composite structures) [N/m2]
fn Natural frequency [Hz]
frez Resonant frequency [Hz]
g Gravitational acceleration [m/s2]
I Moment of inertia of beam’s cross-section [kg ∙ m2]
k Stiffness constant [N/m] or [kg/s2]
K Cross-sectional stiffness [kg ∙ m3/s2]
L Beam length [m]
m Beam total mass [kg]
Q Quality factor [ND]
Vpk Peak voltage [V/g]
ζ Damping ratio [ND]
η Damping loss factor [ND]

344
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION

In the past few years, using energy harvesting as power supply for small-scale electronics has attracted large
research interest. Piezoelectric devices present potential due to their highly efficient energy conversion from
mechanical vibrations, and to the challenge of finding ways to boost the output power, which so far has been
quite feeble. Piezoelectric harvesters use the direct piezoelectric effect for the conversion of mechanical
vibrations into electrical energy, and exploit this energy by connecting the electrodes with an electric circuit.
Piezoelectric energy harvesting [1] has emerged at the beginning of this century, and is being researched since
[2, 3], due to piezoelectric materials’ efficient mechanic-to-electric conversion of kinetic energy. Advances in
materials science make possible the continuous improvement of the energy conversion devices [4, 5].
Nevertheless, most researches focus on new ways of boosting the electric output without modifying the materials
[6, 7, 8]. The purpose is to harness enough energy to power low-power electronics such as wireless sensors,
rendering them autonomous.
Piezoelectric harvesters use the direct piezoelectric effect. When subjected to pressure (or mechanical stress),
they produce electric charge, rendering an alternating current response when vibrating. They generate maximum
voltage output at resonance frequency. Therefore, for resonating structures, the natural frequency of the energy
harvester has to match the fundamental frequency of the vibrations source.
A vibrating piezoelectric device generates alternating current during excitation, which needs to be converted
into direct current. Due to the high impedance and capacitance, piezoelectric devices generate low output power
across the terminals. A full-bridge rectifier converts alternating current into pulsating direct current, but also
suffers from power loss due to the forward voltage drop across the diodes, so the diodes must be carefully chosen
with minimum voltage drop.
Previous researches [9, 10] have shown that for piezoelectric cantilevers with higher width to height ratios, a
bending of the free tip occurs in yOz plane, besides the bending that generates charge in xOz plane (Figure 1).

Figure 1. Vibrations direction for charge generation and curving tendency of the free tip

The present work is conducted on a physical piezoelectric harvester with a considerable width to height ratio,
aiming to prove both through numerical simulations and experimentally that the discovered effect of
constraining the free tip curvature in transversal plane has the desired effect of boosting the electric output. No
literature references have been found about this effect, hence the novelty contribution of our research.

2. THEORY AND PIEZOELECTRIC HARVESTER PROPOSED

Midé PPA-4011 harvester [11] analysed is a quadmorph cantilever beam, comprising four PZT-5H piezoceramic
wafers. The beam consists of 17 very thin layers, its overall dimensions being 71.0 mm x 25.4 mm x 1.32 mm
(L x W x H), as one can see in the CAD model in Figure 2a. Thus, the width to height ratio is 19.24, enough to
exhibit differences in output voltage due to the curvature in vertical transversal plane.
The 0.15 mm thick piezoelectric layers (Figure 2b) are placed between copper electrodes (0.03 mm). Each of
the four such sandwiches is insulated and protected by separating FR4 dielectric layers (0.08 mm). The PZT-5H
wafers are shorter in length and width than the other layers (46 mm x 20.8 mm). Very thin epoxy resin layers
(<0.02 mm) are used to affix all the wafers together. According to manufacturer specifications [11], the epoxy

345
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

layers can be neglected for simulation purposes. The total mass of the piezoelectric harvesting beam is 7.6 grams.
The beams are clamped on a test support as shown in Figure 2c.

Figure 2. Piezoelectric beam: (a) isometric view; (b) tip corner magnification; (c) Support mounting [10]

The piezoelectric harvester operates in d31 mode, meaning that when stress is applied on z axis, charge is induced
on x direction. Resonant frequencies induce large vibration amplitudes in the structure even if the stress inputs
are low. That is why, usually, resonances ought to be avoided. However, when speaking about energy harvesters,
resonant structures like piezoelectric cantilevers need to be excited very close to their resonant frequency for
giving maximum electric output due to high tip amplitudes induced.
Since natural frequency depends on the structure’s stiffness and mass, for cantilevers of rectangular cross-
section, constrained at one end, we have the following relation:

3𝐸𝐼
1 𝑘 1 √ 𝐿3 (1)
𝑓𝑛 = √ =
2𝜋 𝑚 2𝜋 𝑚

In a previous research work [12] the cross-sectional stiffness of K = 0.223 kg∙m3/s2 was calculated for the
discussed piezoelectric harvester. The equivalent density of the multilayer beam was found to be ρeqv = 4469
kg/m3, the efficient length subjected to vibrations in the rear clamping position on the support is Leff = 54 mm,
and A is the beam’s cross-section area. The analytical fundamental frequency was calculated, with the formula
in [13], to be:

1 𝐾
𝑓𝑛 = ∙ 1.8752 √ = 241.68 𝐻𝑧 (2)
2𝜋 𝜌𝑒𝑞𝑣 ∙ 𝐴 ∙ 𝐿𝑒𝑓𝑓 4

3. FEM SIMULATION

The FEM simulations were conducted in COMSOL Multiphysics 5.3a, using Structural Mechanics –
Piezoelectric Devices. The behaviour at microscale is evaluated interconnecting two physics, namely:
 Solid Mechanics: for materials assignment and setting initial and boundary conditions. An acceleration of 1g
(9.81 m/s2) is applied at the base of the inferior clamping bar, and the PZT-5H layers are declared
piezoelectric.
 Electrostatics: Electric charge conservation condition is set for the piezoelectric layers. Terminal and ground
are set on the faces of the PZT-5H wafers. The terminals are declared with zero charge, assuming infinite
impedance.
The physics are coupled through Multiphysics – Piezoelectric effect, allowing simultaneous numerical solving
of all equations from solid mechanics, electrostatics and the constitutive equations of the piezoelectric material.
An Eigenfrequency study was conducted first in order to determine the interval comprising the resonant
frequency for Frequency Domain study. Eigenfrequency is a static study only showing the structure’s natural
frequency, whereas Frequency Domain is dynamic, being influenced by excitation loads and damping
coefficients. That is why through Eigenfrequency study, the electric potential across piezoelectric terminals
cannot be correctly assessed, as it disregards the base acceleration magnitude declared.
Damping has a prominent effect when a structure is subjected to a harmonic excitation at a frequency close to
its natural frequency. For undamped systems oscillating at resonant frequency, the vibration amplitude tends to
infinite. Practically, such systems do not exist, and the actual amplitude at resonance is controlled solely by the

346
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

amount of damping [14]. Typically, for harvesters operating in air, the external damping excitation is negligible
when compared to the inertial excitation term [3].
In certain systems, such as resonators, and for energy harvesting applications, the aim is to get as much
amplification as possible. An alternative damping measure is the quality factor or Q factor, defined as the
amplification at resonance [14]. Relying on the PZT-5H quality factor Q = 30 given in product specifications
[11], we calculate the damping ratio using the following relation:

1 1 1
𝑄= ≈ ⇒𝜁= ⇒ 𝜁 ≈ 0.017 (3)
2𝜁√1 − 𝜁 2 2𝜁 2𝑄

Usually, the conversion between the damping ratio and loss factor damping is considered at a resonant
frequency, For first resonance, η ≈ 2ζ . The predictions at the second resonance differ significantly, but our
interest falls on the fundamental frequency. Therefore, we introduced an isotropic loss factor of ηs = 0.034 as
mechanical damping declared for the piezoelectric material in COMSOL Multiphysics (Figure 3), in order to
assess correctly the behaviour in Frequency Domain study.

Figure 3. COMSOL Multiphysics interface and simulation of piezoelectric cantilever with constrained tip

All cases with concentrated mass and constraining tip bars are simulated with the same initial and boundary
conditions, with the same mesh with tetrahedral elements of default normal size.

Table 1. Simulation results in COMSOL Multiphysics

Maximum electric
No. Case Eigenfrequency [Hz] Frequency Domain [Hz]
potential at 1g [V]
1 Free tip 283.94 [283, 0.1, 287.5] 6.886
2 Concentrated mass 146.03 [145.5, 0.1, 148] 14.401
3 Constrained tip 146.64 [145.8, 0.1, 148.3] 14.895

The frequency increases slightly when constraining the tip, due to adding thickness to the beam (see Equation
(1)), therefore increasing the moment of inertia of the cross-section. The electric potential is expected to increase
more in real experimental conditions. A more noticeable frequency increase is also expected. The electric
potential increases comparing to concentrated mass, due to a higher strain induced in the piezoelectric material
when blocking its curvature on transversal plane.

4. EXPERIMENTAL TESTS

For validating the FEM simulations, experimental tests have been conducted on the cantilevered piezoelectric
harvester. The experimental setup is presented the Figure 4a. Three cases have been analysed, namely: no load

347
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(free tip – Figure 1c), with concentrated proof mass of 4.0 grams (Figure 4b) and with elements constraining tip
curvature (Figure 4c), also weighing 4.0 grams, for the relevance of the experiment.

Figure 4. (a) Experimental setup; (b) Concentrated mass configuration; (c) Tip curvature constrained

The experimental setup consists of:

 Dynamic FFT spectrum analyser with embedded functions generator;
 Power amplifier with sine output power (120 VA RMS value), high signal-to-noise ratio, and signal
amplification in direct current (up to 20 kHz);
 Electrodynamic shaker driven by the signal analyser via power amplifier and connected to the output
channel of the analyser;
 Piezoelectric energy harvester clamped on its support, attached with strong double-sided adhesive tape
on the shaker table, with terminal output connected to input channel 2 of the analyser;
 Accelerometer placed on the shaker table, used as reference input signal for the vibrations amplitude,
and connected to input channel 1 of the analyser.
The data acquired from the signal analyser was saved and represented on the graph in Figure 5, for the three
cases.

10 Experimental voltage response graph [V/g]

Voltage response/acceleration

6
[V/g]

0
50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250
Frequency [Hz]Concentrated mass
Tip constraining mass

Figure 5. Voltage response graph based on experimental data

The peak voltage recorded in free tip case was 3.28 V/g, at the resonant frequency of 214.03 Hz. When adding
the concentrated mass elements, due to increasing tip amplitude, the voltage output increased to 7.611 V/g.
The resonant frequency consequently decreased to 108.38 Hz. Constraining the tip, with blocking elements of
the same mass as the concentrated one, induces higher mechanical von Mises stresses into the piezoelectric
material of the cantilever harvester, not letting the beam bend its corners up and down in vertical plane. Hence,
as expected, a voltage increase is recorded, up to 8.205 V/g. The resonant frequency increases slightly up to
110.58 Hz (with 2.2 Hz), when constraining the tip, because the two bars are jointed with the beam tip, thus
increasing the tip cross-section and its moment of inertia (Equation (1)).

348
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In order to validate the simulation, where the theoretical Quality Factor Q = 30 was used, we calculated the
experimental Q factor. Considering the peak voltage Vpk at resonant frequency frez, we find the left and right
frequencies (fleft and fright) where the electric potential is 0.707 of the peak voltage (or on logarithmic scale 3 dB).
The Quality factor is calculated based on the frequency response acquired data from the signal analyser, using
the following relation:

𝑓𝑟𝑒𝑧
𝑄= (4)
𝑓𝑟𝑖𝑔ℎ𝑡 − 𝑓𝑙𝑒𝑓𝑡

Table 2. Calculated damping based on experimental acquired data

No Case frez [Hz] Vpk Vpk ∙ 0.707 Frequencies Q Dampin Damping
. [V/g] [V/g] for
fleft = 210 factor g ratio ζ loss factor
1 Free tip 214.03 3.280 5.373 V pk ∙ 0.707 26.75 0.0187 0.037
η
fright = 218
[Hz]
Concentrated fleft = 107
2 108.38 7.611 28.964 27.09 0.0185 0.037
mass fright = 111
fleft = 109
3 Constrained tip 110.58 8.205 33.661 27.75 0.0180 0.036
fright = 113

The output voltage was measured considering 1 MΩ + 50 pF input impedance of the signal analyser. Since the
resonant frequency is low, the influence of the capacitance is negligible. The power output plotted in Figure 6
was calculated knowing the voltage output V and the electrical resistance R of the analyser (𝑃𝑅𝑀𝑆 = 𝑉𝑝𝑘 2 /2𝑅).

40 Experimental power output graph [μW]

30
Power output [μW]

20

10

0
50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250
Frequency [Hz]
Tip constraining mass Concentrated mass No load

Figure 6. Power plot based on experimental data

The peak power outputs recorded at resonance for the three cases are highlighted in Table 3 below.

Table 3. Peak power calculated based on experimental voltage

No. Case Peak voltage [V/g] Peak power [μW]
1 Free tip 3.28 5.373
2 Concentrated mass 7.611 28.964
3 Constrained tip 8.205 33.661

5. CONCLUSIONS

The paper proves that constraining the tip curvature on vertical plane, transversal on the charge generation one,
improves the electric output of piezoelectric cantilever harvesters with high enough width to height ratio. The
Frequency Domain simulations in COMSOL Multiphysics proved reliable, since the experimental tests are in
dynamic conditions, and an Eigenfrequency study does not take into consideration the vibrations excitation
loads. The experimental tests showed even more obvious differences than FEM simulations results, between
constraining the tip curvature between blocking elements and cantilever with concentrated proof mass.
A tip mass increases vibrations amplitude and hence the output voltage. Thus, for the results to be comparable
and relevant, the electric output assessment had to be made between the case of cantilever with constrained tip
(with blocking elements) and cantilever with an equivalent concentrated proof mass (4.0 grams).

349
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The piezoelectric transducer, in no load conditions, renders a maximum output voltage of 3.28 V/g at the
resonant frequency of 214 Hz. With the concentrated tip mass, the electric potential increases with 132%, up to
7.611 V/g. at 108.38 Hz. By constraining the tip curvature in vertical transversal plane, 8.205 V/g were generated
at 110.58 Hz.
Thus, the peak voltage at resonance increased with 7.8% comparing to additional concentrated tip mass of the
same value, with a total increase from the free tip beam with 150%. An almost 8% electric potential boost might
not be remarkable for other systems, but for a piezoelectric crystal, this is quite a significant improvement
without bringing any material or structural modifications.

ACKNOWLEDGMENT
We would like to thank Prof. Dr. Engr. Alexandru Morega from Electrical Engineering Faculty within
Politehnica University of Bucharest for his valuable support and for providing access to COMSOL Multiphysics
server.

REFERENCES

[1] Bizon, N, Mahdavi Tabatabaei, N, Blaabjerg, F, Kurt, E. Energy Harvesting and Energy Efficiency. Springer
International Publishing, 2017.
[2] Sodano, H, Inman, D, Park, G. A Review of Power Harvesting from Vibration Using Piezoelectric Materials. The Shock
and Vibration Digest 2004, 36(3), pp. 197-205. DOI: https://doi.org/10.1177/0583102404043275.
[3] Erturk, A, Inman, D. Piezoelectric energy harvesting. New York, USA: John Wiley & Sons, 2011.
[4] Jaitanong, N, Zeng, H, Li, G, Yin, Q, Chaipanich, A. Effect of Nickel Doping on Dielectric, Piezoelectric Properties
and Domain Configurations of PZT Based Ceramics. Materials Science Forum 2017, 883, pp. 27-31. DOI:
https://doi.org/10.4028/www.scientific.net/msf.883.27.
[5] Yu, G, Shao, Y, Liu, J, Huo, R, Li, Y, Duan, Z. Influences of Preparation Process on the Orientation and Properties of
PZT Piezoelectric Thick Film Generation Materials. Advanced Materials Research 2014, 924, pp. 29-35. DOI:
https://doi.org/10.4028/www.scientific.net/amr.924.29.
[6] Rezaeisaray, M, Raboud, D, Moussa, W. Optimizing Energy Output of Piezoelectric Microcantilevers. ASME 2011 9th
International Conference on Nanochannels, Microchannels, and Minichannels, Volume 1 2011 DOI:
https://doi.org/10.1115/icnmm2011-58053.
[7] Usharani, R, Uma, G, Umapathy, M, Choi, S. Design of High Output Broadband Piezoelectric Energy Harvester.
Journal of Mechanical Science and Technology 2017, 31(7), pp. 3131-3142. DOI: https://doi.org/10.1007/s12206-017-
0603-5.
[8] Güleç, H, Gurbuz, M, Toktas, A, Gul, M, Koc, B, Dogan, A. A New Design to Improve Bandwidth of Piezoelectric
Energy Harvester. Journal of the Australian Ceramic Society 2019, 56(1), pp. 117-126. DOI:
https://doi.org/10.1007/s41779-019-00425-4.
[9] Borzea, C I, Comeagă, C D, Uddin, M N, Hrițcu, R D, Ringheanu, V L. Improving the Electric Response of a Cantilever
Piezoelectric Energy Harvester by Constraining Tip Curvature. In The 9th International Conference on Advanced
Concepts in Mechanical Engineering (ACME 2020). 4-5 June 2020, Iași.
[10] Borzea, C, Comeagă, D. Analysis and Optimization of a Piezoelectric Energy Harvester. E3S Web of Conferences
2019, 112, p. 04001. DOI: https://doi.org/10.1051/e3sconf/201911204001.
[11] Midé Technology. PPA PRODUCTS Datasheet & User Manual, pdf, 2017.
[12] Borzea, C, Comeagă, D, Stoicescu, A, Nechifor, C. Piezoelectric Harvester Performance Analysis for Vibrations
Harnessing. U.P.B. Scientific Bulletin, Series C Electrical Engineering and Computer Science 2019, Vol. 81, Iss. 3, pp.
237-248, ISSN 2286-3540.
[13] Vidyapeetham, A V. Free Vibration of a Cantilever Beam.
http://vlab.amrita.edu/?sub=3&brch=175&sim=1080&cnt=1.
[14] Sönnerlind, H. COMSOL Blog. Damping in Structural Dynamics: Theory and Sources. COMSOL Multiphysics, 2019,
https://www.comsol.com/blogs/damping-in-structural-dynamics-theory-and-sources/.

350
PERFORMANCE OF THE FIRE-FIGHTING SYSTEM: ALGERIAN
CASE STUDY
Marref Souad
Laboratory of Research in Industrial Prevention (LRPI), Institute of Health and Industrial Safety, University of Batna 2,
Algeria, [email protected], ORCID: 0000-0001-8872-1539

Chettouh Samia
Laboratory of Research in Industrial Prevention (LRPI), Institute of Health and Industrial Safety, University of Batna
2, Algeria, [email protected], ORCID: 0000-0003-4202-5073

Marref Souad, Chettouh Samia. Performance of the fireproof system: Algerian case Study. 8th
Cite this paper as:
Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract : This paper reflects an investigation about the use of fire passive protection based on the concept
of fireproofing. The use of this technique obeys to international standards especially the 2218 API
used as reference in order to achieve the purposes of this approach. Indeed, the present study attests
to the methodology of the fireproofing used in the complex of "Ourhoud" and to the identified
discrepancies in some of its installations and this in comparison to the recommendations of the
standard and the requirements of the safety philosophy of this project.

Keywords: Risk management, fire, fireproofing, petroleum industry, Algerian industry, 2218 API.

© 2020 Published by ECRES

Nomenclature
Rm 1 breaking constraint
t time [min]
Re elastic real limit
temperature at the instant t [° C]
T Ambient temperature at the instant t = 0 min [° C]
Re 0, 2 plastic elongation of 0.2%
1. INTRODUCTIONT0
A elongation in% of the specimen
NIST National Institute
The Industrial and technological of Standards
development and Technology
continues to generate several types of potential risks from
different origins, such as pressure risk, electrical risk, fire and explosion, etc. The oil and gas industry, like many
others, is subject WTC
to theseWorld Trade Centre
risks, sometimes even catastrophic when they are combined, mainly Fire and Explosion
as regards their combination, including fire and explosion [1]. The presence of products used in such units,
which are highly flammable, combined with combustion can cause explosions and fires of colossal size [1, 2].
These risks generally lead to accidental fires in Oil and Gas facilities, which can have a significant potential for
severe consequences, endangering personnel safety, environment, asset integrity, production continuity and
company reputation [3]. Fire may also cause damages to structures, which may trigger escalation resulting in
severe secondar scenarios [4]. To avoid these events, particularly dangerous for people and assets, there are
many old techniques that have achieved a certain level of safety. Nevertheless, it is still possible to improve this
safety through the use of new technologies such as the investigation on the use of passive protection against the
fire and which is commonly known as the fireproofing; which is a crucial safety barrier in preventing the
escalation of fire scenarios. Fireproofing improves the capacity of process items and of support structures to
maintain their structural integrity during a fire, preventing or at least delaying the collapse of elements of the
structure [5].

351
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

As a reminder, this technique is very important in the oil and gas industries and its uses are governed by
international standards especially the 2218 API used as a reference to elaborate this work.
The present paper reports the results obtained about quality of the fireproofing products used in this oil and gas
company. The work is carried out with an HSE officer to identify discrepancies in some installations and
compare the recommendations with the standard and the requirements of the safety philosophy of this project.

2. TECHNIQUE FOR PROTECTING THE FACILITIES BY THE METHOD OF FIREPROOFING

The Function of the fireproofing

The Equipment design, the location and the spacing are very important, they play a considerable role to limit
the effects of a probable fire that could occur accidentally. another way involved in limiting the damage caused
by the fire, is the application of a passive protection (PFP protection by Fire proof) called fireproofing, as
opposed to the active protection provided by mobile devices such as monitors and fire fighting networks. This
coverage protects operating systems when exposed to fire.
Fireproofing is a passive fire protection based on the application of a protective coating that delays the
temperature raise of structural elements exposed to fire [4, 6, 7]. The primary role of the fireproofing occurs at
the beginning of the incident to ensure the thermal insulation of the equipment, time to allow the shutdown of
the unit, the insulation of the flow fuel on fire, the switching on of the fire-fighting equipment. The installations
to fight against the fire such as fixed water spray systems, the monitors, the fire-fighting trucks or the mobile
hoses also provide an active protection but may, if the fire is very severe, not be able to fight it.
In the absence of the fireproofing, and at the beginning of a fire, the steel structure supports and under the effect
of heat flux, lose their characteristics and may collapse even before the release of the above-mentioned actions.
The implementation of a fireproofing protection can help significantly to reduce human and material losses and
gives the time needed to the response team to intervene with the adequate firefighting means.
The role of the fireproofing therefore is limited to the following three actions [8]:

 The stability of the characteristics of the steel, particularly its resistance to breakage;
 Flame resistance;
 Thermal insulation.

3. CRITICAL STUDY OF THE MATERIALS TO ISOLATE (PETROLEUM CENTENARIANS “

PIPELINE“)

Figure 1: Diagram of the deformation curve of a steel

Rm: breaking constraint

Re: real elastic limit
Re 0, 2 corresponding to a plastic elongation of 0.2%
A: elongation in% of the specimen

352
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Reaction of steel in front of fire

The reaction of structures exposed to fire, depends on the speed at which these elements are heated and the
temperature distribution into the sections and over the length of the elements. The mechanical properties of the
materials decrease with the increasing temperature and, consequently, the mechanical resistance of the elements
therefore decrease. The collapse occurs when the resistance is equal to the action of the applied loads. This can
happen in a very short time when the temperature increases quickly. In this regard, the unprotected metallic
elements present, in some situations, cause an unfavourable reaction due to the very high conductivity of the
steel. In comparison, a mixed element or protected by insulation system, have a better performance, because the
thermal inertia of the elements is greater and the thermal conductivity of the insulating material is lower [9].

Indeed, for the bare steels, studies have determined that:

 At 500 ° C the resistance of iron is halved;

 At 700 ° C the resistance is reduced by one third;
 At 800 ° C the resistance is zero, the iron bends under its own weight.
Classic tests on tensile test pieces illustrate the reaction of the steel, that means the variation of the stress
according to the temperature changes (figure 1). The curve below (figure 2) shows the loss of the mechanical
characteristics of steel in the presence of fire:

Figure 2: Mechanical properties of a common construction steel according to the temperature on tensile specimen
(extracted curves from «« Techniques de l’Ingénieur » “)

Legend: σe: elastic limit stress

σr: ultimate breaking stress
E: modulus of elasticity
δ: Δl / l: coefficient of linear expansion
The values of the characteristics on the ordinate are indicated in percentage of the corresponding to the ordinary
temperature.

353
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3: The twin towers on fire Figure 4: Steel with fireproof

The process of heat transfer

The use of the fireproofing in the industry and in other fields of activities including construction becomes a
means of protection against the inevitable fire [10]. For example the pictures of the twin towers of the World
Trade Centre (WTC) in New York, struck and collapsed on September 11, 2001 (figure 3) and (figure 4) will
remain etched in everyone's memory. After more than three years of investigation which saw the analysis of
200 metal pieces, the "National Institute of Standards and Technology (NIST)" concluded the following:
The steel columns structure supporting the two towers have been weakened by the failure of fireproof coatings
and therefore if the fire insulation had resisted to the airplane impacts, fires would not have weakened the main
elements of the building which eventually collapsed.

Evaluation Method of the resistance to fire

During a fire, the temperature of the steel increases in the same way as the temperature of hot gases but with a
time lag. This time lag depends on the thermal inertia of the element. If the element is protected (fire resistant),
the time lag is greater (8t+1).
The method for evaluating the fire resistance of a metal once it has been flame retarded, refer to fire exposure
conditions and to the performance criteria set by the ISO 834 standard. The element’ fire exposure is conducted
according to the conventional thermal program represented by the function (figure 5): T = 345 -To log10 (8t +
1) [11].
Where t = time [min]
T = temperature at the instant t [° C]
To = Ambient temperature at the instant t = 0 min [° C]
The fire resistance performance are expressed in degrees directly related to the times during which the element
fulfils the role assigned to it. Three levels of resistance to fire are then defined [12]:
• Stability to fire: Only a criterion of mechanical resistance is required (structural elements in general);
• Flame arrestors: In addition to the fire stability performance, the element must meet the criteria for
sealing against flames and hot or flammable gases.;
• Fire-stop: in addition to the Flame arrestors performance, the element must meet thermal insulation
criteria
The degrees of resistance to fire are 1/4 h - 1/2 h - 3/4 h - 1 h - 1 h 1/2 - 2 h - 3 h - 4 h - 6 h
The classification imposed on the fireproofing systems are based on the performance requirements of the
crossed walls.
The different systems of fireproofing available on the market can certify performances ranging from a
protection of 1/2 h to 3 h.

354
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 5 : Evaluation Curve of the resistance to fire of a metal

Case Study of the complex "Hassi Messaoud"

For our installation (figure 6) the exposure time to fire where determined of two hours as per the hazard study,
the fireproofing product must be able to withstand a fed fire during the time determined by the formula T = 345
-t0 fire log10 (8t + 1).

Figure 6: Overview of oil production facilities: General View of the CPF

Under the conditions in southern Algeria, T0 must be equal to 55 ° C, value adopted by Sonatrach for the
equipment size (figure 7).
t = 2 x 60 minutes = 120 minutes
T = 345 log10 (8X120 + 1)
T = 1049 ° C
This value (1049 ° C) is approximately that of the heat flux required by the API 2218 standard and that should
expose the fireproofing product chosen during performance tests.

Figure 7: Evolution of temperatures (extracted curves from " « Techniques de l’Ingénieur »")

355
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. DETERMINATION OF FIREPROOFING REQUIREMENTS

The identification of needs in fireproofing is expressed for an oil or petrochemical plant based on a risk
assessment which includes the occurrence of fire scenarios from which the needs analysis evolves. An
approach for the selection of the fireproofing system includes the following elements:

Hazard assessment

This assessment is made by processes such as the HAZOP, the What if, the QRA, etc. This step enables the
needs assessment in fireproofing, the identification of the locations and the types of areas at risk of fire. The
factors considered include the quantities, the pressures, temperatures and the chemical composition of the
potential fuel sources [11].

• High Fire Potential Equipment;

• Medium Fire Potential Equipment;
• Low Fire Potential Equipment.

Define the fire development scenario (Fire-Scenario Envelope)

This part can be developed either by feedback learned or by analysing the possible accidents that may occur, it
consists in identifying potential release rates and the size of the fire.

It is necessary to know the location of the equipment that may be affected in the presence of a likely fire, the
nature of the product, the size and the potential scope of the fire. The Fire-Scenario Envelope is the three-
dimensional space in which the equipment has a potential of fire which may release flammable liquids forming
a pool fire which can burn quite longer and with sufficient intensity to cause substantial equipment damage. The
definition of this scenario is based on the nature and the severity of the potential fires within the envelope [13].
The nature of this severity is the basis for determining the extent of the passive fireproofing and therefore
determining the type and degree of fire resistance of the fireproofed materials used. The dimensions of Fire-
Scenario Envelope can be founded on the design rules based on the experiences or the qualitative or quantitative
assessments. For the hydrocarbon liquid fuels, and through the experience feedback, the protected area used
frequently for the Fire-Scenario Envelope is that which extends from 20 to 40 feet (6 to 12 meters) horizontally
and from 20 to 40 feet (6 to 12 meters) vertically from the source of liquid fuel.

Requirements for analysis performance

Classification of probable scenarios, probable duration of the fire, the importance of heat flux, the
vulnerability of the installation or equipment.
Selection of the chosen system

This choice is made on the basis of the conditions provided by the standards relating to the equipment or the
installation, the specific conditions of the equipment supplier and their level of protection and the feedback for
the same type of equipment or installation. API 2218 standard has classified (lesson learned) all facilities
according to their vulnerability and for which the provision of the fireproofing is identified by determining the
areas to be protected.
Fireproofing in process areas (treatment areas) (figure 8) and (figure 9):

356
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 8 : Support Structure fire -potential equipment and no fire-potential equipment, in an area of Scenario of fire.

Figure 9: Pipe rack with large pumps potential fire installed

Audit of existing installations

During the time allocated to us, we have carried out the verification of the fireproofing conformity to some
existing installation that are subject to the requirements of the standard relating to fireproofing. It should be
noted that the equipment in subjected to this study are those mentioned by the insurers and for which they have
indicated that there is a need to review their compliance. The audit focused on the following equipment:
Air coolers
This type of equipment which is designed to contain liquid hydrocarbons and which is protected by a fireproofing
product. It is found that this equipment is protected up to a height of 4.5 meters. In the safety philosophy adopted
by the manufacturer, this height was chosen but doesn’t match with the standard. Due to the highly flammable
liquid product conveyed by the air coolers, and the fact that it can spread in all directions, this protection is not
complete and for that reason, there is should be a horizontal protection from 6 to 12 meters and an application
for protection throughout the vertical support and according to Table 1 of the API 2218 standard (6.1.2.2 and
6.1.3) and Figure 10. The absence of such compliance, in case of fire event, may result of a collapse of the whole
structure supporting these air coolers.

357
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 10: air coolers.

Pipe support rack

The API 2218 standard provides for this kind of facilities located in the zone of a fire influence a protection
according to the case:

The protection is performed on a horizontal range of 6 to 12 meters and vertically to the first level of piping and
specifically 15 meters for racks located around balloons containing LPG (Liquefied Propane Gas) and this
because of its highly flammable nature. If the last level of the pipes racks contains pipes with diameters of more
than 6 meters, it must be fireproofed at this level or until 9 meters high (See figure 11).
The summary of the audit is presented as follows:
It is clear that the racks are fireproofed up to a height of 4.5 meters and that the horizontal beams are not even
protected although they are in an area that has a high potential for fire and that in these cases the standard has
clearly recommended its protection according to the principle retained of the fire scenario. This method of
protection that has been retained by the adopted safety philosophy does not conform with the recommendations
of the API 2218 standard. Indeed, and as it was found, the protection of the pipes racks is provided by the
applicable standard and consequently algerian oil and gas company Sonatrach shall remedy this gap according
to the required standard.

Figure 11: Pipe rack without pumps in a hazardous zone

5. CONCLUSION

The opportunity that was given to us to share with Sonatrach the concern to support the issue of conformity of
the fireproofing was very beneficial for us. Indeed, this study has allowed us to become familiar with such
installations that deal with a very sensitive area: the safety of installations. In auditing, it was confirmed that the
HSE Department of Sonatrach will hold an action of control and a thorough check on all metal structures
supporting equipment containing combustible liquids, the ESDV and the wiring control of the safety valves
located inside the Fire Scenario Envelope. This audit will make it possible to determine the gaps with the API
2218standard and to determine the equipment to be fireproofed.

358
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] Chettouh S, Hamzi R, & Benaroua K.. Examination of fire and related accidents in Skikda Oil Refinery for the period
2002–2013. Journal of loss prevention in the process industries, 41, 186-193, 2016.
[2] Lees F. Lees' Loss prevention in the process industries: Hazard identification, assessment and control. Butterworth-
Heinemann, 2012.
[3] Tugnolia A, Cozzania V, Di Padovab A, Barbaresib T, & Talloneb F. Reducing the consequences of accidental fires in
oil & gas facilities: A risk-based procedure for identification of the fireproofing zones. CHEMICAL
ENGINEERING, 32, 2013
[4] Di Padova, A., Tugnoli, A., Cozzani, V., Barbaresi, T., & Tallone, F. Identification of fireproofing zones in Oil&Gas
facilities by a risk-based procedure. Journal of hazardous materials, 191(1-3), 83-93, 2011.
[5] Tugnoli, A, Cozzani V, Di Padova A, Barbaresi T, & Tallone F. Mitigation of fire damage and escalation by fireproofing:
A risk-based strategy. Reliability Engineering & System Safety, 105, 25-35, 2012.
[6] Center for Chemical Process Safety (CCPS), Guidelines for Fire Protection in Chemical, Petrochemical, and
Hydrocarbon Processing Facilities, New York CCPS/AIChE, 2003.
[7] Lees, F. Lees' Loss prevention in the process industries: Hazard identification, assessment and control. Butterworth-
Heinemann, 2012.
[8] Matar, M. Exploration of new avenues for the fireproofing of polymers. (Doctoral dissertation), 2016.
[9] Benlakehal N, Kada A, Lamri B, & Achour, B. Behavior of metallic structures under fire effect: application of heat
transfer in metallic profiles according to Eurocode, Algeria 3. 1 st National Civil Engineering Seminar on Materials and
Environmental Protection, 2011.
[10] Wooden buildings with fire protection north treat. https://www.nordtreat.com/fr/insights/quels-types-de-batiments-
necessitent-des-produits-en-bois-avec-classe-de-resistance-au-feu-b-s1-d0
[11] Purser D. A, & McAllister J. L. Assessment of hazards to occupants from smoke, toxic gases, and heat. In SFPE
handbook of fire protection engineering (pp. 2308-2428), New York, Springer, NY. 2016.
[12]Resistance to fire of products, construction elements and works. http://cs.pontdecheruy.free.fr/livres/livre3/3220.htm
[13] inflammable and combustibles liquides hazards. https://www.guard-x.com/b/le-danger-des-liquides-inflammables-et-
des-combustibles

359
 THE STUDY AND OPTIMIZATION OF
PRODUCTION/FERMENTATION PROCESSES IN BIOFUEL
PRODUCTION FOR MAXIMIZING ENERGY OUTPUTS

Amardeep Singh
University Of Toronto, Canada, [email protected], ORCID: 0000-0003-0052-0885

Singh, Amardeep. The Study and Optimization Of Production/Fermentation Processes In

Cite this paper as: Biofuel Production For Maximizing Energy Outputs. 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: The production process involved in the creation of biofuels consists of a number of operations and
steps that require a meticulous understanding of the parameters and metrics. The production
techniques again differ depending on the pre-treatment systems, source material, the methods used
for extraction, types of nutrients used, cell cultures employed, time undertaken and temperature.
Coupled with confirmatory methods such as Principal Component Analysis, researchers can help
narrow down the extent or degree to which the parameters affect the final outcome and even
configure inputs that may not play a definitive role in greater outputs.
The project first tackles through some conventional case studies involving biofuel production using
an FIS(Fuzzy Interface System) and provides certain insights into the ways in which fuel yields
can be enhanced depending on the particular cases. For the purpose of analysis, tools such as
MATLAB, Python and WEKA have been employed. Python and WEKA have been used
extensively in building principal component analysis reviews for the purpose of this project while
MATLAB has been used for building the FIS models.
Keywords: Biofuels, Energy, Optimization, Process Modelling
© 2020 Published by ECRES

1. INTRODUCTION

Numerical analysis techniques such as the Taguchi method, Signal to Noise Ratio test and ANOVA have allowed
industries to further narrow down their approaches in shaping production system around appropriate conditions
while also cutting down on costs and resources [1]. Before being processed by such algorithms, the data is are
often standardized or normalized to better help the system learn and train the models. This is done in order to
eliminate system complications that can arise when working with datatypes that differ in units and ranges
between maximum and minimum values.[2]
Energy production processes are highly benefitted by the ability of algorithms to predict and perform necessary
changes that have to be implemented as a result of external variabilities and altering production cycles. Adding
filters such as the Kalman filter [3] can help capture yields for multiple complex outputs dependent on various
inputs.
In the realm of biofuel production, such process design methodologies have been used to better interpret the
conditions that give rise to better yields as well as comparing the economic advantages of certain unit operations
along with various feedstock materials. ANNs are particularly invested for process data where deep learning
strategies are required, such as for design cases with voluminous rows and columns [4]. Due to their potential
in tracking out variations and dealing with process outliers, fuzzy logic-based controllers are often implemented
alongside standard units in industries, often after learning and interpreting the data immensely. These can be
considered to be predictive algorithms that track changes in data by comparing testing to training ratios, rather
than using integrative and derivative functions to study how parameters will change.[5][6]
On-site sensors further provide industries with the data necessary to correlate production parameters with
process yields, helping save time and resources. Deep learning models too have been studied for predicting the
yield of different types of feedstocks, leveraged against economic conditions and constraints.[2]

360
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

These variables are first represented as transfer functions in the software system and then inputted to the PID
framework. The system generates plots of the parameters, allowing users to track deviations and strategically
alter the transfer function coefficients to bring the system to stability.[7]
However, during the production phase, such updates have to be fed instantaneously in the model so that they
can generate possible future values. A common issue that Fuzzy Logic systems help tackle are cases of redundant
inputs generating multiple outputs. As the user generates the rules to capture these redundancies and sets
boundary conditions for each instance, predicting them becomes simpler.[7]

2. PRELIMINARY ANALYSIS AND EXPERIMENTS

It has been noted that design experiments are devised using softwares like Qualitek which narrows down the
possible number of combinations for optimizing the production outputs. This plays an important role in
removing recurring data steps and avoids repeating calculations for the learned model. Eventually as the model
learns the Experiments better, an optimized design standard can be obtained [2]. Research has suggested that
sugar derived from molasses is a cost-effective feedstock source, as a result of its higher sucrose content.
Optimization studies have favored S.cerevisiae as the main industrial microorganism for creating ethanol as a
result of its high sucrose efficiency. In the literature discussed [8], sugarcane bagasse was processed using a
pretreatment step involving sulphuric acid with various solid loadings and acid concentrations. The effects of
all these parameters was were noted along with reaction time to determine the digestibility and the glucose yield.
This is a standard MIMO (multiple input, multiple output) neural network where it was observed that the acid
concentrations were producing insoluble hemicellulose which decreased with an increase in sulphuric acid.
Higher solid loadings had a reasonable impact on the digestibility of the biomass as well as heterogeneous
reactions catalyzed by the cellulase.

(a)
Figure 1. (a) PCA Simulation Results from data obtained from literature[8]

In the above graph for example, the normalized data points for the output yields have been plotted as numbers
to show how the output yields are related. The inputs t P/min(time) and Yield % are pointed in opposite
directions, indicating that they may in fact share an inverse relationship for this particular production
experimentation. The axes represent the ‘scores’ of the principal component analysis which reduces the modular
equations for the normalized inputs into a set of linear coefficients. In the next case example [9], experimental
factorial design and response surface analysis were used to optimize ethanol production. The parameters
A,B,C,D in the PCA chart below indicate the initial pH, substrate concentration, organic nitrogen concentration
and salt content respectively. The results indicate that glycerol fermentation for ethanol production is best suited
using E. coli SS1, especially when considering a batch reactor. It is imperative that more than 90% of the initial
glycerol be converted into ethanol which can be achieved by building predictive models. This particular case
uses anaerobic conditions. Parametric studies help reduce the number of feasible inputs to just 4 which simplifies
the model further. Experimentally it has been observed that as the glycerol concentration increased from 20 g/L
to 45 g/L, ethanol yields doubled.
A fermentation period of 72 hours allows for cells to proliferate to109 CFU/ml. Higher cell densities help
contribute to better yields even when glycerol concentrations are high. The direction of the arrows B and D
indicate that they may have an inverse relationship while A and C are closely related.

361
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a)
Figure 2. (a) PCA Simulation Results from E.coli Strain Yield[9]

In the next example, we take a contrasting look at the parameters for maximizing yields from lignocellulosic
biomass. The complex nature of the steps involved in the conversion of cellulose to its monomeric sugars,
necessitates the requirement for accurate predictive models.
This experiments in another paper[10], discuss the composite design for biomass that undergoes pre-treatment,
saccharification and fermentation. The modifications in terms of functional groups were studied using Fourier
transform infrared (FTIR) spectroscopy. Researchers used the KBr pellet technique where the samples were
dried and pressed into a potassium bromide (KBr) disc.
Fermentation was conducted using S. cerevisiae and as per the PCA chart, we can see that while Temperature
may not have a major effect on yields, solid loading and pH have a partially inverse relationship to the output
(Reducing Sugars). Data does suggest that the yields may be higher than acid pretreated and hydrolyzed
sugarcane tops. The next case example[11] used an RSM and CCFD matrix for modelling four variables-
incubation period, initial pH, incubation temperature, and molasses concentration. The experiment again used
locally isolated Saccharomyces cerevisiae Y-39 with maximum yields predicted to be 253 and 255 g/L,
respectively. The design also helped narrow down optimal conditions for Experiments to be centered at 71 h,
pH 5.6, 38°C, molasses concentration 18% wt.% for 100 rpm. The Experiments used additives such as
ammonium sulphate, magnesium sulphate and monopotassium phosphate. pH and temperature were adjusted to
allow for maximal growth of the cultures. Cultures for the medium were created in the medium for 48 h at 30
degree C in a shaking incubator at 150 rpm. Cells were then resuspended in another solution following cleaning
with saline water. Liquid chromatography was used to determine the type of molasses present. Experimental
runs were carried out to a full 24 hour factorial design with the four variables. The number of runs was
interestingly determined by the number of independent variables. Under RSM methodologies, the variables were
designated levels (ranging from -1 to 1). The PCA plot for the variables obtained from Python show that initial
pH and incubation temperature have an inverse influence on the yield. Similar relationships can be seen for the
molasses concentration and the incubation period. To further understand the use of variable designs in industries,
we take a look at the data obtained [12] which shows the variable effects on non-catalytic biodiesel production
from supercritical methanol (SCM) and supercritical ethanol (SCE). A spiral reactor was used for this
experimentation which was formatted using an artificial neural network (ANN) model in order to predict
biodiesel yield. With a low error rate and mean square error, the model can become vulnerable to overfitting and
fail to properly predict future input values. The model showed that the highest yield for biodiesel was determined
at an SCM concentration of 1.01 mol/mol corresponding to the actual biodiesel yield of 1.00 mol/mol. This was
obtained at 350 °C, 20 MPa within 10 min. Similarly, the highest possible yield for biodiesel from SCE was
0.97 mol/mol achieved at 400 °C, 20 MPa within 25 min. The PCA plots obtained show how using a spiral
reactor can have varying effects on the correlational statistics between the inputs. In the plots obtained from
Python, all variables for SCM and SCE are directed in the same direction as the yield showing a net positive
relationship.
The change in direction from left to right in Figure 1. (a) to Figure 2. (a) are indicative of the relationships the
PCA scores have with the standardized inputs and outputs. This still shows that having a higher pressure,
chemical to oil ratio and reaction time promise higher yields. The project is being worked currently to improve
upon the fuzziness of the process by adding in a neural network component through the ANFIS toolbox. The

362
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

implications of such modelling are extremely positive for the energy industry to reduce capital expenses and
waste caused by erroneous predictions.

3. MODELLING STUDIES AND DESIGN

Modelling in the fuzzy interface system first begins by specifying the number of inputs and outputs in the FIS
window. The ranges of values are inputted as well as the type of membership functions they will have. This
could range from triangular, trapezoidal, pentagonal, gaussian and so on. On generating the rules, users can then
view the model schema which shows how the system is learning the model and the output it is generating. In the
case example discussed here, the reducing sugar value is obtained after the system interprets the rules fed through
the interface. The yellow gaussian functions in the figure show the overlap between various inputs. To avoid
redundancies, the system defuzzifies on a limit that is determined as per the number of rules used to generate
the model. Creating more rules will thus lead to great accuracy and reduce repeating output values.
A multidimensional plot for all the inputs and outputs can be generated to show the regions covered by the
algorithm and how they vary with relation to each other. It is important to note that the lack of rules will only
cause the system to generate output values based on previously defined rules resulting in errors. It is thus suitable
to include a rule for every input so that the model can learn to incorporate and capture all sorts of values. A
commonly known issue that arises in such model predictive control techniques is when an input value is entered
that does not fit within the specified range articulated to the system. In the fuzzy logic modelling example, the
user specifies the maximal and minimal range the input and output values can take which are still vulnerable to
errors. However, with the use of automated systems in industries, such models can be taught to ‘update’ these
ranges when outliers arise so as to eliminate such inefficiencies. In the case that multiple outputs share inputs
that fall within similar ranges, the system defines a cut-off so as to prevent redundant outputs from entering the
model.

(a)
Figure 3. (a) Multidimensional plot for process parameters and output from MATLAB.

4. MODELLING SIMULATIONS AND RESULTS

For the fuzzy models generated, the model outputs have been plotted against the experimental outputs to
illustrate how they can be used to model accurate predictive systems. The simulation results for two of the five
data sets studied is seen below relating to the PCA plots. The plots compare the actual and model value, shedding
light on the preciseness of fuzzy logic modelling. The model results for these figures show that the system has
learnt to ‘track’ the values and can adapt well to new inputs. Similar plots were generated for all of the studied
datasets and were better interpreted through PCA analysis to better connect the dynamic nature of the inputs and
the outputs. Within the scope of biofuel production, fuzzy enhanced sensors can become a reality in industries,
moving away from pilot plants and small-scale manufacturing facilities. While the crisp nature of the inputs and
outputs may deter some from considering them as primary predictors in industries, they can still be used to test
out relationship matrices in design Experiments to maximize yields and optimize operating conditions. These

363
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

results also derive consistency with the design methodologies discussed in literature that use dilute acids and
various mixing mechanisms.[3]

100
Digestibility
(Actual Vs Model)
50

RMSE 2.929014

0 Actual Model
0 5 10 15 20 25 30 35 40
(a)
Figure 4. (a) Results from data obtained from experiments[2]

100 Gyield(Actual Vs Model)

50

RMSE 2.189529
0 Actual Model
0 5 10 15 20 25 30 35 40
(a)
Figure 5. (a) Results from data obtained from experiments[8]

The exceptional points in the plots where the expected output and model output have large differences are
indications of the membership functions failing to properly capture the points to a greater accuracy which can
be offset by creating more rules to make defuzzification easier. The lowest RMSE achieved on all of the studied
datasets was 0.040164.

5. CONCLUSION AND DISCUSSIONS

The intricacies that affect the appropriate yields for ethanol and other biofuels tend to share a non-linear
relationship with some of the inputs. Principal component studies of these parameters would indicate that they
all can be modelled to fit equations using advanced levels of design studies such as FIS(Fuzzy Logic Interfaces).
When used with a neuro-fuzzy design mechanism, the accuracy can be improved even further. The method of
using design Experiments is quite crucial to industries as parameters that have significant influence on the yields
can be identified. For fuzzy logic designs, this is quite important as the effects of variables must be included in
a quantifiable way. The results obtained from parametric designs from experiments in literature[13][14] show
that removing certain variables can have a better influence on the models that can predict the values with greater
accuracy.
The experiments show that higher yields can be achieved with better optimal processing conditions that are the
cornerstone of model predictive methodologies. When dealing with variables that are more intrinsic such as
reaction time or pH values combined with yeast inoculum concentrations and nutrient concentrations, such
models help deal with minor changes to a better degree than standard predictive steps [15]. This has been
discussed extensively in the model cases in the paper. As per the fuzzy logic models that have been used here,
the conclusion to be made is that specifying the nature of the functions for the inputs as well as the ranges can
produce accurate models to great degrees. Preliminary analysis too can help understand how the variables are
related so as to frame the rules in more dynamic ways. Certain models show great accuracies such as the one
developed for the data in parametric experiments. While others may show great error rates, these can be reduced
by creating better interface rules and making more membership functions that can correlate the inputs and

364
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

outputs better. CCD and RSM methodologies are particularly of great interest to the industry as they help
develop Experiments from a large batch of runs and focus on the ones that promise great yields. In the case
examples discussed in this paper. A majority of the experiments take use of modelling studies with a predictive
approach which shows how tools like fuzzy modelling can help industries meet product objectives by optimizing
inputs. Biofuel demands will likely see great increases in the future owing to the transient nature of conventional
fuels and the risks they pose. As a result, model predictive analogies such as fuzzy logic will become popular to
better understand the process and help industries meet optimal production levels.
REFERENCES
[1] Ezzatzadegan, L., Morad, N., & Yusof, R.. PREDICTION AND OPTIMIZATION OF ETHANOL
CONCENTRATION IN BIOFUEL PRODUCTION USING FUZZY NEURAL NETWORK. Jurnal Teknologi
2016.8-20.
[2] Nielsen, J., Larsson, C., van Maris, A., & Pronk, J.. Metabolic engineering of yeast for production of fuels and
chemicals. Current Opinion in Biotechnology 2013, 1-18.
[3] Sohpal, V., & Singh, A.. Optimization of alkali catalyst for transesterification of jatropha curcus using adaptive neuro-
fuzzy modeling. Biofuel Research Journal 2015, 3-25.
[4] Cardona, C., & Sánchez, Ó. . Fuel ethanol production: Process design trends and integration opportunities.
Bioresource Technology 2007, 5-10.
[5] Singh, H., Gupta, M., Meitzler, T., Hou, Z., Garg, K., Solo, A., & Zadeh, L.. Real-life applications of fuzzy logic.
Advances in Fuzzy Systems 2013, 1-25.
[6] Suganthi, L., Iniyan, S., & Samuel, A.. Applications of fuzzy logic in renewable energy systems - A review.
Renewable and Sustainable Energy Reviews 2015, 7-25.
[7] López-Zapata, B., Adam-Medina, M., Álvarez-Gutiérrez, P., Castillo-González, J., Hernández-de León, H., & Vela-
Valdés, L.. Virtual sensors for biodiesel production in a batch reactor. Sustainability (Switzerland) 2017, 6-15.
[8] Rivera, E., Plazas Tovar, L., Regina, M., Maciel, W., Bonomi, A., & Filho, R.. Neural Network Modeling to Predict
the Total Glucose Yield after Enzymatic Saccharification of H 2 SO 4 -Catalyzed Hydrothermally Pretreated
Sugarcane Bagasse. Chemical Engineering Transactions 2015, 1-18.
[9] Adnan, N., Suhaimi, S., Abd-Aziz, S., Hassan, M., & Phang, L. . Optimization of bioethanol production from glycerol
by Escheric h ia coli SS1. Renewable Energy 2014,3-8.
[10] Sherpa, K., Ghangrekar, M., & Banerjee, R.. Optimization of saccharification of enzymatically pretreated sugarcane
tops by response surface methodology for ethanol production. Biofuels 2017, 5-19.
[11] El-Gendy, N., Madian, H., & Amr, S.. Design and Optimization of a Process for Sugarcane Molasses Fermentation
by Saccharomyces cerevisiae Using Response Surface Methodology . International Journal of Microbiology 2013, 1-
15.
[12] Farobie, O., Hasanah, N., & Matsumura, Y.. Artificial Neural Network Modeling to Predict Biodiesel Production in
Supercritical Methanol and Ethanol Using Spiral Reactor. Procedia Environmental Sciences 2015, 1-20.
[13] Mohan, P., Ramesh, B., & Reddy, O. . Production and Optimization of Ethanol from Pretreated Sugarcane Bagasse
using Saccharomyces bayanus in Simultaneous Saccharification and Fermentation. Microbiology Journal 2012, 2-20.
[14] Timung, R., Naik Deshavath, N., Goud, V., & Dasu, V.. Effect of Subsequent Dilute Acid and Enzymatic Hydrolysis
on Reducing Sugar Production from Sugarcane Bagasse and Spent Citronella Biomass. Journal of Energy 2016, 4-
28.
[15] Punsuvon, V.. Optimization of Delignification and Enzyme Hydrolysis of Steam Exploded Oil Palm Trunk for
Ethanol Production by Response Surface Methodology. In V. Punsuvon, Sustainable Degradation of Lignocellulosic
Biomass - Techniques, Applications and Commercialization 2013, 4-18.

365
 A MULTİ-AGENT APPROACH FOR DECENTRALİZED
VOLTAGE REGULATİON İN ELECTRİCAL DİSTRİBUTİON
SYSTEM WİTH RENEWABLE GENERATİON

Ömer Gül
Istanbul Technical University, Istanbul , Turkey, [email protected], ORCID: 0000-0003-0379-5446

Emre Köksal
Istanbul Technical University, Istanbul , Turkey, [email protected], ORCID: 0000-0002-7492-130X

Gül, Ö, Köksal, E. A Multi-Agent Approach for Decentralized Voltage Regulation in

Cite this paper as: Electrical Distribution System with Renewable Generation. 8th Eur. Conf. Ren. Energy Sys.
24-25 August 2020, Istanbul, Turkey,

Abstract: Voltage regulation systems for wind generators that connected to an electrical energy system are
centralized. The connection of a large number of wind generators between the existing variable
loads can cause a serious voltage regulation problem, for which the traditional control techniques
may be insufficient. New digital technologies provide monitoring, automatic control and two-way
communication to improve the overall performance of the electrical energy system. With these
technologies, distributed control algorithms applied that are capable of providing correct voltage
regulation in a smart electrical energy system.
In this study, a decentralized, multi-agent voltage regulation algorithm proposed for an electrical
energy system with variable loads and distributed generations. The decentralized, multi-agent
algorithm and flow diagram developed to regulate voltage in the smart electrical energy system,
which includes a load agent, a reactive power control agent, and an on-load tap changer agent. A
medium voltage distribution network modeled in MATLAB/Simulink and the developed
algorithm tested for three different situations.
Keywords: Electrical distribution; voltage regulation; distributed control; multi-agent systems; renewable
generations
© 2020 Published by ECRES

1. INTRODUCTION

Fossil fuel depletion and fossil fuel based power plants threaten the environment and human health day by day.
For these reasons, the importance of renewable energy resources such as wind and solar energy is increasing.
Distributed generation is defined as power plants that have smaller installed capacity than centralized power
plants and capable of connecting from almost every point of the power system [1]. In recent years, the number
and power of distributed generation resources connected to existing electrical distribution systems is increasing
rapidly. Various technical difficulties face in operating the electrical distribution systems designed without
considering this increase in the use of distributed generation resources. New approaches to network planning
and operation needs efficiently utilize the positive effects of distributed production in existing network
integrations and avoid potential adverse effects. Distributed generation has positive effects such as reducing grid
losses, adjusting the voltage profile, and improving the quality of the power. However, before integration and at
the moment of operation some issues required to be considered in detail such as; system grounding, transformer
connection, short circuit effects, bidirectional energy flows and the related protection coordination, voltage
fluctuations, harmonics and the risk of islanding [2].
In the operation of radial networks fed from one side, voltage regulation carry out by centralized control systems.
Conventional control methods aimed to prevent possible voltage drop at line ends due to load currents. The
dissemination of distributed generation requires taking into account the generated power values of distributed
generation, in addition to the voltage control due to load estimates. For this reason, studies aiming at voltage

366
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

regulation by controlling factors such as distributed generation, under-load tap-changers (OLTC) and static
reactive power compensators in the literature [3-10]. The study in [7] controlled the voltage by controlling the
wind turbines in power factor control and voltage control modes, and as a last resort, they also tried to curtail
active power generation. The study in [8] performed the active power curtailment using the sensitivity matrix
obtained from the power flow analysis. The reference number [9] provides voltage regulation by controlling the
capacitors in the system and OLTC with set values previously determined by offline simulations. In [10], voltage
regulation performs by controlling the equipment such as the step voltage regulator (SVR), shunt reactor-
capacitor, static VAR compensator (SVC) with artificial neural networks. In [11], a study success in which the
voltage is controlled within the desired limits by controlling distributed generation and OLTC. In [12], the
distribution network voltage in which distributed generations are located is regulated only by using OLTC, [13]
has done this using SVC. The study in [14] has developed an algorithm for active voltage control for many
distributed generations.
The distribution system is getting structure that is more complicated day by day. In recent years, it has become
common to investigate fast and coordinated operation of distributed generation and voltage control equipment
for voltage regulation. It has become a preferred method in these researches to provide voltage control in
distribution networks by using multi-agent systems. In order to provide distributed control in distribution
networks, MAS (multi-agent systems) used and local voltage control devices operated in coordination with each
other to keep the voltage within the permissible limits [15-17]. The study in [18] controlled the voltage by
defining the control relay and line relay as agent. The study in [19] has established a structure that defined
distribution transformers as voltage control agents and curtail active power generation as a last resort. The study
in [20] defined one manager agent and multiple local agent in order to regulate the voltage changes caused by
photovoltaic distributed generation. The studies in [21, 22] used load agents to monitor changes in the system
and defined OLTC and DGs as control agents. In [23] study provided that an agent communicates with the agent
that was identified as compensating agent when it realized the voltage exceeded the limits, and then demand
voltage support from other distributed generations through its. The study in [24] identified the STATCOM
devices in the system as intelligent agents for voltage regulation with MAS it also used the JADE (Java Agent
Development) software which provides the FIPA (Foundation for Intelligent Physical Agents) agent interaction
platform. The studies in [25-27] preferred to allocate the system to the sub-regions in order to solve similar
voltage problems (modular architecture), thus aiming to make the distributed control more effective. The study
in [28] utilized MAS to minimize losses in addition to voltage regulation used communication agents and
implemented in MATLAB/Simulink.
In this study, MAS based voltage control method preferred for voltage regulation in smart electrical distribution
networks. After giving general information about the agent and the multi-agent system, a decentralized multi-
agent voltage regulation algorithm proposed for an electrical energy system with variable loads and distributed
generations. The decentralized multi-agent algorithm and flow diagram regulate voltage in the smart electrical
energy system, which includes a load agent, a reactive power control agent, and an on-load tap changer agent.
A medium voltage distribution network modeled in MATLAB/Simulink and the developed algorithm tested for
three different situations.

2. VOLTAGE REGULATION

Over the last decade, distributed generation systems connected to the power distribution network increasingly.
However, existing electricity distribution networks designed to feed only the connected loads. Due to the
increasing penetration of distributed generation, voltage fluctuations have become a major problem for
distribution network operators.
The integration of electrical wind power generation into the distribution network can cause problems in voltage
regulation. The main reason for this is that the existing distribution network not designed to include wind power
generation. In fact, the connection of the wind generator to the electric distribution network can cause local
overvoltage [20].
Most of the distribution networks designed as passive networks with a radial structure. Therefore, active and
reactive power flows on one line are point-to-point, where voltage ranges from high to low. The ratio of
reactance and resistivity in the conduction system is X / R ≥ 10, while in the distribution system, this ratio is X
/ R ≤ 0.5; the resistance in the distribution system is higher than in the transmission system. This high resistance
causes a voltage drop from the substation to the consumer in the distribution system.

367
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig. 1. Single line diagram of a sample radial network

The voltage drop is the same as presented in Eq. [1] for the system shown in Fig. 1.
(𝑅𝐿𝑁 𝑃𝐿 +𝑋𝐿𝑁 𝑄𝐿)−𝑗(𝑋𝐿𝑁 𝑃𝐿−𝑅𝐿𝑁 𝑄𝐿 )
△𝑈 = | 𝑈2
| (1)
𝑈1 : Voltage of the secondary side of the transformer
𝑈2 : Voltage of the end of the line
𝐼 : Line current ş
𝑅𝐿𝑁 : Line resistance
XLN : Line reactance
𝑃𝐿 : Active power of the load
𝑄𝐿 : Reactive power of the load

Distributed generation systems connect to the network via synchronous or asynchronous generators, or power
electronic systems. Distributed generation systems have the ability to work in different power factors due to
their design. They can absorb reactive power from the network and provide reactive power to the network. In
this regard, distributed generation has become an important part of voltage control in distribution systems. A
single line diagram of a distribution system with distributed generation and load presents in Fig. 2.

Fig.2. A single line diagram of a distribution system with distributed generation and load

The voltage drop regulated, as demonstrated in Eq. [2], by integrating the distributed generation using the Fig.
2.
𝑅 (𝑃 −𝑃 )+ 𝑋 (𝑄 −(±𝑄𝐷𝐺 ))
△ 𝑈 = 𝐿𝑁 𝐿 𝐷𝐺 𝑈 𝐿𝑁 𝐿 Eq. (2)
2
𝑃𝐷𝐺 : Active power of distributed generation
𝑄𝐷𝐺 : Reactive power of distributed generation
According to Eq. (2), the distributed generation will reduce the voltage drop across the line if the plant generates
reactive power or does not exchange power with the grid. If the generated energy is higher than the load at the
connection point, the power flow will be from the point of distributed production to the substation, and the
voltage in the distributed generation bus will rise. In addition, if the distributed generation absorbs reactive
power from the network, it will be able to act to increase or decrease the voltage according to the relationship
between the active and reactive power of the distributed generation, the active and reactive power of the load,
and the X / R ratio of the line. If the distributed generation is to be included in the network, the worst-case
scenarios considers that would allow the voltage level to exceed the permitted limits for trouble-free integration.
Worst scenarios summarized in terms of connection point voltage: minimum load and maximum generation;
maximum load and minimum generation. Currently, voltage regulation systems for wind generators that
connected to the distribution network are centralized. A decentralized control technique needs to regulate the
distribution profile of the distribution network buses to which the distributed generation system is connected.

3. PROPOSED MULTI-AGENT SYSTEMS BASED VOLTAGE REGULATION ALGORITHM

There is no certain definition of agent. For this reason, different researchers have made different definitions of
agent. The simplest and most valid definition that an agent is anything that can be viewed as perceiving its
environment through sensors and acting upon that environment through actuators in [30].

368
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In general, manner, the structure formed by combining two or more agents called multi-agent system. Agents
can solve the problems that individual knowledge and capabilities are not enough.
Agent-based systems have been one of the most interesting areas of research in the field of computers since the
early 1990s. MAS has been widely used to design and improve software systems. The fields of computer science
such as software engineering, programming, control systems, decision-making and data acquisition are
increasing day by day. In addition to these uses, MAS also can use in electrical engineering fields. Usage areas
in power systems listed as monitoring, distributed control, modelling and simulation, protection [31].
In this section, the simulation study about the voltage variation in the distribution system mentioned. In the test
system modelled in Simulink environment, steady-state voltages out of limits tried to regulate by a multi-agent
system. Agent with different job descriptions try to solve the problem by communicating with each other using
the communication abilities of the smart grid. Hierarchical MAS structure is preferred among MAS
architectures.
The agents at different points of the system increase the flexibility of the network and the distributed control
capability. The proposed control algorithm continuously monitors the voltage value in the different regions of
the system. Taking advantage of the agent communication capabilities, it tries to keep the voltage within the
permissible limits by relevant agents’ effects. The developed multi-agent system structure has three types of
agents, load agent, reactive power control agent and OLTC agent.
The agents in the proposed MAS perform their tasks in a specific order to regulate voltage. Load agents used as
passive factors in this study as well as in many studies. The main task of load agent is monitoring the bus voltage
at each sampling time interval of the simulation. Load agent compares the bus voltage with permissible limit
values. If the bus voltage exceeds the permissible value, related load agent will communicate with corresponding
reactive power agent. If the voltage regulated by reactive power control agent, then the next load agent is
activated.
Reactive power control agent refers to equipment that injects reactive power into the system or absorbs reactive
power from the system. Modern distribution networks include equipment such as shunt reactors, shunt
capacitors, SVC, STATCOM with reactive power capacity. Distributed generations such as wind turbines with
a converter structure can also perform reactive power control.
The main task of the reactive power control agents is to influence the system in order to regulate the voltage in
the system when requested. In this study, this effect is to switch from reactive power control mode to voltage
control mode of distributed generations. At this point, reactive power support provided to the system within the
reactive power capacity of the wind turbine to keep the voltage within the specified set values, or reactive power
can be absorbing from the system. Another reactive power control equipment used in the application is a shunt
capacitor. Shunt capacitor preferred before the wind turbine as a reactive power control agent. In practice, multi-
step capacitor banks are generally preferred for optimal control. In this way, reactive power support provided
by taking the capacitor group into the required amount. However, there is a constant shunt capacitor in this
study. In case of demand, it connects or disconnects with the system.
There are two main tasks of the OLTC agent in the system. The first is to change the step of the transformer to
regulate the voltage of the bus, which exceeded the limit. It compares the voltage value with the voltage value
at the bus. The second task is to keep the number of tap change minimum. Unnecessary tap changes will increase
the maintenance costs of the tap changer.
OLTC agent is the last resort to regulate voltage. Prior to the need to intervene with the OLTC, the reactive
power control agents in the system, such as the distributed generation agent and the shunt capacitor agent etc.,
attempt to regulate the voltage. If after the effects of these agents, the system voltage is still exceeding the limits,
help requested from the OLTC agent.
Three different types of agents utilized in the distribution system that includes a transformer with OLTC, loads,
distributed generations and shunt capacitors. In Fig. 3, the flow diagram gives the coordination and decision-
making mechanism between the factors. The steps of the flow diagram given below.
Load agents continuously monitor the bus voltages to which they relate at all times of the simulation.
Load factors are activated respectively and check if there is an over voltage or under voltage at the voltage level.
If a load agent monitors a voltage violation, the control cycle will stop on that load agent and the relevant reactive
power control agent defined on the load factor will inform.
If a load agent requires the increase of the voltage and there is a capacitor in the influence area of the load agent,
the capacitor agent will have activated in the first place. After the effect of the capacitor agent, the capacitor
factor compares the busbar voltage to the set values. If the busbar voltage regulated, the control cycle will have
transferred to the next load factor.

369
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig.3. MAS Algorithm Flow Diagram

If the voltage regulated after the effect of the capacitor agent, the load agent demand that the wind turbine in the
region will switched operation mode to voltage control mode. If the distributed generation is capable of
regulating voltage within its capacity, it transfers the control cycle to the next load agent.
If a load agent detects that the voltage is above the limits, it directly communicates with the wind turbine as a
reactive power control agent. The turbine expected to reduce the voltage by absorbing reactive power from the
network. If the distributed generation regulates the voltage, it continues the control cycle from the next load
agent.
If a reactive power control agent has an effect and the agent has reached the limits of its capacity, but the voltage
is still out of the limits, the reactive power control agent will activate OLTC agent.
If the reactive power control agent activates the OLTC agent, the OLTC agent will takes voltage value from
related load agent. It compares this value with the defined limit values. It tries to regulate the busbar voltage by
minimum tap changing. If the voltage regulated, the control cycle continues by activating the next load factor.
If the control cycle control all of load agents and does not observe any violation, it will return to the first load
factor and restarts the control cycle.

4. CASE STUDIES

In order to perform MAS based voltage regulation for electrical distribution system, test network used in [32]
that has six busbars adapted to the voltage levels used in Turkey, also some additions made as shown in Fig. 4.

370
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig.4. Single line diagram of test network.

Network Block:
Phase-Phase Voltage: 154 kV, X/R: 10, 3 phase short circuit power: 1760 MVA
Transformers:
1 pcs 154/34.5 kV, 100 MVA, Yd11, 12x±%1.25 OLTC
Connection of wind parks to the network 4 pcs 12 MVA, 34.5/0.575 kV, Dyn
Wind Parks:
4 pcs, 9 MW, includes DFIG type wind turbine
Shunt Capacitor:
34.5 kV, 5 MVAr
Loads:
20 pcs 0.4 kV, 1.4+j0.7 MVA
For power plants internal loads 500 kW load added to simulation system for each plant
Lines:
1272 MCM conductor for overhead lines with pi equivalent circuit
95 mm2 copper XLPE conductor cable between wind park and network connection
The generation amounts of the wind farms and the demand amounts of the loads vary in the test system as in
real networks. According to different generation and demand power amounts, the steady state voltage values
may exceed the limits of ± 10%. Within the criterion mentioned in the regulation [33], a simulation study
executed in the computer environment with the algorithm developed to keep the voltage within ± 10% range.
Within the criterion mentioned in [33] a simulation study in the computer environment with the algorithm
developed to keep the voltage within ± 10% range.

Voltage regulation when all power plants are disabled and load increase in the bus 4
Each of the modelled wind farms produces 1.87 MW of active power at an average wind speed of 8 m / s. It is
assumed that the power plants are located in close regions and the wind speed has increased from 8 m / s to 25
m / s. This increase effects all power plants. We know that wind turbines deactivated a very low and very high
wind speed. Because of increased wind speed, all power plants are disabled and bus voltages reduce as expected.
The bus voltages seen in table 1 column I when wind farms switched off and the MAS is not active.
As shown in table 1 column I, the voltage of the 4th bus has fallen below 10% of its nominal voltage (0.894 pu).
In order to regulate the voltage with MAS, the control diagram given in Figure 3 is included in the system. The
system is simulated again, then the voltage drops in the 4th bus is detected by the 4th load agent and informed
the reactive power control agent of 4th bus. The shunt capacitor in the 4th bus is activated.
When the shunt capacitor switches on, the busbar voltage regulated. In the latter case, the voltage 4th bus is
increased above the permissible lower limit (0.945 pu). The bus voltages of the system with MAS voltage
regulation given in table I column II.

Table 1. All wind farms switched off

I II III IV
Bus 1 0.951 0.966 0.951 0.963
Bus 2 0.922 0.950 0.923 0.934
Bus 3 0.903 0.943 0.904 0.915
Bus 4 0.894 0.945 0.895 0.906

In the case where the wind turbines are not active and the shunt capacitor on the 4th bar is activated, the steady-
state voltages are within the permissible limits. To make the situation worse, it is assumed that Load 4 requires
twice the load to test the control capability of the system. Under these conditions, the voltage of 4th bus is 0.895
pu as seen from Table 1 column III. If it is desired to control the voltage with MAS, the flow diagram is
implemented as follows. 4. the load agent detects the voltage drop.

371
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig.5. Test network modelled in Simulink.

The flow diagram is transferred to the next step because the shunt capacitor is already active and another reactive
power control agent of 4th bus (distributed generation) is disabled. Voltage support is requested from the next
step. OLTC increases the voltage by changing its tap position the "-1". In the final analysis, the busbar voltages
are regulated to the appropriate values as seen in Table 1 column IV.

Voltage regulation in case of one power plant failure

In the second scenario, a single power plant is disabled. It has been assumed that WP-3 is disabled for a fault or
similar reason. The system is analysed under the condition that WP-3 is deactivated. Because of the analysis, it
was observed that the busbar voltages decreased as expected. However, because the voltage drop is not over
10%, the load agents do not detect a violation. Load-2, Load-3, Load-4 in the network were increased to 1.6, 2
and 2.2 times respectively and it was aimed to create a decrease in voltage in the system. In the re-analysed
system, it was observed that the voltage drops of the 4th bus was over 10%. The bus voltages in the case where
WP-3 is deactivated and the loads demand more load can be seen in Table 2 column V.

Table 2. Only a wind farm switched off

V VI
Bus 1 0.936 0.946
Bus 2 0.917 0.928
Bus 3 0.903 0.915
Bus 4 0.889 0.902

As shown in Table 2 column V, the voltage of the 4th bus has fallen below the limit values (0.889 pu). If voltage
regulation is performed with MAS, the MAS algorithm will follow these steps. 4th load factor detects a voltage
violation and reports the reactive power control agent. In this case, the WP-4 distributed generation in the 4th
bus is switched from the reactive power control mode to the voltage control mode to regulate the voltage. WP-
4 boosts the busbar voltage by injecting 0.727 MVAr of reactive power into the grid. In the last case, the voltage
of 4th bus is increased to 0.902 pu as seen in table 2, column VI.

Voltage regulation in the case of increased production of all power plants

As in the previous two scenarios, it was assumed that the wind speed of 8 m / s was increased to 12 m / s for all
power plants. In this case, the generation amounts of power plants increase from 1.87 MW to 6.38 MW. The
increase for production in the power plants leads to an increase in bus voltage. In addition to this effect, Load-
1, Load-2 and Load-3 are also deactivated. The value of the Load-4 was reduced to 2/5. The decrease for
consumption in the busbars has a negative effect on the voltage. As a result, the voltage increase in the 3rd bus
to 1.102 pu as seen in Table 3, column VII.

372
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 3 All wind farms switched on

VII VIII

Bus 1 1.093 1.083

Bus 2 1.099 1.088
Bus 3 1.102 1.089
Bus 4 1.100 1.089

If voltage regulation is performed with MAS, the 3rd load recognizes that the voltage increase is above the limits
and informs the related reactive power control agent. Distributed generation in the 3rd bus is known as it has the
ability to absorb reactive power from the network with its DFIG structure. In order to regulate the bus voltage,
WP-3 operation mode switched to voltage control mode. WP-3 absorbs 3.56 MVAr reactive power to from the
network and regulates the voltage by this way. The voltage of 3rd bus is moved to the allowable range (Table 3,
column VIII).

5. CONCLUSION

Due to the changes in the energy generation of renewable energy sources, centralized voltage control is
inadequate in keeping the voltage in the permitted range in the operation of electrical distribution systems and
it is inevitable to use the distributed control approach. In this study, multi-agent system based algorithm and
flow diagram developed for the voltage regulation in the smart electrical distribution system, which includes
load agent, reactive power control agent, OLTC agent. A medium voltage distribution network modelled in
MATLAB / Simulink and the proposed algorithm tested for three different situations.
In case 1, it was observed that the wind power plants in the system were disabled and the demand load decreased
and the voltage exceeded limits. The designed MAS was able to increase the voltage above 0.9 pu by activating
the shunt capacitor and determining the appropriate tap value of OLTC.
In case 2, failure of a single power plant and an increase in the demand load simulated. The voltage dropped to
below 0.9 pu. The MAS algorithm demanding support from the wind farms and using the reactive power
capacities of the turbines can regulate this voltage violation.
In case 3, contrary to other cases, the most important negative effect of distributed generation in literature has
been examined. Increased energy generation on distributed generation and reduced demand load increased bus
voltages of the system. In order to regulate the voltage, the MAS algorithm benefited from the reactive power
capacity of the plants, as in case 2, and was able to regulate the voltage by absorbing reactive power from the
system.
In the simulation study, the changes in the distribution system due to generation and consumption can cause
voltage irregularities, and the voltages may exceed ± 10% as stated in the regulations. By using MAS approach,
it is possible to regulate the voltage with minimum communication and operational change. In this study, MAS
modeled and applied on MATLAB software with some assumptions. In order to eliminate the deficiencies
detected by the load flow and dynamic analyzes in the distribution networks, the necessary equipment
installation and coordinated control are important for system reliability.

ACKNOWLEDGMENTS

We would like to thank the ITU BAP unit funding support for "Investigation of the Working Conditions of
Renewable Energy Sources and Energy Storage Systems Connected to Electricity Distribution Network", project
number 39015.

REFERENCES

[1] Dondi, P., Bayoumi, D., Haederli, Julian D., Suter, D. (2002). Network Integration of Distributed Power Generation, Journal of Power
Sources, Vol. 106, s.1-9.
[2] IEEE, IEEE Application Guide for IEEE Std 1547™, IEEE Standard for Interconnecting Distributed Resources with Electric Power
Systems, IEEE Std 1547.2TM-2008.
[3] Carvalho, P., Correia, P., F., Ferreira, L.A. (2008). Distributed Reactive Power Generation Control for Voltage Rise Mitigation in
Distribution Networks, IEEE Transactions on Power Systems,Vol. 23, No. 2, s.766-772

373
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[4] Hashim, T., T., Mohamed, A., Shareef, H. (2012). A Review on Voltage Control Methods for Active Distribution Networks, Przeglad
Elektrotechniczny (Electrical Review): Warszawa, Poland, s.304–312.
[5] Tonkoski, R., Lopes, L., A., El-Fouly, T.H. (2011). Coordinated Active Power Curtailment of Grid Connected PV Inverters for
Overvoltage Prevention, IEEE Transactions Sustainable Energy, Vol. 2, s.139–147.
[6] Viawan, F., Karlsson, D., (2007). Combined Local and Remote Voltage and Reactive Power Control in the Presence of Induction
Machine Distributed Generation, IEEE Transactions on Power Systems, Vol. 22, s.2003–2012.
[7] Sansawatt, T., Ochoa, L., F., Harrison, G., P. (2010). Integrating Distributed Generation Using Decentralised Voltage Regulation,
Power and Energy Society General Meeting, IEEE , vol., no., s.1-6.
[8] Zhou Q., Bialek, J., W. (2007). Generation Curtailment to Manage Voltage Constraints in Distribution Networks, IET Generation,
Transmission & Distribution, Vol.1, No.3, s.492-498.
[9] Viawan, F., A., Karlsson, D. (2007). Combined Local and Remote Voltage and Reactive Power Control in the Presence of Induction
Machine Distributed Generation, IEEE Transactions on Power Systems, Vol.22, No.4, s.2003-2012.
[10] Sampaio,R. F., Melo,L. S., Leão,R. P.S., Barroso,G. C., Bezerra,J. R., Automatic restoration system for power distribution networks
based on multi-agent systems, IET Generation, Transmission & Distribution, Volume 11, Issue 2, January 2017, p. 475 – 484.
[11] Toma, S., Senjyu, T., Yona, A., Sekine, H., Funabashi, T., Chul-Hwan Kim, (2008). Optimal Control Of Voltage In Distribution
Systems By Voltage Reference Management, Power and Energy Conference, 2008. PECon 2008. IEEE 2nd International, vol., no.,
s.1239-1244.
[12] Baerthlein, E., M., Hartung, M., Panosyan, A. (2014). Variable Voltage Set Point Control of Tap Changers in Distribution Grids,
Innovative Smart Grid Technologies Conference Europe (ISGT-Europe), 2014 IEEE PES , s.1-6.
[13] Hatta, H., Kobayashi H. (2008). Demonstration Study On Centralized Voltage Control System For Distribution Line With Sudden
Voltage Fluctuations, SmartGrids for Distribution, 2008. IET-CIRED. CIRED Seminar, s.1-4.
[14] Kulmala, A., Repo, S., Järventausta, P. (2014). Coordinated Voltage Control in Distribution Networks Including Several Distributed
Energy Resources, IEEE Transactions on Smart Grid, Vol. 5, No. 4, s.2010-2020
[15] Baran, M., E., El-Markabi, I., M. (2007) A Multiagent-Based Dispatching Scheme For Distributed Generators For Voltage Support
On Distribution Feeders. IEEE Trans. Power Syst., Vol.22, No.1, s.52–59.
[16] Farag, H., E., El-Saadany, E.F. (2013). A Novel Cooperative Protocol For Distributed Voltage Control in Active Distribution
Systems, IEEE Trans. Power Syst. 2013, Vol. 28, s.1645–1656.
[17] Elkhatib, M., E., El-Shatshat, R., Salama, M. (2011) Novel Coordinated Voltage Control For Smart Distribution Networks With
DG. IEEE Trans. Smart Grid 2011, Vol. 2, s. 598–605.
[18] Aquino-Lugo, A., A., Klump R., Overbye, T., J. (2011). A Control Framework for the Smart Grid for Voltage Support Using Agent-
Based Technologies, IEEE Transactions on Smart Grid, Vol.2, No.1, s.173-180
[19] Lang, S., Niannian C., Mitra, J. (2013). Multi-Agent System Based Voltage Regulation In A Low-Voltage Distribution Network,
North American Power Symposium (NAPS), s.1-6.
[20] Yorino, N., Zoka, Y., Watanabe M., Kurushima, T. (2015) An Optimal Autonomous Decentralized Control Method for Voltage
Control Devices by Using a Multi-Agent System, IEEE Transactions on Power Systems, Vol. 30, no.5, s. 2225-2233.
[21] Sajadi, A., Farag, H., E., Biczel P., El-Saadany, E., F. (2012) Voltage Regulation Based On Fuzzy Multi-Agent Control Scheme in
Smart Grids, Energytech, 2012 IEEE , s.1-5.
[22] Farag, H. E., El-Saadany, E., F., Seethapathy, R. (2012). A Two Ways Communication-Based Distributed Control For Voltage
Regulation in Smart Distribution Feeders, IEEE Trans. Smart Grid, Vol. 3, No. 1, s. 271–281.
[23] Zad,B. B., Jacques Lobry,J., Vallée, F., (2018). Impacts of the model uncertainty on the voltage regulation problem of medium-
voltage distribution systems, IET Generation, Transmission & Distribution, Volume 12, Issue 10, May 2018, p. 2359 - 2368.
[24] Tousi, M., R., Hosseinian, S., H., Jadidinejad A., H., Menhaj, M., B. (2008). Multi-agent Based Voltage Control of STATCOMs to
Enhance Elimination of Voltage Disturbances in Power System Contingencies, Universities Power Engineering Conference, UPEC
2008, 43rd International, s.1-5.
[25] Wang,X., Wang,C., Xu,T., Guo,L., FanS.,, Wei,Z., (2018) Decentralised voltage control with built-in incentives for participants in
distribution network, IET Generation, Transmission & Distribution, Volume 12, Issue 3, February 2018, p. 790 – 797.
[26] Pachanapan, P., Anaya-Lara O., Lo, K., L. (2009). Agent-Based Control For Power Quality Enhancement In Highly Distributed
Generation Networks, 2009 Proceedings of the 44th International Universities Power Engineering Conference (UPEC), , s.1-5.
[27] Nguyen, P., H., Myrzik J., M., A., Kling, W., L. (2008). Coordination of Voltage Regulation in Active Networks, Transmission and
Distribution Conference and Exposition, 2008. T&D. IEEE/PES , s.1-6.
[28] Arshad A. (2015). Multi-Agent System Based Distributed Voltage Control in Distribution Systems. (Master Thesis). Aalto
University, School of Electrical Engineering, Espoo, Finland.
[29] Mahmud, M., A., Hossain, M., J., Pota, H., R. (2011). Analysis of Voltage Rise Effect on Distribution Network with Distributed
Generation, Preprints of the 18th IFAC World Congress, Milano, Italy.
[30] Ekwue, A., Cheng, D. T., Macqueen J. F., (1997) Artificial Intelligence Techniques in Power Systems, 1997, IET Digital Library
Book DOI: 10.1049/PBPO022E
[31] McArthur, S., D., J., Davidson, E., M., Catterson, V., M., Dimeas, A., L., Hatziargyriou, N., D., Ponci, F., Funabashi, T. (2007).
Multi-Agent Systems for Power Engineering Applications—Part I: Concepts, Approaches, and Technical Challenges, IEEE
Transactions on Power Systems, Vol.22, No.4, s.1743,1752.
[32] Nguyen, H., P. (2010). Multi-Agent System Based Active Distribution Networks. (Doctoral Thesis). Department of Electrical
Engineering, Eindhoven University of Technology, Eindhoven, Holland.
[33] Regulation on Quality of Service for Electricity Distribution and Retail Sales. Published in Official Gazette 21 December 2012 dated
(28504). <www.epdk.gov.tr>, date of access: 06.04.2018.

374
 NUMERICAL MODELING OF A NEW WIND FARM BASED ON
LOCAL MEASURED ENVIRONMENTAL PARAMETERS

Victorita Radulescu
University Politehnica of Bucharest, Bucharest, Romania, [email protected], ORCID: https://orcid.org/0000-0002-
8707-1914

Radulescu C. Victorita, Numerical modeling of a new wind farm based on local measured
Cite this paper as: environmental parameters, 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,
Turkey

Abstract: Since the Romanian solar and wind potential proved very attractive to foreign investors, a massive
campaign of monitoring the environmental data, in favorable areas, began in 2008. Areas as the
Danube meadow, Southern Moldova, and the Black Sea coast have been selected. After a few
general considerations, the measurements campaign from the Danube meadow, the place selected
for a new wind farm, is briefly mentioned. The monitoring poles of 80 m high, their components,
and a selection of the monitored environmental data over more than two years are considered the
starting point for further analysis. Based on the recorded values, a selection of the turbines models
possible to be installed in this area was established, associated with the future wind farm capacity.
The programs, methods, and used measuring equipment are presented. The vertical distribution of
wind velocity is strongly affected by ground roughness. Based on the recorded wind velocities, the
power curves of the aggregates correlated with the terrain roughness and efficiency coefficients of
wind turbines were determined. Based on the average recorded velocity, the energy possible to be
produced was estimated, correlated with the allowable operating range wind velocity. Finally,
some conclusions and references are also presented.
Keywords: Energy efficiency, Environmental monitoring, Numerical simulation, Wind farms
© 2020 Published by ECRES

1. INTRODUCTION

Energetic domain, especially based on the utilization of renewable resources, represents one of the main
objective strategic sectors for any country, regardless of its degree of development. Nowadays it is one of the
most dynamic domains due to the exponential increase of electric consumption, especially if we consider that,
in the not very far future, most of the new automobiles will be "clean", and charged with electric power. For
Romania and not only, the wind resources have a vital role in realizing this goal. At the end of 2008, the installed
capacity of wind power farms was 65 GW in the EU-27, producing 142 TWh hours of electricity and
representing 4.2% of the total EU electricity demand, [1]. In 2019, the electricity produced in wind power farms
was in the EU of about 9%, [2]. The power sector, especially the domain using renewable resources is one of
the strategic sectors for any country, regardless of the degree of development. Due to the present Romanian
economic and political context, a national objective is to secure energy for a "wider Europe", competitively
based on the Treaty signed by all EU countries, known as the Energy Community Treaty, [3]. A new objective
for the EU is to extend internal energy markets beyond the boundaries of the European Union, by creating a
Pan-European Energy Market.
Renewable resources as hydro, wind, and solar represent the most important Romanian energetic potential and
offer unlimited possibilities of utilization on the local and national energetic system, [4]. In Southeast Europe,
Romania is considered the country with the highest wind potential, with a capacity of more than 14000 MW
possible to be produced. In January 2019, based on over 8 billion euros investments in wind power farms have
been recorded 3025 MW produced energy by wind resource, representing around 30% of the total Romanian
produced energy, [5].
To project a future wind farm is essential to consider the accessibility at the emplacement, availability of the
electric network, wind potential of the region, and the acceptability of the total costs.

375
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The actual performance of the wind turbines extends the domain of the wind scale utilization, from the necessary
values of wind intensity of higher than 8 m/s, when the blades of the turbines started to rotate eight years ago,
to around 5 m/s nowadays. That represents an increase with more than 60% of the total wind available resources.
In Romania in present is developed a huge project at Cogealac, Constanta County, representing one of the biggest
projects developed by the Czech part, of around 600 MW. Based on this wind power farm, the existent hydro-
electric power plants, and new future wind farms that will be developed, Romania has answered at the main EU
target, of assuring a produced energy from renewable resources of 20% from the total produced energy, until
2020.

2. EXPERIMENTAL ENVIRONMENTAL DATA ACQUISITION

In the present context in Romania, referring to the development of new wind and solar power farms was held a
constant campaign of monitoring the environmental data, to locate the most economically favorable zones.
Figure1-a presents the distribution of the wind intensity at 50 m from the ground, reported by ANM, National
Authority of Meteorology. The environmental data acquisition, into the selected area, Figure 1-b, c was
accomplished for more than two years. The wind farm will be equipped with turbines having the hub at
approximately 100 m high. In these conditions, the masts for monitoring were selected as having 80 m high,
Figure 2-a, fully equipped with anemometers and wind vanes placed at 70 m, 50 m and 40 m from the ground,
and sensors for temperature, humidity, pressure, and temperature at 10 m high.

(a) (b) (c)

Figure 1. (a) Wind intensity map in Romania, (b), (c) Selected area for the future wind farm.

Figure 2-b has illustrated the system of anchoring on four directions, at 7 different levels, at 10 m from each
other, oriented after the cardinal points. At the top of the tower is placed a red lamp of illumination ON-OFF for
the night and a lightning conductor. A data-logger with a solar panel for registering the measured data is also
part of the tower Figure 2-c, d. The data-logger is equipped with a GSM data transmission system, but also with
the possibility of storage for up to 30 days.

376
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b) (c) (d)

Figure 2. (a) The monitoring mast, (b) The system of anchoring, (c), (d) Data-logger.

The Data-logger registers at each 10" (seconds) the wind speed and direction. Those values are mediated at 10'
(minutes), together with the standard deviation, air temperature, pressure, and humidity, into binary files. These
files are converted into ASCII text files with a specialized acquisition soft. With the mediated values at 10’ was
elaborated the wind rose, for each anemometer. Supplementary calculations referring at the influence of the pole
itself over the wind velocity distribution and its fluctuations, associated with the influences induced by each
anemometer to the others, over the vertical distribution of the horizontal component of the wind direction, must
be also considered. Table 3 and Table 4 present part of the recorded measurements for the wind intensity and
turbulence intensity, for a few months from 2018-2019. Table 5 synthesizes the maximum recorded values,
between June 2018 and February 2020. Figure 3 shows a comparison between the actual measured results and
the Weibull diagram, used in the wind farm modeling, as an example for an altitude of 60 m. It may be observed
that some very small differences appear, given that the Weibull diagram is a statistical analytical approximation,
based on the actual measured data.

Table 1. Daily average velocity

Month December 2018 January 2019 February 2019 Mar.19 Apr.10
Day 70m 50m 40m 70m 50m 40m 70m 50m 40m 70m 50m 40m 70m 50m 40m
01 16.4 11.8 10.1 12.1 9.96 5.94 12.7 6.90 6.15 13.4 9.29 8.27 12.2 7.73 6.46
02 12.9 8.46 7.06 17.7 13.8 12.2 11.5 7.48 6.41 12.3 8.40 7.25 10.8 6.88 5.65
03 11.7 7.47 6.15 16.5 14.2 10.9 12.9 8.77 7.45 13.4 9.37 8.16 12.1 7.76 6.53
.... .... .... .... .... .... .... .... .... .... .... .... .... .... .... ....
20 15.5 11.8 10.5 16.5 12.4 11.4 10.5 6.36 5.29 12.1 7.98 7.01 14.5 10.3 9.11
21 13.9 9.87 8.56 15.4 11.3 9.93 18.8 14.5 13.2 11.5 7.42 6.44 16.2 12.4 10.5
.... .... .... .... .... .... .... .... .... .... .... .... .... .... .... ....
Average 13.8 9.60 8.24 13.1 9.69 7.48 12.5 8.48 7.33 19.5 15.1 13.7 13.1 9.02 7.76

Table 2. Turbulence intensity

Int. N NNE NE ENE E ESE SE SSE S SSW SW WSW W WNW NW
.... .... .... .... .... .... .... .... .... .... .... .... .... .... .... ....
8 0.12 0.121 0.118 0.116 0.118 0.117 0.111 0.109 0.118 0.122 0.124 0.119 0.1 0.109 0.111
8.5 0.12 0.128 0.124 0.119 0.121 0.111 0.12 0.114 0.118 0.121 0.125 0.118 0.1 0.113 0.118
9 0.13 0.127 0.124 0.118 0.116 0.128 0.119 0.114 0.118 0.124 0.123 0.102 0.1 0.111 0.112
9.5 0.12 0.129 0.125 0.123 0.128 0.129 0.124 0.118 0.136 0.134 0.131 0.127 0.1 0.122 0.121
10 0.13 0.135 0.123 0.121 0.119 0.185 0.163 0.152 0.154 0.161 0.153 0.153 0.2 0.141 0.139
10.5 0.12 0.123 0.128 0.121 0.111 0.123 0.121 0.119 0.116 0.127 0.129 0.116 0.2 0.112 0.121
11 0.12 0.127 0.124 0.122 0.111 0.104 0.156 0.149 0.132 0.129 0.131 0.128 0.2 0.127 0.128
.... .... .... .... .... .... .... .... .... .... .... .... .... .... .... ....

377
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 3. Maximum values of the velocity intensity

Max. Velocity 10' (m/s)
Month 70 m 50 m 40 m
06.2018 18.17 17.88 16.63
… …. …. ….
04.2019 28.13 17.62 16.71
05.2019 45.8 19.39 19.13
06.2019 36.27 21.2 20.84
07.2019 36.17 19.37 18.6
08.2019 33.33 14.05 13.27
…. …. …. ….
02.2020 28.42 19.11 17.6
Figure 3. Comparison between real measurements and
the Weibull distribution

By numerical modeling, using the Matrix method, it can be seen and graphically check the speed estimation in
each direction, Figure 4-a. This forecast may show the correlation rate of the R coefficient, which was the basis
of estimation of the long-term wind speed. This forecast confirms as been true the estimate of 98.9% of trust for
the correlation of the data by the method of matrices and the method of least squares. This correlation is based
on wind speed (acceleration-deceleration) and its direction. Mediation corrections are introduced between long-
term values and concrete data measured in the area of interest. They represent changes to the modeling through
the common distribution that uses the elements of the wind speed and direction from the respective matrix.
Figure 4-b presents the frequency of the occurrence by velocities classes. The long-term values recorded in the
area are represented in green and blue the values recorded with the monitoring poles. Figure 4-c presents the
estimation of the energy production in each sector.

(a) (b) (c)

Figure 4. (a) Wind intensity, (b) Wind frequency, (c)Energy production by direction.

3. NUMERICAL MODELING OF THE WIND TURBINES

Based on the previous wind turbines used in Cogealac-Fantanele wind farm, for modeling the new wind park
farm were selected five types of wind turbines, all of them have almost the same power: VESTAS V100-1.8
MW, VESTAS V90- 2MW, GE 2.5xl, GE 2.75 MW, and VESTAS V112 - 3MW, Table 4. Their main
characteristics are presented in Table 4.
Wind turbines were selected based on some characteristics: the axis of the rotor should be at high enough level
from the ground as to minimize the influence of the terrain roughness, the average values of the wind velocity
at the level of the turbine hub to be in its optimal operation area, and to be competitive on the price market.

Table 4. Main characteristics of the selected wind turbines

Type Rotor Diameter [m] Hub high [m] Installed Power [kW] IEC
V100 100 105 1800 III
V90 90 105 2000 I+III
GE2.5x 100 100 2500 II+III
GE2.75 103 103 2750 II+III
V112 112 119 3000 II+III

378
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

For each type of turbine, depending on the height at which the nacelle is, the power curve was determined and
the efficiency in operation was established. In Figure 5 are presented the obtained coefficients of efficiency and
trust, as it follows: in Figure 5-a turbine GE 2.5xl with nacelle at 100 m, Figure 5-b turbine GE 2.75 with nacelle
at 100 m, Figure 5-c turbine V90 with nacelle at 80 m and 95 m, and in Figure 5-d turbine V100 with nacelle at
84 m, 94 m, and 119 m. These coefficients were estimated for each type of turbine used in modeling
but considered isolated. In the analysis of the wind farm, corrections will be made, taking into account the
mutual influence during the normal operation, due to the parasitic vortices and speed fluctuations.

(a) (b) (c) (d)

Figure 5. (a) Wind turbine GE 2.5xl, (b) Wind turbine GE 2.75, (c) Wind turbine V90, (d) Wind turbine V100.

For the analyzed surface, regarding the location of the future wind farm, topography, roughness, and dimensions
of the turbines, the allowable speed intervals have been determined. Figure 6 presents a selection of the obtained
results.

(a) (b) (c)

Figure 6. (a) Vestas V100, (b) GE2.75, (c)GE 2.5xl.

4. MODELING THE WIND FARM

For modeling the wind farm, the soil roughness was estimated and respected the criterion of the minimum
necessary distance between the wind turbines to minimize the reciprocal influence and to assure an optimum
circulation of the wind to all turbines. Figure 7-a, shows the imposed restricted area in conformity with the
international normative, for part of the wind park, and in Figure 7-b the entire park, with 27 turbines, estimated
to be realized in the first stage.
For each of the five selected turbines, the energy production was calculated based on the multi-annual
measurements of the wind intensity and direction. It was estimated the increase of energy due to the presence of
the hill but also the losses due to the ground roughness and parasitic currents, for each of the twelve directions,
N-NNE-ENE-E-ESE-SSE-S ... Table 5 partially shows the results obtained for the GE2.5xl MW turbine, where
BE-Base Energy, RE-Resultant Energy, *)- Increase of energy due to the hill presence, **)-losses. Figure 7-c
shows the turbulence by sectors and in Figure 7-d the main parameters at the nacelle level.

379
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b) (c) (d)

Figure 7. (a)-(d) Results of the wind farm modeling for turbine GE 2.5xl.

Table 4. Produced energy by sectors with turbines GE 2.5xl

Sector N NNE ENE E ESE SSE .. Total
BE [MWh] 37714,5 7374,15 5918,85 5836,5 6574,05 24221,7 .. 244861,2
*) [MWh] 656,1 265,95 180,9 120,15 71,1 -132,75 .. 2866,05
**) [MWh] -1129,05 -12,15 -47,7 -18 -99 -1505,25 .. -6890,4
RE [MWh] 37242 7627,95 6052,05 5938,6 6546,15 22583,7 .. 240836,9

Calculations were similarly made, for all selected turbines; Table 5 shows the obtained results for each turbine.

Table 5. Produced energy by sectors with turbines GE 2.5xl

Turbine D H P Pyear Pav/T Fc D E
- [m] [m] [MW] [MWh] [MWh] [%] [h] [%]
V90 90 105 54 211350.2 7827.8 25.7 30496.5 93.8
GE 2.5xl 100 100 67.5 240836.9 8919.9 23.2 34636.5 91.8
GE 2.75 103 98.3 74.25 261114.3 9670.9 22.3 34587 91.5
V112 112 119 81 292392 10829.3 28.1 34956 91.7
V100 100 105 48.6 226989.9 50442.2 31.1 34690.5 93.3
Here: D- Rotor diameter, H-nacelle height, P- the power of the wind farm, P year-Annual production, Pav/T-
Average production per turbine, Fc- capacity factor, D - Operating hours, and E-Efficiency.
Each turbine of the wind farm operates with certain efficiency, given by the relative position of the other
turbines, geographical position, air currents, disturbances, etc. In estimating the wind farm operation efficiency,
the energy production must be analyzed with different probabilities for each turbine, taking into account the
general uncertainty of the park. It should also be noted that, in close connection with the geographical conditions,
there are some favorable directions concerning energy production, [6], [7]. To estimate the wind park’s
efficiency, the energy production on each direction must be analyzed, associated with the influence of the hills
present in the area. The probability for each model of the turbine model is calculated.

5. DISCUSSIONS

In estimation the uncertainties and assessing the margins of error, some aspects must be taken into account:
- Mutual interference of the turbines, well known as the wake-effect which increases with the decrease of the
distance between turbines
- Incorrect positioning of the turbines related to the wind flow. In this case, there are unjustified losses due to
the parasitic currents in the wind farm. The air current is disturbed in the main direction of the predominant wind
and areas of disturbed current appear which can influence the correct operation of the turbines in the park.
- Statistical processing of the recorded data; even in the case of using high-performance software, it is also based
on statistical calculations, which in turn induce a margin of errors
- Evaluations on the theoretical and gross potential may be modified over time, affected by equipment
performance
- Uncertainties related to the current operation, management, energy utilization, etc.
To estimate the values of these uncertainties, a specific module of the WindPRO program was used, which
applies a mathematical model for establishing and describing the interference effect between the wake-effect of
the turbines. The values obtained are centralized in Table 6.

380
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 6. Uncertainty values

Uncertainty Value
Model wake-effect 1.28%
Un-proper positioning of the wind turbine 0.64%
Transfer in location 0.44%
Perturbed currents at the nacelle level 0.37%
Power curve of the turbine 1.08%
Human errors 0.52%
Errors due to the statistical calculation 0.23%
Total 5.16%

The uncertainty determined in the production estimation is about 5.6%. It represents a very low value for wind
farms, mainly due to the existence of long-term correlations. At this obtained value, a reduction of the estimated
production by 1.8% is recommended by all wind turbine manufacturers, which includes the unavailability of the
operation due to the frost, possible interruptions, etc.

ACKNOWLEDGMENT
I thank my colleagues from SC Hidroelectrica SA who funded the measurements and finally, co-funded the
realization of the wind power farm.

REFERENCES

[1] Wind-Energy -in-Europe-Scenarios-for-2030, September 2017

[2] EWEA, Europe's onshore and offshore wind energy potential, Technical report, 2019
[3] EWEA European Wind Energy Association, Wind in power: European statistics, 2018
[4] G. Dragomir, A. Șerban, G. Năstase, A. I.Brezeanu, Wind energy in Romania: A review from 2009 to 2016, Renewable
and Sustainable Energy Reviews, vol.64, pp. 129-143, 2016
[5] Catalin Lupoaie, CEZ face energie eoliana cat un reactor de la Cernavoda (Romanian), Ziarul Financiar, 2009
[1] [6] IEC 61400-12-1: Wind turbines, Measnet classification
[2] [7] IEC 61400-25-1, 2006-12, Wind turbine, Communications for monitoring and control of wind power plants - Overall
description of principles and models

381
 ENERGY AND EXERGY ANALYSIS OF SOLAR BASED
HYDROGEN PRODUCTION BY HELIOSTAT FIELD AND
PARABOLIC TROUGH COLLECTOR

Alper Karakoç
Karabük University, Department of Energy Systems Engineering, Karabük, Turkey, [email protected],
ORCID: 0000-0002-3692-5895

Yalım Gültekin
Başkent University, Anadolu OSB Vocational School, Ankara, Turkey, [email protected], ORCID: 0000-
0001-5868-0166

Birol Kılkış
Polar Technology, Hacettepe Teknokent, Ankara, Turkey, [email protected], ORCID: 0000-0003-2580-3910

Karakoç, A, Gültekin Y, Kılkış B. Energy and exergy analysis of solar based hydrogen
Cite this paper as: production by heliostat field and parabolic trough collector. 8th Eur. Conf. Ren. Energy Sys.
24-25 August 2020, Istanbul, Turkey

Abstract: Today, hydrogen production is often based on renewable energy sources like solar energy. In this
paper, a conceptual thermodynamic cycle regarding hydrogen production either with a heliostat
field or a parabolic trough collector system was examined and compared. Comparisons were
carried out concerning steam generation and Organic Rankine cycle for satisfying the electricity
demand of electrolyzer of the fuel cell. The impact of hydrogen generation methods was
investigated in two parts. In the first part, hydrogen generation methods were examined based on
energy and exergy analysis. The results showed that energy and exergy efficiency of the heliostat
field were 27.48% and 28.19% respectively and for parabolic trough collector systems, these
efficiencies were 29.71% and 30.48%, respectively. In the second part of the study, the amount of
hydrogen production from a heliostat field with an installed capacity of 100 MW and a parabolic
trough system of the same capacity were compared. According to the results obtained in this study,
the hydrogen production and indirect CO2 emission rate regarding the heliostat field and the
parabolic trough collector system were 109416.49 and 118295.62 standard liters per minute, 0.359
and 0.343 kg CO2/kW, respectively. Therefore, it has been concluded that the parabolic trough
collector system is a better technology for hydrogen production compared to a heliostat field under
given conditions.
Keywords: Hydrogen Production, Heliostat Field, Parabolic Trough Collector, Energy Analysis, Exergy
Analysis
© 2020 Published by ECRES

1. INTRODUCTION

Ever-increasing energy consumption and attributable CO2 emissions must be decoupled from the so-called
economic development such that truly sustainable development is achieved for assuring the deceleration of the
climate warming trend that the humanity is facing as the most important challenge. It is becoming increasingly
difficult to meet the energy demand among the technological achievements. In recent years the use of hydrogen
as a renewable and sustainable energy storage medium and energy carrier has become one of the focal points of
researchers. Studies are concentrating on membrane technology and hydrogen production methods and energy
resources. Devrim and Bilir [1], studied the performance parameters of the proton exchange membrane. In their
study, the hydrogen and oxygen equilibrium equations in the fuel cell were examined and a model was developed
in which the fuel cell was used with other energy sources. During fuel cell operation, electricity demand of the

382
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

electrolyzer was satisfied by wind turbines and photovoltaic arrays. In the study conducted by Atif et al., energy
and exergy analysis of solar tower plant with supercritical carbon dioxide cycle were completed [2]. According
to their study, carbon dioxide cycle is a candidate for the power block of heliostat tower systems. Parabolic
collectors are used as an alternative to heliostat towers to satisfy the energy demand of the electrolyzer. Marefati
et al. [3] worked on fuel cell and parabolic trough collector hybrid system.
The findings obtained in this study were evaluated according to the energy and exergy performance of the system
and they observed that parabolic collectors may be a rational source of power for the electrolyzer. The literature
review indicates that the comparison of hydrogen production methods by heliostat field and parabolic trough
collectors has not been reported till now in terms of energy, exergy and indirect emission analysis. In this regard,
the objective of this study is to construct an energy and exergy framework for overall hydrogen production
system. In this way, the losses and causes of the system were determined by energy and exergy analysis and the
necessary optimizations were made in this modeling. The above-mentioned modelings are shown in Figures 1
and 2 respectively.

Figure 1. Fuel cell operation with heliostat tower

Figure 2. Fuel cell operation with parabolic trough collector

2. MATHEMATICAL MODELLING

Conventional thermodynamically analysis models are generally limited to the first law of thermodynamics.
Energy analysis consists on first law of thermodynamics. With energy analysis, the various losses that occur for
each unit of a power plant cannot be qualitatively assessed. Within this regard, to make more accurate modeling
and comparative studies the second law of thermodynamics applies too in order to determine inefficiencies and
losses. In order to meet such a need, a two-stage energy and exergy model was developed. In the first model,
the energy demand of PEM was met by using the heliostat tower and in the second model; the electricity
requirement of PEM is supplied by parabolic collectors.
First, the principle of conservation of energy is given below:
𝑄̇ − 𝑊̇ = 𝑚̇ (ℎ𝑜 − ℎ𝑖 ) (1)
In this equation, 𝑄̇ (kW) is the heat transfer rate to the system and 𝑊̇ (kW) is the work generated from the
system. ℎ𝑜 and ℎ𝑖 are inlet and exit enthalpies of the fluid respectively. Similar to the first law, exergy balance
is written as follows;
∑ 𝐸̇𝑥,𝑖 + ∑ 𝐸̇𝑥 𝑄̇ = ∑ 𝐸̇𝑥,𝑜 + 𝑊̇ + 𝐸̇𝑥,𝑑 (2)
Here, exergy inputs, outputs and destructions are denoted by 𝐸̇𝑥,𝑖 , 𝐸̇𝑥,𝑜 and 𝐸̇𝑥,𝑑 , respectively. The performance
of parabolic trough solar collectors centered on the receiver of the collectors. The energy equilibrium on the
receiver will give necessary data for heat gain. The net heat gain in parabolic trough collector was evaluated as
follows;
𝑄̇𝑛𝑒𝑡 = 𝑄̇𝑢 = 𝑄̇𝑠 − 𝑄̇𝑂𝑝𝑡 − 𝑄̇𝐶𝑜𝑛𝑣 − 𝑄̇𝑅𝑎𝑑 (3)
In this equation, net heat gain is found by subtracting optical, convective and radiation losses from the amount
of heat from the sun. The net amount of heat transferred to the heat transfer fluid is obtained by the calculation
of the net heat obtained from the solar energy in the system. In this way, the amount of heat transferred to the
power block through the heat transfer fluid and the amount of electricity generation can be determined. The
amount of heat transferred to the heat transfer fluid has been calculated as follows [4];
𝑄̇𝐻𝑇𝐹 = 𝑄̇𝑢 = 𝑚̇𝐻𝑇𝐹 (ℎ𝑜𝑢𝑡 − ℎ𝑖𝑛 ) = 𝑚̇𝐻𝑇𝐹 Δ𝑐𝑝 Δ 𝑇 (4)
In this formula, ℎ𝑖𝑛 and ℎ𝑜𝑢𝑡 refers inlet and leaving enthalpy of the heat transfer fluid in the receiver. Efficiency
analysis in the system consists of two main steps. In the first step, the second law efficiencies of parabolic

383
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

collectors and the heliostat system were calculated separately. The equations used in the efficiency calculations
are shown below;
𝑊̇𝑛𝑒𝑡
𝜂𝑂𝑣𝑒𝑟𝑎𝑙𝑙 (𝐼𝐼) = (5)
𝐸̇ 𝑥𝑆
Here, the exergy efficiency of the system, including the power block, is found by the ratio of the net amount of
work to the exergy of the solar thermal radiation absorbed by the receiver. In this way, parabolic collectors and
heliostat tower can be compared according to the second law of thermodynamics. Hydrogen production from
solar energy source while the power obtained from solar energy systems electrolyzer is transferred, and thus
could lead to the production of hydrogen. The amount of hydrogen produced can be found by the following
equation;
𝐼 𝐼 𝑃𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑙𝑦𝑧𝑒𝑟
𝑁𝐻2 = 𝑛𝑐𝑒𝑙𝑙 = (6)
2∗𝐹 2 ∗ 𝐹 𝑃𝑐𝑒𝑙𝑙 𝑝𝑜𝑤𝑒𝑟
In this equation, 𝐹 is Faraday constant (96485 s A /mol), 𝑛𝑐𝑒𝑙𝑙 is the number of cells in the fuel cell, 𝑃𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑙𝑦𝑧𝑒𝑟
and 𝑃𝑐𝑒𝑙𝑙 𝑝𝑜𝑤𝑒𝑟 are electrolyzer and cell power respectively.
In the energy economy, any power plant destroys exergy during the production of electricity. This exergy
destruction has to be compensated by other power plants. The mentioned additional activity has an unavoidable,
indirect CO2 emission. This indirect CO2 emission is expressed as follows [5];
∆𝐶𝑂2 = 𝜀𝑓 𝑥 Ė𝑥,𝑑𝑒𝑠𝑡𝑟𝑜𝑦𝑒𝑑 (7)
Here, 𝜀𝑓 is an emission factor and it is included in the calculations as 0.27, which is based on natural gas fuel
operated-thermal plant for heat generation.

3. RESULTS AND DISCUSSION

In this study, energy and exergy analysis of solar and based hydrogen production with the heliostat field and
parabolic trough collector were performed. In addition to energy and exergy efficiencies, indirect emission
values of the systems were also taken into account during the evaluation of the findings. The results of energy
and exergy analysis of the solar systems, power blocks, and overall systems are provided in Table 1 and Table
2. With the use of the energy analysis, it is found that the installed power of the 100 MW heliostat field has first
law efficiency of 27.48%. On the other hand, the same installed power for the parabolic trough collector system
first law efficiency of 29.71%. The second law analysis for these systems shows that parabolic trough collector
system second law efficiency 30.41% outruns the heliostat field which has the second efficiency of 28.19%.

Table 1. Energy and exergy results of power generation from the heliostat field
Subsystem Energy Analysis Exergy Analysis
Received (MW) Loss Efficiency (%) Received (MW) Loss Efficiency
(MW) (MW) (%)
Heliostat field 100 25 75 176.98 44.27 75
Power block 67.5 58.88 12.77 66.11 49.51 25.12
Overall 100 72.52 27.48 176.98 127.09 28.19
Table 2. Energy and exergy analysis of power generation from parabolic trough collector
Subsystem Energy Analysis Exergy Analysis
Received (MW) Loss Efficiency (%) Received (MW) Loss Efficiency
(MW) (MW) (%)
PTCS 100 20 80 191.35 38.27 80
Power block 72 63.17 12.77 71.47 53.51 25.12
Overall 100 70.29 29.71 191.35 133.03 30.48
In the second main phase of the study, the amount of hydrogen production SLPM (standard liter per minute) and indirect
CO2 emission amounts were calculated. The findings are listed in Table 3.

Table 3. Hydrogen production and indirect emission analysis of power generation from solar systems
Solar System Heliostat field PTSC
Hydrogen Production (SLPM) 109416.49 118295.62
Indirect CO2 Emission (kg CO2/kW) 0.343 0.359

384
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. CONCLUSION

In this study, energy and exergy analysis of solar and based hydrogen production with heliostat field and
parabolic trough collector was completed as the first main part of the study. The results show that parabolic
trough collector system has better performance in terms of energy and exergy efficiency. In the second part of
the study, the amount of hydrogen production and indirect CO 2 emission amount of the evaluated methods of
hydrogen production were compared. The emission analysis revealed that direct and indirect emission amounts
are lower in parabolic systems. According to the results obtained, the production of hydrogen by parabolic
collectors is more advantageous than the heliostat system concerning the amount of hydrogen produced and the
emissions released in addition to the first and second laws of thermodynamics.

REFERENCES

[1] Atif, M, AlSulaiman, FA. Energy and exergy analyses of solar tower power plant driven supercritical carbon dioxide
recompression cycles for six different locations. Renewable and Sustainable Energy Reviews 2017; 68:153-167
<https://doi.org/10.1016/j.rser.2016.09.122>
[2] Devrim, Y, Bilir, L. Performance investigation of a wind turbine–solar photovoltaic panels–fuel cell hybrid system
installed at İncek region – Ankara, Turkey. Energy Conversion and Management 2016; 126: 759-766
<https://doi.org/10.1016/j.enconman.2016.08.062>
[3] Marefati, M, Mehrpooya, M, Shafii, MB. A hybrid molten carbonate fuel cell and parabolic trough solar collector,
combined heating and power plant with carbon dioxide capturing process. Energy Conversion and Management 2019;
183: 193-209 <https://doi.org/10.1016/j.enconman.2019.01.002>
[4] AlZahrani, AA, Dincer, I. Energy and exergy analyses of a parabolic trough solar power plant using carbon dioxide
power cycle. Energy Conversion and Management 2018; 158: 476-488
<https://doi.org/10.1016/j.enconman.2017.12.071>
[5] Kilkis, B. Exergy: game changer or game maker. In: CLIMA 13th REHVA World Conference Proceedings, 26-29 May
2019: Paper No: 467, Session 4F, pp.1-3 < https://doi.org/10.1051/e3sconf/2019111060 201 E3S 111 CLIMA 9 6024
24>

385
 PRELIMINARY EXPERIMENTAL CALIBRATION AND
VALIDATION OF A DYNAMIC SIMULATION MODEL FOR
FAULT DETECTION AND DIAGNOSIS OF AIR-HANDLING
UNITS

Antonio Rosato
Department of Architecture and Industrial Design, University of Campania Luigi Vanvitelli, Aversa, Italy,
[email protected], ORCID: 0000-0001-5482-3529

Francesco Guarino
Department of Architecture and Industrial Design, University of Campania Luigi Vanvitelli, Aversa, Italy,
[email protected], ORCID: 0000-0003-3655-7065

Vincenzo Filomena
C.I.R.A. (Italian Aerospace Research Centre), Capua, Italy, [email protected], ORCID: 0000-0003-4163-9858

Sergio Sibilio
Department of Architecture and Industrial Design, University of Campania Luigi Vanvitelli, Aversa, Italy,
[email protected], ORCID: 0000-0002-8750-5401

Luigi Maffei
Department of Architecture and Industrial Design, University of Campania Luigi Vanvitelli, Aversa, Italy,
[email protected], ORCID: 0000-0003-4130-5065

Rosato, A, Guarino, F, Filomena, V, Sibilio, S, Maffei, L. Preliminary Experimental

Calibration and Validation of a Dynamic Simulation Model for Fault Detection and
Cite this paper as:
Diagnosis of Air-Handling Units. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: Automated Fault Detection and Diagnostics (FDD) could provide a cornerstone for predictive
maintenance of Air-Handling Units (AHUs). An innovative multi-sensorial laboratory, called
SENS-i Lab, has been set-up at the Department of Architecture and Industrial Design of the
University of Campania Luigi Vanvitelli (Italy). The laboratory is equipped with an AHU (nominal
cooling/heating capacity of 5.0/5.0 kW) aiming to control the thermo-hygrometric comfort inside
a 4.0x4.0x3.6 m test room; the AHU is fully instrumented in order to monitor and control its
operation. Several experiments for assessing the performance of the AHU have been carried out;
a detailed dynamic simulation model has been developed by means of the software TRNSYS and
then calibrated and validated with respect to the measured data. The results demonstrated a very
good agreement between the predicted outputs and the experimental observations, highlighting the
suitability of the developed model to be used in combination with FFD methods for implementing
HVAC predictive maintenance programs.
Keywords: Air-handling units; Model calibration and validation; Fault detection and diagnosis; Predictive
Maintenance
© 2020 Published by ECRES

1. INTRODUCTION

HVAC (Heating, Ventilation and Air-Conditioning) systems consume nearly 40% of the world energy demand.
Fault Detection and Diagnosis (FDD) [1, 2] methods can monitor the operation of HVAC components as well
as detect and predict the presence of the defects (deviations from normal or expected operation) causing a faulty

386
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

operation. Ideally, FDD systems could also resolve (diagnose) the type of problem and/or identify its location,
giving instructions for undertaking corrective actions. Efficient FDD methods could detect faults before the
building occupants notice the effects and they would reduce the repair and maintenance costs of the plants. In
addition, these methods could provide manufacturers and dealers with feedback about the design and sales of
systems in order to identify where any improvements could be made. Moreover, improving the operation of
HVAC systems could significantly reduce energy consumption and related emissions. FDD methods are based
on the use of accurate instruments measuring key-parameters, the variation of which in comparison to a
“nominal/healthy” trend is assumed as a symptom of defects; once the test quantity reaches some predetermined
levels that reflects the seriousness of the defect requiring corrective actions to take place, the test quantity is set
into an alarm state (symptom) and reasoning is started to find the cause of the “alarm - symptom – fault” chain.
In a survey of United Kingdom buildings, the data showed how 25-50% of energy waste was due to faults in the
HVAC units; this range could be reduced below 15% in the case of the faults being detected and identified early
in the premature stage before unacceptable damage occur [3]. FDD techniques have been used for decades in
aerospace, nuclear and industrial sectors, and their use in building operation and control applications is becoming
more widespread [1-3]. In order to apply FDD methods to HVAC units or components, it is necessary to compare
real behavior of the systems to the “nominal/healthy” operation without faults that can be modeled by means of
simulation programs and/or artificial intelligence techniques [3]. Therefore, models able to accurately simulate
the HVAC behavior have to be developed and used in combination with FFD methods for implementing
predictive maintenance programs. In this paper, four experimental tests for assessing the performance of the
AHU have been carried out; a detailed dynamic simulation model has been developed by means of the software
TRNSYS and validated by contrasting the predictions with the experimental observations.

2. DESCRIPTION OF THE LABORATORY

The SENS i-Lab is an innovative, multi-sensorial and multi-purpose laboratory, located at the Department of
Architecture and Industrial Design of the University of Campania Luigi Vanvitelli (Aversa, Italy, latitude:
40°58′21″ N, longitude: 14°12′26″ E). The laboratory consists of an Integrated Test Room, allowing to carry
out, in vivo or in virtual, subjective tests where the human experience of urban / rural or industrial environments,
architectures and products, can be measured. The lab is served by an Air-Handling Unit (AHU) able to control
the indoor air temperature, relative humidity, velocity and quality inside the Integrated Test Room. The room is
characterized by a floor area of 16 m2 with a height of 3.6 m. It is composed of four internal vertical walls, a
horizontal ceiling as well as a horizontal floor. Figure 1 reports the schematic of the AHU serving the Integrated
Test Room.

RAF: Return air fan

SAF: Supply air fan
HUM: Humidifier
PreHC: Pre-heating coil
CC: Cooling coil
PostHC: Post-heating coil
RAFil: Return air filter
SAFil: Supply air filter
OAFil: Outside air filter
HRS: Cross flow static heat recovery system
DOA: Outside air damper
DEA: Exhaust air damper
DHRS: Dumper of heat recovery system
DRA: Return air damper
VPreHC: Valve regulating the flow entering PreHC
VCC: Valve regulating the flow entering CC
VPostHC: Valve regulating the flow entering PostHC
VHUM: Valve regulating the flow entering HUM
RS: Refrigerating system
HP: Heap Pump

Figure 1. Schematic representation of the AHU serving the integrated test room.

The AHU is composed of the following main components: supply air fan (SAF), return air fan (RAF), pre-
heating coil (PreHC), cooling coil (CC), steam humidifier (HUM), post-heating coil (PostHC), cross flow static
heat recovery system (HRS), vapor-compression air-to-water electric refrigerating unit (supplying the CC),
vapor-compression air-to-water electric heat pump (supplying both the PreHC and PostHC), valves (VPreHC,

387
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

VPostHC, VCC, VHUM) regulating the heat carrier fluid flow rate entering, respectively, the PreHC, PostHC, CC and
HUM, outside air damper (D OA), return air damper (D RA), exhaust air damper (DEA), damper of heat recovery
system (DHRS), outside air filter (OAFil), return air filter (RAFil), supply air filter (SAFil). Two 0.08x0.18 m
return air intake vents are installed on the south-oriented wall and two 0.08x0.18 m return air intake vents are
installed on the north-oriented wall; a 0.60x0.60 m square ceiling swirl diffuser is mounted on the ceiling of the
room. Table 1 describes the main characteristics of the main AHU components. The heat carrier fluid is a mixture
of water and ethylene glycol (90/10% by volume).

Table 1. Characteristics of main AHU components

Nominal heating capacity (kW) 4.1
Pre-heating coil (PreHC) Nominal heat carrier fluid / air flow rate
0.710 / 600
(m3/h)
Cooling coil (CC) Nominal cooling capacity (kW) 5.0
Nominal heat carrier fluid/air flow rate (m3/h) 0.860 / 600
Humidifier (HUM) Steam capacity (kg/h) / Nominal power (kW) 5.0 / 3.75
Nominal heating capacity (kW) 5.0
Post-heating coil (PreHC)
Nominal heat carrier fluid/air flow rate (m3/h) 0.860 / 600
Vapor compression electric Heat Pump (HP) - Nominal heating / cooling capacity (kW) 14.0 / 13.4
model ANL 050HQ [4] / Refrigerating System Nominal input power (kW) 4.75 / 4.48
(RS) – model ANL 050Q [4] Nominal fluid flow rate (m3/h) 2.41 / 2.31

The hot heat carrier fluid (with a highest temperature up to 45°C) supplying both the pre-heating coil as well as
the post-heating coil is obtained thanks to the operation of the heat pump, while the refrigerating system is used
to provide the cold heat carrier fluid (up to a lowest temperature of 7 °C) flowing inside the cooling coil. Both
the refrigerating unit and the heat pump (located outside) are equipped with a 75 liters thermal energy tank where
the heat carrier fluid is stored. According to the manufacturer data [4], the Coefficient of Performance COP (i.e.
the ratio between the thermal output and the power input) ranges between 2.11 and 4.06 for a supply fluid
temperature of 45 °C, while the Energy Efficiency Ratio EER (i.e. the ratio between the cooling output and the
power input) of the refrigerating system is in the range 2.53÷5.73 for a supply fluid temperature of 7 °C. The
AHU is fully instrumented in order to monitor, control and record the main operating parameters of the system.
The main characteristics of the sensors are reported in Table 2.

Table 2. Characteristics of the sensors used for the AHU monitoring

Model Monitored parameter Measuring range Accuracy
Siemens QFM2160 [5] Return air temperature TRA 0 ÷ 50 °C ± 0.8 K
Return air relative humidity RHRA 0 ÷ 100 % ±3%
Siemens QFM2160 [5] Supply air temperature TSA 0 ÷ 50 °C ± 0.8 K
Supply air relative humidity RHSA 0 ÷ 100 % ±3%
Siemens QAM2161.040 [5] External air temperature TOA - 50 ÷ 50 °C ± 0.75 K
Siemens QAM2161.040 [5] Air temperature at outlet of the CC TA,out,CC - 50 ÷ 50 °C ± 0.75 K
TSI 7575, 982 IAQ [6] Air temperature around the room TBEA - 10 ÷ 60 °C ± 0.50 K
Air relative humidity around the room RHBEA 5 ÷ 95 % ±3%

The end-users can manually set: the desired targets of both the indoor air temperature (TSP,Room) and relative
humidity (RHSP,Room) to be achieved inside the test room, the deadbands DBT and DBRH for both TSP,Room and
RHSP,Room, respectively, the velocity of both the return air fan (OLRAF) and the supply air fan (OLSAF), the opening
percentages of the return air damper (OPDRA), the outside air damper (OPDOA), the exhaust air damper (OPDEA)
and the static heat recovery system damper (OPDHRS). The air flow rate moved by the supply air fan can be varied
between 0 (OLSAF=0%) and 4800 m3/h (OLSAF=100%), while the air flow rate of the return air fan is in the range
from 0 (OLRAF=0%) to 2050 m3/h (OLRAF=100%). The parameters OPDRA, OPDOA and OPDEA can be varied in
the range 0÷100%, where 100% means that the dampers are fully open. The parameter OPDHRS can be set to
100% (the heat recovery does not occur) or 0% (the heat recovery from return air flow takes place). A specific
control logic has been developed by the manufacturer in order to control the operation of the system and achieve
the desired targets. Table 3 describes the conditions controlling the activation and deactivation of the main
components of the AHU. Even if the AHU is equipped with a PreHC, this component is never used during
normal operation (always de-activated). The operation of the humidifier as well as the flow rate of the heat
carrier fluid entering the post-heating coil or the cooling coil can be continuously adjusted between 0% and

388
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

100%. In particular, the heat carrier fluid flow rate can ben varied between 0 (OPV_CC/OPV_PostHC=0%) and 0.860
m3/h (OPV_CC/OPV_PostHC=100%), while the steam mass flow rate can be modulated from 0 (OP V_HUM=0%) up to
5 kg/h (OPV_HUM=100%). The signals managing the operation of the valves are generated by PID (proportional-
integral-derivative) controllers. The operation of both the refrigerating unit and heat pump is controlled in order
to maintain the desired temperatures TCT and THT (with a deadband of 1 °C) inside the cold and hot tanks,
respectively.
Table 3. Control strategies of the AHU
ON OFF
Humidifier RHRA ≤ (RHSP,Room - DBRH) RHRA ≥ (RHSP,Room + DBRH)
Cooling coil TRA ≥ (TSP,Room + DBT) OR TRA ≤ (TSP,Room - DBT) AND
RHRA ≥ (RHSP,Room + DBRH) RHRA ≤ (RHSP,Room - DBRH)
Post-heating coil TRA ≤ (TSP,Room - DBT) TRA ≥ (TSP,Room + DBT)
Vapor compression electric heat pump THT < (45 °C - 1°C) THT ≥ (45 °C + 1 °C)
Vapor compression electric refrigerating system TCT > (7 °C + 1°C) TCT ≤ (7 °C – 1 °C)

3. EXPERIMENTAL TESTS

Four experiments have been carried out to investigate the AHU behavior during steady-state and transient
operations. Table 4 describes the operating conditions of the tests in terms of target indoor air temperature
TSP,Room, target indoor air relative humidity RHSP,Room, initial air temperature TRoom_initial and relative humidity
RHRoom_initial in the test room.

Table 4. Boundary conditions of experimental tests

Test number 1 2 3 4
TSP,Room (° C) 22 → 24 → 26 26 → 24 → 22 22 28
TRoom_initial (° C) 19.8 28.1 22.0 28.1
RHSP,Room (%) 50.0 50.0 60 → 70 55 → 45
RHRoom_initial (%) 52.5 51.1 51.4 64.8

The experiments were carried out by measuring the parameters indicated in Table 2 every minute and
maintaining constant the following conditions: DBT=1°C, DBRH=5%, OLRAF=50%, OLSAF=50%, OPDEA=20%,
OPDOA=20%, OPDRA=100% (outdoor fresh air and return air are mixed) and OPDHRS=100%. Figure 2 reports the
experimental values of T RA, TSA, TOA, TA,out,CC, TBEA, RHRA, RHSA, RHBEA measured during the tests described
in Table 4 as a function of the time. In addition, this figure indicates the values of T SP,Room and RHSP,Room. The
figures highlight that the AHU is able to maintain the desired temperature and relative humidity inside the test
room. It could be underlined that the percentage of time during which indoor air temperature is within the given
deadband (  1°C) is 70.32%, 78.27%, 81.40% and 100 %, respectively, for the tests 1, 2, 3, 4; the percentage
of time during which indoor air relative humidity is within the given deadband (  5%) is 91.57%, 92.20%,
65.12% and 79.31% during the tests 1, 2, 3, 4, respectively.

389
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

TRA
TRA T
TSA
SA T
TOA
OA T
TA,out,
A,out,CC T
T_BEA
BEA
T
TRA
RA T
TSA
SA T
TOA
OA T
TA,out,
A,out,CC T
T_BEA
BEA
T
TSP,Room
SP,Room RHRA
RHRA RH
RHSA
SA RHBEA
RH_BEA RH
RHSP,Room
SP,Room
T
TSP,Room
SP,Room RH
RHRA
RA RHSA
RHSA RH
RH_BEA
BEA RH
RHSP,Room
SP,Room
40 100 40 100
a) b)
37 90 37 90

34 80 34 80

Relative humidity (%)

Relative humidity (%)

31 70 31 70
Temperature ( C)

Temperature ( C)
28 60 28 60

25 50 25 50

22 40 22 40

19 30 19 30

16 20 16 20

13 10 13 10

10 0 10 0
29-May-9.30 29-May-10.33 29-May-11.36 29-May-12.39 29-May-13.42 29-May-14.45 29-May-15.48 04-Jun-8.52 04-Jun-10.04 04-Jun-11.16 04-Jun-12.28 04-Jun-13.40 04-Jun-14.52
T
TRA
RA T
TSA
SA TOA
TOA T
TA,out,
A,out,CC T
T_BEA
BEA T
TRA
RA TSA
TSA T
TOA
OA T
TA,out,
A,out,CC T
T_BEA
BEA
T
TSP,Room
SP,Room RHRA
RH RA RH
RHSA
SA RHBEA
RH_BEA RH
RHSP,Room
SP,Room TTSP,Room
SP,Room RH
RHRA
RA RH
RHSA
SA RH_BEA
RHBEA RH
RHSP,Room
SP,Room
40 100 40 100
c) d)
37 90 37 90

34 80 34 80

Relative humidity (%)

Relative humidity (%)

31 70 31 70
Temperature ( C)

Temperature ( C)
28 60 28 60

25 50 25 50

22 40 22 40

19 30 19 30

16 20 16 20

13 10 13 10

10 0 10 0
04-Jun-15.15 04-Jun-15.21 04-Jun-15.27 04-Jun-15.33 04-Jun-15.40 04-Jun-15.46 04-Jun-15.52 04-Jun-15.58 05-Jun-11.34 05-Jun-11.42 05-Jun-11.49 05-Jun-11.56 05-Jun-12.03 05-Jun-12.10 05-Jun-12.18 05-Jun-12.25 05-Jun-12.32

Figure 2. Experimental data measured during the test 1 (a), test 2 (b), test 3 (c) and test 4 (d).

4. SIMULATION MODELS

The software of choice for this current study is the TRaNsient SYStems (TRNSYS) software platform 17 [7].
In this program, sub-systems are modelled as individual components (called Types) which can be cross-linked
to one-another.
Detailed models have been applied to take into account for: (i) cooling/thermal loads, (ii) the part load
characteristics of components, (iii) interaction between loads and AHU outputs, and (iv) system management
and control. Table 5 lists the main TRNSYS Types used for modeling the system components. They have been
selected from the TRNSYS libraries and enhanced by manufactures performance data or information available
in current scientific literature. Type 56 models the thermal behavior of the room; types 753e and 508c use the
bypass fraction approach to solve for the outlet air and fluid conditions; type 941 allows to simulate the
performance of RS and HP by taking into account the effects of both outside and supply air temperatures
according to the manufacturer data; a specific weather data file (EnergyPlus) has been considered for modelling
the outside climatic conditions (type 15).

Table 5. TRNSYS Types used in the simulation model

HP
Test 3-way Climatic
Component PostHC CC HUM / Fans Dampers Controls PID
Room valves conditions
RS
Type 56 753e 508c 641 941 930 647 648 2 23 15

5. MODEL VALIDATION

The model of the AHU developed in TRNSYS environment has been validated by contrasting the simulation
results with the experimental data described in the previous section. The whole experimental database consists
of 869 points. The simulations have been performed by assuming the following inputs equal to the measured
data: desired targets of indoor air temperature (T SP,Room) and indoor air relative humidity (RHSP,Room), velocity of
return air fan (OLRAF), supply air fan (OLSAF), opening percentages of return air damper (OPDRA), outside air
damper (OPDOA), exhaust air damper (OPDEA), heat recovery system damper (OPDHRS), external air temperature
(TOA), deadbands (DBT and DBRH) of targets of both indoor air temperature and relative humidity. The
simulations have been carried out with a time-step equal to 1 minute (according to the measurement frequency).
Figures 3a, 3b, 3c, 3d compare the predicted and experimental outputs in terms of return air temperature
(corresponding to the temperature inside the test room) and return air relative humidity (corresponding to the
relative humidity inside the test room) for the tests 1, 2, 3, 4 (described in Table 4 and Figure 2), respectively.

390
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

30 90 30 90
a) TRA…….
TRA, exp TRA, pred
T_RA RHRA….......
RHRA, exp RH_RA
RH b) T
TRA….000
RA,exp T
T_RA,pred
RA,pred RH
RHRA000000
RA,exp RH
RH_RA,pred
RA,pred
RA, pred
29 85 29 85

28 80 28 80

27 75 27 75

Relative humidity (%)

26 70 26 70

Relative humidity (%)

Temperature ( C)
Temperature ( C)

25 65 25 65

24 60 24 60

23 55 23 55

22 50 22 50

21 45 21 45

20 40 20 40

19 35 19 35

18 30 18 30
29-May - 9.30 29-May - 10.27 29-May - 11.25 29-May - 12.23 29-May - 13.20 29-May - 14.18 29-May - 15.15 29-May - 16.13 04-Jun - 8.52 04-Jun - 10.04 04-Jun - 11.16 04-Jun - 12.28 04-Jun - 13.40 04-Jun - 14.52
30 90 30 90
T T RH RH d) T
TRA T
T_RA RH
RHRA00000 RH
RH_RA
29
c) TRA
RA,exp T_RA
RA,pred RHRA
RA,exp RHRA
RA,pred
85 29
RA,exp RA,pred RA,exp RA,pred
85

28 80 28 80

27 75 27 75

Relative humidity (%)

26 70 26 70

Relative humidity (%)

Temperature ( C)

Temperature ( C)
25 65 25 65

24 60 24 60

23 55 23 55

22 50 22 50

21 45 21 45

20 40 20 40

19 35 19 35

18 30 18 30
04-Jun - 15.15 04-Jun - 15.22 04-Jun - 15.29 04-Jun - 15.36 04-Jun - 15.42 04-Jun - 15.49 04-Jun - 15.56 05-Jun - 11.34 05-Jun - 11.42 05-Jun - 11.49 05-Jun - 11.56 05-Jun - 12.03 05-Jun - 12.10 05-Jun - 12.18 05-Jun - 12.25 05-Jun - 12.32

Figure 3. Comparison of the experimental results and the predicted data: (a) test 1, (b) test 2, (c) test 3, (d) test 4.

The experimental results have been compared with the simulation outputs to assess the accuracy of the
calibration by using the following metrics quantifying the instantaneous differences: the average error ̅;
ε the
average absolute error |ε̅|; the root mean square error εRMS. These parameters are defined as follows:

εi = g pred,i − g exp,i (1)

ε̅ = ∑ εi /N (2)
i=1

|ε̅| = ∑|εi |/N (3)

i=1

(εi −ε̅)2
εRMS = √∑N
i=1
(4)
N

where gpred,i and gexp,i are, respectively, the predicted and measured values at time step i and N is the number of
points. Table 6 summarizes the values of 𝛆̅|𝛆̅| RMS.

Table 6. Errors between the model predictions and measurements

Test Number of experimental points Parameter 𝛆̅ |𝛆̅| 𝛆𝐑𝐌𝐒
Test 1 409 TRA -0.11 °C 0.32 °C 0.39 °C
RHRA -1.64 % 3.07 % 3.68 %
Test 2 359 TRA -0.07 °C 0.26 °C 0.30 °C
RHRA -3.12 % 4.06 % 3.84 %
Test 3 43 TRA 0.62 °C 0.72 °C 0.58 °C
RHRA -0.95 % 1.39 % 1.39 %
Test 4 58 TRA 0.32 °C 0.33 °C 0.23 °C
RHRA 1.05 % 3.11 % 3.96 %
Whole database 869 TRA -0.03 °C 0.31 °C 0.40 °C
RHRA -2.04 % 3.40 % 3.85 %

The results reported in Table 6 highlight that the maximum instantaneous errors are about 7% and about 26%,
respectively, for T RA and RHRA; these deviations are mostly related to a few points occurring during transient
operation caused by a change of the desired targets. With reference to the whole database, the values of
𝛆̅|𝛆̅| RMS are equal to -0.03 °C, 0.31 °C and 0.40 °C, respectively, for TRA and equal to -2.04%, 3.40% and

391
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3.85%, respectively, for RHRA. The lowest values of RMS in terms of TRA and RHRA correspond, respectively, to
the tests 4 and 3; the largest values of RMS in terms of TRA and RHRA are associated, respectively, to the tests 3
and 4. The deviations between measured and simulated data are fully coherent with the accuracy of the
instruments, demonstrating that predicted outputs agree very well with experimental observations. Therefore,
even if in the future it would be desirable to extend this work by considering wider ranges of operating
conditions, it can be stated that the model gives an accurate representation of the dynamic and steady-state
performance and it can also be usefully adopted in combination with FFD methods for the detection of any non-
optimal states of HVAC systems under predictive maintenance programs.

ACKNOWLEDGMENT
This work was undertaken as part of the program “PON FSE-FESR Ricerca e Innovazione 2014-2020” of the
Italian Ministry of Education, University and Research, Action I.1 “Dottorati Innovativi con caratterizzazione
industriale”.

REFERENCES
[1] Katipamula, S, Brambley, M. Review article: methods for fault detection, diagnostics, and prognostics for building
systems - a review, part I, HVAC&R 2005, 11, 3-25. https://doi.org/10.1080/10789669.2005.10391123.
[2] Katipamula, S, Brambley, M. Review article: methods for fault detection, diagnostics, and prognostics for building
systems - a review, part II, HVAC&R 2005, 11, 169-187. https://doi.org/10.1080/10789669.2005.10391133.
[3] Yu, Y, Woradechjumroen, D, Yu, D. A review of fault detection and diagnosis methodologies on air-handling units,
Energy and Buildings 2014, 82, 550-562. https://doi.org/10.1016/J.ENBUILD.2014.06.042.
[4] AERMEC, https://h2r4r4a5.ssl.hwcdn.net/assets/Series/Attachments/8787/aermec-anl-depliant.pdf.
[5] SIEMENS, https://mall.industry.siemens.com/mall/en/it/Catalog/Products/10177433?tree=CatalogTree.
[6] TSI,https://www.tsi.com/getmedia/d2a8d1d1-7551-47fe-8a0f-3c14b09b494b/7575_QTrak_A4_UK_5001356-
web?ext=.pdf.
[7] TRNSYS, “The transient energy system simulation tool.” [Online]. Available: http://www.trnsys.com.

392
 PRELIMINARY ASSESSMENT OF IMPACT OF TYPICAL
FAULTS ON ENERGY PERFORMANCE OF AIR-HANDLING
UNITS BY MEANS OF A DYNAMIC SIMULATION MODEL: A
CASE STUDY

Antonio Rosato
Department of Architecture and Industrial Design, University of Campania Luigi Vanvitelli, Aversa, Italy,
[email protected], ORCID: 0000-0001-5482-3529

Francesco Guarino
Department of Architecture and Industrial Design, University of Campania Luigi Vanvitelli, Aversa, Italy,
[email protected], ORCID: 0000-0003-3655-7065

Vincenzo Filomena
C.I.R.A. (Italian Aerospace Research Centre), Capua, Italy, [email protected], ORCID: 0000-0003-4163-9858

Sergio Sibilio
Department of Architecture and Industrial Design, University of Campania Luigi Vanvitelli, Aversa, Italy,
[email protected], ORCID: 0000-0002-8750-5401

Luigi Maffei
Department of Architecture and Industrial Design, University of Campania Luigi Vanvitelli, Aversa, Italy,
[email protected], ORCID: 0000-0003-4130-5065

Rosato, A, Guarino, F, Filomena, V, Sibilio, S, Maffei, L. Preliminary Assessment of Impact

of Typical Faults on Energy Performance of Air-Handling Units by means of a Dynamic
Cite this paper as:
Simulation Model: a Case Study. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: A widespread adoption of predictive maintenance procedures using automated Fault Detection and
Diagnostics (FDD) methods could be used to reduce the energy waste of poorly maintained and/or
improperly controlled Air-Handling Units (AHUs). A detailed dynamic simulation model of the
AHU (nominal cooling/heating capacity of 5.0 kW) serving the SENS-i Lab of the Department of
Architecture and Industrial Design of the University of Campania Luigi Vanvitelli (Italy) has been
developed, calibrated and validated based on several experimental tests by means of the dynamic
simulation software TRNSYS. In this paper, the model has been used to analyze the performance
of the AHU while serving a typical Italian building office. In greater detail, the simulations have
been carried out with and without the occurrence of typical faults of AHUs, such as the positive
and negative offsets of both room temperature sensor and relative humidity sensor. The results
allowed to investigate the dynamic variations of key-parameters associated to fault operation,
identify simplified rules for detection of any non-optimal states of HVAC and implementation of
effective FDD procedures, evaluate the impact of faults on thermo-hygrometric comfort conditions
and electric energy consumption, assess the potential energy and economic benefits of automated
FDD methods.
Keywords: Air-handling units; Fault detection and diagnosis; Predictive Maintenance; Energy Efficiency;
Model calibration and validation.
© 2020 Published by ECRES

393
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION

Air-handling units (AHUs) represent one of the main contributors to the world’s energy use. Faults can develop
during the AHU installation or routine operation, resulting in excessive energy waste or inefficient usage of
energy as well as uncomfortable environment. Companies follow different HVAC maintenance programs. The
approach of reactive maintenance is often not convenient due to the fact that repairing the damaged parts after
failure could be extremely expensive, cause safety issues and/or discomfort of the occupants. One of the main
challenges of preventive maintenance is to determine when the maintenance has to be performed; it has to be
conservative in order to prevent safety issues as well as reduce the costs of failures, but scheduling the
maintenance very early could mean wasting system life that is still usable. Fault Detection and Diagnosis (FDD)
[1] methods can monitor the operation of AHU components as well as detect and predict the presence of defects
(deviations from normal or expected operation). Ideally, FDD techniques could also resolve (diagnose) the type
of problem and/or identify its location, giving instructions for undertaking corrective actions. Efficient FDD
methods could also reduce the repair and maintenance costs of the plants, reduce energy consumption and related
emissions. FDD methods are based on the use of accurate instruments measuring key-parameters, the variation
of which in comparison to a “nominal/healthy” trend is assumed as a symptom of defects; once the test quantity
reaches some predetermined levels that reflects the seriousness of the defect requiring corrective actions to take
place, the test quantity is set into an alarm state (symptom) and reasoning is started to find the cause of the
“alarm - symptom – fault” chain. In a survey of United Kingdom buildings, the data showed how 25-50% of
energy waste was due to faults in the AHU units; this range could be reduced below 15% in the case of the faults
being detected and identified early in the premature stage before unacceptable damage occur [2]. FDD
techniques have been used for decades in aerospace, nuclear and industrial sectors, and their use in building
operation and control applications is becoming more widespread [1,2]. In order to apply FDD methods to HVAC
units or components, it is necessary to compare the real behavior of the systems to the “nominal/healthy”
operation without faults that can be modeled by means of simulation programs and/or artificial intelligence
techniques [1,2]. Therefore, models able to accurately simulate the AHU behavior have to be developed and
used in combination with FFD methods for implementing predictive maintenance programs. The authors
previously calibrated and validated a model simulating the operation of an AHU by contrasting the predictions
with measured data [3]; in this paper, this model has been used to (i) analyze the specific behaviors of key-
parameters, and (ii) assess the impacts on thermo-hygrometric comfort as well as electric energy consumption
in the cases of 4 typical soft faults being intentionally introduced.

2. EXPERIMENTAL SET-UP

The SENS i-Lab is an innovative, multi-sensorial and multi-purpose laboratory, located at the Department of
Architecture and Industrial Design of the University of Campania Luigi Vanvitelli (Aversa, Italy, latitude:
40°58′21″ N, longitude: 14°12′26″ E). The lab is served by an Air-Handling Unit (AHU) able to control the
indoor air temperature, relative humidity, velocity and quality inside the Integrated Test Room. Figure 1 reports
the schematic of the AHU serving the Integrated Test Room.

RAF: Return air fan

SAF: Supply air fan
HUM: Humidifier
PreHC: Pre-heating coil
CC: Cooling coil
PostHC: Post-heating coil
RAFil: Return air filter
SAFil: Supply air filter
OAFil: Outside air filter
HRS: Cross flow static heat recovery system
DOA: Outside air damper
DEA: Exhaust air damper
DHRS: Dumper of heat recovery system
DRA: Return air damper
VPreHC: Valve regulating the flow entering PreHC
VCC: Valve regulating the flow entering CC
VPostHC: Valve regulating the flow entering PostHC
VHUM: Valve regulating the flow entering HUM
RS: Refrigerating system
HP: Heap Pump
Figure 1. Schematic representation of the AHU serving the integrated test room.

394
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1 describes the main characteristics of the main AHU components. The heat carrier fluid is a mixture of
water and ethylene glycol (90/10% by volume). The AHU is fully instrumented in order to monitor, control and
record the return air temperature T RA, return air relative humidity RHRA, supply air temperature T SA, supply air
relative humidity RHSA, external air temperature T OA, air temperature at the outlet of the cooling coil TA,out,CC,
air temperature around the room T BEA, and air relative humidity around the room RHBEA. The end-users can
manually set: the desired targets of both the indoor air temperature (TSP,Room) and relative humidity (RHSP,Room),
the deadbands DBT and DBRH for both TSP,Room and RHSP,Room, respectively, the velocity of both the return air fan
(OLRAF) and the supply air fan (OLSAF), the opening percentages of the return air damper (OPDRA), the outside
air damper (OPDOA), the exhaust air damper (OPDEA) and the cross flow static heat recovery system damper
(OPDHRS). The parameters OPDRA, OPDOA and OPDEA can be varied in the range 0÷100%, where 100% means
that the air dampers are fully open. The parameter OPDHRS can be set to 100% (heat recovery does not occur) or
0% (heat recovery from return air flow takes place).

Table 1. Characteristics of main AHU components

Pre-heating coil (PreHC) Nominal heating capacity (kW) 4.1
Cooling coil (CC) Nominal cooling capacity (kW) 5.0
Humidifier (HUM) Steam capacity (kg/h) / Nominal power (kW) 5.0 / 3.75
Post-heating coil (PreHC) Nominal heating capacity (kW) 5.0
Heat Pump (HP) -model ANL 050HQ / Nominal heating / cooling capacity (kW) 14.0 / 13.4
Refrigerating System (RS) – model ANL 050Q Nominal input power (kW) 4.75 / 4.48

A specific control logic has been developed by the manufacturer in order to control the operation of the system
components and achieve the desired targets. Table 2 describes the conditions controlling the activation and
deactivation of the main components of the AHU. Even if the AHU is equipped with a PreHC, this component
is never used during normal operation (always de-activated). The operation of the humidifier as well as the flow
rate of the heat carrier fluid entering the PostHC or the CC can be continuously adjusted between 0% and 100%
by means of PID (proportional-integral-derivative) controllers. The operation of both the refrigerating unit and
heat pump is controlled in order to maintain the desired temperatures TCT and THT (with a deadband of 1 °C)
inside the cold and hot tanks. Additional details about the experimental set-up can be found in [3].

Table 2. Control strategies of the AHU

ON OFF
Humidifier RHRA ≤ (RHSP,Room - DBRH) RHRA ≥ (RHSP,Room + DBRH)
Cooling coil TRA ≥ (TSP,Room + DBT) OR TRA ≤ (TSP,Room - DBT) AND
RHRA ≥ (RHSP,Room + DBRH) RHRA ≤ (RHSP,Room - DBRH)
Post-heating coil TRA ≤ (TSP,Room - DBT) TRA ≥ (TSP,Room + DBT)
Vapor compression electric heat pump THT < (45 °C - 1°C) THT ≥ (45 °C + 1 °C)
Vapor compression electric refrigerating system TCT > (7 °C + 1°C) TCT ≤ (7 °C – 1 °C)

3. EXPERIMENTAL TESTS AND SIMULATION MODELS

Four experiments have been carried out in order to investigate the AHU behavior during steady-state and
transient operations. They are described in Table 3 in terms of target indoor air temperature T SP,Room, target
indoor air relative humidity RHSP,Room, initial air temperature TRoom_initial and relative humidity RHRoom_initial in the
test room.

Table 3. Boundary conditions of experimental tests

Test number 1 2 3 4
TSP,Room (° C) 22 → 24 → 26 26 → 24 → 22 22 28
TRoom_initial (° C) 19.8 28.1 22.0 28.1
RHSP,Room (%) 50.0 50.0 60 → 70 55 → 45
RHRoom_initial (%) 52.5 51.1 51.4 64.8

The experiments were carried out while maintaining constant the following conditions: DBT=1°C, DBRH=5%,
OLRAF=50%, OLSAF=50%, OPDEA=20%, OPDOA=20%, OPDRA=100% and OPDHRS=100%.

395
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The software of choice for modeling the AHU is the TRaNsient SYStems (TRNSYS) software platform 17 [4].
In this program, sub-systems are modelled as individual components (called Types) which can be cross-linked
to one-another.
Detailed models have been applied to take into account for: (i) cooling/thermal loads, (ii) the part load
characteristics of components, (iii) interaction between loads and AHU outputs, and (iv) system management
and control. The TRNSYS Types used for modeling the system components have been selected from the
TRNSYS libraries and enhanced by manufactures performance data or information available in current scientific
literature according to the common characteristics of the components used in practice. The model of the AHU
developed in the TRNSYS environment has been validated by contrasting the simulation results with the
experimental data in order to assess the accuracy of the calibration by using the following metrics quantifying
ε the average absolute error |ε̅|, and the
the instantaneous difference in the model predictions: the average error ̅,
root mean square error εRMS. The results of the empirical validation allowed to conclude that the model gives an
accurate representation of dynamic and steady-state performance and it can also be used in combination with
FFD methods for the detection of any non-optimal states. Additional details about the experimental tests and the
simulation model can be found in [3].

4. FAULTS ANALYSIS

In this section, the experimental performance (described in the section 3) have been simulated by intentionally
introducing 4 different typical soft faults into the operation of the AHU; the simulations have been performed
by running the calibrated and validated model (described in the previous section 3), while assuming the values
of the following inputs equal to the experimental data measured during the tests performed on the AHU operating
under normal condition: external air temperature T OA, air temperature TBEA and relative humidity RHBEA around
the room, target room temperature TSP,Room and relative humidity RHSP,Room, DBT, DBRH, OLRAF, OLSAF, OPDEA,
OPDOA, OPDRA and OPDHRS. Then, the experimentally measured performances of the AHU operating without faults
have been compared with those associated to the operation in the cases of faults occurrence. The comparison
has been performed in order to (i) analyse the specific behaviours of key-parameters associated to each fault,
and (ii) assess the differences in terms of thermo-hygrometric comfort hours as well as electric energy
consumption caused by the fault occurrence. In particular, the following 4 soft faults, which are independent of
each other, have been taken into account: 1) fault 1: positive offset of return air temperature sensor (+2 °C); 2)
fault 2: negative offset of return air temperature sensor (-2 °C); 3) fault 3: positive offset of return air relative
humidity sensor (+10 %); 4) fault 4: negative offset of return air relative humidity sensor (-10 %). Fault 1 means
that the measured return air temperature is 2.0 °C higher than the true value; fault 2 means that the measured
return air temperature is 2.0 °C lower than the right value; fault 3 means that the measured return air relative
humidity is 10% higher than the true value; fault 4 means that the measured return air relative humidity is 10%
lower than the true value. Figures 2a-h report the values of the return air temperature T RA, supply air temperature
TSA, return air relative humidity RHRA and supply air relative humidity RHSA as a function of the time. Figures
2a, 2c, 2e, and 2g refer to the operating conditions of the experimental tests 1 and 2 (see Table 3), while figures
2b, 2d, 2f and 2h correspond to the boundary conditions of the experimental tests 3 and 4 (see Table 3). Each
figure reports the experimental trends obtained in the case of operation without faults in comparison to the trends
of the same parameters while only one of the above-mentioned 4 faults is occurring. The target values of both
the indoor air temperature TSP,room and relative humidity RHSP,room are also reported in the same figures.
The trends of key-parameters under the fault 1 are reported in Figures 2a (tests 1 and 2) and 2b (tests 3 and 4).
With respect to the operation without faults, it can be noticed that, whatever the test is, in the case of fault 1 the
controllers require a substantially longer operating time of both cooling coil and humidifier, causing an increase
in terms of electric energy consumption associated to the operation of the refrigerating system as well as the
preparation of steam flow. However, this case is also characterized by a shorter operating time of post-heating
coil (thanks to the fact that the measured T RA is larger than the real one), reducing the electric energy consumed
by the heat pump.
The trends associated to key-parameters under the fault 2 are indicated in Figures 2c (tests 1 and 2) and 2d (tests
3 and 4). With respect to the operation without faults, the results highlight that, whatever the test is, in the case
of fault 2 the controllers require a significantly longer operating time of post-heating coil, causing a greater
electric energy consumption associated to the heat pump; however, in this case, the operating time of both
cooling coil and humidifier are reduced (thanks to the fact that the measured T RA is lower than the real one),
lowering the related electric energy consumption. The trends of key-parameters under the fault 3 are shown in

396
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figures 2e (tests 1 and 2) and 2f (tests 3 and 4). With respect to the operation without faults, the simulations
indicate that, whatever the test is, in the case of fault 3 the controllers require a substantially shorter operating
time of humidifier (thanks to the fact that the measured RHRA is higher than the real one); however, in this case
the operating time of the cooling coil is longer, causing a larger electric energy consumption of the refrigerating
system. The trends associated to key-parameters under the fault 4 are indicated in Figures 2g (tests 1 and 2) and
2h (tests 3 and 4). With respect to the operation without faults, the simulation data underline that, whatever the
test is, in the case of fault 4 the controllers require a negligible variation of operating time of post-heating coil,
whereas there is a significant increase in the operating time of the humidifier (causing an increment of the
associated electric energy consumption), together with a decrease in the operating time of the cooling coil
(thanks to the fact that the measured RHRA is higher than the real one). The differences between the operation
with and without faults are significantly consistent, so that specific rules could be identified in order to easily
detect the presence of non-optimal states of the HVAC performance.
42 100 42 TT_RA_w/o_fault
RA, w/o_fault T
T_RA_fault1
RA_fault1 TSA,w/o_fault
T_SA_w/o_fault TSA_fault1
T_SA_fault1 TT_SP
SP,Room 100
a) TT_RA_w/o_fault TT_RA_fault1
RA_fault1 TSA,w/o_fault
T_SA_w/o_fault T
T_SA_fault1
SA_fault1 TT_SP
SP,Room b)
RA, w/o_fault RH
RH_RA_w/o_fault
RA, w/o_fault RH
RH_RA_fault1
RA_fault1 RH
RH_SA_w/o_fault
SA, w/o_fault RH
RH_SA_fault1
SA_fault1 RH
RH_SP
SP,Room
39 RH
RH_RA_w/o_fault
RA, w/o_fault RH
RH_RA_fault1
RA_fault1 RH
RH_SA_w/o_fault
SA, w/o_fault RH
RH_SA_fault1
SA_fault1 RH
RH_SP
SP,Room 90 39 90

36 80 36 80

Relative humidity (%)

Relative humidity (%)
33 70 33 70
Temperature ( C)

Temperature ( C)
30 60 30 60

27 50 27 50

24 40 24 40

21 30 21 30

18 20 18 20

15 10 15 10

12 0 12 0
0 100 Test
200 1 300 400 500 Test 2
600 700 800 0 20 Test 3 40 60 Test804 100
42 100 T T T T TT_SP
TT_RA_w/o_fault TT_RA_fault2 TSA,w/o_fault T TT_SP 42 T_RA_w/o_fault
RA, w/o_fault T_RA_fault2
RA_fault2 T_SA_w/o_fault
SA,w/o_fault T_SA_fault2
SA_fault2 SP,Room 100
c) RA, w/o_fault RA_fault2 T_SA_w/o_fault T_SA_fault2
SA_fault2 SP,Room
d) RH RH RHSA_fault2
RHRA, w/o_fault
RH_RA_w/o_fault RH
RH_RA_fault2 RH
RH_SA_w/o_fault RH_SA_fault2 RH
RH_SP RH
RH_RA_w/o_fault
RA, w/o_fault
RH_RA_fault2
RA_fault2 RH_SA_w/o_fault
SA, w/o_fault RH_SA_fault2 RH
RH_SP
SP,Room
39 RA_fault2 SA, w/o_fault RH SA_fault2 SP,Room 90 39 90
36 80 36 80
Relative humidity (%)

Relative humidity (%)

33 70
Temperature ( C)

33 70
Temperature ( C)

30 60 30 60

27 50 27 50

24 40 24 40

21 30 21 30

18 20 18 20

15 10 15 10

12 0 12 0
0 100 Test
200 1 300 400 500 Test 2
600 700 800 0 20 Test 3 40 60 Test
80 4 100
42 100 42 TT_RA_w/o_fault T RA_fault3
T_RA_fault3 T
T_SA_w/o_fault T
T_SA_fault3 T
T_SP 100
e) T
T_RA_w/o_fault
RA, w/o_fault TT_RA_fault3
RA_fault3 T
T_SA_w/o_fault TSA_fault3
T_SA_fault3 T
T_SP
SP,Room
RA, w/o_fault SA,w/o_fault SA_fault3 SP,Room
SA,w/o_fault f) RH RH RH RH RH
39 RH
RH_RA_w/o_fault
RA, w/o_fault RH
RH_RA_fault3
RA_fault3 RH
RH_SA_w/o_fault
SA, w/o_fault RH
RH_SA_fault3
SA_fault3
RH
RH_SP
SP,Room 90
RH_RA_w/o_fault
RA, w/o_fault RH_RA_fault3
RA_fault3 RH_SA_w/o_fault
SA, w/o_fault RH_SA_fault3
SA_fault3 RH_SP
SP,Room
39 90
36 80 36 80
Relative humidity (%)

Relative humidity (%)

33 70
Temperature ( C)

33 70
Temperature ( C)

30 60 30 60

27 50 27 50

24 40 24 40

21 30 21 30

18 20 18 20

15 10 15 10

12 0 12 0
0 100 Test
200 1 300 400 500 600 Test 2 700 800 0 20 Test 3 40 60 Test
80 4 100
42 TT_RA_w/o_fault
RA, w/o_fault TT_RA_fault4
RA_fault4 T
T_SA_w/o_fault
SA,w/o_fault T
T_SA_fault4
SA_fault4 T
T_SP
SP,Room 100 42 TT_RA_w/o_fault
RA, w/o_fault
T
T_RA_fault4
RA_fault4 TT_SA_w/o_fault
SA,w/o_fault TT_SA_fault4
SA_fault4 T
T_SP
SP,Room 100
g) RH RH RHSA_fault4 RH h) RH RH RH RH RH
RH_RA_w/o_fault
RA, w/o_fault RH
RH_RA_fault4
RA_fault4
RH_SA_w/o_fault
SA, w/o_fault RH_SA_fault4 RH_SP
SP,Room RH_RA_w/o_fault
RA, w/o_fault RH_RA_fault4
RA_fault4 RH_SA_w/o_fault
SA, w/o_fault RH_SA_fault4
SA_fault4
RH_SPSP,Room
39 90 39 90

36 80 36 80
Relative humidity (%)
Relative humidity (%)

33 70 33 70
Temperature ( C)

Temperature ( C)

30 60 30 60

27 50 27 50

24 40 24 40

21 30 21 30

18 20 18 20

15 10 15 10

12 0 12 0
0 100 Test
200 1 300 400 500 Test 2
600 700 800 0 20 Test 3 40 60 Test
80 4 100

Figures 2. Comparison between normal operation and faulty operation:(a) fault 1 – Tests 1,2, (b) fault 1 – Tests 3,4, c)
fault 2 – Tests 1,2, d) fault 2 – Tests 3,4, (e) fault 3 – Tests 1,2, (f) fault 3 – Tests 3,4, (g) fault 4 – Tests 1,2, (h) fault 4 –
Tests 3,4.

397
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

5. CASE STUDY

A typical small-size office building located in the city of Naples (southern Italy, latitude: 40°51′22″ N, longitude:
14°14′47″ E) is assumed as case-study to be served by the air-handling unit described in the section 2. The
building is characterized by a floor area of 18 m2 and a volume of 54 m3, with a total area of windows equal to
1.8 m2. The conductive thermal resistance of building elements has been defined (0.71 m2K/W for windows, 3.1
m2K/W for the roof, 2.7 m2K/W for the floor, 2.9 m2K/W for external vertical walls) according to the threshold
values imposed by Italian legislation requirements. The office is empty during the weekends, while in weekdays
there is a constant number of 3 people working from 8:00 am up 6:00 pm. The working activity is characterized
by an internal sensible and latent gain of 65 W/occupant and 55 W/occupant, respectively; the heat gains due to
3 PCs (420 W), a laser printer (110 W) and artificial lighting systems (3.75 W/m2) are taken into account. The
target of indoor air temperature (TSP,Room) is set to 20 °C (±1 °C) during the heating period (November 1st – March
31st) and 26 °C (±1 °C) during the cooling period (April 1st – October 31st); the target of relative humidity
(RHSP,Room) is always set to 50 % (±5 %) during the entire year. The AHU system can operate to achieve the
targets only when at least one occupant being inside the office. Figure 3 highlights the thermal comfort hours
(i.e. the percentage of time during which the target of indoor air temperature is achieved), the hygrometric
comfort hours (i.e. the percentage of time during which the desired target of indoor air relative humidity is
achieved), for both the case without faults and the cases when one the four above-mentioned faults occurs in
order to facilitate the comparison among the different scenarios. The same figure is also showing, for all
scenarios, the overall electric energy consumption due to the operation of the refrigerating system, heap pump,
humidifier, supply and return air fans. In particular, Figures 3a, 3b and 3c refer to the whole year, the heating
period only and the cooling period only, respectively. These figures highlight that the thermal comfort hours and
hygrometric comfort hours represent 85.5% and 82.5%, respectively, of the entire operation time of the AHU
with reference to the scenario without faults. In comparison to the performance under normal operation:
 whatever the period is, the occurrence of faults 1 and 2 greatly reduce the thermal comfort hours (together
with a decrease of hygrometric comfort hours), significantly deteriorating the comfort of the occupants; the fault
1 is characterized by a lower electric energy consumption thanks to the reduced operating time of post-heating
coil and heat pump (due to the fact that the measured TRA is larger than the real one); the fault 2 causes a larger
electric energy consumption due to a longer operating time of post-heating coil (associated to the fact that the
measured TRA is lower than the real one);
 whatever the period is, the occurrence of faults 3 and 4 greatly reduce the hygrometric comfort hours (while
the thermal comfort hours remain almost constant); the fault 3 is characterized by a lower electric energy
consumption thanks to a shorter operating time of humidifier (due to the fact that the measured RHRA is larger
than the real one); the fault 4 causes a larger electric energy consumption due to a longer operating time of the
humidifier (associated to the fact that the measured RHRA is lower than the real one).
100 100
a) Thermal Comfort Hygrometric Comfort Electric energy consumption b) Thermal Comfort Hygrometric Comfort Electric energy consumption
90
Electric energy consumption (MWh)

90 86.0
Electric energy consumption (MWh)

84.5 82.5 84.71 80.39 85.09 84.19

83.3 88.47
78.2
80 80
Comfort hours (%)

67.9 70
Comfort hours (%)

70
61.0
60 60
53.82
50 50
40.70 40.50
40 40

30 26.6 30
21.0 22.3
17.8
20 20
13.1 12.70 9.23
9.6 10.2 6.29 4.75 4.47 8.56
10 4.8 10
4.34
0 0
Without fault Fault1 Fault2 Fault3 Fault4 Without fault Fault1 Fault2 Fault3 Fault4
100
c) Thermal Comfort Hygrometric Comfort Electric energy consumption
Electric energy consumption (MWh)

90 86.59
84.43 83.92 95.52 82.75
80 75.58
Comfort hours (%)

70
60 53.32
50
40
30
20 16.56
11.50 12.45 13.07
7.42 8.31 5.89
10 5.11
0
Without fault Fault1 Fault2 Fault3 Fault4

Figures 3. Comfort hours and electric consumption with/without faults: (a) whole year, (b) heating period, (c) cooling
period.

398
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

ACKNOWLEDGMENT

This work was undertaken as part of the program “PON FSE-FESR Ricerca e Innovazione 2014-2020” of the
Italian Ministry of Education, University and Research, Action I.1 “Dottorati Innovativi con caratterizzazione
industriale”.

REFERENCES

[1] Katipamula, S, Brambley, M. Review article: methods for fault detection, diagnostics, and prognostics for building
systems - a review, part I, HVAC&R 2005, 11, 3-25. https://doi.org/10.1080/10789669.2005.10391123.
[2] Yu, Y, Woradechjumroen, D, Yu, D. A review of fault detection and diagnosis methodologies on air-handling units,
Energy and Buildings 2014, 82, 550-562. https://doi.org/10.1016/J.ENBUILD.2014.06.042.
[3] Rosato, A, Guarino, F, Filomena, V, Sibilio, S, Maffei, L. Preliminary Experimental Calibration and Validation of a
Dynamic Simulation Model for Fault Detection and Diagnosis of Air-Handling Units. In: 8th Eur. Conf. Ren. Energy
Sys. 24-25 August 2020, Istanbul, Turkey
[4] TRNSYS, “The transient energy system simulation tool.” [Online]. Available: http://www.trnsys.com.

399
 A SIMPLE METHOD FOR PARAMETER EXTRACTION OF A PV
MODULE
Nader Anani
School of Engineering, University of Wolverhampton, Telford, Shropshire, TF2 9NT, UK, [email protected].
ORCID: 0000-0002-0944-8915

Anani, N., A Simple Method for Parameter Extraction of a PV Module, 8th Eur. Conf.
Cite this paper as:
Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The aim of this paper is to present a fast and simple method for characterization and modelling of
a PV module with due account taken of its application as a teaching tool for undergraduate courses.
Thus, the paper presents a simplified mathematical analysis that underpins the modelling process
of a PV module, which is based on the single-diode model (SDM). Subsequently, the paper
explains a fast and straightforward method for extracting the parameters of the single-diode model
from experimentally obtained I-V (current-voltage) characteristic of a PV module. The parameter
extraction strategy uses a spreadsheet approach to solve the nonlinear equations of the model,
which alleviates the need for any nonlinear equation solvers. The method was used, as a laboratory
exercise for engineering students, to extract the parameters of a small monocrystalline silicon PV
module and the results concurred well with the manufacturer’s datasheet and with experimental
results. The differences were within the manufacturer’s specified tolerance of 10%.
Keywords: Ideality factor, parallel resistance, parameter extraction, photon current, photovoltaic systems,
saturation current, series resistance.
© 2020 Published by ECRES

1. INTRODUCTION
When integrating photovoltaic systems into the undergraduate teaching of renewable energy technologies, it is
necessary that students have a good understanding of the modelling and characterization process of a PV
generator. However, this can be a complex process and requires a sound understanding of numerical and/or
analytical methods of solving systems of nonlinear equations. Therefore, the aim of this paper is to present a
concise mathematical formulation of the single-diode model equations of a PV module and demonstrates a fast
and simple method of solving these equations.

The design, analysis, and simulation of a PV system requires a lumped circuit parameter model of the PV
generator. A PV generator could be as simple as a single PV cell, a PV module consisting of several cells
connected in series, or more typically an array of several modules arranged in series-parallel combinations to
deliver the rated power at the required terminal current and voltage levels. For this purpose, the single-diode
model has been commonly used to model a PV generator due to its simplicity and accuracy [1]. The SDM for a
PV generator is shown in Figure 1 and the terminal I-V and P-V (power-voltage) characteristics of a typical PV
generator are shown in Figure 2. The P-V curve exhibits a maximum power point (MPP) where a PV generator
must be operated for maximum efficiency. However, the I-V characteristic and hence the location of the MPP
changes with variations in irradiance and temperature [2]. Therefore, the power processing system in a PV plant
normally includes a maximum power point tracker (MPPT). The problem of tracking the MPP is compounded
due to partial shading (PS), which can have adverse effects on the performance of a PV system [3]. Mitigation
of the effects of PS involves the use of bypass diodes, which can lead to multiple peaks in the P-V curve [4],
[5]. Therefore, the design and simulation of MPPTs, exploring the effects of partial shading, and the need to

400
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

efficiently size a PV generator require an equivalent circuit model of the PV generator. Numerous analytical [6]
and numerical [7] schemes to derive such an equivalent circuit have been reported in the literature with varying
degrees of complexity and accuracy [8]. In this paper, a simple method for extracting the five parameters of the
SDM model, namely the photocurrent, the series and parallel resistances, the saturation current, and the ideality
factor is demonstrated.

Figure 1. The single-diode equivalent circuit model

Figure. 2 Normalised I-V and P-V curves of a typical
of a PV cell.
PV module.

The method uses several proven simplifications and approximations and three nonlinear equations, which are
consequently solved iteratively using a spreadsheet as explained in section 3. The simplicity of the method
makes it attractive as a teaching tool and a laboratory exercise for electrical engineering students. Following this
introduction, the mathematical background for extracting the parameters of the single-diode model is presented
in section 2. The method is explained with the aid of an example for extracting the single-diode model parameters
of a small PV module in section 3. Finally, section 4 presents the conclusions of the work.

2. MATHEMATICAL ANALYSIS
The SDM of a single PV cell, shown in Figure 1, consists of an insolation-controlled current source 𝐼𝑝ℎ , a diode,
and the resistances Rs1 and Rsh1 which represent the series and shunt parasitic resistances of the cell respectively.
The terminal current, is given by

I  I ph  I d  I sh (1)

where Id , is the diode current and Ish is the shunt resistance current. The diode current Id is given in terms of
the diode voltage Vd 1 , the reverse saturation current I o and the ideality factor a of the diode, by the Shockley
equation as

Id  Io [exp(Vd1 / aVt )  1] (2)

where Vt (V) is the thermal voltage which is defined in terms of the temperature T in Kelvin (k), the electronic

charge q1.6021019 (C) and the Boltzmann’s constant k  1.380  10-23 (J/k) as

Vt kT / q (3)

Note that the subscript 1 indicates a single cell. The current in the parallel resistance Ish is given by

Ish Vd1/ Rsh1 (4)

The diode voltage may be expressed as

Vd1V1 IRs1 (5)

401
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Therefore, the terminal current I and voltage V1 of the SDM of a single cell are related by the implicit nonlinear
and transcendental equation

(V  IRs1) (V  IRs1)
I  I ph  I o [ exp( 1 ) 1]  1 (6)
aVt Rsh1

The single-diode model can be easily adjusted to model a PV module consisting of Ns identical cells connected
in series as shown in Figure 3 [9]. The terminal current I, the photocurrent I ph and the diode current Id of the
module are the same as those for any single cell within the module. The diode voltage of a single cell is given
in terms of the module current I, and voltage V, as Vd1  V / N s  IRs1 which may be written as

Vd1  (V  IN s Rs1 ) / N s (7)

Figure.3. The single-diode model of a PV panel consisting of Ns series-connected cells.

Therefore, the terminal module current, which is the same as the single-cell current, may now be written as

V  INs Rs1 V  INs Rs1

I  I ph  I o [exp( )  1]  (8)
NsaVT Ns Rsh1

However, the total series resistance of the module, which is typically less than 1  , is

Rs  N s Rs1 (9)

Similarly, the total shunt resistance of the module, which is typically in the order of few hundred ohms, is

Rsh  Ns Rsh1 (10)

Therefore, the terminal current and voltage of the module are related by

V  IRs (V  IRs )
I  I ph  I o [ exp( ) 1]  (11)
NsaVt Rsh

402
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

To model a PV panel, the five parameters Rs , Rsh , a, Io , and I ph must be determined. These parameters are not
included in manufacturers’ datasheets of PV modules. However, they can be extracted from information
normally provided in the datasheet of a PV module. A typical datasheet includes the I-V data at three vital points:
the short-circuit point (0, Isc ) , the open-circuit point (Voc ,0) , and the maximum power point (Vmp , I mp ) . Some
datasheets also include the full I-V characteristic. However, the data provided by a datasheet is always specified
2
at only one operating condition, namely the standard test condition (STC). (STC: insolation G  1000 W/m , Air
o
Mass=1.5, T  25 C ). Therefore, parameters extracted from STC data are only valid at STC and hence, for any
other operating condition of temperature and insolation, these parameters must be adjusted accordingly [2].

Extraction of the Photocurrent

This parameter can be estimated from the SC point by substituting V 0 and I  I sc in Eq. (11)

I R I R
Isc  I ph  Io [exp( sc s )  1]  sc s (12)
NsaVt Rsh

Since the series resistance is very small, typically less than 1, the diode voltage IscRs is also very small and
comparable to the thermal voltage, that is exp( IscRs / NsaVt )1 . Therefore, the diode and shunt resistance
currents are too small compared to the short-circuit current and hence, they can be neglected [10] [11].
Consequently Eq. (11) reduces to I ph  Isc . To estimate the photocurrent at any other condition of temperature
T (k) and insolation G (W/m 2 ) , the following expression may be used [12]

G
I ph   Isc,STC   I,STC (T  TSTC )  (13)
GSTC

where Isc,STC is the short-circuit current at STC and  I,STC is the temperature coefficient of the short-circuit
current at STC, which are always available in the datasheet.

Extraction of the Shunt Resistance

The effect of the shunt resistance is dominant at higher currents, i.e. near the short-circuit point. Therefore, the
derivative of the I-V curve at the short-circuit point may be used to estimate the shunt resistance. The derivative
of Eq. (11) at any point along the I-V curve can be readily derived as

dI I o   (V  IRs )  dI  1 R dI
 exp  1 Rs   R  s (14)
dV aN sVT   aN sVt  dV  sh Rsh dV

Solving for the derivative

dI I (V  IRs ) 1 Rs I R (V  IRs )
  [ o exp( ) ] [1  o s exp( )]
dV aNsVt aNsVt Rsh Rsh aN sVt aNsVt (15)

403
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

At the short-circuit point, the derivative becomes

dI I Isc Rs 1 R I R Isc Rs
 [ o exp( ) ] / [1  s  o s exp( )] (16)
dV sc aNsVt aN sVt Rsh Rsh aNsVt aNsVt

The term Io exp( IscRs / aNsVt ) represents the diode current and this is too small compared to the short-circuit current
and thus it can be neglected [10] [11]. In addition, since Rs  Rsh , the term Rs / Rsh may also be neglected [13].
Therefore, the derivative can be approximated as

dI
  1/ Rsh (17)
dV sc

Hence, the shunt resistance can be estimated from the slope of the I-V curve at the short-circuit point.

Extraction of the Series Resistance

The effect of the series resistance is significant at higher voltages and low currents. Therefore, the value of the
series resistance may be estimated from the slope of the I-V curve at the open-circuit point. Substituting the OC
point (Voc ,0) in the general expression for the derivative in (15)

dI Io Voc 1 Rs I R Voc
 [ exp( ) ] [1   o s exp( )] (18)
dV OC aN V
s t aN sVt Rsh Rsh aN sVt aN sVt

The term Io exp(Voc / aNsVt ) represents the diode current and since under open-circuit condition, the output current
is zero, the diode current is the same as the photon current (less the negligibly small current in the shunt
resistance) and the photon current is the same as the short-circuit current, i.e.

Voc
I o exp( )  I sc (19)
aN sVT

Hence the diode current is the same as the short-circuit current. The derivative becomes

dI  I 1  Rs I R
   sc   (1   sc s ) (20)
dV OC  aN sVt Rsh  Rsh aNsVt

In practice, 1/ Rsh Isc / aNsVT and hence, this term and the term Rs / Rsh may be neglected. The series resistance
may, therefore, be estimated from

dV aN sVt
Rs   (21)
dI oc Isc

The first term can be estimated from I-V curve. However, the ideality factor in the second term is unknown and
must be determined.

Extraction of the Reverse Saturation Current

404
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The reverse saturation current may be estimated at the open-circuit point where its effects are most significant.
At this point, the output current I  0 and I ph  Isc . Substituting in Eq. (11) and using the approximation
exp(Voc / aNsVT 1) , the saturation current may be estimated from

V V
I o  ( Isc  oc ) exp( oc ) (22)
Rsh aNsVt

The open-circuit voltage Voc and Ns can be obtained from the datasheet. However, the ideality factor a is
unknown needs to be determined.

Extraction of the Ideality Factor

Substituting the MPP point in the general equation of the module current, i.e. in Eq. (11)

(Vmp  I mp Rs ) (Vmp  I mp Rs )
I mp  I sc  Io [exp( )1]  (23)
aNsVt Rsh

To determine the parameters Io , Rs , and a , the three Eqs. (21), (22) and (23) may be solved simultaneously using
a nonlinear equation solver, such as MATLAB. However, numerical solutions can be very sensitive to the initial
guess and conversion can be problematic depending on the initial guess amongst other factors [14]. An easier
approach, is to use a spreadsheet solution as explained in the next section.

3. EXPERIMENTAL SETUP AND RESULTS

The experiment uses readily available apparatus consisting of a variable 400 W light source that simulates the
spectrum of the sunlight, a solar insolation meter, a cooling fan, a thermometer, and a variable resistance used
as a load. The 1 W Parallax 750-00030 monocrystalline silicon solar panel whose specifications are: Ns 12 ,
Voc  6.9 V , I sc  0.18 A , Vmp  6 V , and Vmp 6 V was used to illustrate the parameter extraction method. The
datasheet did not include the I-V curve; therefore, this was obtained in the laboratory. The variable light source
2
was adjusted until the measured insolation at the surface of the panel was 1000 (W.m ). The fan helped
maintaining the surface temperature around 25 o C
to approximate a STC operating conditions. Using the variable
resistance, the terminal voltage across the PV panel was swept from zero to Voc . The resulting I-V curve and
the corresponding P-V curve are shown in Figure 4. These were used in extracting the parameters of the single-
diode model as illustrated in the next subsections.

Extraction of the Photocurrent and the Shunt Resistance

405
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The photon current is the same as the short-circuit current. This was obtained from the datasheet and confirmed
by experiment as I ph 0.18 A The shunt resistance was estimated from the slope of the I-V curve at the short
circuit condition.

dI 1
  Rsh 385  (24)
dV sc Rsh

Extraction of the Ideality Factor, Series Resistance and Saturation Current

Using a Microsoft Excel spreadsheet these parameters were obtained as follows: The ideality factor a was swept
from 1 to 2 in steps of 0.1 and each step the saturation current Io was calculated using (22) and thus, the series
resistance calculated using Eq. (21) as illustrated in the spreadsheet in Table 1. Then, for each set of values of
a , I o , and Rs , an I-V curve for the module was derived as follows: Using another spreadsheet, the diode voltage

Vd 1 was swept from 0 to 0.6 V in steps of 0.05 V and at each step the diode current and the shunt resistance
current were calculated using

qVd1 / akT
I d  Io (e 1) (25)

And

Ish Vd1/( Rsh / Ns ) (26)

The module current was consequently obtained from

I Isc Id Ish (27)

The module voltage was calculated using

V  Ns [ Vd1 I ( Rs / Ns ) ] (28)

Table 1. Sweeping the value of the ideality factor and calculating the values of and Rs .

Finally, the set of values of a , Io , and Rs that resulted in the closest match between the estimated I-V curve
and the datasheet/experimental data, was chosen. This is illustrated in Table 2 for the case of a=1.6, i.e. when
I o 1.3483107 A and Rs 0.761823  . This was the set that resulted in the best match as illustrated in Figure 5,

406
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

which shows a comparison between the laboratory measured I-V characteristic and the estimated
characteristics using the extracted parameters of the PV module when the ideality factor is a=1.6.

Figure 4. Measured I-V and P-V curves at 1000 W/m2. Figure 5. The measured and extracted I-V curves.

4. CONCLUSION
The paper presented the mathematical analysis and a fast and simple method for extracting the parameters of
the single-diode model of a PV module form the experimentally obtained I-V curve. The simplicity of the
method lends itself as a useful tool for introducing PV module characterisation and parameters extraction to
engineering and science students who may not be conversant with numerical methods. Results obtained were
within 10% of the experimental and manufacturer’s specified data. For example, the estimated MPP was
(5.7 V, 0.16 A) compared to the manufacturer’s point of (6 V, 0.17 A), and the estimated open-circuit voltage
was 7.3 V compared to the manufacturer’s value of 6.9 V.

Table 2. Sweeping the diode voltage and calculating the module current and voltage.

REFERENCES

[1] Weidong, X, Dunford, W, Capel, A, A Novel Modeling Method for Photovoltaic Cells. In: Proceedings of the IEEE
35th Annual Power Electronics Specialists Conference, 2004, Aachen, Germany, pp. 1950-1956 Vol.3.
[2] Ibrahim, H, Anani, N, Variations of PV module parameters with irradiance and temperature, Energy Procedia, vol. 134,
no. 1, pp. 276-285, 2017. https://doi.org/10.1016/j.egypro.2017.09.617.
[3] Anani, N, Renewable Energy Technologies and Resources, London: Artech House, 2020.
[4] Kim, KA, Krein, PT, Reexamination of Photovoltaic Hot Spotting to Show Inadequacy of the Bypass Diode, IEEE
Journal of Photovoltaics, vol. 5, no. 5, pp. 1435-1441, Sept. 2015.

407
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[5] Ibrahim, H, Anani, N, Variation of the performance of a PV panel with the number of bypass diodes and partial shading
patterns. In: Power Generation Systems and Renewable Energy Technologies, PGSRET, Istanbul, 2019.
[6] Papathanassiou, EI, Batzelis, SA, A Method for the Analytical Extraction of the Single-Diode PV Model Parameters,
EEE Transactions on Sustainable Energy, vol. 7, no. 2, pp. 504-512, 2016.
[7] Lim, LHI, Ye, Z, Ye, J, Yang, Du, D, A linear method to extract diode model parameters of solar panels from a single
I–V curve, Renewable Energy, vol. 76, pp. 135-142, 2015.
[8] Gaevskii, A, A Method for Determining Parameters of PV Modules in Field Conditions. In: 6th IEEE International
Conference on Energy Smart Systems, Kyiv, Ukraine, 2019.
[9] N. Femia, G. Petrone, G. Spagnuolo and M. Vitelli, Power and Control Techniques for Maximum Energy Harvesting
in Photovoltaic systems, London: CRC Press, 2012.
[10] Kennerud, KL, Analysis of performance degradation in CdS solar cells, IEEE Trans Aerospace and Electronic Systems,
vol. 5, no. 6, pp. 912-917, 1969.
[11] Brano, VL, Orioli, A, Ciulla, G, Gangi, AD, An improved five-parameter model for photovoltaic modules, Solar
Energy Materials and Solar Cells, Solar Energy Materials and Solar Cells, vol. 94, no. 8, pp. 169-178, 2010.
[12] Villalva, MG, Gazoli, GR, Filho, ER, Comprehensive approach to modeling and simulation of photovoltaic arrays,
IEEE Trans. Power Electronics, vol. 24, no. 5, pp. 1198-1208, 2009.
[13] Chan, DSH, Phang, JCH, Analytical methods for the extraction of solar-cell single- and double-diode model parameters
from I-V characteristics, IEEE Transactions on Electronic Devices, vol. 34, no. 2, pp. 286-293, 1987.
[14] H. Ibrahim and N. Anani, Evaluation of Analytical Methods for Parameter Extraction of PV modules, Energy Procedia,
vol. 134, pp. 69-78, 2017. https://doi.org/10.1016/j.egypro.2017.09.601.

408
 ADAPTIVE OVERCURRENT PROTECTION FOR ELIMINATE
TO MISOPERATION IN ELECTRICAL DISTRIBUTION
NETWORKS

Ömer Gül
Istanbul Technical University, Istanbul, Turkey, [email protected], ORCID: 0000-0003-0379-5446

Anıl Çağlar Doğancı

Istanbul, Turkey, [email protected], ORCID: 0000-0002-7247-733X

Gül, Ö, Doğancı, A Ç, Adaptive overcurrent protection for eliminate to misoperation in

Cite this paper as: Electrical distribution networks. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: Electrical distribution system operators should be able to provide their customers with electrical
energy with economically acceptable cost and the minimum possible duration and frequency of
interruption. In recent years, thanks to outstanding technological developments on numerical
relays' communication interface, they became more intelligent and flexible.
In this study, it will emphasize that, in order to prevent misoperations, instead of using classic
protection systems, adaptive protection system that has a dynamic structure should use. Moreover,
it will argue that, adaptive protection systems are more effective than the classic protection
systems. However, in this study problem of misoperations in real electric distribution system,
which has a classic protection system and has different feeding scenarios discussed. To solve this
problem, the whole operation scenarios of the electric distribution system scenarios create and
short-circuit fault analysis performed for all scenarios by using SINCAL software. According to
results of the analysis and the catalog information of the equipment, protections relays’ setting
groups created. These setting groups also defined in supervisory control and data acquisition
(SCADA) software. Thus, when the defined scenarios’ realized, SCADA could automatically
change the setting groups of protection relays.
Keywords: Adaptive protection, misoperations, overcurrent protection, smart grids
© 2020 Published by ECRES

Nomenclature
SCADA Supervisory control and data acquisition
RTU Remote Terminal Unite

1. INTRODUCTION

In order to ensure uninterrupted and high quality electrical energy distribution to end users, network components
must protected against malfunctions. Protection systems, electrical distribution network faults detected quickly
and reliably, thus ensuring that the fault is cleared [1]. In case of high load variation in electricity distribution
networks, seasonal or different shifts etc., breakers make unnecessary trips despite the failure [2].
In case of failure in interconnected electricity network that feds industrial distribution network, standby
generator systems activated and fed to the industrial distribution network. In this case, where standby generator
activated, short circuit currents decrease because of increased source impedance. When the setting values of the
overcurrent relays are not changed, the problem of the protection appears due to the failure of the faults to be
detected [3-5]. With the widespread distribution of distributed generation sources connected to the radial
electrical distribution networks fed from a single source, the topology of the radial electrical distribution
networks has changed. The protection problems that emerged because of the topology change of the network

409
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

have been one of the most studied research topics [6-13]. In addition, in open ring network operation, load
changes due to load transfers during post-failure restoration work can cause erroneous trips in overcurrent
protection [14-16].
Due to the changes occurring in the topologies, loads, etc. of the electricity distribution networks, a reliable
protection system required by preventing false trips.
In recent years, with the advent of digital protection relays with communication possibilities, the protection
relays have become more flexible, smart and superior features [1]. The matching protection systems, which
realized by using the new possibilities of digital relays, allow the adjustment of the adjustment parameters
according to the status of the protection zone thanks to its multiple adjustment groups [1]. Thus, the protection
system adapts to different operating situations and minimizes the possibility of malfunction. Compliant
protection systems offer more selective and reliable protection compared to conventional protection systems
thanks to this dynamic structure [17]. With each passing day, electricity distribution companies and industrial
enterprises generating their own electricity prefer protection systems that can adapt instead of classical
protection systems. Another smart electricity distribution system that is rapidly becoming widespread is
SCADA, which provides network operators with significant convenience by enabling information-based
decisions and remote maneuvering [18].
In this study, SCADA based adaptive protection methodology is proposed to prevent malfunction of the
protection system in electricity distribution systems. As an application of the proposed methodology, an
electrical distribution network adaptive protection application given in detail, which fed by two backed ring
structures and three backup power generators that are activated by load change. Depending on the operating
maneuvers, working scenarios are created for different nutritional situations and short circuit failure analysis of
these scenarios are made with SINCAL software, protection parameters are determined, necessary verification
tests are made by defining relay and SCADA system.

2. ADAPTIVE PROTECTION METHODOLOGY

The task of the protection systems in the electricity distribution networks is to detect the faults and types that
occur, quickly and reliably with the help of the observed electrical quantities, and to ensure that the energy is as
continuous as possible by deactivating the faulty operating element (line, transformer, generator, motor, etc.)
when necessary. [1]. Often, the reliability of a protection system defined as not causing faulty trips and ensuring
that it will operate at the time of failure [2]. In electrical distribution network protection schemes, it can be
grouped in three different groups as faulty operating conditions, not tripping, faulty opening and unnecessary
tripping:
In addition, unnecessary tripping to the breakers may occur in cases where the load change in seasonal or
different shifts is high, although there is no malfunction. For the negativity that causes the protection system to
malfunction, different protection parameters created according to the season or shift change and activated when
needed [19, 20].
Since the short circuit currents will change in the case of the island where the industrial distribution networks
fed from the electricity grid are fed by the redundant power systems, the set values of the relays need to be
changed [21-23]. Otherwise, its protection may disappear due to malfunction or failure to detect.
In order to get rid of the faulty trip condition described above, different setting group values created in the
protection relay and activated in the protection relay. The change of the setting groups done manually from the
top of the protection relay or automatically with the SCADA, which has time information and control authority
[23].
The protection relays that operate by being aware of the positions of the equipment in the region they protect
and the status of the system can change the current protection parameters according to different situations. Thus,
protection systems can adapt to changes in the network structure and operating conditions. It is defined as the
protection that can adapt to the changing conditions of the electricity distribution system, which is capable of
adapting the protection structure aimed at the correct use of different relay stages [5, 6,18]. The greatest
advantage of adaptable protection systems compared to conventional protection systems is that selectivity
provided much well by dynamically changing the opening settings of their protection relays according to the
distribution system's status.
In adaptive protection coordination, the relays must respond to changing system conditions and adapt the new
settings to new valid conditions, such as operational or topological changes [9]. Improper adjustment of the

410
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

relays, a much larger part of the network disabled with the unnecessary operation of any relay. This reduces
system reliability and increases the number of users who have no energy [6].
In order to increase the reliability of the protection relays, there is a need for communication systems that will
rapidly transmit the data at the remote points of the electricity distribution network to the relays. It is also an
economic approach to obtain the data regarding the load or network topology required by adaptive protection
systems over the existing SCADA. The methodology to be followed in order to implement adaptive protection
systems based on SCADA is summarized below.
• Create working scenarios of the electricity distribution system.
• Protection parameters are calculated and grouped separately by making fault analysis for each scenario.
• The number of setting groups is obtained as much as the number of scenarios.
• Setting groups defined to protection relays.
• Define all scenarios’ criteria to the SCADA system.
• Created all required commands at SCADA system, in order to change setting groups of protection relays.
• Commands parameterized on related RTUs to communicate with SCADA and protection relays.
• Thus, when the predetermined scenarios occur, it sends commands to the protection relays and changes the
protection-setting group.
The features of the SCADA system used in the recommended adaptive protection system summarized below.
There should be two redundant servers in the SCADA system. Having collected all numerical and analog values
from the field, RTU sends this information to the active SCADA server. The active server sent this information
to two operator computers running SCADA software. With the SCADA software installed on the operator
computers, these signals cause changes to the SCADA pages, giving the user an audible and visual warning.
Changes to the pages of the existing SCADA system made through the "Engineering Computer", which is
authorized to make changes in the project. The changes made here and then sent from the engineering computer
to the active server. The backup server and operator computers receive these updates automatically and update
their SCADA pages. There is a common local network connection among all the elements in the SCADA system.

3. AN APPLICATION OF ADAPTIVE OVER CURRENT PROTECTION BASED ON SCADA

3.1 Determination of the electrical distribution system scenario

In this section, the recommended SCADA based protection application is given to enable the setting group
values of an electrical distribution system protection relays operated as a ring network to be activated in the
protection relays appropriately depending on the system's feeding status and the number of diesel generators in
operation.
The electricity distribution system has three 1600 kVA powered diesel generators to ensure the continuity of the
energy in case of mains power failure. These diesel generators can be activated one, two or three at the same
time, when the energy system separated from the network and switches to island mode. Two interconnected feed
feeders working with interconnected grids and one bus powered by these feeders, 34.5 kV voltage level busbar,
two redundant transformers connected to this bar, each with 34.5 / 6.3 kV 10 MVA power transformer, one at
each transformer output. It is a ring network connected to two 6.3 kV busbars, 6.3 kV voltage substations
connected to each other by coupling.

411
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig. 1. A view of the electricity distribution system on SCADA

Many special pages designed for monitoring and control, which is the main feature of SCADA. In this study,
only the setting groups of the protection relays and the electrical distribution system single line diagram given
in Figure 1.

3.2 Determine the protection parameters for scenarios and grouping

Electricity distribution system three phase short circuit current values calculated by SINCAL software.
Calculation results for all different scenarios given in Table 1.

In Table 1, in some substations, it is seen that the rated current measured by the protection relay when the energy
system is connected to the grid is larger than the short circuit current value in a malfunction that may occur in
the case of feeding with a generator working independently from the grid. This situation makes it difficult to
parameterize the protection system by causing the nominal current and short circuit currents intertwined in
different operating scenarios. Therefore, the need to determine a different overcurrent and short-circuit current
setting for each operating condition has arisen. The cable of existing electricity distribution network can carry
maximum 526 A according to its catalog information. The maximum current value that a generator with 1600
kVA power given in Annex-C can provide at 6.3 kV voltage level has been calculated as 146 A.

412
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table. 1. Three-phase short-circuit currents for different supply conditions

Fault Location Fault Network G3 G2 Jen G1
Supply
Max Min Max Min Max Min Max Min
[A] [A] [A] [A] [A] [A] [A] [A]
TR-01 Network-1 10542 9534 2220 2010 1516 1375 776 705
TRM-04 111 101 23 21 16 15 8 7
TR-04 TRM-01 8161 7334 2009 1816 1393 1262 725 658
TRM-05 517 465 127 115 88 80 46 42
TR-05 TRM-04 7498 6718 1938 1749 1351 1223 707 641
TRM-06 635 569 164 148 114 104 60 54
TR-06 TRM-05 6380 5683 1798 1621 1266 1145 669 607
TRM-07 844 752 238 214 167 151 89 80
TR-07 TRM-06 5948 5285 1736 1564 1228 1110 652 592
TRM-09 930 827 272 245 192 174 102 93
TR-09 TRM-07 4564 4021 1498 1346 1076 971 581 527
TRM-11 1251 1103 411 369 295 266 159 144
TR-11 TRM-09 3599 3152 1280 1148 929 838 509 461
TRM-12 1560 1367 555 498 403 363 221 200
TR-12 TRM-11 2935 2563 1092 978 799 720 441 399
TRM-13 1874 1637 698 625 510 460 282 255
TR-13 TRM-12 2637 2300 995 891 729 657 404 366
TRM-14 2071 1807 781 700 573 516 317 287
TR-14 TRM-13 2075 1810 783 701 574 517 318 288
TRM-16 2631 2296 993 889 728 656 403 365
TR-16 TRM-14 1798 1571 664 595 485 437 267 242
TRM-17 3071 2684 1134 1016 828 747 457 414
TR-17 TRM-16 1642 1436 593 531 431 389 237 215
TRM-18 A 3399 2975 1227 1100 893 805 490 444
TR-18 TRM-18 A 1317 1159 441 396 318 287 172 156
TRM-19 4326 3806 1449 1302 1043 942 565 513
TR-18 A TRM-18 3754 3292 1319 1183 956 862 522 473
TRM-17 1502 1316 527 473 382 345 209 189
TR-19 TRM-18 1216 1072 395 355 283 256 153 138
TRM-20 4694 4140 1524 1370 1092 986 589 534
TR-20 TRM-19 1132 1000 357 321 255 230 137 124
TRM-21 5032 4447 1587 1427 1133 1024 608 552
TR-21 TRM-20 968 858 287 258 203 184 108 98
TRM-22 5765 5118 1709 1539 1211 1095 644 585
TR-22 TRM-21 762 680 208 187 146 132 77 70
Network-2 6803 6075 1854 1673 1300 1177 685 621

For the determination of the substation, which will be taken as basis for the protection parameterization process,
it is necessary to calculate the fault currents that will flow to the right and left of each substation at the time of
failure. These required current values calculated by SINCAL software and given in Table 2.

413
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 2. Minimum of 3-phase short-circuit current of transformer substations

Fault Fault Network 3 Jen 2 Jen 1 Jen
Location Supply
Min Difference Min Difference Min Difference Min Difference
[A] [A] [A] [A] [A] [A] [A] [A]
TR-01 SEBEKE-1 9534 9433 2010 1989 1375 1360 705 697
TR-04 101 21 15 7
TR-04 TR-01 7334 6868 1816 1700 1262 1182 658 616
TR-05 465 115 80 42
TR-05 TR-04 6718 6149 1749 1601 1223 1120 641 587
TR-06 569 148 104 54
TR-06 TR-05 5683 4931 1621 1407 1145 994 607 527
TR-07 752 214 151 80
TR-07 TR-06 5285 4459 1564 1320 1110 937 592 499
TR-09 827 245 174 93
TR-09 TR-07 4021 2919 1346 977 971 705 527 382
TR-11 1103 369 266 144
TR-11 TR-09 3152 1785 1148 650 838 475 461 261
TR-12 1367 498 363 200
TR-12 TR-11 2563 926 978 354 720 260 399 144
TR-13 1637 625 460 255
TR-13 TR-12 2300 494 891 191 657 141 366 78
TR-14 1807 700 516 287
TR-14 TR-13 1810 486 701 188 517 139 288 77
TR-16 2296 889 656 365
TR-16 TR-14 1571 1113 595 421 437 309 242 171
TR-17 2684 1016 747 414
TR-17 TR-16 1436 1539 531 569 389 417 215 230
TR-18 A 2975 1100 805 444
TR-18 TR-18 A 1159 2648 396 905 287 655 156 357
TR-19 3806 1302 942 513
TR-18 A TR-18 3292 1976 1183 710 862 518 473 284
TR-17 1316 473 345 189
TR-19 TR-18 1072 3068 355 1015 256 731 138 396
TR-20 4140 1370 986 534
TR-20 TR-19 1000 3447 321 1106 230 793 124 428
TR-21 4447 1427 1024 552
TR-21 TR-20 858 4260 258 1281 184 911 98 487
TR-22 5118 1539 1095 585
TR-22 TR-21 680 5396 187 1486 132 1045 70 552
SEBEKE-2 6075 1673 1177 621

Failure currents that will flow from neighboring substations TR-13 and TR-16 in TR-14, which is dark in Table
2, have the lowest difference between fault currents in all substations in the electricity distribution system. In
other words, the substation taken as basis for the protection parameterization process is TR-14. In Table 3, the
numerical values of the currents that will flow towards TR-14 by TR 13 and TR 16 to which connected as a ring
are given in case of failure in TR-14.

Table. 3 Short-circuit current to flow through TR13 and t R16

Minimum short circuit Minimum short circuit Load current supplied Cable ampacity
current to flow from TR 13 current to flow from TR 16 by the generator [A] rating [A]
[A] [A]
G1 288 365 146 526
G2 517 656 292 526
G3 701 889 438 526
Network 1810 2296 526

414
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In order to detect even the lowest fault current that will occur, the lowest value taken as reference from the
values in Table 2. When the lowest short circuit current values multiplied by the safety factor of 0.75, it does
not cause any error in the selectivity of the protection system. Since the highest current value that the power
distribution cable can withstand is 526 A, the value of “Setting Group D” is the same as the “Setting Group C”
value. For this reason, the four adjustment groups decrease to three.
Another magnitude needed after determining the electrical distribution system short circuit current value is the
overcurrent value. The maximum current that a generator can give is according to 146 A, and the safety factor
must take 1.1. The calculated protection parameters of the electricity distribution system given in Table 4.

Table 4. Electrical distribution system protection parameters

Minimum short Minimum short Load current Initial Short circuit Cable
circuit current circuit current supplied by value of current value ampacity
to flow from TR to flow from TR the generator inverse (Setting Group rating [A]
13 [A] 16 [A] [A] time curve Value) [A]
[A]
G1 288 365 146 160 216 526
G2 517 656 292 322 388 526
G3 701 889 438 482 525 526
Network 1810 2296 482 525 526

3.3 Identifying Scenarios in SCADA

The first process required to introduce different working scenarios to the existing SCADA system is to select
numerical signals that will allow the interpretation of the operating conditions in the field by the SCADA. If the
signal value is “0 (zero)” for generators, it indicates that it is not active, and “1 (one)” is active. 34.5 kV / 6.3
kV transformer feeder circuit breaker position is the most suitable solution for mains connection. In addition,
the circuit breaker position of the transformer output feeder 1210, which operates redundantly with the
transformer output feeder 1107, is also necessary for determining the energy supply state.

3.4 Electrical distribution system islands operation scenarios

Digital circuits created for the previously determined A, B, and C scenarios given in Figure 2.

a) b) c)
Figure 2. Scenario A, B and C set group digital circuit

As seen in Figure 2, after it is understood that the signal value of the cutters 1107 and 1210 from the selected
digital signals is zero (0), that is, it is disabled, the electricity distribution system has no connection with the
network. By obtaining the information that only one of the three generators connected to the system activated,
the SCADA system decides that the electricity distribution system is appropriate for the setting group A. This
numerical process introduced to the system in SCADA. B and C scenarios introduced to the system in a similar
way as scenario A.
After this process, four signals must defined for each protection relay in the current system, in order to send the
command groups to the corresponding protection relays, which are appropriate for the scenarios previously
defined to SCADA. 3 of these signals are digital signals and one is analog output signal. The setting is used to

415
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

send zero (0), one (1) and two (2) values to the group analog output signal protection relay. If the relay sends
zero (0), the setting group A will be active, if it sends one (1), the setting group B will be active, when the two
(2) sends, the setting group C will be active.
In order to send the analog output signal zero (0), one (1) and two (2) values to the protection relay, special
functions defined in SCADA. Functions as many as the number of setting groups introduced for each protection
relay. Commands functions sent automatically or manually to protection relay. While creating custom scripts
for sending commands to protection relays automatically, a special setting group page created to be sent to
protection relays at any time.
In order to do this process automatically, the previously determined scenarios are used. The output of each
scenario assigned to a signal that is not connected to the field in SCADA. When the result of logic operation of
the scenario is one (1), the value of this signal is one (1), and it ensures the operation of a specially assigned
Script.

CONCLUSION

In this study, SCADA based protection methodology has been developed and the application of the developed
methodology related to over current protection of a ring distribution electricity distribution system has been
implemented. The tests performed showed that SCADA correctly determines the operating status of the system
and sends correct commands to RTUs. RTUs activates the correct setting group in protection relays according
to the commands they received after checking the availability of the field. A protection system tests that can
monitor the electricity distribution system and ensure that the protection relays operate in the most suitable
setting group according to the condition of the system completed.
Using RTU and SCADA is not obligatory to ensure an adaptive protection system. For instance, if the adaptive
protection system applied for a small number of substations, taking advantage of the IEC61850 communication
protocol is smarter solution. However, if the electrical distribution system is huge and protection relays are
located different areas, using 61850 may not be efficient enough to ensure a reliable adaptive protection system.

REFERENCES

[1.] Kezunovic M, Ren J, Lotfifard S, Design, Modeling and Evaluation of Protective Relays for Power Systems, Springer,
Switzerland, 1-5, 2015.
[2.] Bian J J, Slone, A D, Tatro P J, Protection system misoperation analysis, IEEE PES General Meeting | Conference &
Exposition, 1-5, Washington,27-31 July 2014.
[3.] Özdemir A et al, Impacts of protection system misoperation on the reliability of Turkish National Power Transmission
System, 2011 International Conference on Advanced Power System Automation and Protection, 16-20 October.
2011 10.1109/APAP.2011.6180475
[4.] Yazdanpanahi H, Wilsun Xu, Yun Wei Li, A Novel Fault Current Control Scheme to Reduce Synchronous DG's Impact
on Protection Coordination, IEEE Transactions on Power Delivery, 29, 2, 542-551, 2014
[5.] Liang C, Khodayar M E, Shahidehpour M, Adaptive Protection System for Microgrids: Protection practices of a
functional microgrid system, IEEE Electrification Magazine, 2, 1, 66-80, 2014.
[6.] Haj-ahmed M A, Illindala M S, Investigation of protection schemes for flexible distribution of energy and storage
resources in an industrial microgrid, IEEE Trans. Ind. Appl., vol. 51, no. 3, pp. 2071–2080, May 2015
[7.] Contreras A F, Ramos G A, Rios M A, Methodology and Design of an Adaptive Overcurrent Protection for Distribution
Systems with DG, International Journal of Electrical & Computer Sciences, 12, 5, 1-10, 2012.
[8.] Baran M, El-Markabi I, Adaptive Over Current Protection for Distribution Feeders with Distributed Generators, Power
Systems Conference and Exposition, 715-719, New York, 10-13 October 2004.
[9.] Shen S. et al., An adaptive protection scheme for distribution systems with DGs based on optimized Thevenin
equivalent parameters estimation, IEEE Trans. Power Del., vol. 32, no. 1, pp. 411–419, Feb. 2017.
[10.] Brahma S, Girgis A., Development of adaptive protection scheme for distribution systems with high penetration of
distributed generation, IEEE Transactions on Power Delivery, 19, 1, 56 - 63, 2004.
[11.] Shih-Chieh Hsieh, Chao-Shun Chen, Cheng-Ta Tsai, Cheng-Ting Hsu, Chia-Hung Lin, “Adaptive Relay Setting for
Distribution Systems Considering Operation Scenarios of Wind Generators”, IEEE Transactions on Industry
Applications, 50, 2, 1356-1363, 2014.
[12.] Jain R, Lubkeman D L, Lukic S M, Dynamic Adaptive Protection for Distribution Systems in Grid-Connected and
Islanded Modes, IEEE Transactions on Power Delivery, Vol. 34, No. 1, February 2019, 10.1109 /
TPWRD.2018.2884705
[13.] Jing M, Wang X, Zhang Y, Yang O, Phadke A K, A novel adaptive current protection scheme for distribution systems
with distributed generation, International Journal of Electrical Power & Energy Systems, 43, 1, 1460-1466, 2012.

416
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[14.] Urdaneta, R. Nadira, and L. Jimenez P., Optimal coordination of directional overcurrent relays in interconnected
power systems, IEEE Transactions on Power Delivery, 3, 3, 903-911, 1988.
[15.] Mahari A, and Seyedi H, An analytic approach for optimal coordination of over current relays", IET Generation,
Transmission & Distribution, 7, 7, 674-680, 2013.
[16.] Noghabi A S, Mashhadi H R, Sadeh J, Optimal coordination of directional overcurrent relays considering different
network topologies using interval linear programming, IEEE Transactions on Power Delivery, 25, 3, 1348-1354,
2010.
[17.] Sachdev M S, Sidhu T S, Talukdar B K, Topology detection for adaptive protection of distribution networks,
International Conference on Energy Management and Power Delivery, 445-450, Westin Stamford, 21-23 November.
1995
[18.] Wadi M, Baysal M, Shobole A, Tur M R, Reliability Evaluation in Smart Grids via Modified Monte Carlo Simulation
Method, 7th International Conference on Renewable Energy Research and Applications (ICRERA), Paris, France,
14 - 17 October 2018, ss.841-845
[19.] Abdelaziz A Y, Talaat H, Nosseir A I, Hajjar A A, An adaptive protection scheme for optimal coordination of over
current relays, Electric Power Systems Research, 61, 1, 1 - 9, 2002.
[20.] Shah K R, Detjen E D, Phadke A G, Feasibility of adaptive distribution protection system using computer overcurrent
relaying concept, IEEE Transactions on Industry Applications, 24, 5, 792 - 797, 1988.
[21.] Moirangthem J, Krishnanand K R, Dash S S, Ramaswami R, Adaptive differential evolution algorithm for solving
non-linear coordination problem of directional over current relays, IET Generation, Transmission & Distribution, 7,
4, 329 - 336, 2013.
[22.] Coffele F, Booth C, Dysko A, An Adaptive Overcurrent Protection Scheme for Distribution Networks, IEEE
Transactions on Power Delivery, 30, 2, 561 - 568, 2014.
[23.] Shah P H, Bhalja B R, New adaptive digital relaying scheme to tackle recloser-fuse miscoordination during distributed
generation interconnections, IET Generation, Transmission & Distribution, 8 , 4, 682 – 688, 2014.

Appendix

Characteristics of the electrical distribution system

11 kV and 7 busbar electricity distribution system
Loads 0.75 MW, power factor: 0.9, Total load amount: 6 MVA
Generator power 600kW, power factor: 0.9
Line impedances: r = 0.48 Ω, x = 0.2859 Ω,
Relay trip value calculation:
Minimum short circuit current value 2150A, maximum load current 325A
Maximum load current < Relay trip value < Minimum short circuit current)
650 <Relay trip value <720 Relay trip value = 700A

Figure. Generator characteristics

417
 THD ANALYSIS FOR A PMSG WITH DIFFERENT MAGNET
STRUCTURE

Adem Dalcalı
Bandırma Onyedi Eylül University, Department of Electrical and Electronics Engineering, Faculty of Engineering and
Natural Sciences, Bandırma, Balıkesir, Turkey, [email protected], ORCID: 0000-0002-9940-0471

Erol Kurt
Gazi University, Department of Electrical and Electronics Engineering, Faculty of Technology, Ankara, Turkey,
[email protected], ORCID: 0000-0002-3615-6926

Dalcalı, A, Kurt, E. THD Analysis for A PMSG with Different Magnet Structure. 8 th Eur.
Cite this paper as:
Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In the present paper, the effects of different magnet structures to the cogging torque (CT) and total
harmonic distortion (THD) of the output waveforms from a permanent magnet generator has been
explored and reported theoretically. The generator is a radial flux machine and four different
structures are considered for the magnet arrangement in the rotor component. This three-phase
machine exhibit different behaviors toward various magnet structures, i.e. rectangular (R), inclined
slotted rectangular (ISR), skewed double rectangular (SDR) and inclined slotted skewed double
rectangular (ISSDR), respectively. It has been proven that both CTs and THDs vary significantly
for all types. Strictly speaking, the CT values change between 89.95 mNm and 436.75 mNm, and
the lowest CT is measured for the ISSDR magnet geometry, while the conventional R magnet
geometry gives the worst value with 436.75 mNm. Besides, the THD values vary between 1.63
and 3.55 for different magnet orientations. While the worst THD value is encountered for ISR
magnet, the best THD is obtained for SDR. We expect that those results would give important
information to the scientists and engineers in order to have more efficient machines.
Keywords: PMSG, Total harmonic distortion, Magnet structure
© 2020 Published by ECRES

1. INTRODUCTION

Energy consumption is increasing depending on the energy requirements arising from the increase of the world
population and the developing industry. This situation causes a limited reduction of fossil fuel reserves. It is
evident that those sources are not sustainable due to the fact that the reserves of fossil fuels will expire in the
near future and they even cause environmental damage right now during the energy production process world-
widely [1]. At the meantime, the popularity of cheap, local and environmentally friendly renewable energy
sources is increasing day by day. Wind energy, among other renewable energy sources, plays an important role
in ensuring energy reliability, reducing foreign dependency and solving environmental problems especially for
windy climatic countries [2]. That also increases the motivation on designing and producing new and efficient
energy conversion mechanisms, especially generators, energy harvesters, converters, and other power electronic
components [3,4].
Many types of generators are used for wind energy turbines. These have many advantages and disadvantages
compared to each other [5]. Whatever are they, the rare earth magnets made by NdFeB have been provided new
horizons on the way of having efficient, and high energy density machines. If they are cooled efficiently, radial
flux machines can have better torque values, and high-power density within a compact volume. The permanent
magnet (PM) machines can meet the needs for higher torque in less weight than asynchronous and switched
reluctance machines according to the literature [6,7]. PM generators can be used directly in wind turbines
without any need of gear system. These advantages are increasingly used in permanent magnet synchronous
generators (PMSGs) in renewable energy applications [8,9]. There are some studies in the literature on magnet

418
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

optimization by using the numerical and finite element method of PMSGs [10-12]. Those studies have generally
focused on optimizing cogging torque [13].
In the present study, the effects of 2.5 kW rated power have been examined. For that, initially different magnet
surface mounted PMSGs, which are used in wind turbines have been designed and simulated. Each simulation
procedure includes the analysis on the basis of magneto-static, and magneto-dynamic manners. While the
stabilities of the designs have been proven in finite element method analyzes, time-dependent electromagnetic
solutions have revealed the phase voltage waveforms, and cogging torques. Besides, total harmonic distortions
have been analyzed from the resulting waveforms. Using the graphical method, harmonic components and
angles of voltages are obtained and THDs in percentages are subtracted from the patterns.
The paper is organized as follows: Section 2 introduces the proposed designs with the required physical
parameters. The next section gives the main findings on the CTs and THDs. A comprehensive discussion is also
given in this section. Finally, the concluding remarks are reported in conclusions section.

2. DESIGNED PM GENERATOR SPECIFICATION

The generator, which is subject to the current study is a 2.5 kW powered, three phases, PM included machine
having a direct drive and an inner rotor. These generator types are frequently used in wind turbines to convert
the mechanical rotation energy to the electricity, since they do not require any complicated gear systems [14-
16]. The design parameters of the initial design are given in Table 1.

Table 1. The design parameters of the generator [17]

Parameter Value
Rated power (W) 2500
Frequency (Hz) 50
Stator outer diameter (mm) 280
Stator inner diameter (mm) 202
Stack length (mm) 54
Pole number 14
Number of slots 84
Magnet material N35
Stator, rotor material M19
Stator slot fill factor (%) 68.4

The PM materials can be mainly classified as Alnico, Ceramic and rare earth magnets. The use of rare earth
magnets is increasing day by day due to the decreasing costs with the developing production technology. NdFeB
type magnets have high magnetic flux density (B) and maximum magnetic energy given by the multiplication
of flux density Bmax and field strength Hmax [18]. Therefore, this type of magnet has been preferred in many
electric machine studies with lower volumes and high-power densities. The 3D model of the first design of the
generator is given in Figure 1.

Phase A
Phase B
Phase C
S
Magnet N
Magnet S

Figure 1. 3D model of initial designed generator.

419
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In this study, the model of the generator with four different magnet structures, whose performances have been
previously examined in Ref. [17], is taken as reference. In Fig. 2, rotor and magnet models of generators are
given.

(a) (b)

(c) (d)

Figure 2. The designed types: (a) Rectangular (R), (b) inclined slotted rectangular (ISR), (c) skewed double rectangular
(SDR) and (d) inclined slotted skewed double rectangular (ISSDR), respectively.

The design abbreviated with R as a genuine and traditional structure has the magnets with N-S-N-S orientation
without any slot or skewing. In the second design (i.e. ISR), a slot given in Fig. 2(b) is opened on the part of the
magnet by facing the stator surface. The slot is 1 mm depth and 2.15 mm thickness. In Design 3, the magnet is
divided into two equal parts with a certain inclination angle of 2.14⁰ and parts are skewed to each other. The last
design, Design 4 is obtained by opening two slots in the magnet.
According to the output performances of the designs, both rated powers and cogging torque values are
summarized as in Table 2. While Design 1 gives the highest power with 2.5 kW, the powers of others drop to
2.46, 2.45 and 2.43 for designs 2,3, and 4, respectively. That only coincides to 2.76 % maximum decrease in
other designs, too. That low decrease in power does not give any overall side effect, when the THD and CT are
taken into account in the following part of the study.

Table 2. The rated power and cogging torque of designs

Design Output power (W) Cogging torque (mNm)
Design 1 (R) 2500 436.75
Design 2 (ISR) 2462 434.58
Design 3 (SDR) 2455 159.60
Design 4 (ISSDR) 2431 89.95

3. RESULTS AND DISCUSSION

In power systems, it is desired that electrical energy is continuous and high quality. Harmonics are full multiples
of the fundamental frequency in energy systems, causing distortions in the main voltage and load current

420
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

waveforms. The presence of harmonic components causes a distortion in the waveform. These distorted
waveforms can be analyzed by applying a well-known method, i.e., Fourier Analysis. It is important for the
system to work in a desired way by performing these analyzes. When evaluated in terms of electrical machines,
harmonics cause heating in the magnetic components [19], voltage distortions in generators, heat production and
acoustic noises in motors [20]. In general, harmonics in electrical machines are exposed in the form of heat and
cause losses in the machine [21].
When periodic functions are expressed in the Fourier series, the first term can be written as a constant, and other
terms as a series of cosine and sinuses of the floors of a variable [22]. Using the graphical method, harmonic
spectra of nonsinusoidal waves can be determined from Fourier equations. In this method, the wave is divided
into equidistant vertical parts and the average value of each is determined [23]. Fourier coefficients are expressed
using Eqs. 1 and 2.
𝑚
2
𝐴𝑛 = ∑(𝑦𝑘 × 𝑠𝑖𝑛( 𝑛 × 𝜃𝑘 )) (1)
𝑚
𝑘=1

𝑚
2
𝐵𝑛 = ∑(𝑦𝑘 × 𝑐𝑜𝑠( 𝑛 × 𝜃𝑘 )) (2)
𝑚
𝑘=1

Here 𝑚 denotes the vertical division number of non-sinusoidal wave, 𝜃𝑘 denotes the mean value of the divided
parts in 𝑥-axis (𝑘 = 1, 2, … , 𝑚), 𝑦𝑘 gives the values of voltage and current, which is corresponding to 𝜃𝑘 and 𝑛
denotes the harmonic order [24]. The voltage can be expressed in Eq. 3 as it consists of the basic component and
harmonic components [21].

𝑉𝑠 (𝑡) = 𝑉𝑠1 (𝑡) + ∑ 𝑉𝑠ℎ (𝑡) (3)

ℎ=1

Here, 𝑉𝑠1 (𝑡) denotes the main component, and 𝑉𝑠ℎ (𝑡) shows the harmonic component. The waveforms of the
phase voltages obtained from the designed generators are given in Fig. 3.

Figure 3. Phase voltage waveforms of designs.

421
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The function of voltage amplitude can be expressed in terms of Eqs. 1 and 2 as follows by using the Fourier
coefficients [21,23]:

𝐵1 𝐵𝑛
𝑉(𝜃) = (√𝐴21 + 𝐵12 ) 𝑠𝑖𝑛 [𝜃 + 𝑡𝑎𝑛−1 ( )] +. . . + (√𝐴2𝑛 + 𝐵𝑛2 ) 𝑠𝑖𝑛 [𝜃 + 𝑡𝑎𝑛−1 ( )] (4)
𝐴1 𝐴𝑛

In Table 3, amplitude of voltage and phase angles are given following the graphical method.

Table 3. Main and harmonic components of the output voltage waveforms for four designs
Generator The degree of harmonic Amplitude of the Phase angle of the
components (n) voltage (V) voltages (degree)
1 (fundamental) 189.508 -16.73
5 2.461 67.06
7 3.079 49.82
Design 1
11 3.644 -79.17
13 0.970 54.05
15 0.717 -48.95
1 (fundamental) 186.480 -17.08
5 3.708 2.30
7 4.205 37.07
Design 2
11 2.396 -81.08
13 2.283 14.63
15 1.308 4.42
1 (fundamental) 189.098 -10.68
5 1.989 23.98
7 1.374 44.17
Design 3
11 1.528 -5.52
13 0.977 13.58
15 0.642 10.23
1 (fundamental) 183.841 -10.57
5 2.516 21.16
7 1.584 28.83
Design 4
11 1.068 6.92
13 1.015 1.31
15 0.820 23.43

The distortions in current or voltage waveforms are expressed in % by THD correlation. The voltage distortion
components expressed in Eq. 4 is put into Eq. 5, and the THD of the corresponding voltage is obtained:

𝑉𝑑𝑖𝑠 (𝑡) = 𝑉𝑠 (𝑡) − 𝑉𝑠1 (𝑡) ∑ 𝑉𝑠ℎ (𝑡) (5)

ℎ=1

𝑉𝑑𝑖𝑠
%𝑇𝐻𝐷 = 100 × (6)
𝑉𝑠1

The total harmonic distortion values of designs are given in Table 4.

Table 4. THD values of designs

Design THD (%)
Design 1 (R) 2.898
Design 2 (ISR) 3.554
Design 3 (SDR) 1.633
Design 4 (ISSDR) 1.853

422
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In the study, the harmonic spectrum have been extracted for each design with the help of MATLAB (License
number: 40692431) software following the use of main and harmonic components, whose amplitudes and angles
are given in Table 4.

(a) (b)

(c) (d)
Figure 4. Harmonic spectra of (a) design 1, (b) design 2, (c) design 3, (d) design 4, respectively.

Considering Table 4 and Fig. 4 together, the Design 2 has a relatively higher THD value compared to other
designs. As it is clear from Fig. 3, especially the 5th and 7th harmonics are higher in amplitude.

4. CONCLUSIONS

When the total harmonic distortions are examined, it is seen that the best design is Design 3 with 1.633%. When
the performance of this design is analyzed, it is seen that it has the lowest CT among the proposed models with
159.60 mNm torque. The generator with the lowest CT (89. 95mNm) is observed to have a THD of 1.853%.
However, that gives a 69 W drop in the output power of this machine compared to the genuine design (i.e.
Design 1). CT, which is an undesirable situation in PM machines, is expected to be as low as possible. The
%THD value is expected to be as low as possible. By considering those criteria, Designs 3 and 4 become the
most suitable ones. Indeed, the Design 4 provides an 80% improvement in CT compared to the conventional
machine type. It is obvious that design 4 offers the best performance among the all designs and that technique
can be used for other machines to have better CT and THD values.

REFERENCES

[1] Pareek, A, Dom, R, Gupta, J, Chandran, J, Adepu, V, Borse, PH. Insights into renewable hydrogen energy: Recent
advances and prospects. Materials Science for Energy Technologies 2020; 3: 319-327. Doi.org/10.1016/j.mset.2019.12.002
[2] IEA. World Energy Outlook 2014. International Energy Agency 2014; Paris, France.
[3] Mutlu, B, Kurt, E, Bizon, N, Lopez Guede, JM. Design and Fabrication of a New Micro-Power Scaled Electromagnetic
Harvester. Journal of Energy Systems 2019; 3(2): 51-66. DOI: 10.30521/jes.554900.

423
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[4] Bouloukza, I, Mordjaoui, M, Kurt, E, Bal, G, Ökmen, C. Electromagnetic design of new radial flux permanent magnet
motor. Journal of Energy Systems 2018; 2(1): 13- 27. DOI: 10.30521/jes.397836
[5] Polinder, H, van der Pijl, FFA, de Vilder, GJ, Tavner, P. Comparison of direct-drive and geared generator concepts for
wind turbines. In: Internetional Conference on Electric Machines and Drives; May 2005, pp. 543-550.
[6] Zeng, Y, Cheng, M, Liu, G, Zhao, W. Effects of magnet shape on torque capability of surface-mounted permanent
magnet machine for servo applications. IEEE Transactions on Industrial Electronics 2020; 67(4): 2977- 2990. DOI:
10.1109/TIE.2019.2910025
[7] Ma, Q, El-Refaie, A, Lequesne, B. Low-cost interior permanent magnet machine with multiple magnet types. IEEE
Transactions on Industry Applications 2020; 56(2): 1452-1463. DOI: 10.1109/TIA.2020.2966458
[8] Kurt, E, Gör, H. Electromagnetic design of a new axial flux generator. In: 6th International Conference on Electronics,
Computers and Artificial Intelligence; Oct. 2014, pp. 39-42.
[9] Zala, BO, Pugachov, V. Methods to reduce cogging torque of permanent magnet synchronous generator used in wind
power plants. Elektronika Ir Elektrotechnika 2017; 23 (1): 43-48. DOI: 10.5755/j01.eie.23.1.12714
[10] Jia, L, Lin, M, Le, W, Li, N, Kong, Y. Dual-Skew magnet for cogging torque minimization of axial flux PMSM with
segmented stator. IEEE Transactions on Magnetics 2020; 56(2). DOI: 10.1109/TMAG.2019.2951704
[11] Dajaku, G, Gerling, D. PM wind generator with reduced amount of rare-earth magnet material. In: International
Conference on Clean Electrical Power 2019, pp. 15-21.
[12] Habib, A, Che, HS, Rahim, NA, Tousizadeg, M, Sulaiman, E. A fully coreless multi-stator multi-rotor (MSMR) AFPM
generator with combination of conventional and halbach magnet arrays. Alexandria Engineering Journal 2020; 59: 589–
600. https://doi.org/10.1016/j.aej.2020.01.039
[13] Fang, H, Song, R, Cai, X. Analysis and reduction of cogging torque for magnetic-gear PMSG used in wave energy
conversion. In: 22nd International Conference on Electrical Machines and Systems 2019.
[14] Gul, W, Gao, Q, Lenwari, W, Optimal design of a 5-mw double-stator single-rotor PMSG for offshore direct drive
wind turbines. IEEE Transactions on Industry Applications Jan/Feb. 2020; 56(1): 216- 225. DOI:
10.23919/ICEMS.2018.8549413
[15] Zhang, Z, Zhao, Y, Qiao, W, Qu, L. A discrete-time direct torque control for direct-drive PMSG-based wind energy
conversion systems. IEEE Transactions on Industry Applications July/August 2015; 51(4): 3504- 3514. DOI:
10.1109/TIA.2015.2413760
[16] Zeng, X, Liu, T, Wang, S, Chen, Z. Comprehensive coordinated control strategy of PMSG-based wind turbine for
providing frequency regulation services. IEEE Access 2019; 7: 63944- 63953.DOI: 10.1109/ACCESS.2019.2915308
[17] Dalcalı, A, Kurt, E, Çelik, E, Öztürk, N. Cogging torque minimization using skewed and separated magnet geometries.
Journal of Polytechnic 2020; 23(1): 223-230. https://doi.org/10.2339/politeknik.552273
[18] Dalcalı, A. Influence of rotor magnet material and stator winding geometry on output torque in spherical actuator.
Journal of Engineering Sciences and Design 2019; 7(1): 145-151. DOI: 10.21923/jesd.437980
[19] Ingale, R. Harmonic analysis using FFT and STFT. International Journal of Signal Processing, Image Processing and
Pattern Recognition 2014; 7(4): 345-362. DOI: 10.14257/ijsip.2014.7.4.33
[20] Lo, WC, Chan, CC, Zhu, ZQ, Xu, L, Howe, D, Chau, KT. Acoustic noise radiated by PWM-controlled induction
machine drives. IEEE Transactions on Industrial Electronics 2000; 47(4): 880–889. DOI: 10.1109/41.857968 ·
[21] Ocak, C, Dalcalı, A. A Comparison of the THD values of four-pole BLDC motors with different slot and winding
arrangements. In: 8th International Conference on Advanced Technologies; August 26-30, 2019, pp. 70-73.
[22] Efe, S.B., “Harmonics in power systems and analysis of harmonics”, MSc Thesis, İnönü University, Malatya, 2006.
(In Turkish)
[23] Kocatepe, C, Uzunoğlu, M, Yumurtacı, R, Karakas, A, Arıkan, O. Harmonics in Electric Plants. Turkey: Birsen
Publishing House, 2003. (In Turkish)
[24] Mohan, N, Undeland, TM, Robbins, VP. Power Electronics. USA: John Wiley & Sons, 2003.

424
 1D MODELING OF OLIVE OIL MILL WASTEWATER
EVAPORATION WITH A PARABOLIC TROUGH SOLAR
COLLECTOR

Özel Binboğa
Gazi University Graduate School of Natural and Applied Sciences, Ankara, Turkey, [email protected]
ORCID: 0000-0003-3222-5637

M. Zeki Yılmazoğlu
Gazi University Graduate School of Natural and Applied Sciences, Ankara, Turkey, [email protected]
ORCID: 0000-0001-7874-768X

Binboga, Ozel, and Yilmazoglu M. Zeki, 1D Modeling of Olive Oil Mill Wastewater
Cite this paper as: Evaporation with a Parabolic Trough Solar Collector. 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: In this study, it is thought to circulate the working fluid through the solar parabolic collector and
a lagoon to increase the evaporation rate of the olive oil mill wastewater. In the conventional
approach, the wastewater is expected to evaporate naturally. However, with increasing rain, the
wastewater mixes with the subterranean water. It is planned that the working fluid, which is heated
in the solar collector, will flow into the lagoon and transferred its heat to increase the evaporation
rate of the wastewater. The model was simulated using MATLAB Simulink code. The effect of
different types of collectors on the working fluid, the evaporation effect of the lagoon area and the
lagoon area heat transfer coefficient were investigated. As a result, the working fluid reached the
highest temperature (boiling temperature) in the model#1, then in model#3 and model#2,
respectively. The highest temperature difference was obtained to be 5.6oC between the 1st and 2nd
models. The lagoon area is simulated as 225m2, 400m2, 625m2, and the highest value reached by
the lagoon temperature is respectively; 332.6 K, 335.7 K, 345.9 K. The lagoon heat transfer
coefficient is simulated at 20, 60, 90W/m2K and the highest value reached by the lagoon
temperature is respectively; 335.7 K, 353.1 K, and 358.8 K. Considering the amount of the
wastewater after evaporation, it is observed that this system would be more efficient when used in
large lagoons and with a collector structure as in model#1.
Keywords: Solar Energy, Parabolic Collector, Wastewater Treatment
© 2020 Published by ECRES
Nomenclature
OMW Olive Mill Wastewater
COD Chemical oxygen demand
𝑚̇ Working fluid mass flow rate (kg/s)
𝐶𝑝 Specific heat capacity (J/g°C)
𝐴𝑟𝑜 Outer surface area of receiver (m2 )
𝐴𝑐𝑖 Outer surface area of cover (m2 )
σ Stefan Boltzman constant (W/m2 K4)
𝑇𝑟 Receiver temperature (K)
𝑇𝑐 Cover temperature (K)
ԑ𝑟 Emissivity of receiver
ԑ𝑐 Emissivity of cover
𝐸𝑠 Solar Exergy
𝑄𝑠 Heat generation from sun (Q)
𝑇𝑎𝑚 Ambient temperature (K)
𝑇𝑠𝑢𝑛 Sun temperature (K)
𝑇𝑓𝑚 Mean temperature of fluid (K)
𝜌𝑓𝑙𝑢𝑖𝑑 Density of the working fluid (kg/m3)

425
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑚̇𝑖 Lagoon inlet mass flow rate of working fluid (kg/s)

𝑚̇𝑜 Lagoon outlet mass flow rate of working fluid (kg/s)
𝐸𝐴 Energy of pump inlet (W)
𝐸𝐵 Energy of pump outlet (W)
P Power of the pump (W)
ℎ𝐷 Mass transfer coefficient
𝑇𝑖𝑛 Inlet lagoon temperature (K)
𝑇𝑜𝑢𝑡 Outlet lagoon temperature (K)
𝑇𝑤𝑤 Temperature of wastewater (K)

1. INTRODUCTION

Olive is believed to have an eight thousand year history. Olive, which is one of the economically and
commercially important products for producer countries, is also a culture symbolizing of the Mediterranean
region. This product has a periodicity. For this reason, it shows the feature of giving a product for one year and
giving little or not. 97% of olive production in the world is grown in the Mediterranean region. In Turkey, the
average olive oil production in the last 5 years was approximately 175 thousand tons and the average olive
production was 413 thousand tons and olive oil production was 1 million 640 thousand tons. With the record
growth of young trees in 2017/18, total olive production was 2 million 100 thousand tons in the Turkey [1].
Olive oil is an olive by product produced from the succulent part (mesocarp) of the olive fruit without any
chemical treatment. European Union countries are at the top of olive oil production. Olive oil production in
Turkey, in 2017/18, is estimated to be 263 thousand tons [2]. Olive oil extraction rate is quite high. Olive
vegetation water, called olive OMW (olive mill waste water), consists of a mixture of water from olive structure,
olive washing water and added water during olive oil production. Olive oil production processes are divided into
two according to the operating mode of the process. The first one is the conventional pressing system known as
conventional, and the other is continuous systems called modern system. In addition, olive oil production
processes are divided into two according to the variety of products. As a result of the 3 phase production process,
the systems that produce olive oil, OMW and pomace and processes that produce a mixture of olive oil and
pomace OMW after the process are called 2-phase production processes. Olive oil mill wastewater, which occurs
in olive oil enterprises, causes environmental pollution due to the organic pollution it contains. The presence of
polyphenol in the land, leads to the pollution of groundwater and prevents the development of flora, which
performs biological treatment processes. Industrial and agricultural wastes are among the most important
environmental problems with their ionic metals and organic components. One of the environmental problems
that are generally experienced in the Mediterranean region is that olive OMW is left unconscious and purified
to nature or mixed with soil and groundwater. OMW is negatively affects environmental health. OMW is a
phenolic component that results from olive oil production. The properties of this by product vary according to
the olive oil production method. The composition of this water changes according to the type of olive, the
maturity of the olives, the content of the olive water, the cultivation of the soil and the climatic conditions.
OMW, which is discharged into the streams and soil without being purified, causes the formation of unbearable
odors, the destruction of agricultural production, contamination of groundwater, and increase of insects [2]. The
OMW issue is handled in all its dimensions by the Ministry of Agriculture and Forestry and the Ministry of
Environment and Urbanization. However, the systems set up for disposal did not provide a definitive solution
in terms of both cost and operation. Three methods can be mentioned for the solution of the problem. The first
is the conversion of existing 3-phase plants into 2-phase and the establishment of a rice-integrated plant for the
processing of aqueous waste resulting from 2-phase processing (the current structure forces it to 2-phase
production). The second approximation is, continue working in 3 phases, to process the wastewater in the
integrated facility, and to evaporate the resulting OMW in evaporation lagoons with sufficient capacities.

426
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)
Figure 1. (a.b) Olive Oil Mill Wastewater Natural Evaporation Lagoons

Another method is to establish a treatment plant and to purify the land water that has formed and to use it in
other areas (as a liquid and solid fertilizer as a result of appropriate processes due to the plant nutrients and
organic substances such as nitrogen and potassium it contains) [1]. OMW have a primary environmental
importance. These wastes can be considered both as a source to be recycled and as a waste to be treated.
Wastewater in olive oil factories has high organic matter, suspended solids and lipid content and is produced at
high flow rates. Many scientists are working on efficient and cost-effective recycling alternatives. Accordingly,
various alternatives and their combinations including chemical, mechanical, physical, biological and thermal
methods have been tested [3]. With the UF process, 99.9% oleic acid, 60.2% polyphenol and 65% COD removal
were achieved. In pre-treatment + Anaerobic treatment, 64.4-85% COD was removed [4]. 28% COD removal
was performed with ozone removal and 88% COD removal was achieved with aerobic treatment [5]. 80-90%
COD removal was achieved with advanced oxidation processes [6]. 70% COD and 90% phenol removal was
achieved by aerobic biological treatment [7]. 80-90% COD removal efficiency was achieved by chemical
treatment and 90% COD removal efficiency was achieved by biological treatment [8]. 41% and 85% total and
simple phenol, 70% COD removal was obtained [9]. 60% phenol and 23% COD were removed by Fenton
process, 86% COD and 70% phenol were removed by aerobic treatment [10]. Electro-Fenton + sedimentation
provided 53% COD, 77% AKM, 78% polyphenol, 92% fat removal [11]. Electrochemical treatment removed
COD 43%, color 90%, phenol 76%, turbidity 75%, AKM 71% [12] . In plasma treatment, COD and BOD rates
decreased by 94.42% and 95.37%, respectively. The amount of dissolved oxygen was increased from 0.36 to
6.97 mg / l [13]. Vacuum evaporation method was applied in a study conducted in 1997 in Italy for wastewater
treatment. This work was carried out in a vacuum evaporator, under 5kPa vacuum, at 380 ° C. Distillation
efficiency is about 90%. The chemical oxygen demand of the colorless distillate sent to the storage tank is 3000
- 4000 mg / L, and the high pollution load is caused by alcohol products consisting of fermenting sugar and
volatile organic matter contained in OMW [14].

The average solar energy density, i.e. solar constant, which is approximately 160 km from the earth's surface,
perpendicular to the arrival of the sun's rays at the upper limit of the atmosphere, is about 1.37 kW / m². The
region that makes best use of solar energy is the part of the equator located between 35° North and South
Latitudes. This region is called 'World Solar Belt'. It receives 2000-3500 hours of sun in a year and its solar
potential varies between 3.5-7 kWh / m² / day. The average annual amount of solar radiation in the world varies
between 2000-2500 kWh / m² in arid regions and 1000-1500 kWh / m² in higher latitudes. Turkey's territory lies
between 36°- 42° North latitude and 26 ° -45 ° East meridians. The actual area it covers, including lakes and
islands, is 814,578 km2. The projection area is 783.562 km2. Turkey is in a very advantageous position in the
solar energy potential due to its geographical location.

427
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)
Figure 2. (a)Solar Energy Atlas of Turkey (b) Olive Production Regions in Turkey

Solar collectors and thermal energy storage components are two core subsystems in solar thermal applications.
While solar collectors need to have good optical performance, thermal storage subsystems require high thermal
storage density
and excellent heat transfer rate. Solar collectors must have as much absorption as possible. It is important that
the storage system is small in volume so that it can absorb the heat at the required speed. In solar thermal
applications, solar radiation is absorbed by a solar collector as heat and then transferred to the working fluid (air,
water or oil). The heat carried by the working fluid can be used to provide domestic hot water / heating or to
charge a thermal energy storage tank from which heat can be drawn (for overnight or cloudy days). Parabolic
trough solar concentrators, the fluid passing through the absorber pipe placed along the axis of focus, the
temperature rises by taking the absorbed energy. Reflective surfaces designed in parabolic shape reflect the solar
radiation to the absorbent tube located along the axis of the focus. The energy coming on the absorbent pipe is
given to the fluid in the pipe and the temperature of the fluid is increased. High temperatures (above 300 oC) can
be achieved with parabolic trough solar concentrators. Çolak [15], designed mathematically modeled and
optimized a parabolic trough type solar collector that can follow the sun. El Fadar et al. [16], designed a cooling
system with parabolic trough type collectors. According to their results, they achieved higher performance with
a parabolic trough type collector than flat collectors. Fernandez et al. [17], examined the historical development
of the parabolic trough solar concentrator and the construction and marketing of this concentrator system over
the last century. They especially focused on systems producing electricity using steam at 400 oC. Ceylan and
Ergün [18], tested the temperature controlled parabolic corrugated solar collector system. Experiments were
carried out at 40-100ºC. In the study, the highest energy efficiency was found as 61.2% at 100ºC and the highest
exergy efficiency was 63% at 70ºC. Ghasemi and Ranjbar [19], investigated the effect of nanofluids on the
thermal performance of the parabolic solar collector, and it was stated that the nanofluid collector used better
than pure water.

2. MATERIAL AND METHOD

In this study, it is considered that system structure in which OMW do not move from lagoons and pollute soil
and groundwater by increasing the amount of evaporation from a lagoon supported by solar energy instead of
natural evaporation. For this purpose, a portable (mounted on the truck) solar energy system is considered,
considering that the system can be used in many tightening plants. It is intended to pass work fluid through the
parabolic trough collector and transfer the energy of the radiation from the sun to the fluid in this pipe. Heating
fluid heated in the collector is planned to enter the lagoon model and transfer its energy to OMW. The model
was simulated according to three different collector structures using MATLAB Simulink / Simscape program
(Fig.3).

428
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. 1D MATLAB Model of the System

2.1 Mathematical Modelling

Useful heat generation (𝑄𝑢 ) in parabolic trough collectors can be calculated by the energy balance in the working
fluid volume as shown below. 𝑇𝑖 and 𝑇𝑜 is inlet and outlet collector temperature of working fluid.

𝑄𝑢 = 𝑚̇ 𝐶𝑝 (𝑇𝑜 − 𝑇𝑖 ) (1)

The thermal efficiency (𝜂𝑡ℎ ) of the solar collector is equal to the ratio of the useful heat produced (𝑄𝑢 ) to the
current solar energy(𝑄𝑠 ). Eq. [2] is the definition of collector thermal efficiency.

𝑄𝑢
𝜂𝑡ℎ = (2)
𝑄𝑠

Solar energy absorbed (𝑄𝑎𝑏𝑠 ) by the absorber tube can be expressed with beneficial heat (𝑄𝑢 ) and thermal
losses (𝑄𝑙𝑜𝑠𝑠𝑒𝑠 ) as shown Eq. [3].

𝑄𝑎𝑏𝑠 = 𝑄𝑢 + 𝑄𝑙𝑜𝑠𝑠𝑒𝑠 (3)

Thermal losses of the solar collector can be calculated by examining the heat transfer mechanism from the
absorber to the glass tube. Due to the vacuum between the absorber and the glass tube, the phenomenon of
convection is neglected and only radiation thermal losses are taken into the calculations

𝐴𝑟𝑜 𝜎(𝑇𝑟 4 − 𝑇𝑐 4 )
𝑄𝑙𝑜𝑠𝑠𝑒𝑠 = (4)
1 1 − 𝜀𝑐 𝐴𝑟𝑜
𝜀𝑟 + 𝜀𝑐 ( 𝐴𝑐𝑖 )

Exergy analysis of the solar collector is carried out by making a detailed analysis including the exergy of solar
radiation, useful exergy production (𝐸𝑢 ), exergy losses and exergy destruction. In exergy analysis of parabolic
trough collectors, it is assumed that only rays that can be assumed to be diluted are used. Thus, the Petela model
can be used to estimate the exergy flow in incoming solar radiation. This model takes into account that the sun
is a radiation temperature depot estimated to be 5770 K in the outer layers. Eq. [5] shows the exergy flow of
undiluted solar radiation.

4 𝑇𝑎𝑚 1 𝑇𝑎𝑚 4
𝐸𝑠 = 𝑄𝑠 [1 − ( )+ ( ) ] (5)
3 𝑇𝑠𝑢𝑛 3 𝑇𝑠𝑢𝑛

Useful exergy output can be calculated according to the Eq. [6].

429
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑇𝑜 ∆𝑃
𝐸𝑢 = 𝑄𝑢 − 𝑚̇ 𝐶𝑝 𝑇𝑎𝑚 𝑙𝑛 [ ] − 𝑚̇ 𝑇𝑎𝑚 (6)
𝑇𝑖 𝜌𝑓𝑙𝑢𝑖𝑑 𝑇𝑓𝑚

Eq. [6] can be applied to both liquid and gas working fluids. Especially for liquids, the pressure drop (ΔP) is
usually small and can be neglected. On the other hand, working with gas working fluids is associated with high
pressure losses due to the low density of gases and needs to be considered. The exergy efficiency (𝜂𝑒𝑥 ) of the
solar collector is defined as the ratio of its useful production to exergy input, as the Eq. [7] shows.

𝐸𝑢
𝜂𝑒𝑥 = (7)
𝐸𝑠

The following continuity equations can be used for the circulation pump to be used in the system. 𝑃𝑖 and 𝑃𝑜 is
power inlet and outlet of the pump. 𝜌𝑖 and 𝜌𝑜 inlet and outlet density of the working fluid.

𝑚̇𝑖 + 𝑚̇𝑜 = 0 (8)

𝐸𝐴+ 𝐸𝐵 + 𝑃𝑝𝑢𝑚𝑝 = 0 (9)

𝑃𝑜 − 𝑃𝑖
𝑃𝑝𝑢𝑚𝑝 = 𝑚̇ 𝑖 𝜌 + 𝜌 (10)
𝑖 𝑜
2

The amount of heat transfer from the pipe bundle to the OMW can be expressed by the Eq. [11].

𝑞 = ℎ𝐴(𝑇 − 𝑇𝑠 ) (11)

A= Area of the pipe bundle (m2)

h= Heat transfer coefficient (W/m2K)
T= Pipe bundle temperature(K)
Ts = Surrounding temperature (K)

The amount of water vapor transfer (ṁs ) from the OMW to the air per unit time is given Eq. [12]. Cy is the
OMW concentration and Cb is the air concentration.

𝑚̇𝑠 = ℎ𝐷 𝐴(𝐶𝑦 − 𝐶𝑏 ) (12)

The following Eq. [13] can be written for the working fluid;

𝑚̇𝑖 = 𝑚̇𝑜 (13)

Conservation equation for moist sample is given Eq. [14]. Wi and Wo are described as lagoon inlet and outlet
work done by the working fluid. 𝑚̇𝑤𝑤 is the mass flow rate of OMW.

𝑊𝑖 𝑚̇𝑖 + 𝑚̇𝑤𝑤 = 𝑊𝑜 𝑚̇𝑜 (14)

The amount of energy flow inlet (𝑄̇𝑖𝑛 ) and outlet (𝑄̇𝑜𝑢𝑡 ) the lagoon.

𝑄̇𝑖𝑛 = 𝑚̇ 𝐶𝑝 (𝑇𝑖𝑛 − 𝑇𝑤𝑤 ) (15)

𝑄̇𝑜𝑢𝑡 = 𝑚̇ 𝐶𝑝 (𝑇𝑜𝑢𝑡 − 𝑇𝑤𝑤 ) (16)

430
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2.2 Economic Analysis

The main purpose of this study is to perform the disposure of the aqueous waste economically and quickly after
the olive oil production process. For this reason, the 3D mechanical design of the system is generated so that
olive oil factories can use it integrated and through rental. The current approach in Turkey is converting 3 phase
systems to 2 phase systems. By converting the olive oil production method to 2 phase system transporting of
olive mill waste water become problematic for the olive pomace companies. Therefore, this offered system
eliminates the transportation problem and enables using 2 phase systems by evaporating most of the water with
solar energy. For both uses, a lagoon should be built in the olive oil plant to meet the annual amount of waste.
All equipment in the system must be deployed within the olive oil factory for integrated use. In the rental method,
only the construction of the lagoon will be sufficient. In the design of the system, a parabolic trough solar
collector, circulation pump, solar panel and inverter that will provide the energy of the pump and the collector
solar tracking mechanism are used. In order to circulate the heating fluid over the lagoon, it is envisaged to use
the type of pipe known as HDPE in the literature and industry.

(a) (b)
Figure 4. (a) 3D Model of the System (b) Top view of 3D Model of the System
In Turkey, generally operating in the Marmara Region, lagoon areas have been selected as follows when viewed
in the direction of production to factories in different sizes.
Table 1. Types of Evaporation Lagoons
Small Scale Medium Scale Large Scale
Annual Production 205 tons 569 tons 1.025 tons
Daily Production 561 kg 1,6 tons 3 tons
OMW Amount (Daily) 1,8 tons 4,8 tons 9 tons
3
Lagoon Volume 1.609m 4.589m3 8.604m3
Lagoon Dimensional 15x15x3 20x20x3 25x25x3
Number Of Lagoons 2 4 5

For 3-phase systems, during the olive oil processing process, the approximate output for 1 ton of olive input is
considered as 500 kg of solid waste. Approximately 50% of this solid waste is assumed to be waste water. For
a system like this study, more than one lagoon can be built into the olive oil factory. However, each lagoon can
be subjected to evaporation in one cycle. The purpose of this is to reduce the initial investment cost and to use
solar energy more efficiently. The high temperature obtained from the sun can complete its cycle in the
circulation pipe without losing its heat. Using the 2020 catalog data, the installation costs of these 3 plants for
an integrated system can be calculated as follows.
Table 2. 1st Investment Cost For Integrated System
225m2 400m2 625m2
Lagoon Area Built 10.800,00TL 19.200,00TL 30.000,00TL
Wire Cloth 2.347,00TL 4.174,00TL 6.523,00TL
Parabolic Collector 23.154,00TL 23.154,00TL 23.154,00TL
Circulation Pipe 2.145,00TL 2.145,00TL 2.145,00TL
Circulation Pump 1.191,00TL 1.191,00TL 1.191,00TL
Solar Energy Sys. 7.487,00TL 7.487,00TL 7.487,00TL
Inv. Cost / Lagoon 47.124,00TL 57.351,00TL 70.500,00TL
TOTAL 94.248,00TL 229.400,00TL 352.500,00TL

431
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The amount gained from the annual solid pomace sales can be calculated by looking at the amount of annual
solid waste and evaporation rate.
Table 3. Total Gained From Olive Pomace Sales
Small Scale Medium Scale Large Scale
Olive Oil Production 205 ton 569 ton 1.025 ton
Amount of Solid Waste 615 ton 1.7 ton 3.075 ton
Amount of OMW 308 ton 850 ton 1.537 ton
Olive Pomace Price 750 TL 750 TL 750 TL
TOTAL 231.000,00TL 637.500,00 TL 1.152.750 TL

For a large-scale factory, an economic analysis was performed assuming that five lagoons installed. In this case,
the initial investment cost will be approximately 352.500,00 TL. For a factory of this size, the payback period
of the system is about 4 months. The system payback period for a medium-sized factory is about 4 months.
Similar calculations were made for factories with small-scale production. The system payback time for a small-
scale factory is about 5 months. As can be seen from the calculations above, it is seen that the longest period is
in the large scale factory since the payback periods are compared. Therefore, if the integrated system is found
to be costly, it is also possible to use the study by rental. This means that a new sector is created in the olive
industry for use. Factories that avoid the initial investment cost will choose the rental model of the system. Thus,
they will avoid the cost of the pump, collector, solar energy system and other fittings. Only the cost of wire mesh
and lagoon construction will be sufficient to use this system. In Table x, the first investment cost results are
given for the rental system in a large-medium-small scale companies.

Table 4. 1st Investment Cost For Rental System

Small Scale Medium Scale Large Scale
Lagoon Area Built 10.800,00TL 19.200,00TL 30.000,00TL
Wire Cloth 2.347,00TL 4.174,00TL 6.523,00TL
Number of Lagoons 2 4 5
TOTAL 26.294,00TL 93.496,00TL 182.615,00TL

3. RESULTS

Working Fluid Temperature (K) Lagoon Temperature (K)

370 335
360 330
350 325
340 320
315
330
310
320 305
310 300
300 295
290 290
15:30
16:30
09:30
10:00
10:30
11:00
11:30
12:00
12:30
13:00
13:30
14:00
14:30
15:00

09:30
10:00
10:30
11:00
11:30
12:00
12:30
13:00
13:30
14:00
14:30
15:00
15:30
16:30

1st Model 2nd Model 3rd Model 1st Model 2nd Model 3rd Model

(a) (b)
Figure 5. (a) Working fluid temperature (b) Lagoon temperature

It is seen that in Fig. 5a., the model with the highest working fluid temperature is the 1 st model. Then, it is seen
as 3rd model and 2nd model, respectively. One of the advantages of the model 1 over model 2 is that the diameter
of the absorbent tube and glass tube is larger. Thus, it showed more absorption feature of solar radiation.
Likewise, it has been seen that the 3rd model has a higher heat transfer ability than the 2nd model. It can be said
that the absorption of the 3rd model is less than that of the 1st model. In other words, in addition to the collector
absorbency feature in model 3, the reflectance feature of the sun rays increased. In the second model, the heat

432
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

losses to the environment have decreased with the shrinkage of the outer surface area of the pipe. In order for
evaporation to be efficient, the system changed the flow when the work fluid reached 363 K and kept the work
fluid below 363K. It is assumed that the working fluid, which is transported by the pipe connections between
the collector and the lagoon, does not suffer from thermal loss during transportation. Considering the lagoon
temperatures, in parallel with the collector temperatures, the highest temperature is in the 1 st model, then
respectively; It was seen in the 3rd and 2nd models (See Fig. 5b.). The model with the highest temperature increase
in the morning hours is the 1st model. This will have a positive effect on the evaporation of olive waters in the
lagoon. After the simulations, there is no high temperature differences were observed between the 2 nd and 3rd
models in the morning hours. A stable increase is observed in the lagoon temperature in the section until 13.30.
After this hour, especially in the 1st model, the work fluid has reached the boiling temperature. Therefore, the
flow rate has been reduced by the system. However, the circulation of the work fluid continued just below the
boiling temperature. Based on this, the lagoon temperature continues to increase. After 13.30, the temperature
increase in the lagoon in the 1st model is less than the 2nd and 3rd models. Because the lagoon temperature in
model 1 is closer to the equilibrium temperature when the work fluid reaches the boiling temperature.

Working Fluid Pressure (MPa) Mass Flow Rate (kg/s)

3.50E+08 3.5
3.00E+08 3
2.50E+08 2.5
2.00E+08 2
1.50E+08 1.5
1.00E+08 1
5.00E+07 0.5
0.00E+00 0

11:00

13:30

16:30
09:30
10:00
10:30

11:30
12:00
12:30
13:00

14:00
14:30
15:00
15:30
11:00

12:00
09:30
10:00
10:30

11:30

12:30
13:00
13:30
14:00
14:30
15:00
15:30

1st Model 2nd Model 1st Model 2nd Model

3rd Model 3rd Model
(a) (b)
Figure 6. (a) Fluid pressure (b) Mass flow rate

Fig. 6b. shows that the model with the highest-pressure value was the 3rd model, then the 1st and the 2nd models,
respectively. The main reason of the 3rd model pressure values are higher than other models can be considered
to be 2 times longer than the other models. The diameters of the model 1 and 3 are larger than the model 2.
Therefore, the surface area affecting the work fluid is more than the 2 nd model. This causes the pressure value
to be higher in 1st and 3rd. Looking at the pressure graph, all three models increase up to certain hours of the day.
There is a decrease after a certain hour. The reason for this increase and decrease is related to the increase and
decrease of the flow rate. The flow is primarily increased to increase the amount of heat transfer of the work
fluid. The pressure increased as the flow rate increased. When the work fluid reached the boiling temperature,
the flow was kept below the boiling temperature by reducing the flow. In order to bring it below the boiling
temperature, the flow rate also caused the pressure drop.

433
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Exergetic Efficiency Thermal Efficiency

100% 100%

80% 80%

60% 60%

40% 40%

20% 20%

0% 12:00 0%

15:30

11:00

15:30
09:30
10:00
10:30
11:00
11:30

12:30
13:00
13:30
14:00
14:30
15:00

09:30
10:00
10:30

11:30
12:00
12:30
13:00
13:30
14:00
14:30
15:00

16:30
1st Model 2nd Model 3rd Model 1st Model 2nd Model 3rd Model

(a) (b)
Figure 7. (a) Thermal Efficiency (b) Exergetic Efficiency

Considering exergetic efficiency, the most efficient model was calculated as the 1 st model and then the 2nd and
3rd models, respectively. In model 1, thermal losses are less than other models, so heat transfer to the working
fluid is higher than other models. 3. exergetic efficiency is lower than other models since thermal losses are high
in the model. When the thermal efficiency graph is examined, it is seen that the 1 st model has the highest
efficiency. One of the reasons for this, as mentioned earlier, is the low thermal losses in this model. The
efficiency of all systems is low in the morning hours. As the solar radiation increases, the system efficiency
increases.

Lagoon Temperature (K) OMW Volume (m3)

(1st Model)
1800
350 1700
340 1600
330 1500
1400
320
1300
310
1200
300 1100
290 1000
15:30
16:30

11:30

15:00
09:30
10:00
10:30
11:00
11:30
12:00
12:30
13:00
13:30
14:00
14:30
15:00

09:30
10:00
10:30
11:00

12:00
12:30
13:00
13:30
14:00
14:30

15:30
16:30

A=225m2 A=400m2 A=625m2 A=225m2 A=400m2 A=625m2

(a) (b)
Figure 8. (a) Lagoon Temperature (b) Amount of OMW Evaporation

As the lagoon area increases, the temperature inside the lagoon also increases (See Fig. 8a.). One of the main
reasons for this is the increase of the heat transfer surface of the lagoon area. At the same time, as the area of the
lagoon grows, the frequency of the pipe bundle on the lagoon will increase, and the existing area will be subject
to more fluid circulation. The collector model was chosen as the first model and the system was simulated in
different area sizes. In the graphic above, it is assumed that all 3 lagoon models have 1720m3 OMW in their
reservoir in the first place. All lagoon models are simulated under the same conditions. Looking at the results,
the model with the highest evaporation is 625m2, then 400m2, 225m2 respectively. It is stated in the previous
graphs that as the lagoon area increases, the temperature inside the lagoon also increases. As the temperature
inside the lagoon increased, evaporation in OMW increased. Thus, the amount of water vapor separated from
the pomace will increase accordingly. The amount of evaporation is high, it is important in terms of evaporation

434
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

time. Rapid evaporation time will be important in terms of cycle time in factories with high production volume.
This is a subject of high importance for olive oil producers for the pomace trade. In the morning, lagoons with
areas of 625m2 and 400m2 are undecided as the amount of evaporation. The reason for this is that the area is
large and the heat is not trapped inside the lagoon. As the lagoon temperature increases, stability increases.

Lagoon Temperature (K) ΔV

(A=400m2) (m3)
370 6
5
350
4
330 3
2
310
1
290 0
13:00

14:30

16:30
09:30
10:00
10:30
11:00
11:30
12:00
12:30

13:30
14:00

15:00
15:30

09:30

12:00

14:30
10:00
10:30
11:00
11:30

12:30
13:00
13:30
14:00

15:00
15:30
16:30
h=20 h=60 h=90 20-60 20-90 60-90

(a) (b)
Figure 9. (a) Lagoon Temperature (b) Amount of OMW Evaporation

In order to decide on the right lagoon model, the first collector model was simulated according to different heat
transfer coefficients within the lagoon. The purpose of this simulation is to increase the heat transfer coefficient
of the environment and to observe the heat transfer relationship between the high temperature fluid in the pipe
bundle and the OMW in the lagoon. As the heat transfer coefficient, respectively; 90, 60, 20 W/m2K values were
selected. When looking at the graph, the heat transfer coefficient value with the highest temperature in the lagoon
is 90 W/m2K. Then, respectively; 60 and 20 W/m2K. In Fig. 9a., lagoon temperature differences between heat
transfer coefficients are given during the day. Although there is no significant difference between 60 and 90
W/m2K in the morning hours, this difference has opened up as the temperature increases. To increase the heat
transfer coefficient, a fan can be placed between the bundle and the lagoon. Thus, the speed of high temperature
air will increase and Reynolds number will increase accordingly. The Nusselt value of the air with rising
Reynolds number will also increase. Thus, the heat transfer coefficient will increase and the amount of heat
transfer will increase accordingly.

Small Scale Facility Large Scale Facility

OMW Transport Cost OMW Transport Cost
Transport Price (TL)
Transport Price (TL)

30.000.00 150.000.00
20.000.00 100.000.00
10.000.00 50.000.00
0.00 0.00
12.5 25 50 100 12.5 25 50 100
Trasport Distance (km) Trasport Distance (km)

Solid Pomace Aqueous OMW Solid Pomace Aqueous OMW

(a) (b)
Figure 10. (a,b) Small and Large Scale Facility OMW Transportation Cost

As can be seen from Fig. 10a,b., transporting aqueous OMW is more costly than transporting in solid form. One
of the biggest aims of the study is to reduce the OMW transportation cost of olive oil plants. The transportation
cost after the evaporation process is 2 times decreased according to the Fig. 10. The distance of the pomace plant
from the olive oil facility is important in terms of transportation cost. The transportation cost increases as the
distance increases. However, transportation costs increase depending on the amount of waste generated after the

435
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

olive oil process. The highest transportation cost is in the large scale facility, then in the medium and small scale
facility respectively.

4. CONCLUSION

The effect of different types of collectors on the working fluid outlet temperature, the effect of the lagoon area
and the lagoon area heat transfer coefficient to the evaporation of OMW were investigated. As a result, the
working fluid is at the highest temperature (boiling temperature), first in the 1st model and then in order; It
reached in the 3rd and 2nd models. The highest temperature difference was seen between the 1 st and 2nd models.
The lagoon area is simulated as 225 m2, 400 m2, 625 m2 and the highest value reached by the lagoon temperature,
respectively; 332,6 K, 335,7 K, 345,9 K. The lagoon heat transfer coefficient is simulated at 20, 60, 90 W/m2K
and the highest value reached by the lagoon temperature is respectively; 335,7 K, 353,1 K, 358,8 K was seen as.
Considering the amount of the wastewater after evaporation, it is observed that this system would be more
efficient when used in large lagoons and with a collector structure as in 1 st model. More than one lagoon can be
built into the olive oil company for a system like this study. However, each lagoon can be subjected to
evaporation in one cycle in itself. Installation costs were calculated for an integrated system using the catalog
data for large, medium and small scale olive oil companies in accordance with the selected area sizes. If large,
medium and small-scale companies establish an integrated facility, the initial investment cost is respectively;
352.500,00 TL; 225.400,00 TL; 94.248,00 TL. However, the payback period of the facilities was found to be;
4; 4; 5 months. The transportation cost after the evaporation process is 2 times decreased. As a result of this
study, by using the system both integrated and by rental; Additional earnings for olive oil companies, the
formation of a new sector (for the rental system) and the prevent OMW release to the nature and not destroying
the flora. It is concluded that the system has a sustainable structure due to the using solar energy.

REFERENCES
[1] T.C. Ticaret Bakanlığı Esnaf, Sanatkarlar ve Koperatifçilik Genel Müdürlüğü. (2019). 2018 Yılı Zeytin Ve Zeytinyağı Raporu.
[2] Kurdaş, İ. (2011). Zeytin Karasuyundaki Kirleticilerin Gideriminin İncelenmesi. İstanbul Üniversitesi, Fen Bilimleri Enst itüsü.
[3] Azbar, N., Bayram, A. Filibeli, A., Müezzinoğlu A., Şengül, F., Özer, A. (2010). A Review Of Waste Management Options In Olive
Oil Production. Critical Reviews in Environmental Science and Technology, 34:209-247.
[4] Beccari, M., Majone, M., Riccardi, C., Savarese, F., Torrisi, L. (1999). Integrated Treatment Of Olive Oil Mill Effluents: Effect Of
Chemical And Physical Pretreatment On Anaerobic Treatability. Water Sci. Technol.
[5] Benitez, F.J., Beltran-Heredia, J., Acero, J.L., Pinilla, M.N. (1997). Simultaneous Photodegradation And Ozonation
Plus Uv Radiation Of Phenolic Acids. Major Pollutants In Agro-Industrial Wastewaters, J. Chem. Technol.
Biotechnol.
[6] Benitez, F.J., Beltran-Heredia, J., Torregrosa, J., Acero, J.L. (1999). Treatment Of Olive Mill Wastewaters By
Ozonation, Aerobic Degradation And The Combination Of Both Treatments. Journal of Chemical Technology and
Biotechnology.
[7] Beltrán-Heredia, J., Torregrosa, J., García, J., Domínguez, J.R., Tierno, J.C. (2001). Degradation Of Olive Mill Wastewater By The
Combination Of Fenton’s Reagent And Ozonation Processes With An Aerobic Biological Treatment, Water Science and
Technology.
[8] Bressan, M., Liberatore, L., D’alessandro, N., Tonucci, L., Belli, C., Ranali, G. (2004). Improved Combined
Chemical And Biological Treatments Of Olive Oil Mill Wastewaters. Journal of Agricultural and Food Chemistry.
[9] Kotsou, M., Kyriacou, A., Lasaridi, K., Pilidis, G. (2004). Integrated Aerobic Biological Treatment And Chemical Oxidation With
Fenton’s Reagent For The Processing Of Green Table Olive Wastewater. Process Biochemistry.
[10] Bettazzi, E., Morelli, M., Caffaz, S., Caretti, C., Azzari, E., Lubello, C. (2006). Olive Mill Wastewater Treatment:
An Experimental Study. Water Science&Technology.
[11] Khoufi, S., Alouni, F., Sayadi, S. (2006). Treatment Of Olive Oil Mill Wastewater By Combined Process Electro-
Fenton Reaction And Anaerobic Digestion. Water Research.
[12] Khoufi, S., Feki, F., Sayadi, S. (2007). Detoxification Of Olive Mill Wastewater By Electrocoagulation And
Sedimentation Processes, Journal of Hazardous Materials.
[13] Ibrahimoglu B., Yilmazoglu, M. Z. (2018). Disposal Of Olive Mill Wastewater With DC Arc Plasma Method. Journal of
Environmental Management.
[14] LED Italia. (1997). Olive Mill Water Treatment Through Vacuum Evaporation. Technical report 7, LED Italia s.r.l.
[15] Çolak, L. (2003). Güneşi Takip Eden Parabolik Oluk Tipi Güneş Kolektörlerinin Matematiksel Modellemesi, Tasarımı Ve Teknik
Optimizasyonu, Gazi Üniversitesi, Ankara.
[16] El Fadar, A., Mimet, A., Azzabakh, A., Perez-Garcia, M., Castaing, J. (2009). Study Of A New Solar Adsorption Refrigerator
Powered By A Parabolic Trough Collector.
[17] Fernandez-Garcia, A., Zarza, E., Valenzuela, L., Perez, M., (2010). Parabolic-Trough Solar Collectors And Their Applications.
[18] Ceylan İ., Ergün A. (2013). Thermodynamic Analysis of a New Design of Temperature Controlled Parabolic Trough Collector.
[19] Ghasemi, S. E. and Ranjbar, A.A. (2016). Thermal Performance Analysis Of Solar Parabolic Trough Collector Using Nanofluid As
Working Fluid: A CFD Modelling Study, Journal of Molecular Liquids.

436
 HIGH VOLTAGE SOLAR INVERTER STRUCTURES

Ahmet Yildiz
Natural and Applied Sciences , Solar Energy Institute , Ege University, Bornova, Izmir TR 35100 , Turkey,
[email protected] ORCID: 0000-0001-6797-8511

Onder Ozgener
Solar Energy Institute, Ege University, Bornova , TR 35100 Izmir, Turkey, [email protected]
ORCID: 0000-0001-6792-2198

Leyla Ozgener
Department of Mechanical Engineering, Faculty of Engineering, Manisa Celal Bayar University, Muradiye, TR 45140
Manisa, Turkey [email protected], Solar Energy Institute, Ege University, Bornova , TR 35100 Izmir, Turkey,
[email protected] ORCID: 0000-0002-7670-0842

Yildiz A., Ozgener O., Ozgener L., Solar inverter structures with high voltage use. 8th Eur.
Cite this paper as:
Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: This study aims to achieve high voltage AC levels in a unique inverter structure to create a more
compact, easy to use, reliable and greater powered structure. Inverter structures with high voltage
outputs are questioned in regards to technical and commercial aspects. Inverter designs at high
voltage levels for a certain power, are compared with each other and relevant low voltages. The
computer aided simulation programs were used for this purpose. The LCL filter structure is an
important component that is significantly affected by the variation of voltage. This situation
critically affects inverter design and size. The obtained results were questioned in terms of
advantages, disadvantages, opportunities and threats. Despite the disadvantages of the proposed
structures in terms of cost, size and weight, they can be feasible above certain power levels.
Increasing the voltage level can also contribute to the inverter size and quantity optimization,
especially in GW sized photovoltaic sites. The inverter output voltage level can be optimized with
installed power and number of inverters to be used.

Keywords: Solar inverters, high voltage use, inverter size optimization, SST inverter structures
© 2020 Published by ECRES

Nomenclature
AC Alternating Current
DC Direct Current
GW Giga Watt
LCL Inductance – Capacitance – Inductance Type Low Pass Filter
LCOE Levelized Cost of Energy
SPWM Sinusoidal Pulse Width Modulation
SST Solid State Transformer
THD Total Harmonic Distortion
USD United States Dollar

1. INTRODUCTION

Solar photovoltaic systems and their components are getting more and more important in recent years. The main
factors which provide this, are the reduced costs and increased installations. Both cases decrease the LCOE
(Levelized Cost of Energy). Likewise increased installations lead to higher power and robust inverter designs.
Inverter is the secondary part of whole system in terms of efficiency and cost. The main element of the system
is the photovoltaic panel itself. On the other hand, it is clear that new inverter technologies are developing

437
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

rapidly. The main goals for solar inverter innovations, are increasing efficiency, decreasing costs and providing
reliability. Robust designs just like multi string inverter configurations and central inverter configurations may
be used for this purpose.
This study aims to provide one of these innovative designs, using high voltage AC levels in a compact inverter
structure. It is expected that new designs may also be compact, easy to use, reliable and greater powered.
Different inverter structures are considered for this purpose with regards to their technical and commercial
aspects. Inverter designs are compared with each other at different voltage levels.
LCL filter structure is the most effected element against voltage variation at the design stage. Increasing both
DC and AC voltage allows the LCL part of inverter to grow physically. This growth will be critically important
with affecting the cost and size of inverter. With the decision to use high voltage, optimization will be required
when considering its advantages, disadvantages, opportunities and threats. The aim of this study is to assist
researchers, market participants and investors to perform optimized analysis including output voltage level,
installed power and quantity of inverters.
This study is different from others when considering output voltage levels primarily and also important when
highlighting the relationship between output voltage level and solar inverter characteristics.

2. THEORY AND GENERAL VIEW

The most important matter in the field of renewable energy is the rise of photovoltaics. As mentioned before,
installation and energy costs have both decreased critically. This aspect is explained with using Figure 1 and
Figure 2. The data in the figures are obtained from global reports. [1-11]

Figure 1 Global renewable power installations

The annual increase of three main renewable power installations are shown in Figure 1. The success of
photovoltaic systems is undeniable. In recent situation annual solar power installation is higher than the summary
of both hydraulic and wind power. On the other hand photovoltaic systems have lower capacity factors and
inefficient conversion rate compared to other renewable systems. These factors determine the photovoltaic
energy costs which are described as LCOE, which is a little bit expensive than others. Although photovoltaic
power installations are unrivaled in cost, the cost of photovoltaic energy is comparable. Geographical location
and the size of installation are the main factors when aiming to achieve lower LCOE levels. The comparison of
LCOE values by years are shown in Figure 2.

438
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2 Levelized cost of electricity for photovoltaic systems

Recent investments aim to increase photovoltaic power in a single area. The lands in desert climate, which
cannot be cultivated, are used for this purpose. Along with the increasing installations, the need for a new,
efficient, and greater powered inverter is also emerging. Solar inverter cost has secondary importance after
photovoltaic panel in total cost. Considering a market which has annual investment of one hundred billion
dollars, the cost and efficiency of each component becomes critical. New approaches for inverter design have
potential to realize new inverter models which are smaller, effective, cheaper and more reliable. Multi level
inverter topologies are used for this purpose. [12-17]
In literature research, direct usage of high voltage in a compact inverter structure was not questioned as
comprehensive before. This study aims to compare new high voltage inverter structures to low voltage
equivalent models when considering costs, efficiency, reliability, weight and size.

3. METHODOLOGY

Before comparing the different inverter models, we need to examine the definition of high voltage. According
to IEC and BS standards potential difference above 1000 V alternating current and 1500 V direct current, is
assumed as a high voltage level. This definition is identified among phase conductors. This study aims to
compare solar inverters above 1 kV output voltage levels with their low voltage relevants. It is also noted that
voltage differs when a single phase system is used.

3.1 Method

This study explains the comparison of same powered solar inverter models at different voltage levels. Compared
inverter models have the same power values and are assumed the equivalent structures. The voltage levels are
selected as one as a low voltage 0.4 kV and the other at high voltage 3 kV. Discussed models are both mono
phase full H Bridge structures. They have 3 kW output power, 10 kHz SPWM switching signal frequency, and
50 Hz grid frequency. Both models have LCL filter structure with specified designs. Driver parts are used the
same for both structure. Simulation programs Matlab Simulink, Simscape and Express PCB were used to get
the numerical results.

3.2 Assumptions

Output powers were selected at the same value and 3 kW for both. Inverter structures are both monophasic and
made with H Bridge topology with LCL filter. Modulation types are selected SPWM with the same grid
frequency and switching frequency. Hence topologies, modulation types and frequencies were assumed the
same. Working conditions were set to 45 °C instead of room temperature when considering the heats effects on
electronic switches.

439
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3.3 Uncertainty Analysis

In our uncertainty analysis we have considered each electronic component by its specific features. Passive
electronic components were set to %1 accuracy to eliminate negative effects. Catalog values were reduced to
working values by linear interpolation for active electronic components. By the way an acceptable accuracy was
made for electronic structure.
Cost analysis were made by using USD currency without any interest or inflation effect.

3.4 Formulas

The comparisons for each models contain different LCL type low pass filter structures. This LCL filter was
designed for each individual structure. The calculation for this design was possible by using Equation [1 – 10].

Zb = Vgrid2 / PN Equation 1

Cb = 1 / wgrid Zb Equation 2

wgrid = 2 π fgrid Equation 3

Imax = PN √2 / Vgrid Equation 4

∆IL = %10 of Imax Equation 5

Linv = VDC / (6 fsw ∆IL) Equation 6

Cf = %3 of Cb Equation 7
1
√( 2 +1)
𝑘𝑎
Lgrid = 2 Equation 8
𝐶𝑓 𝑤𝑠𝑤

(Linv+Lgrid)
fres = 1 / 2 π √𝐿𝑖𝑛𝑣 𝐿𝑔𝑟𝑖𝑑 𝐶 Equation 9
𝑓

Rs = 1 / 3 2 π fres Cf Equation 10

Equivalent of letters are also shown below with their SI units;

Zb: Base Impedance (Ω), Cb: Base Capacitance (F), Vgrid: Grid Voltage (V), fgrid: Grid Frequency (Hz), fsw:
Switch Frequency (Hz), fres: Resonance Frequency (Hz), Wgrid: Grid Radial Speed (rad), PN: Nominal Power
(W), Imax: Maximum Current (A), ∆IL: Current Variation (A), Linv: Inverter Side Inductance (H), L grid: Grid Side
Inductance (H), Cf: Filter Capacitance (F), Rs: Serial Resistor (Ω)

3.5 Working Philosophy and Simulations

The electronic switches of any inverter structure should be controlled in a logical order to achieve pure sinus
wave at the output. In this study, first and fourth of H Bridge switches were controlled with sinus waves in grid
frequency, and other second and third were controlled with SPWM in switching frequencies. Switched PWM
signal is transmitted to the load in sinus form by passing through the LCL filter. According to simulation results,
desired sinus wave form was achieved.
Modern inverter structures need a different switching frequency which is higher than grid frequency. The usage
of high frequency effects the harmonic behavior of the system. Suppression of this noise generation is important
for system reliability. Total Harmonic Distortion (THD) shouldn’t exceed above 5% which is described in IEEE
519-2014 standards.
Two different solar inverter models are identical in terms of control circuits. But they include fundamental
differences in operating voltages, electrical impedances and LCL filter values. The main factors that create these
differences is the decreasing current.

440
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. RESULTS AND DISCUSSION

Table 1 shows the obtained results of the comparison for two different inverter models with simulation studies.
According to results, it is possible to design solar inverter structures that provide energy output at high voltage
level.

Table 1 Comparison of Low Voltage Structure and High Voltage Structure

Solar Inverter With High Solar Inverter With Low
Voltage Output Voltage Output
Assumptions
Output Voltage 3 kVp-p (1732 Vp-n) 0.381 kVp-p (220 Vp-n)
Phase Amount Mono Phase Mono Phase
Modulation Type SPWM SPWM
Grid Frequency 50 Hz 50 Hz
PWM Frequency 10 kHz 10 kHz
Main Topology Full H Bridge Full H Bridge
Filter Type LCL Filter LCL Filter
Output Power 3 kW 3 kW
Simulation Results
η 74.54 % 66.79 %
THD 0.86 % 1.45 %
Vdc 2457 V 359 V
Vinsulation 4500 V 1500 V
Rload 1 kΩ 16.13 Ω
Linv 167 mH 3.09 mH
Cfilter 96 nF 5.9 uF
Lgrid 16 mH 0.26 mH
Size 32 dm3 1 dm3
Weight 20 kg 2 kg
Cost 580 USD 220 USD
Uncertainty Analysis
DC Source Constant Constant
Mosfet Driver Examined Separately Examined Separately
Active Electronic Components
IXTK20N150 /
MOSFET Code IXTL2N450
IXTX20N150
MOSFET Package ISOPlus i5-Pak TO264 / Plus247
MOSFET Rd Catalog Value 1.01 Ω (1.65 A ve Tj 25 C) 1.02 Ω (12.7 A ve Tj 25 C)
MOSFET Rd Catalog Value 2.15 Ω (1.65 A ve Tj 125 C) 2.38 Ω (12.7 A ve Tj 125 C)
With Interpolation 1.238 Ω (1.65 A ve Tj 45 C) 1.292 Ω (12.7 A ve Tj 45 C)
Passive Electronic Components
Cdc 1% 1%
Linv 1% 1%
Cfilter 1% 1%
Lgrid 1% 1%
Rload 1% 1%

It is acceptable that decreasing current value provides some positive effects to system efficiency. This is an
important benefit when high voltage is used in solar system design. Additionally higher power values can be
achieved by this way, which cannot be achieved when using low voltage outputs.
On the other hand, the filter structures of the high voltage inverters will contain inductance elements with much
higher values than equivalent low voltage models. It is a major failure for high voltage inverter designs.

5. CONCLUSION AND ANALYSIS

The comparison made considering two basic inverter models. One of them uses high voltage output and the
other uses low voltage output. It is aimed to provide a simple explanation. Figure 6 shows the realized SWOT
analysis for two different compared models with high voltage output and low voltage output.

441
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

STRENGTH WEAKNESS
* Developing more efficient inverter structures * Designed products need to be developed for
commercialization
* Obtaining a compact structure * Technology has not been sufficiently tested
* It is easy to handle and easy to use * Cost may increase
* It is possible to design higher power inverters * The use of high voltage requires additional
than as usual insulation
* Lack of low voltage connection will result in * Device dimensions and weight increase
fewer joints, ease of operation and less critically with increasing inductance values in
downtime the LCL filter
* It is possible to benefit from domestic * There is a certain installation cost to ensure
production incentives mass production

OPPORTUNITIES THREATS
* Photovoltaic installations continue to * The use of high voltage poses a threat to
increase exponentially in a global scale occupational health and safety
* Considering market trends, a higher power * HV is proposed for industrial installations
inverter model can be advantageous rather than residential installations
* Additional domestic contribution incentives * The Solar Market is highly dynamic, this
can be provided for Solar sites occurs competitive difficulties
* Export potential is very high * Standardization and certification may take a
long time
Figure 3 Realized SWOT Analysis

This study is rare in terms of discussing inverter characteristics and output voltage levels together. Also it
explains the relationship between the variations of voltage levels and the inverter characteristics.
The formation of inverter structures with high voltage output can give us the advantages of a compact product
with reduced electrical port, allowing more efficient designs, higher power output and a domestic product.
It also includes the disadvantages that has not yet been commercialized and proven. Additionally it might be
more costly, need more developing and need additional insulation. Also it has significant defects in robustness
and weight.
With this new approach, some opportunities may be created just like introducing a new product with great power
for the market. Although, there are major risk factors such as shock and fire, which arise from lack of safety due
to the use of high voltage levels. These high voltage inverter devices should be improved with research and
development studies. The technology must be proven with sample site installations and long term experiences.
It is also clear that new researches are needed to provide technical suggestions for designers.
Unfortunately new inverter models will not have a price advantage because high voltage electronic switches are
normally rare. It is possible that developments in electronic technology are expected to offer more opportunities
at this field in the near future.

REFERENCES

[1] Renewables 2009 Global Status Report, REN21 2009, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat).
[2] Renewables 2010 Global Status Report, REN21 2010, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat).
[3] Renewables 2011 Global Status Report, REN21 2011, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat).
[4] Renewables 2012 Global Status Report, REN21 2012, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat).
[5] Renewables 2013 Global Status Report, REN21 2013, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat). ISBN 978-3-9815934-0-2
[6] Renewables 2014 Global Status Report, REN21 2014, Renewable Energy Policy Network for the 21st
century (Paris: REN21 Secretariat). ISBN 978-3-9815934-2-6

442
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[7] Renewables 2015 Global Status Report, REN21 2015, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat). ISBN 978-3-9815934-6-4
[8] Renewables 2016 Global Status Report, REN21 2016, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat). ISBN 978-3-9818107-0-7
[9] Renewables 2017 Global Status Report, REN21 2017, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat). ISBN 978-3-9818107-6-9
[10] Renewables 2018 Global Status Report, REN21 2018, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat). ISBN 978-3-9818911-3-3
[11] Renewables 2019 Global Status Report, REN21 2019, Renewable Energy Policy Network for the 21st century (Paris:
REN21 Secretariat). ISBN 978-3-9818911-7-1
[12] Babaei E., Kangarlu M.F., Mazgar F.N., Symmetric and asymmetric multilevel inverter topologies with reduced
switching devices, Electric Power Systems Research 86 2012, 122-130.
[13] Chinnaiyan V.K., Jerome J., Karpagam J., An experimental investigation on a multilevel inverter for solar energy
applications, Electrical Power and Energy Systems 47 2013, 157-167.
[14] Parabaharan N., Jerin A.R.A., Palanisamy K., Umashankar S., Integration of single phase reduced switch multilevel
inverter topology for grid connected photovoltaic system, Energy Procedia 138 2017, 1177-1183.
[15] Kumar Y.V.P., Ravikumar B., A simple modular multilevel inverter topology for the power quality improvement in
renewable energy based green building micro grids, Electric Power Systems Research 140 2016, 147-161.
[16] Sridhar V., Umashankar S., Sanjeevikumar P., Ramachandaramuthy V.K., Popa L.M., Fedak V., Control architecture
for cascaded H bridge inverters in large scale PV Systems, Energy Procedia 14 2018, 549-557.
[17] Sastry J., Bakas P., Kim H., Wang L., Marinopoulos A., Evaluation of cascaded H bridge inverter for utility scale
photovoltaic systems, Renewable Energy 69 2014, 208-218.

443
 NUMERICAL SIMULATION OF AN ON-GRID NATURAL GAS
PEMFC - SOLAR PHOTOVOLTAIC MICRO CHP UNIT: AN
ENERGY AND FINANCIAL ANALYSIS FOR RESIDENTIAL AND
INDUSTRIAL APPLICATIONS

Renato de Oliveira Gabriel

Department of Mechanical Engineering – Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro, Brazil,
[email protected], ORCID: 0000-0001-9254-8281

Sergio Leal Braga

Department of Mechanical Engineering – Pontifical Catholic University of Rio de Janeiro,, Rio de Janeiro, Brazil,
[email protected], ORCID: 0000-0003-4741-676X

Florian Pradelle
Department of Mechanical Engineering – Pontifical Catholic University of Rio de Janeiro,, Rio de Janeiro, Brazil,
[email protected], ORCID: 0000-0003-4306-8083

Eduardo Torres Serra

ES+PS Consultoria, Rio de Janeiro, Brazil, [email protected], ORCID: 0000-0002-7076-9363

Cesar Luiz Coutinho Sobral Vieira

Guascor, São Paulo, Brazil, [email protected], ORCID: 0000-0002-6464-8506

Gabriel, RO, Braga, SL, Pradelle, F, Serra, ET, Vieira, CLCS. Numerical simulation of an
on-grid natural gas PEMFC - solar photovoltaic micro CHP unit: an energy and financial
Cite this paper as:
analysis for residential and industrial applications. 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: As global natural resources depletion and concern on emission of greenhouse gases intensify, the
interest for low emission technologies and the use of renewable energy increased in the world. In
this context, this paper aims to present an hybrid energy system for on-grid micro residential and
industrial applications. The system is composed of a micro combined heat and power (CHP) unit,
including a natural gas (NG) reformer coupled with proton-exchange membrane fuel cell
(PEMFC), solar photovoltaic modules (PV) and a bank of batteries (B), connected to the grid
through a bidirectional inverter. Different system configurations (PEMFC + PV + B, PEMFC + B,
PEMFC + PV), with or without cogeneration of the heat from the CHP system, were investigated
to assess the NG and electricity flows and economic cost during operation of 10 years in an 2020-
2040 horizon. The impact of fuel cell sizing and two electricity tariffs was also assessed for both
applications. Afterwards, the cash flow in terms of net present value of a twenty years operation
period was simulated within the Brazilian context. The results showed that the proposed system
were economically viable and more efficient than traditional PV systems.
Keywords: Combined heat and power (CHP) unit; Proton-exchange membrane fuel cell (PEMFC); solar
photovoltaic (PV) modules; Batteries (B); Natural gas (NG); Simulation
© 2020 Published by ECRES

1. INTRODUCTION

Electric energy demand is expected to increase 1.3% per year up to 2050 in non-OECD countries, according to
the U.S. Energy Information Administration (EIA) [1]. At the same time, the world searches for alternative low
emission energy sources and conventional solar photovoltaic (PV) systems are considered some of the most

444
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

promising technologies for the next decades. In spite of its great advantages in both economic and environmental
viewpoints, high area is required for modules installation which and its high dependence on the reverse metering
factor may limit its applications in large cities. Other technologies such as hybrid systems with fuel cells are
been studied as alternative possibilities to conventional PV systems.
Reckioua et al. [2] presented a detailed review on hybrid systems configuration with proton-exchange membrane
fuel cell (PEMFC) and demonstrated the feasability of a PEMFC+PV system with hydrogen storage for small
applications. Maclay et al. [3] studied a PV system coupled with reverse fuel cell RFC, battery bank and ultra-
capacitors as storage components. An economic assesment was performed to identify the best combination in
terms of efficiency and cost. Later, Maclay et al. [4] carried out and experimental study to investigate the
dynamic behavior of the same system under rapid changes in load demand. Ozgirgin et al. [5] performed a
simulation analysis on an integrated on-grid hybrid system with PV pannels, PEMFC and battery bank. They
studied the impact of the seasons and the potential for cogeneration. An energetic and exergetic study of an
hybrid system with PEMFC and solid oxide fuel cell fed with natural gas (NG) was conducted by Barelli et al.
[6] for small residential applications. Zuliani et al. [7] compared the differences in electric and thermal
efficiencies of low and high temperature PEMFC for small applications. Finally, Marcoberardino et al. [8] made
a comparative economic assessment for a 5 kW PEMFC system and traditional CHP systems. The impact of NG
composition and electric energy prices were evaluated in four countries of the European Union and always
showed environmental benefit even if the system cannot be competitive with subsidies from the governments.
In this context, this paper presents a numerical simulation in Matlab (version R2019b ) of an hybrid system
consisting of a PEMFC unit combined with solar PV and battery bank for small on-grid residential and industrial
applications in Brazil. Several system configurations with cogeneration possibility. Two types of tariffs
(conventional (Conv.) and alternative (Alt.)) and two types of consumers (residential and industrial) were
investigated as to their impacts in system’s operation, efficiency and financial analysis.

2. METHODOLOGY

The proposed on-grid system consists of a micro combined heat and power CHP unit (NG reformer + PEMFC),
PV pannels and a battery bank (B) connected to a bidirectional inverter to provide electric energy to a consumer
in the AC terminal. Furthermore, cogeneration can be achieved for supplying an additional thermal demand. As
allowed by Brazilian legislation, extra electric energy production within the system is sent to the grid, generating
energy credits for future use when demand exceeds total generation. The reverse metering factor used in this
work was 100%, according to Brazil’s current scenario. A schematic representation of the modeled system and
the profile of consumers’s electric demand are shown in Figure 1 (a) and (b), respectively.

(a) (b)
Figure 1. (a) Schematic representation of the modeled system and (b) Consumers power demand for one day [9].

Power flow through each system component was determined via a control algorithm introduced in Appendix A
for conventional tarriff. The PV array works as the primary source of energy during the day, charging the battery
and exporting extra energy production to the grid. When demand exceeds total generation within the system,
PEMFC unit provides electrical output to sustain the deficit just after a brief warming-up period of 30 minutes,

445
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

during which it consumes 550W from the grid. Once power demand is too high, the system starts to buy
electricity directly from the grid.
Three climate options for photovoltaic power input has been considered, accounting for sunny, partially cloudy
and cloudy days. Climate for each day is randomly generated into the program for the simulation.
Current flowing in and out of the battery bank is determined dividing electric power by the tension (48V). State
of charge (SOC) limits are fixed at 20% minimum and 90% maximum Later, total charge remaining in the
battery and battery’s SOC are updated according to battery’s capacity [10].
Molar flow of hydrogen going into the fuel cell is given by equation (1) [11]. Number of single cells within fuel
cell stack (𝑁𝐹𝐶 ) and active surface area (𝐴𝑎 ) were assumed as 75 and 200 cm2, respectively [12]. F is the Faraday
constant (96,485 C/mol).
𝐼𝑁𝐹𝐶 𝐴𝑎
𝑛𝐻2 =
2𝐹𝐴𝑎 (1)

Next, molar flow of methane (CH4) is defined as a quarter of the molar flow of hydrogem (H 2) [11] and total
gas consumption in m3/min within CHP unit is calculated in (2) by assuming 100% metane composition and
standard ambient conditions.

𝑣 = 1.316𝑚𝐶𝐻4 (2)

Finally, thermal power rejection caused by the exotermic reactions at high temperature operation of the fuel cell
is given by (3) [12]. Electric and thermal efficiencies are calculated according to equations (4) and (5) [12] and
were found to be 44.52% and 42.43%, respectively. 𝑉𝑠𝑡 and 𝐼𝑐𝑒𝑙𝑙 are the stack total voltage and current through
each single fuel cell. Hydrogen stoichiometry ( 𝑆𝐻2 ) was assumed as 1.15, as in [12].

𝑃𝑡ℎ = 𝑁𝐹𝑐 (1.25 − 𝑉𝑐𝑒𝑙𝑙 )𝐼𝑐𝑒𝑙𝑙

(3)
𝑉𝑐𝑒𝑙𝑙 1 𝑉𝑐𝑒𝑙𝑙
𝜂𝑒𝑙 = 𝜇𝑓 =𝑆
1.25 𝐻2 1.25 (4)
1.25−𝑉𝑐𝑒𝑙𝑙 1 𝑉𝑐𝑒𝑙𝑙
𝜂𝑡ℎ = 𝜇𝑓 = (1 − ) (5)
1.25 𝑆𝐻2 1.25

Energy input operation tariffs for electric energy and natural gas in 2020 are chosen equal to the current cost in
the city of Rio de Janeiro, Brazil (as shown in Appendix B). Equipment costs are presented in Appendix C.
For the financial assessment, first a net present value evaluation NPV of the system’s operation is calculated
according to equation (6). The NPV represents all cash flows of a project (incoming and outgoing) evaluated at
present (t = 0) [13]. In equation (6), I corresponds to the total investment on the project, representing a negative
input cashflow at the beginning of system’s operation. 𝐶𝐹(𝑖) is the cashflow at year 𝑖 , which can be a positive or
negative value and 𝑟 is the interest rate (taken as 2.5% a year) [13]. NPV must be positive for the system to be
viable.
(𝑖) 𝐶𝐹
𝑁𝑃𝑉 = −𝐼 + ∑𝑁
𝑖=0 (1+𝑟)𝑖 (6)

Secondly, the cumulative cost of each system configuration during ten years operation in the 2020 to 2040
horizon is introduced as another method of financial analysis. For this to work, future aquisition costs of each
power source in the system needed to be estimated for the next two decades. A linear decay function was used
to interpolate IRENA’s estimates on battery decaying prices up to 2040 [14]. Equations (7) and (8) were used
to estimate PV and PEMFC decaying acquisition costs, acording to [15] and [16], where 𝑦 corresponds to the
year. No reduction costs were considered in inverters aquisition.

𝐹𝐶(𝑦) = 𝐹𝐶(𝑦 = 2020) 𝑒𝑥𝑝(0.3182 (𝑦 − 2020))−0.2367

(7)
𝑃𝑉(𝑦) = 𝑃𝑉(𝑦 = 2020) [(1 + 0.09)(𝑦−2020) ]−0.2875 (8)

446
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In order to evaluate system’s feasibility and impact on each possible configuration (FC + PV + B, FC + PV and
FC + B), a financial analysis was performed for both types of consumers in the Brazilian scenario. A total of ten
PV pannels and three batteries were used for each consumer. Electric energy tariff was assumed to increase 2%
above inflation per year against 0% increase in NG tariff after linear decay to 60% of its original cost during
three consecutive years. The rapid decay in the predicted natural gas tariff was due to the decentralization
program on the Brazilian NG market, initiated in 2018 [17].

3. RESULTS AND DISCUSSIONS

Figures 2 (a) and 3 (a) show the cashflow in net present value (NPV) through a twenty-year operation starting
in 2020 for two residential and industrial consumers, respectively. Results correspond to the use of conventional
tariff. On the other hand, figures 2 (b) and 3 (b) estimate the total cumulative cost of a ten-year system operation
in the 2020 to 2040 horizon. Tables 3 through 6 present detailed data collected from NPV analysis for each type
of consumer.
All system configurations achieved payback before the year 2040 and required inverters substitution after fifteen
years operation. Only two inverters and two battery bank substitutions were required in all residential cases. In
cogeneration cases, a discount proportional to recovered heat from CHP and gas tariffs was made for each type
of consumer, explaining the major gas bill cost reductions in these cases. Also, electric energy costs were very
close to the minimum energy consumption of 100 kWh/month stated for three-phase facilities connected to the
grid, acknowledging for control system’s algorithm success in deliverying power through system’s components.
Percentage time operation of fuel cell varied between 52% in residential PEMFC+PV systems to 92% in
industrial PEMFC+B cases and final profits up to US$ 82,000 were achieved, accounting for system’s feasibility
in all scenarios. Very little difference was observed in the final profit between PEMFC+PV and PEMFC+B
configurations with the use of cogeneration, where the later presented slighty reduced costs in the industrial
case. On the other hand, without cogeneration, PEMFC+PV configuration tended to show better cost
effectiveness (especially on residential case) because of lower NG consumption.

(a) (b)
Figure 2. (a) Cashflow in net present value (NPV) for twenty-year system’s operation and (b) Total cost for ten-year
system’s operation in an 2020-2040 horizon for two residential consumers with conventional tariff.

Table 3. Technical data for twenty-year system’s operation for two residential consumers with conventional tariff.
With Cogeneration Without Cogeneration
Parameters
FC + PV + B FC + PV FC + B FC + PV + B FC + PV FC + B
NG flow (𝑚3 /month) 286.14 285.58 398.33 286.29 285.41 398.33
Percentage of time 51.80 51.70 72.11 51.83 51.67 72.11
PEMFC operation (%)

447
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 4. Economic data for twenty-year system’s operation for two residential consumers with conventional tariff.
With Cogeneration Without Cogeneration
Parameters
FC + PV + B FC + PV FC + B FC + PV + B FC + PV FC + B
Total Investment (US$) 83,675 79,085 79,775 83,675 79,085 79,775
Electric Energy (US$) 7,136 7,339 7,101 7,138 7,335 7,101
Natural Gas (US$) 5,134 5,154 1,190 44,602 44,500 57,566
Maint. & Replc. (US$) 17,723 11,610 15,723 17,723 11,610 15,723
Total System Cost (US$) 113,668 103,188 103,789 153,138 142,530 160,165
Final Profit in NPV (US$) 42,892 52,354 51,615 10,903 20,467 5,922
Payback (yr.) 2033 2032 2032 2038 2037 2039

(a) (b)
Figure 3. (a) Cashflow in net present value (NPV) for twenty-year system’s operation and (b) Total cost for ten-year
system’s operation in an 2020-2040 horizon for two industrial consumers with conventional tariff.

Table 5. Technical data for twenty-year system’s operation for two industrial consumers with conventional tariff.
With Cogeneration Without Cogeneration
Parameters
FC + PV + B FC + PV FC + B FC + PV + B FC + PV FC + B
NG flow (𝑚3 /month) 396.63 396.10 512.05 396.30 396.34 512.05
Percentage of time 71.80 71.71 92.70 71.74 71.75 92.70
PEMFC operation (%)
Bank substitutions 1 - 0 1 - 0
Required inverters 2 2 1 2 2 1

Table 6. Economic data for twenty-year system’s operation for two industrial consumers with conventional tariff.
With Cogeneration Without Cogeneration
Parameters
FC + PV + B FC + PV FC + B FC + PV + B FC + PV FC + B
Total Investment (US$) 83,675 79,085 76,545 83,675 79,085 76,545
Electric Energy (US$) 7,127 7,144 7,103 7,140 7,133 7,103
Natural Gas (US$) 22,308 22,294 25,461 57,331 57,336 70,725
Maint. & Replc. (US$) 14,551 11,610 6,380 14,551 11,610 6,380
Total System Cost (US$) 127,661 120,133 115,489 162,697 155,164 160,753
Final Profit in NPV (US$) 71,658 78,602 82,491 43,262 50,209 45,805
Payback (yr.) 2031 2030 2030 2034 2033 2033

To investigate the impact of different electric energy tariffs and fuel cell sizing, a complete system (FC + PV +
B) was simulated under the same circunstances as before under conventional and alternative tariffs for five
residential and industrial consumers. In these cases, however, only cogeneration scenarios were evaluated as to

448
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

obtain the maximum profit as possible for the consumers. While running on alternative tariff, batteries were
forced to charge to its maximum state of charge just before 5 PM and then fully discharge between 6 PM and 9
PM, exporting large amounts of energy to the grid and generating lots of energy credits.
Figure 4 and table 7 show all data acquired. Both residential and industrial consumers required three inverters
while running on conventional tariff and five inverters while on alternative tariff. In parallel, just one battery
substitution was needed for industrial cases and three substitutions for residentional cases with conventional
tariff, while five substitutions were required for both of them on alternative tariffs’s cases. Even though the fuel
cell opperated 100% of the time for all five consumers’s cases (552 𝑚3/month of NG), complementary electric
energy from the grid was needed to supply total demand, except for the residential case with alternative tariff.
In fact, the alternative tariff strategy has a positive effect, allowing major energy cost reductions by forcing the
batteries to shift power generation within the system to around 6 PM and 9 PM. Thus, the system exports large
amounts of energy to the grid when tariff is more expensive, generating more credits for use in other times of
the day. That strategy comes with a price though, requiring the batteries to complete much more cycles than in
the conventional tariff case, meaning more frequent bank substitutions and necessity for more inverters to
perform great power flow during peak hours. Because the most part of consumer demand is concentrated during
peak hours for the residential case, the alternative tariff strategy provided higher reduction on energy cost than
the industrial case, allowing total profit of US$ 213,000 at the end of system’s operation.

(a) (b)
Figure 4. (a) Cashflow in net present value (NPV) for twenty-year system’s operation and (b) Total cost for ten-year
system’s operation in an 2020-2040 horizon for five consumerswith conventional and alternative tariffs.

Table 7. Economic data for twenty-year system’s operation for five consumers with both tariffs.
Residential Industrial
Parameters
Conv. Alt. PV only Conv. Alt. PV only
Total Investment (U$) 99,640 106,100 56,530 99,640 106,100 72,740
Electric Energy (U$) 92,080 7,505 7,109 219,324 169,748 7,113
Natural Gas (U$) 0 0 0 26636 26636 0
Maint. & Replc. (U$) 40,366 62,813 38,380 25,193 62,813 49,840
Total System Cost (U$) 232,086 176,418 102,019 370,793 365,297 129,693
Final Profit in NPV (U$) 171,289 213,036 284,166 162,253 163,877 359,351
Payback (yr.) 2028 2027 2024 2028 2028 2024

Simulation with alternative tariff for two consumers’s cases were performed but no advantage was observed due
to low energy consumption from the grid. In other words, the alternative tarrif strategy could not reduce energy
bill costs while the system was oversized, thus presenting worse results than the conventional tariff.

449
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

5. CONCLUSION

An hybrid system was presented and simulated under different system configurations and two type of tariffs,
allowing for cogeneration from high temperature operation of the CHP unit. Energy and economic assessments
were made and proved for the system’s viability in all discussed scenarios, where great economic advantages
were achieved as compared to direct electric energy consumption from the grid. Additionally, it was found that
the alternative tariff strategy could provide significantly cost reductions only when system was undersized.
The results show that, beyond requiring much less space for equipment installation, the analysed system has the
advantage of power shifting control throughout the day over conventional PV systems. As consequence, less
dependence on reverse metering factor variation is expected from the former. Finally, the battery bank could
sustain energy supply even if connection to the grid was lost at any moment, providing an efficient back-up for
consumer’s power demand, opening up several potential applications for the proposed system.
A control algorithm with the use of neural networks or fuzzy-logic based techniques is considered to be of great
interest for future improvements on this work. Different NG compositions should also be examined as to
extendend the system’s applicability on different regions of the planet.

ACKNOWLEDGMENT

The authors are indebted to CNPq/MCT, CAPES, FAPERJ, and FINEP for the financial support to the
Department of Mechanical Engineering at the Pontifical Catholic University of Rio de Janeiro (PUC-Rio). The
authors gratefully acknowledge Guascor for the financial support.

REFERENCES

[1] U.S. Energy Information Administration Office of Energy Analysis. International Energy Outlook 2019 with projections
to 2050 - September 2019 Available at: https://www.eia.gov/outlooks/ieo/pdf/ieo2019.pdf
[2] Rekioua, D.; Bensmail, S.; Bettar, N. Development of hybrid photovoltaic-fuel cell system for stand-alone application.
International Journal of Hydrogen Energy 2014; 39: 1604-1611 <http://dx.doi.org/10.1016/j.ijhydene.2013.03.040>
[3] Maclay, J. D.; Brouwer, J; Samuelsen, G. S. Dynamic modeling of hybrid energy storage systems coupled to
photovoltaic generation in residential applications. Journal of Power Sources 2006; 163: 916-925
<10.1016/j.jpowsour.2006.09.086>
[4] Maclay, J. D.; Brouwer, J; Samuelsen, G. S. Experimental results for hybrid energy storage systems coupled to
photovoltaic generation in residential applications. International Journal of Hydrogen Energy 2011; 36: 12130-12140
<10.1016/j.ijhydene.2011.06.089>
[5] Ozgirgin, E.; Devrin, Y.; Albostan, A. Modeling and simulation of a hybrid photovoltaic (PV) module-electrolyzer-
PEM fuel cell system for micro-cogeration applications. International Journal of Hydrogen Energy 2015; 40: 15336-
15342 <http://dx.doi.org/10.1016/j.ijhydene.2015.06.122>
[6] Barelli, L.; Bidini, G.; Gallorini, F.; Ottaviano, A. An energetic-exergetic comparison between PEMFC and SOFC-
based micro-CHP systems. International Journal of Hydrogen Energy 2010; 36: 3206-3214
<10.1016/j.ijhydene.2010.11.079>
[7] Zuliani, N.; Taccani, R. Microcogeration system based on HTPEM fuel cell fueled with natural gas: Performance
analysis. Applied Energy 2012; 97: 802-808 <10.1016/j.apenergy.2011.12.089>
[8] Di. Marcoberardino, G.; Manzolini, G. Investigation of a 5 kW micro-CHP PEM fuel cell-based system integrated with
membrane reactor under diverse EU natural gas quality. International Journal of Hydrogen Energy 2017; 42: 13988-
14002 http://dx.doi.org/10.1016/j.ijhydene.2017.02.016
[9] Francisquini, A. A. Estimação de Curvas de Carga em Pontos de Consumo e em Transformadores de Distribuição. MSc,
Universidade Estadual Paulista, São Paulo, Brazil, 2006.
[10] Robyn, A. Jackey. A Simple, Effective Lead-Acid Battery Modeling Process for Electrical System Component
Selection. The Mathworks Inc. 2007 <10.1016/j.jpowsour.2006.09.086>
[11] Inac, S.; Unverdi, S. O.; Midilli, A. A parametric study on thermodynamic performance of a SOFC oriented hybrid
energy system. International Journal of Hydrogen Energy 2019; 44: 10043-10058
<https://doi.org/10.1016/j.ijhydene.2019.01.247>
[12] Chen, X.; Wan, Z.; Gong, G.; Luo, L. Performance analysis of 5 kW PEMFC-based residential micro-CCHP with
absorption chiller. International Journal of Hydrogen Energy 2015; 44: 10647-10657 <
https://doi.org/10.1016/j.ijhydene.2015.06.139>
[13] Simpson, R.; Sastry, S. K. Basic Economic Principles and Deciding Among Alternatives. Chemical and Bioprocess
Engineering, New York: Springer, 2013, pp. 325-347.
[14] Kairies, K. Battery storage technology improvements and cost reductions to 2030: A Deep Dive. In: International
Renewable Energy Agency Workshop 17 March 2017, Dusseldorf, Germany.

450
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[15] IRENA, Renewable Power Generation Costs in 2017, International Renewable Energy Agency, Abu Dhabi, 2018.
[16] E4tech, The Fuel Cell Industry Review, 2019.
[17] Vieira, R. Governo projeta redução de 40% no preço do gás com quebra de monopólio da Petrobras. O Globo 2019 <
https://oglobo.globo.com/economia/governo-projeta-reducao-de-40-no-preco-do-gas-com-quebra-de-monopolio-da-
petrobras-23761162>

Appendix A. Control algorithm for power flow through the system.

451
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

452
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Appendix B. Grid electricity and natural gas costs.

Electric energy Natural gas
Cost Type of Consume range Cost
Type of tariff
(US$/kWh) consumer (m3/month) (US$/m3)
Conventional 0.244 Distributed 0 - 200 1.127
Alternative (low)
Generation 201 – 5,000 0.753
before 17:00 and 0.215 0-7 1.543
after 22:00 8 - 23 2.012
Residential
Alternative (inter.) 24 - 83 2.437
17:00 to 18:00 and 0.297 above 84 2.572
21:00 to 22:00 0 - 200 1.507
Alternative (high) Industrial 201 - 500 1.463
0.445
18:00 to 21:00 above 500 1.420

Appendix C. Equipment Costs.

Aquisition Cost Instalation Cost Maintenance Cost
Equipment
(US$/unit) (US$/unit) (US$/unit/year)
Micro CHP w/ 5 kW PEMFC Unit 63,000 157.5
5625
6.8 kW Bidirectional Inverter 3,230 -
48V / 100 Ah VRLA Battery 765 - -
245 W PV Pannel 162 32.5 5.0

453
 NAVIER–STOKES-MAXWELL EQUATIONS FOR MHD EFFECTS WITH
TURBULENT FLOW IN ADVANCED ARIES-RS FUSİON TECHNOLOGY
Aybaba Hançerliogulları
Kastamonu University,Arts& Sciences Faculty ,Physics Department ,Kastamonu / Turkey ,[email protected]
[email protected], ORCID: 0000-0000-1700-8480

Erol Kurt
Gazi University, Technology Faculty,Department ofElectrical and Electronic Engineering,Ankara
/[email protected], ORCID: 0000-0002-3615-6926

Abbas Ghasemizad
University of Guilan ,Science Faculty, Physics Department,Rasht/Iran, [email protected], ORCID: 0000-0001-
6452-6309

Rezvan Rezaeizadeh
University of Guilan ,Science Faculty, Physics Department,Rasht/Iran,[email protected], ORCID: 0000-0003-
6219-6174

Yosef G. Ali Madee

Sirte University,healty faculty,radiology department,Sirte /Libya, [email protected], ORCID: 0000-0003-4652-
2990

Ettahir El Hadi Ali Omar Swese

Kastamonu University,Arts& EnginneringFaculty ,Material Science Department ,Kastamonu/Turkey,
[email protected], ORCID:000-0001-57-4261

Hançerlioğlları, A., Kurt, E., Ghasemizad, A., Rezaeizadeh, Rezvan., Madee, Y.G.A., Swese,
E.el H.A.O.,Navier–Stokes-Maxwell equations for MHD effects with turbulent low in
Cite this paper as:
advanced ARIES-RS fusion technology,8th Eur. Conf. Ren. Energy Sys.,24-25August 2020,
Istanbul.

Abstract: In the present work,we used aMonte Carlo method(i.e. MCNP5X) together with a nuclear data
library of Endf /V-VI with fusion technology in order to design the tokamak reactor of ARIES-
RS. In the frame of a controlled nuclear fusion reactor, almost few grooves of radioactive fallout
compared to the conventional fission reactors have unlimited fuel reserves and are pre-
programmed by developed states. Many tests for commercialization and continuity of DT fuel has
been essential. The ARIES-RS study explores the use of free flowing liquid surfaces to form the
inner surface of the chamber around an existing fusion plasma. The liquid wall concept has
important advantages compared to the traditional solid wall concept with its feature of renewable
wall: It provides possibility for high power density. The magnetohydrodinamic (MHD) equations
consist of macroscopic transport equations and magnetic induction equation, thereby the turbulent
models based on Navier-stokes-Maxwell equations are suitable for two types of liquids, working
fluid, molten salts and liquid metals. The MHD balance is discussed in terms of radiation heat
transfer conditions, current driving and nuclear performance in that context. It was proven that due
to the low fission rates in the fusion zone, among the types of fusion energy studied, the matrix
material, 232Th and 233U, selected as the fuel becomes more efficient in terms of performance of
the reactor.
Keywords: Fusion,ARIES-RS, magnetohydrodynamics, Liquid Wall,Navier-Stokes-Maxwell
© 2020 Published by ECRES

454
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1.INTRODUCTION

The ARIES team has been conducting researches for various advanced fusion knockers and non-knock models
for at least 25 years. In their studies, the operating performance and DT supply have been the main
concern.However, the same team started work on the ARIES-RS, an innovative, advanced fusion, reversed-
shear torus model in 1996. ARIES-RS reactor is essentially designed with ferritic steel and V-Cr4-Ti4 alloy, self-
cooled mild steel as the building material. The design can be based on the reverse knockout mode of the plasma
cutting and using moderately advanced engineering concepts such as Lithium, the average plasma with a large
radius of 5.52m and a small plasma diameter of 1.38m, with a total thermal power of 2620 MW and a fusion
power of 2170 MW with anexact output electrical power 1000 MW as conceptualized in Refs. [1,3]. It has
revealed that a fusion power plant using the DT mixture as the fuel type is indispensable. The candidate liquids
such as LiBeF 4, Li17Pb83, LiSn and LiNaBeF4 are used in various regions of the reactor, which are the mild
molten salts with various chemical properties providing stable operation of the reactor. The torus type ARIES-
RS reactor seems to provide the best economic performance in terms of engineering and meeting the future
energy demand with refrigerated Vanadium alloy plasma surface components and steady state reversed truss
buckle. It is stated that the ARIES-RS engineering design process reachesto the high-level task requirements
following the collaboration with the ARIES team as stated in Refs. [1,4-7].

Fig. 1 shows the fusion power core cross section in within a 3D appearance and Table 1 summarizes the basic
parameters for the design of ARIES-RS with a moderate aspect ratio.

Figure 1. The ARIES-RS fusion power reactor segments of 3D.

Extensive efforts were made to develop a reliable configuration allowing all sectors to be quickly removed and
subsequently disassembled in hot coals during the plant operation. The use of a self-cooled blanket, a large
design window and a simplified insulated design are considered. One of the largest cost items is the state of
power density and radiation level. Ferritic steel was used to reduce the cost of the system and sustain the long
life of the reactor. Besides, an additional saving was provided in order to use the radial and other large regions
of the blanket. Here, the main concern is to get the highest fusion power from the design.

The structure of the heater and electro-mechanical units with first wall, blanket, shield and the diverter are
important parts to be optimized. ARIES-RS reactor is a hybrid design, where fusion and fission reactions take
place together. This reactor is a new fusion reactor design using a liquid wall between the fusion plasma and the
solid first wall for Tritium production or energy transfer. High-quality fissile fuel can be obtained by using
the fast, energetic neutrons 238U and 232Th as a result of the reaction. That reality brings a key issue: The concept
of liquid first wall (FW) in fusion reactors is a more appropriate choice. Neutrons emitted as a result of fusion
reactions occurring in the plasma are captured by Thorium and Uranium in the blanket resulting in fragmented
fuel production.

455
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

As with previous ARIES designs, the MCNP5X code was included for various simulations of this reactor. The
segments and plasma structure of the Aries-RS fusion power reactor were illustrated in Figure- 1 by the ARIES
team. For the ARIES-RS reactor of reactor design, magnetohydrodynamics (MHD) which is advanced
commercial fusion power created by using liquids such as LiBeF4, Li17Pb83, LiSn and
LiNaBeF4 solutions for various candidate melt coolers with steel structure for both internal and external has
been used. The vapour pressure of the Li17-Pb83 and natural Li %, mole composition was calculated for the
temperature range of 200 K and 1000 K via containing a representative operating temperature range of the Aries-
RS reactor. Li17-Pb83, LiNaBeF4 and pure Li phase diagrams were measured by Aries-RS [4-7]. The molten salts
were also used as high-temperature heat transfer medium in many industries.

In the present work, the first liquid wall and blanket in the ARIES-RS reactor have been studied from the surface
at certain distances by using the mcnp5x simulation program. With the proposed study, we aim to increase the
rector's neutronic performance during the fusion event. Within this scope, the reactor has been investigated in
terms of Tritium reproduction rate (TBR), energy multiplier coefficient (M), fissile fuel production rate, gravity
effect of FW Pb and neutronic wall loading. In Section 2, FW and MHD stability flow has been described for
the reactor. Section 3 reports the main numerical results. Finally, the concluding remarks are given in the last
section.

2.FW/MHD STABILITY FLOW APPROACH

Table 1 shows the various physical properties and ARIES-RS parameters used in the design of the ARIES-RS
fusion reactor. The salt is solid in the liquid phase at the standard temperature and pressure due to the high
temperature. The coolant pressure of the FW is 12MPa and has a total surface area of 542 m2. The FW has an
average neutron wall load (NWL) of 4.1MW/m2 and an average heat flux about 0.48MW/m2.

In ARIES-RS studies, Navier-Stokes and Maxwell equations were used to analyse the effects of various fluids
(i.e. flibe, Lithium and Sn-Li). The studies have been carried out to determine the effect of the plasma aspect
ratio, elongation, pressure and flow profiles and the position of the curved fixing wall. Plasma flow and pressure
profiles are determined to provide a high stable β and large, well aligned boot current at the same time. This
means that the minimum value of the safety factor, qmin is above 2 everywhere, and q95 is kept above 3.5 which
is moved as close as possible to the plasma boundary, a negative cutoff degree given by q0-coming ≥ 0.3
protection and maximizing the match between the boot current profile and the desired MHD constant current
profile. Plasma geometry plays an important role in MHD stability as well as overall power plant design. The
aspect variations showed the well-known trend that b increases as the aspect ratio decreases, but the current
propulsion power also increased. Table-2 shows the hydrodynamics parameters of the working liquid for
ARIES-RS fusion reactors engineering design issues that could not be measured accurately for system code
analysis led to the choice of [9-12]. One of the main concerns about liquid-metal cooled systems is MHD
pressure drops. In typical plant designs, the MHD pressure drop for the internal regime of torus results in
pressure stresses beyond the material limits. In order to reduce this MHD pressure drop, a kind of insulation
wall is required with a direct coating or sandwich structure. With the assumption of reliable insulating
coatings, the MHD pressure drop is no longer a major concern. The design of the first wall, a blanket and shield
can be optimized to improve heat transfer and simplify configuration. The FW and breeding
blanket use a simple, boxlike structure with Li cooler flowing in poloidal road [13-17].

456
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1. Physical scale parameters of ARIES-RS reactor.

Parameters scale
Major radius (m) 5.52
Minor plasma radius (m) 1.38
Aspect ratio 4.00
Plasma vertical elongation (x-point) 1.70
Plasma current (MA) 11.32
Bootstrap current fraction 0.88
Current-drive power (MW) 81
Toroidal field on axis (T) 7.98
Peak field at the TF coils (T) 16
Toroidal b 0.05
Average neutron wall load 3.96
(MW/m2)
Coolant inlet temperature (°C) 330
Coolant outlet temperature (°C) 610
Fusion power (MW) 2170
Total thermal power (MW) 2620
Net electric power (MW) 1000
thermal efficiency 0.46
Reactor Net efficiency 0.38
Recirculating power fraction 0.17
Cost of electricity (KWh/cent) 75.79
Plasma current, Ip (MA) 11.3
On-axis toroidal field (T) 7.98
Triangularity 0.50
Poloidal β 2.28
Toroidal β 4.96
Bootstrap current (MA) 10.0
Driven current, ICD (MA) 1.2

2.1. Thick Liquid of wall concepts

In the ARIES-RS fusion reactor, we offer many advantages such as replacing the first liquid wall with thick
liquids, high fusion power density, high control reliability and usability, and low failure reduced radioactive
waste quantities and increased build a life [15-17]. It contains structurally vanadium alloy self-cooling lithium
design material. It is decided by the ARIES team that this blanket has the potential to perform better in order to
require a moderate extrapolation of today's technology. V-alloy has low activation, low temperature, high
temperature capability and it can handle the high heat flux. Besides, this blanket provides excellent performance.
Figure-2 shows the 2D sketch of ARIES-RS sector [12-15,16-18].

457
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. The sketch of ARIES-RS core 2D sector.

2.2 Current Driver

In the ARIES-RS reference balance, the boot current fraction is 0.88 by using density and temperature profiles
characteristic for the inverted shear plasmas. The external non-inductive techniques are often required to drive
off-axis currents in the plasma center and near the axial cutting zone. A series of non-inductive current drive
techniques have been considered for the inverted shear power plant. Depending on the reference balance, a
combination of these techniques is required to drive currents in different parts of the plasma. Due to the tendency
to radiate radially towards the plasma center, the ICRF fastest waves are best suited for current propulsion on
the magnetic axis. On the other hand, with a suitable launch spectrum, the low-hybrid wave can be made to
move currents off-axis in the cutting reverse region. Scans of the plasma triangularity of 0.2 to 0.6 showed that
a higher value than 0.4 is required for high β. β is increased with an internal deflector requirement (mainly
adequate neutron protection for space requirements) within the limits of 0.5.

Table 2. Hydrodynamic features operating liquid for ARIES-RS fusion reactors.

Property Unit Flibe (550 C0) Lithium(500C0) Tin-Lithium(500C0)
composition Mole*A 66%LiF,34%BeF2 100% Li 80% Sn,20% Li
melting point K 733 459 599
density Kg/m3 2011 485 6800
dynamic viscosity Kg/m.s 0.0116 0.32x10' 1.2x103
6
electrical cond. I/12m 184 2.83x10 1.67x106
thermal cond. W/m.K 1.06 49.6 33.44
specific heat J/Kg.K 2380 4170 317
surface tension N/m2 0.2 0.35 0.53
liquid thickness In 0.45 0.40 0.40
liquid velocity m/s 8.1 10 10
channel 1/2 width t 0.68 0.57 0.07
flow length m 8 8 8
toroidal field T 8 8 8
radial field T 0.2 0.2 0.2
radius of curvature m 6.7 6.7 6.7

Plasma elongation improves b, but it makes the plasma unstable to any vertical movement. The maximum
elongation is limited by passive stabilization conductive structure and feedback control system. Neutral rays can
provide such plasma rotation rates. On the other hand, ARIES-RS only uses RF current drive [6, 9,15-
19]. Figure- 3 shows reverse cut plasma configuration. The highest plasma rotation speeds from RF were
observed for ICH at 3x104 m/s.

458
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Reverse cut plasma configuration.

3. NUMERICAL CALCULATIONS AND RESULTS

The design and calculations of Aries-RS are performed in a 3D torus structure by using the MCNP5X code/Endf
/VI library. MCNP5X is a general target Carlo N-particle code and is also running for neutron, electron, photon
MCNP5X computational history of about 8.6 x105 particles. MCNP5X simulation code and Endf-5b/V-VI
library are used for its design. The parameters of the FW of reactor are examined for different blankets and
molten salt mixtures as shown in Table- 3, where a valid driver requirement for ARIES-RS balance
sits. Specially; tritium production rate, energy multiplication factor, fissile fuel breeding, radiation damage
amount, initial installation cost (COS), heat deposition rate, neutron wall loading (NWL) are explored.

Table 3. Valid Driver Requirements for ARIES-RS Balance.

System Frequency N Power (MW) Launcher
(GHz) Position
ICRF Fastwave 0.091 2.0 5.4 15
High frequency wave 1 2.1 21.4 0
Lower hybrid wave 4.6 1.9 9.5 -15

The compatibility among high temperature materials is the transfer of fluid dynamics such as plasmas to the
MHD model, including the effects of electromagnetic forces inside and inside the blanket.

3.1. The ARIES-RS Reactor Design

The ARIES-RS design process due to the need and generally operates with an inverted plasma. Sufficient power
is required to make a viable diver solution in the scrape-off-layer ( SOL) and submersible areas. In this case, the
sum of the particle and radiation heat flow does not exceed 6 MW/m2. As a function of fuel density n ≥ 2x1020/m3
and temperature 9 ≤ I ≤15 KeV, the contours of the auxiliary power required for steady state plasma power
balance are displayed [2,23,25-28,31]. Transport and radiation losses in areas outside the ignition zone exceed
fusion power and that assists a steady-state operation requirement with a medium aspect ratio (i.e. α = 4.0),
lower plasma current (IP = 11.32 MA) and a preload current ratio (f = 0.88) is quite high. As a result, the
auxiliary power required for the RF current driver is relatively low (80 MW). At the same time, the average beta
(β = 5%) is high, that is providing power densities close to practical engineering limits (peak neutron wall 5.7
MW/m2). The inlet temperature of the refrigerant into the reactor's heart is approximately 330°C and the coolant
outlet melt salt temperature from the reactor is approximately 610 °C It is designed as a Vanadium alloy. Table
-4 shows the dimensionless flow parameters in ARIES-RS reactors.

459
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table.4 Dimensionless numbers of working liquid.

Unit Flibe (550c) Lithium (500c) Tin-Lithium
Reynolds No., Re DVp/pi 6.32x105 6.187x106
Hartmann No., Ha B W(a/p)2 685.14 4.288x105

Modified Hartmann Hap Ha., 453.3 3.0x105

Radial Hartmann, HaR HaRBRW(a/tt) 17.12 1.0x104
Interaction No., 1,1.2 (Hao)2/Re 0.325 1.46x104

Froude No., Fri V2/(gD) 0.836 1.275

Froude No., Fr V2/(gR) 1 1.52
Thermal Diffusivity k/(pC0) 2.2x1e 2.4x10-5
Prandtl No., Pr Cot/Ic 25.56 0.0269
Modified Reynolds Re/(Ha4 1394 20.62

3.2. Simulation of MCNP5X / Blanket Design

The MCNP5X -5 (Monte Carlo) method is a simulation technique based on creating numerical solutions with
the help of the generation of random samples for physical and mathematical problems. Indeed, it is a numerical
process where an analytical solution is possible or not. The statistical simulations run using the Monte Carlo
method by using random generated numbers. The calculated cross-sectional view of Aries-RS with the help of
code is shown in Figure -4.

Table 4. Properties of dimensionless numbers for the working liquid of ARIES-RS

Unit Flibe (550c) Lithium (500c) Tin-Lithium
Reynolds No., Re DVp/pi 6.32x105 6.187x106
Hartmann No., Ha B i W(a/p)2 685.14 4.288x105
Modified Hartmann Hap Ha. 453.3 3.0x105
Radial Hartmann, HaR HaRBRW(a/tt) 17.12 1.0x104
Interaction No., 1,1.2 (Hao)2/Re 0.325 1.46x104
Froude No., Fri V2/(gD) 0.836 1.275
Froude No., Fr V2/(gR) 1 1.52
Thermal Diffusivity k/(pC0 2.2x1e 2.4x10-5

Table -5 shows that all materials by means of compositions for Aries-RS reactor of the regions blanket [5-7,17-
21].

Repeated structure systems for lattice detectors and DXTRAN coordinate transformations, lattice geometries,
weld properties, weld subroutine SRCDX, MCNP5X tally geometry program features for subroutines are
detailed.The tally examples include the FMn, FSn, and FTn cards and the TALLYX subroutine contains the
user-defined tallies using the FUn card. CNP tracks particles through the geometry using bounding surfaces
specified in cell cards, calculates the intersection of the trajectory of a track with each bounding surface by
following the minimum positive distance to a junction. [2,23,25-31].

460
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3.3. Tally Cards

Depending on the Monte Carlo program, MCNP5X has defined the standard TALLYX cards available for
neutron calculations (electron, proton, photon and neutron) for each. 3D torus for fusion reactor can be easily
used by the user depending on these addresses and MCNP5X input data. All are normalized per source particle
unless modified. Similarly, the reactor in the calculation of criticality (KCODE), neutron transport and Euler
relations are defined in the MCNP5X /ENDF library. With the MCNP5X simulation program, different tally
types are shown with particle types in Table -6. The variation of the With the MCNP5X simulation program,
different tally types are shown with particle types in Table -6. The variation of the TBR value versus the mixture
components of MCNP5X-5/ liquid blankets of different thicknesses.

Table.5 The regions of blanket IB material main compositions

Regions Inboard
Blanket 10% V structure + 90% Li
Replaceable shield 15% V structure + 10% Li + 75% V filler
HT shield 5% V structure + 76% Tenelon filler +
4% W structure + 5% Li
LT shield 15% Tenelon structure + 5% Li +
53% B4C filler (90% d.f.) + 27% WC filler (95% d.f.)
Vacuum vessel 35% Tenelon structure + 5% He coolant + 40% B4C filler
Regions Outboard
First wall 100% V structure
Blanket 10% V structure + 90% Li
Reflector 15% V structure + 10% Li + 75% Tenelon filler
HT shield 15% V structure + 5% Li + 73% Tenelon filler + 7% W
structure
LT shield 15% Tenelon structure + 5% Li + 80% borated Tenelon filler
Vacuum vessel 2% Tenelon structure + 5% He coolant + 70% B-Tenelon
filler

3.4.1. Tritium Breeding Rate (TBR) and Energy Multiplication Factor (M)

Tritium growth rate (TBR) is defined as the ratio of tritium production rate in the system to the burnt tritium
ratio in plasma. To provide adequate Tritium cultivation, there must be a medium containing liquid lithium
flowing. By using MCNP5X simulation, Table 7 of the Tritium production rates (TBR) resulting from the natural
Lithium (7.56% 6Li and 92.44% 7Li isotopes) for various blanket layer thicknesses (15, 20, 25, 30 and 50 in cm)
of the ARIES-RS reactor are given in Table-6 [1-3,23-27,30-33]. In the case of the DT fusion reactor, it should
be TBR ≥ 1,05 for self-sufficient Tritium. The amount of Tritium is taken from the reproduction reaction of 6Li
and 7Li isotopes in the blanket. Thermal neutrons with 6Li (n, α) T reaction and fast neutrons Tritium produced
with the help of 7Li (n,α,'n). Exothermic and Endothermic neutron capture reactions of 6Li and 7Li also affect M
values, respectively. These reactions are given as follows:

6
Li +n → 4He+T+Q(4.784) MeV (1)
7
Li +n → 4He+T+n +Q (-2.467 ) MeV (2)

461
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 6. Tally Cards and particle Type of MCNP5X

Mnemonic Tally Description Fn units *Fn units
F1:N or F1:P or F1:E Current integrated over a surface particles MeV
F2:N or F2:P or F2:E Flux averaged over a surface particles/cm2 MeV/cm2
F4:N or F4:P or F4:E Flux averaged over a cell particles/cm2 MeV/cm2
F5a:N or F5a:P Flux at a point or ring detector particles/cm2 MeV/cm2
F6:N or F6:N,P or F6:P Energy deposition averaged over a cell MeV/g jerks/g
F7:N Fission energy depositionaveraged over a cell MeV/g jerks/g
F8:P or F8:E or F8:P,E Energy distribution of pulses created in a detector pulses MeV
+F8:E Charge deposition charge N/A

In Figure- 5, local tritium breeding ratio TBR (atoms/sec) and M (MeV) states of energy multiplication
coefficient change versus % 6Li can be seen per source neutron. In graphic, it is clear that enriching the Li results
in a significant decrease in the TBR and a negligible increase M. In addition, the local TBR and M values are
shown for outboard and inboard regions. The overall TBR and M depend on the neutron coverage fractions
(NCF) of regions surrounding the plasma and the blanket thickness in each region. The relative NCF for the
inboard and outboard regions vary significantly with the aspect ratio. The Tritium breeding ratio (TBR) would
be given, TBR = Tbr6 + Tbr7, where Tbr6 and Tbr7 on Li6 ad Li7 depended, respectively. The Tritium breeding
ratio (TBR) can be stated as follows:

Tbr6= ∬ ∅ ∑ (n,α)T dE dV and Tbr7= ∬ ∅ ∑ (n,n՛,α)T dE dV (3)

Another important neutronic parameter is the energy impact factor M. It is defined as the ratio of the ratio total
energy deposited in the systems to the incident neutron kinetic energy. The energy produced in (D,T) reactor
should be higher than this energy. The plasma multiplication coefficient can be obtained by increasing the fusion
neutron energy, mainly in the blanket with 233U and 232Th fission. In addition to the kinetic energy transfer of
14.1 MeV Fusion-induced neutrons resulting from the Fusion DT reaction also Li is hoped to generate neutrons
in the additional doped fission energy from a blanket of melt salts. In fusion reactors, fission reactions contribute
to energy production, thereby, electrical energy is produced in such a way that it contributes significantly to
energy multiplication (M). The basic equation for the M relation can be expressed with the following
formula. The ARIES-RS fusion reactor has a fusion power of 2170 MW and an energy multiplication of M =
1.2. This produces a total power of ~2600 MW. Due to the low fission rates found in the fusion zone, the energy
ratio is somewhat low [1, 17,31-34,34-38]. Energy multiplication factor (M) can be given as follows:
200<𝑗>
M=1+{ +0,34 (𝑇𝑏𝑟6)+0,175 (𝑇𝑏𝑟7)} and < j >= ∬ ∅∑F dE dV (4)
14.1

3.4.2. Fissile fuel production rate

The fertile isotope 232Th transforms into a high quality fossil fuel (238U) isotope by capturing an energetic
neutron formed as a result of fusion 232Th (n, βe) 233U. However, some of the fissile quantities of this breed
would be reduced due to the burning of the new fossil fuel during the running time. Therefore, more than half
the accumulated nuclear fuel can be burnt after a year of operation. However, some of these breed fissile
quantities would be reduced due to the burning of new fossil fuel in place during the running time. Therefore,
more than half of the accumulated fissile fuel can be burned after a year of plant work. The fissile fuel breeding
reactions are given as follows:

462
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

238
92 𝑈 + 10𝑛 → 239
92𝑈 ∗→ 239
93𝑁𝑝 + 𝛽− →239
94𝑃𝑢 +𝛽
−
(5)
232
90 𝑇ℎ + 10𝑛 → 233
90𝑇ℎ ∗→ 233
91 𝑃𝑎 + 𝛽− →233
92𝑃𝑢 +𝛽
−
(6)

The amount of damage caused by radiation from the reactor refers to the degradation of a mesh crystal lattice to
its current position of high-energy radiation passing through it. Liquid melts walls are especially important in
terms of reducing activation radiation damage to structural materials, as among their important advantages in a
fusion reactor.

Figure 5. Local TBR and M exchange Versus Of % Li

Significant damage to structural materials will result in displacement of Lattice atoms (DPA) as a result
of their interaction with collisions with very fast neutrons. However, radiation effects are the result of radiation
damage to the mechanical and physical properties of the soil, and the mechanism of damage is the production
of metallic gas Hydrogen from nuclear reactions. Hydrogen atoms and isotopes will diffuse outward from the
first wall material with high temperatures, while Helium atoms will be accumulated in the first wall of the fusion
reactor. These damages will restrict the structure of the first wall and the working life of the reactor. Effects of
liquid layer thickness on radiation damage parameters such as atomic displacement (DPA) and proton,
Deuterium, Tritium and Helium production rates [23-27,30-33,34-38]. In Figure- 6 TBR per fission neutron is
shown for MCNP and ANISN. In addition, the ARIES-RS fusion reactor modelling with the MCNP5X code
modification is shown in Figure- 7 from simulation procedure.

463
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 6. Compare of MCNP5X and ANISN simulation.

Figure 7. ARIES-RS fusion reactor with MCNP5X code motification.

3.4.3. Gravity effect of FW Pb

The effect of FW's gravity is 110 kg per Pb / channel weight. The total outer Pb is about 16 kg, and the weight
of Pb produces a maximum of 130 Mpa Von Mises stress in the channel structure. In Figure-8, the
variation of gravitational effects of FW /Pb-Filled duct is shown. The free floating channel is 100 mm from the
center, thereby Pb creates an equivalent tension in channel displacement [11,21]. The heat deposition rate has a
shape similar to the energy multiplication coefficient. In addition, the heat accumulation rate of fission reactions
is vital. In this study, for the Modification ARIES-RS reactor, heat accumulation rate various
mixture compositions of molten salt and heavy metals, thickness of blanket layers, 4MW/m2 neutron wall load
are analysed. Figure- 9 shows that nuclear heat deposition rate in the Li (Li V4Cr4Ti) reflector systems. The
steepness of the profiles in the Li (Li V4Cr4Ti) blanket is much less than in the Flibe/glib-Fs) blanket. The rate
of heat accumulation in the system increases with the increased content of heavy metal salt for the liquid first
wall, a blanket and shield, a selected fluid layer thickness (Fig. 8). In blanket and shield, a selected fluid layer
thickness. In terms of nuclear energy, the fission capture cross-section of 233U is greater than that of 232Th fission
combustion, so the heat accumulation rate of Uranium is much higher than that of Thorium [9,27-30,31-34].

464
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 8. The variation of gravity effects of FW /Pb-filled duct.

With respect to neutron radiation damage, displacement per atom (DPA) and the Helium production rate
are evaluated. The highest material damage under neutron irradiation is expected at the first wall, where
Vanadium is the major constituent. For Vanadium, a damage limit of DPA=200 is suggested [2-4,17-19].

Figure 9. Nuclear heat deposition rate in the Li(Li V4Cr4Ti) reflector.

Figure 10. NWL of ARIES-RS Reactor versus distance from midpoint.

With respect to neutron radiation damage, a displacement per atom (DPA) and the helium production rate is
evaluated. The highest material damage under neutron irradiation is expected at the first wall, where vanadium
is the major constituent. For vanadium, a damage limit of DPA=200 is suggested [2-4,17-19].

465
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3.4.4. Neutron wall loading

For the full assessment of neutron wall loading, it is primarily necessary to determine the amount of radiation
interacting with the structural material of the ARIES-RS reactor. It is important to design the various components
of the neutron wall charge and fusion power core and evaluate the radiation environment and intensity around
the torus geometry. In the neutron wall loading with the change of vertical distance from the middle plane,
Tritium production depends on the blanket structure. Plasma systems is made up 10-20% of the first wall.
Figiure- 10 shows that NWL of ARIES-RS reactor versus distance from midpoint. ARIES-RS of diverter
structure depending on the outboard (OB) and inboard (IB) NWL values are 5.67MW/m2 and 4.03
MW/m2, respectively. Due to its asymmetric geometry, the upper half of the first wall and the diverter are
modelled with the code MCNP5X-5 to create the poloidal distribution. The ARIES-RS forward power fusion
reactor creates a maximum fusion power of 2.170 MW. The total surface area of the fire wall and resulting in
the machine's average peak (4.1 MW/m2). The NWL value reaches a peak in the midpoint of the outboard FW
with the outboard and the diverter plate and exchange outboard [1-3]

4. CONCLUSIONS

In this study, a model Monte Carlo study has been performed for the advanced ARIES-RS fusion reactor. The
advantages of using the proposed model in the mixed mode by adding a fission zone to the system are reported.
The main results of the present study can be summarized as follows: By placing such a fission zone in the
ARIES-RS reactor, a significant amount of high-quality fissile fuel production is possible. In addition, the
emergence of various energies arising from the plasma and the initial setup cost (COS) of the rector allows the
use of candidate solution salts of different thicknesses and spontaneous tritium capable of forming Tritium [1-
3,8-10]. The design and calculations of ARIES-RS have been performed by a 3D ball torus using the MCNP5X
simulation code / Endf -5B/VI library. In order to ensure the continuity of the operation of the reactors, one of
the important tasks in Aries-RS is to provide Tritium self-sufficiency. In particular, the Li17Pb83 composition is
carried out for nuclear heating. Blanket and harvest high-temperature Li17Pb83 composition is a breeding
material, important for both good Tritium reproductive potential and self-cooling conditions. In the case of no
additions to the blanket, TBR ≥ 1.20 values are observed for molten. In the case of no additions to the blanket,
TBR ≥ 1.20 values are observed in molten liquid such as plasmas have been enhanced by compatibility between
high temperature materials, MHD model inside and outside the blanket, and the effects of electromagnetic
forces. It will offer a neutron source to conventional fission type convection reactors. It will also assist in neutron
activation analysis used in various sites. Due to the low fission rates in the fusion zone among the types of fusion
energy studied, the matrix material, 232Th and 233U which are selected as fuel is more efficient in terms
of performance of the reactor. The world present Thorium reserves are many times higher than the Uranium
reserves. This type of ARIES-RS reactors with less cost and ergonomic and high performance will allow more
efficient usage of Thorium in order to provide fissile fuel for conventional fission reactors.

6. REFERENCES

[1] F.Najmabadi R, fusion engineering and design 65, 143-164(2003).

[2] Aries team,M.S. Tillack,X.R.Wang et al., fusion energ and design. 49–50, 689–695(2000).
[3] F. Najmabadi, Fusion Technol. 30 (3) (1996) 1286.
[4] Laila A. El-Guebaly ,The Aries team,fusion engineering anddesign 65 ,263-/284(2003).
[5] T.K.Mau,S.C.Jardin,C.E.Kessel,J.E.Menard,R.L.Miller.et al.,physics basis for the aries-st powerplant,
www.ferp.ucsd.edu.
[6] Bremister,j.,MCNP5X-4a general monte carlo code n-particle transport code,version 4a, la-12625.(1993)
[7] MCNP5X/a general monte carlo n-particle transport code,version 4B, los alamos national laboratory report la-
12625,(1993).
[8] A Hançerlioğullari,M. Cini,journal of fusion energy, doi:10.1007/s10894-013-9637-6, (2013).
[9] M.S.Tillack,X.R.Wang,J.Pulsifer,S.Malang, ,fusion energy and design. 65, 215–261 (2003).
[10] A.Hançerlioğullari,journal of kastamonu education14(2),545-556,(2006).

466
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[11] K.Miyamoto,Controlled fusion and plasma physics,taylor&francis group,crc press29-111,(2007).

[12] W.Reiersen,F.Dahlgren, H.M.Fan ,C.Neumeyer,I.Zatz.,fusion engineering and design.,65,303-322(2003)
[13] K.Starke,L.Buhler, S.Horanyi, Fusion energ des 84,1794–1798 (2009).
[14] Tokamaks,plasma physics and controlled nuclear fusion,springer series on atomic ,38,269-317,(2005
[15] O.Benes,R.J.M.Konings, ,ournal of fluorine chemistry,130:22-29(2009).
[16] T.G.Cowling,M.A.,D.Phil.,F.R.S.,magnetohydynamics,published by Adam hilger ltd(1976).
[17] M.A Abdou,,The APEX team., Fusion engineering and Design, 45 ,67-145( 1999).
[18] M.A Abdou ,A.Ying, et al., Fusion engineering and Design,54,181-247(2001)
[19] R.E.,Rognlien,T.D., Rensink,M.E.,Smolentsev, S.S, M.Z ,Youssef , Sawan,et al.Ar.,Fusion Engineering and Design
,72,181-221,(2004)
[20] Youssef,M.Z.,Morley,.N., El-Azabet,A., X-rays surface and volumetric heat deposition and tritium breeding issues in
liquid-protected FW in high power density devices. Paper presented at the 13. topical meeting on the technology of
fusion power. Nashville, Tennessee, 7-11 ,(1998).
[21] M A.Abdou,.Preface,Fusion engineering and design 72,1-2 ,(2004)
[22] J.Bremister., MCNP5X-4A general monte carlo code n-particle transport code, version 4A, La-12625, (1993).
[23] R.W Moirns., Nuclear Fusion 37:557,(1997).
[24] M A.Abdou,.Fusion Eng.Des.,45,145-167,(1999).
[25] M A.Abdou ,A.Ying,N.,Morle.,Fusion Eng.Des.,54,181-247,(2001)
[26] S.Şahin, M.Übeyli.,Energy Convers,Manage,45,1497-1512,(2004).
[27] X.P Lio T.Y.,Tong .,FDS Group, Fusion Engineering and Design 75-79,1275-1279,(2005).
[28] B.Şarer,M.Günay.,M.E.Korkmaz.,A.Hançerlioğulları,Fusionsciencetechnology., 52(1),107-115,(2007).
[29] K.A.McCarty,D.A.Petti., R.L. Moore.B.J., Merrill.,Fusion Eng.Des.,51-52,549-554(2000)
[30] Sze D.K.,”Argonne National Laboraty,1-16,(2001).
[31] L. Waganer, F. Najmabadi, M.S. Tillack, What MustDemo Do Proc. IEEE 16th Symp. on Fusion Eng.,Champaign,
IL,(1995).
[32] F. Najmabadi, et al., The Starlite Project—Assessment Phase Report, University of California, San Diego report
UCSD-ENG-005, (1996).
[33] J. Kaslow, et al., Criteria for Practical Fusion PowerSystems—Report from the EPRI Fusion Panel, EPRI report BR-
104469, (1994).
[34] F. Najmabadi, R.W. Conn, et al.,The ARIES-I Tokamak Fusion Reactor Study—The Final Report, UCLA
reportUCLA-PPG-1323,(1991)
[35] R.W. Conn, F. Najmabadi,et al., ARIES-I: A Steady-State, First-Stability Tokamak Reactor with Enhanced Safety and
Environmental Features, Proc. 13th Int. Conf. on Plasma Phys. and Controlled Nucl. Fusion Research, Washington,
DC, October1990, International Atomic Energy Agency, Vienna, p. 659.,(1991).
[36] Y. Seki, M. Kikuchi, et al., Concept Study of the Steady State Tokamak Reactor (SSTR), Japan Atomic
EnergyResearch Institute Report JAERI-M-91-081, 1991
[37] First Results and Plasma Current Start-Up in Taban Tokamak,H. R. Mirzaei , R. Amrollahi, journal of Fusion
Energy ,38, 548–556(2019)
[38] A. O. Kovalev, A. R. Polevoi, E. I. Polunovskiy, L. Bertalot, Yu. A. Kashchuk,D. V. Portnov, A. Loarte, M.
Loughlin S. D. Pinches , journal of Fusion Energy , 39, 40–52,(2020).

Appendixes
Table 7. MCNP5X /The Variation Cell of The TBR and M.
% 10 Li Volume TBR M
Cell (Cm3) (Atom/Cm3) (Mev)
2 6,40395E+06 7,51555E-02 8,95714E-01
22 2,34092E+06 2,52927E-02 2,83951E-01
3 1,60390E+08 8,25353E-01 6,92654E+00
33 5,29112E+07 2,51907E-01 2,01376E+00
5 2,29741E+07 0,00000E+00 1,54006E-02
55 4,92448E+06 0,00000E+00 4,25595E-03
Toplam 2,4994E+08 1,22742 1,726970
%20 Li Volume TBR M
(cm3) (Atom/cm3) (MeV)
Cell
2 6,40395E+06 8,00250E-02 9,60119E-01
22 2,34092E+06 2,71252E-02 3,05675E-01
3 1,60390E+08 8,20777E-01 7,08465E+00
33 5,29112E+07 2,49492E-01 2,05482E+00
5 2,29741E+07 0,00000E+00 1,51309E-02

467
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

55 4,92448E+06 0,00000E+00 4,15592E-03

Toplam 2,4994E+08 1,22459 1,746187

% 30 Li Volume TBR M
Cell (cm3) (Atom/cm3) (MeV)
2 6,40395E+06 8,01538E-02 9,96813E-01
22 2,34092E+06 2,67981E-02 3,13868E-01
3 1,60390E+08 8,04062E-01 7,17338E+00
33 5,29112E+07 2,45430E-01 2,06638E+00
5 2,29741E+07 0,00000E+00 1,39139E-02
55 4,92448E+06 0,00000E+00 3,68782E-03
Toplam 2,4994E+08 1,20016
1,755120
%50 Volume TBR M
Cell (cm3) (Atom/cm3) (MeV)
2 6,40395E+06 7,84823E-02 1,03718E+00
22 2,34092E+06 2,69168E-02 3,28712E-01
3 1,60390E+08 7,81582E-01 7,26450E+00
33 5,29112E+07 2,36461E-01 2,08037E+00
5 2,29741E+07 0,00000E+00 1,35684E-02
55 4,92448E+06 2,90189E-02 3,62490E-03
Toplam 2,4994E+08 1,16217 1,764768
%80 Volume TBR M
Cell (cm3) (Atom/cm3) (MeV)
2 6,40395E+06 7,77228E-02 1,08133E+00
22 2,34092E+06 2,66242E-02 3,41285E-01
3 1,60390E+08 7,61577E-01 7,35674E+00
33 5,29112E+07 2,27577E-01 2,08703E+00
5 2,29741E+07 2,63220E-02 1,63678E-01
55 4,92448E+06 8,82426E-03 5,58882E-02
Toplam 2,4994E+08
1,12865 1,774444
%90 Volume TBR M
Cell (cm3) (Atom/cm3) (MeV)
2 6,40395E+06 7,53879E-02 1,09108E+00
22 2,34092E+06 2,58901E-02 3,44724E-01
3 1,60390E+08 7,48911E-01 7,40418E+00
33 5,29112E+07 2,23140E-01 2,09681E+00
5 2,29741E+07 2,57399E-02 1,61458E-01
55 4,92448E+06 8,46106E-03 5,43257E-02
Toplam 2,4994E+08
1,779093
1,10753
%100 Volume TBR M
Cell (cm3) (Atom/cm3) (MeV)
2 6,40395E+06 7,53879E-02 1,09108E+00
22 2,34092E+06 2,58901E-02 3,44724E-01
3 1,60390E+08 7,48911E-01 7,40418E+00
33 5,29112E+07 2,23140E-01 2,09681E+00
5 2,29741E+07 0,00000E+00 1,33043E-02
55 4,92448E+06 2,57399E-02 1,61458E-01
Toplam 2,4994E+08
1,10321 1,779867

468
 THE LOAD SHEDDING PREDICTION IN GECOL USING
ARTIFICIAL NEURAL NETWORK
Ashraf Mohammed Abusida
Department of Computer Engineering, Kastamonu University, Kastamonu-Turkey, [email protected],
ORCID: 0000-0002-9735-5814

Aybaba Hançerlioğullari
Art & science faculty, Physics Department, Kastamonu University, Kastamonu-Turkey, [email protected],
ORCID: 0000-0000-1700-8480

Javad Rahebi
Department of Electrical and Computer Engineering, Altinbas University, Istanbul-Turkey, [email protected],
ORCID: 0000-0001-5418-9601

Erol Kurt
Department of Electrical and Electronics Engineering, Gazi University, Ankara-Turkey, [email protected],
ORCID: 0000-0002-3615-6926

Seçil Karatay
Department of Electrical and Electronics Engineering, Kastamonu University, Kastamonu-Turkey, [email protected],
ORCID: 0000-0002-1942-6728

Yasemin Gültepe
Department of Computer Engineering, Kastamonu University, Kastamonu-Turkey, [email protected],
ORCID: 0000-0002-8684-9907

Rezvan Rezaeizadeh
Science faculty, Physics Department, University of Guilan, Rasht-Iran, [email protected],
ORCID: 0000-0003-6219-6174

Hançerlioğlları, A., Kurt, E. The load shedding prediction in gecol using artificial neural
Cite this paper as:
network, 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul.

Abstract: The conducted experiments produce a prediction model to help the company's administrators and
technicians make the right and quick decision. A supervised machine learning technique has been
adopted and applied for SCADA data. In this study, for the load shedding prediction in general
electricity company of Libya, an artificial neural network with multilayer perceptron is used. The
energy consumption is used for prediction. For inputs of the neural network, four parameters are
used as well. These parameters are power generation, temperature, humidity and wind speed. The
collected data from the SCADA database were pre-processed to be prepared in a suitable format
to be fed to the Neural Network algorithm. A set of experiments has been conducted on this data
to obtain a prediction model. The produced model was evaluated in terms of accuracy and
reduction of loss. It can be concluded that the obtained results are promising and encouraging.
Keywords: Load Shedding Prediction, GECOL, Artificial Neural Network

© 2020 Published by ECRES

469
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION

Nowadays, there exists a lot of information and data that can be handled from business transactions and scientific
data and information retrieval are simply no longer enough for decision making. It is an industrial computer
system that monitors and controls a process all of the transmission and distribution elements of electrical utilities,
and monitoring substations, transformers and other electrical assets for the company. This technique has been
applied for the first time to SCADA data. This data have been recorded from January 2013 until December 2019.
They are in the form of historical data every hour throughout the day on the total energy produced from power
stations, the total network load and the average temperatures, humidity, weather conditions and wind speed for
these historical periods. This technique would additionally predict the expected amount of energy consumed to
manage load shedding according to certain hours of the day represented by peak hours, and according to specific
periods of the year and weather conditions for that period, to maintain the stability of the electrical network and
reduce the cost. Additionally, this can assist the company's decision-makers in taking the appropriate decisions
more quickly and maintain the quality of service provided by the company [1][2].

The general electric company of Libya GECOL was established Under Act No. 17 of 1984 AD and is responsible
for the completion of the projects for operating and servicing the electric networks, stations for energy
production and their distribution and transformation stations. Also, the company is responsible for the energy
transmission lines and their distribution, the electricity control centres and the management of the operation and
servicing of desalination stations in the whole country [1].

Predicting the amount of energy consumed and managing electric load shedding status hourly is a basis for
strategic information and one of the keys to the stability of the electrical networks in this company, and it also
effectively helps in planning and improving the quality of services it provides [3].

Conventional forecasting currently approved is a waste of time and is costly; further, it requires experts,
engineers and technicians with experience in the production, distribution and transmission of electric energy,
and who are often limited and have a historical archive of this data, not to mention they are prior and expected
knowledge of the loads to be provided in the electrical network. Prior and expected knowledge is required
according to the peak hours of the day, according to a specific season of the year and according to specific
climatic conditions, especially in the presence of a vast amount of readings for the amount of electrical energy
produced from power plants, the increasing demand for energy in the produced electrical network and the
temperatures recorded at specific times [4].

To cope with these issues, machine learning algorithms have been proposed to support the automated prediction
of load shedding for the electric network. Among machine learning algorithms, Neural Networks (NNs) have
been applied to SCADA data set of the electricity company [5].

The problem arose in addressing the presence of a huge amount of data stored in the database of General
Electricity Company of Libya (GECOL) and represented in the database of SCADA, where the decision-makers
of the electricity company were not taking advantage of this historical data that was recorded in the SCADA
database to predict electrical loads. The expected energy required by the electrical network under certain climatic
conditions of temperature, humidity, or wind speed can help them in making strategic decisions as quickly as
required. Accordingly, the idea of this article is to invest the available digital format data in producing beneficial
prediction models of electrical loads for the energy required with better accuracy in terms of load shedding,
which in turn can help decision-makers make the right and quick decisions regarding the stability of electricity
network in their company. Additionally, the obtained knowledge would provide the engineers and technicians
in GECOL with the essential information that can be used to draw up plans for the regulation and management
of electrical energy distribution [6].

A program was developed by a well-known software package for data extraction and exportation to external
files which is referred to as Visual Studio 2019 and Microsoft Office 2019. These tools are applied to the
collected data from the SCADA system, that can be dealt with in this study and configured in a format that is
appropriate for this study represented in the artificial neural network.

470
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In this paper, a machine learning technique is applied to a vast amount of historical data stored in the database
of General Electricity Company of Libya (GECOL) and represented in the database of supervisory control and
data acquisition (SCADA).

1.1 Data set

The dataset used for experiments in this paper was historical data on the total electrical energy produced per
hour, the temperature recorded at the same time, humidity, and wind speed within a specified period from
January 2013 until December 2019 are specified as shown in Table 1.
Table 1. GECOL Dataset Properties
Variables Unit Description
Date Date time Represent of Date of day
Time Time Represent the time in the day
Represent The amount of electrical
Energy Generation Numeric
energy produced by (MW)
Temperature Numeric Represent of Temperature
Humidity Numeric Represent a percentage of Humidity
Wind Speed Numeric Represent The wind speed by km/h

As a sample of the collected data used in this study, Figure 1 illustrates in a simplified way to represent the data,
which includes the value of the energy produced, the value of energy consumed, the value of load-shedding in
a specific hour of the day and different periods of the year and under different climatic conditions.

7000

6000

5000

4000

3000

2000

1000

0
Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample7 Sample 8

Energy Generation Energy consumed load-shedding

Figure 1. A sample of the data collected

The obtained knowledge would provide the decision-makers with the essential information that can be used to
draw up plans and to ensure continuity and quality of service.

The rest of the paper is organized as follows: Section 2 reviews artificial neural network concepts, Section 3
provides information on the prediction algorithm and the experimental result, and finally, Section 4 presents
conclusions.

2. Artificial neural network

An artificial neural network is a model for the processing of information inspired by the study of bioelectric
networks in the brains of people and animals, produced by neurons and their synapses [7].

The mathematical analogue of the biological neural network is a set of interconnected simple computational
elements (neurons). Each neuron receives signals from the others (in the form of numbers), sums them up, as

471
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the sum goes through an activation function, and thus determines its activation (excitation degree), which is
transmitted through the output connections to the other neurons. Each connection has a weight which,
multiplying by the signal, determines its significance (strength). The weights of the connections are analogous
to the strength of the synaptic impulses transmitted between biological neurons. A negative value of weight
corresponds to a suppressive impulse and a positive value to an excitatory one [8].

The neuron model shown in simplified form in Figure 2 can also be considered as a threshold volume.

Figure 2. The structure of an artificial neuron

Threshold (θ) unit, which collects the output and hazel Harmandir only produces an output exceeds the sum of
the internal threshold of the entrance as a threshold neuron unit receives signals from the synapse and collects
all the signals that are generated by multiplying the appropriate weight. You will now sign up to the power of
the threshold gathered strong sign stimulus is transmitted along the axon and dendrites of other neurons.
Intersecting compared with all the signs of internal threshold gated neurons with dendrites and synapses from
the axons sign spreads threshold is exceeded. Output depending on whether the result of the sum function value
is above or below the threshold activation functions are formed by the normalizing. ANN, connecting these
simple nodes and the unit is obtained by conversion to a network. ANN block diagram is shown in Figure 3.

Figure 3. ANN general block diagram

To set the threshold value of the signal entering the activation function at this point bias (θ) is called a fixed
input value applied to the input neurons, this allows the threshold to be changed to receive the expected outputs.
ANN as a basic, simple structure and versatile. Each node cells called the nth degree and a non-linear transfer.
Processing elements called nodes and links therebetween. Each connection is involved in the transmission of
one-way mark delay. An unlimited number of processing element inputs and a single output connection.
However, if this needs to be copied and used in many cells of a single output power supply, the output of the
processing element may be any desired mathematical type. Indicate the function of the output of the processing

472
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

element x input value, y output value, Ψ transfer function, f (x) collecting function, wkj k is the connection
weights, θk for sequential neurons, including neurons sequential threshold for k can be expressed as follows [9].

𝑓 (𝑥) = 𝑤𝑘1 𝑥1 + 𝑤𝑘2 𝑥2 +. . . +𝑤𝑘𝑛 𝑥𝑘𝑛 + 𝜃 = ∑𝑛𝑗=0 𝑤𝑘𝑗 𝑥𝑗 ; 𝜃 = 𝑤𝑘0 𝑥0 (1)

𝑦 = 𝜓(𝑓(𝑥)) = 𝜓(∑𝑛𝑗=1 𝑤𝑘𝑗 𝑥𝑗 ) (2)

Expressed in the form of a neuron matrix are as follows.

𝑥0
𝑥1
𝑓𝑘 (𝑥) = [𝑤𝑘0 𝑤𝑘2 𝑤𝑘3 . . . 𝑤𝑘𝑛 ] [ . ] = 𝑤𝑘𝑇 𝑥 (3)
𝑥𝑛

Inputs are information that enters the cell from other cells or the external environment. The information enters
the cell via links on weight and weights, will determine the effect on the corresponding input cell [10].

The architecture of the proposed neural network with backpropagation algorithm is illustrated in figure 4.

Figure 4. Architecture of proposed MLP with backpropagation algorithm

As shown in figure this figure, four inputs are used. for the output the energy consumption is used.

The levenberg Marquardt [11] is used for the training algorithm. In this algorithm, for the estimation of the error,
the mean square error is used for the reducing of the error.

3. Experimental result
For the implementation of the proposed method, the Matlab 2019a with 4GB Ram, Core i7, 8th generation laptop
is used. about 10 times the method was run and the result saved in table 1.

The proposed ANN-MLP is shown in figure 5.

473
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 5. the proposed ANN-MLP

The result of the train data and output neural network is shown in figure 6. The red-coloured result shows the
target real data (energy consumption), and the blue-coloured result shows the ANN result for train data. The
MSE value for this scenario is 151344.4585.

Figure 6. Result for target train data and output of the neural network

474
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The result of the test data and the output neural network is shown in figure 7. The red-coloured result shows the
target real data (energy consumption), and the blue-coloured result shows the ANN result. The MSE value for
this scenario is 155927.63.

Figure 7. Result for target test data and output of the neural network

The performance analyzing of this scenario is shown in figure 8.

Figure 8. performance analyzing for 1000 iteration

As shown in this figure the best MSE is 0.010325 and this value obtained after 20 iterations.

The training state of the model is shown in figure 9.

475
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 9. training state of the proposed model

As shown in this figure the gradient value in each iteration depends on this model. The best gradient value is
2.67e-5 and this value obtained after 1000 iteration.

For the mu the best value obtained after 1000 iteration and this value is 1e-12. This parameter is the training
rate. Finally, the validation fail is 0 for each iteration in the training state.

The regression analyzing of the proposed model is shown in figure 10.

Figure 10. regression analysis

In this figure, the regression is obtained 0.9234, normally this value should be near to 1, but the data used in the
proposed model is nonlinear.

The Mean Square Error for training and testing data is shown in figure 11. The number of neurons that used for
the model are 5, 8, 10,12 and 16. For each scenario, the result is illustrated as well.

476
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

170000.00

165000.00

160000.00

155000.00

150000.00

145000.00

140000.00
Neuron Neuron Neuron Neuron Neuron
number: 5 number: 8 number: 10 number: 12 number: 16
MSETrain 165003.00 158685.55 157273.99 153890.77 151344.46
MSETest 162045.58 162908.42 161913.22 151581.67 155927.63

Figure 11. Mean Square Error for training and testing data

As seen in this figure when the number of the neurons are increasing the MSE value is decreasing. So for the
best model, the number of neurons must increase. With the low number of the neuron, we cannot get a good
result.

The Mean Absolute Percentage Error (MAPE) and R2 score for training data and testing data are shown in figure
12.

90.00%
80.00%
70.00%
60.00%
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
Neuron Neuron Neuron Neuron Neuron
number: 5 number: 8 number: 10 number: 12 number: 16
MAPETrain 5.67% 5.57% 5.45% 5.45% 5.40%
MAPETest 5.78% 5.58% 5.49% 5.55% 5.55%
R2Train 84.15% 84.62% 85.28% 85.45% 85.54%
R2Test 83.45% 84.52% 84.80% 84.60% 84.57%

Figure 12. MAPE and R2 score for training data and testing data

The Root Mean Square value for training and testing data is shown in figure 13.

477
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

415.00
410.00
405.00
400.00
395.00
390.00
385.00
380.00
375.00
Neuron Neuron Neuron Neuron Neuron
number: 5 number: 8 number: 10 number: 12 number: 16
RMSETrain 407.27 400.96 392.26 390.41 388.98
RMSETest 414.73 401.74 397.99 399.68 400.47

Figure 13. Root Mean Square value for training and testing data

4. CONCLUSION
In this study for the prediction of the energy consumption, multilayer perceptron with one hidden layer and multi
neuron numbers are used. Simulation results show that the proposed method has good accuracy for the prediction
of energy consumption. The accuracy of the prediction is 95%, and this value is getting after running ten times.
For the input, four parameters are used. The obtained result shows that this system can be used instead of the
real system.

REFERENCES

[1] W. Alsuessi, “GENERAL ELECTRICITY COMPANY OF LIBYA (GECOL),” Eur. Int. J. Sci. Technol., vol. 4, no.
1, pp. 1–9, 2015.
[2] A. M. Abusida and Y. Gültepe, “An Association Prediction Model: GECOL as a Case Study,” 2019.
[3] F. Ahwide and Y. Aldali, “The current situation and perspectives of electricity demand and estimation of carbon
dioxide emissions and efficiency,” Int. J. Environ. Ecol. Eng., vol. 7, no. 12, pp. 979–984, 2014.
[4] G. Vukojevic, A. Svalovs, K. A. S. Ghadem, and A. O. D. Ali, “Transient analysis of svc response in the south region
of the Libyan transmission network,” in 2011 IEEE Trondheim PowerTech, 2011, pp. 1–6.
[5] M. Khamaira, A. Krzma, and A. Alnass, “Long Term Peak Load Forecasting for the Libyan Network,” 2018.
[6] H. ZHANG, R. YUAN, and W. SUN, “APPLICATION OF DATABASE ACCESS MIDDLEWARE
TECHNOLOGY IN SCADA DATABASE SYSTEM [J],” Power Syst. Technol., vol. 17, 2005.
[7] F. Cardinaux, M. Enenkl, F. Giron, T. Kemp, and S. Uhlich, “Method, system and artificial neural network.” Google
Patents, 18-Feb-2020.
[8] R. Fournier, L. Wang, O. V Yazyev, and Q. Wu, “Artificial Neural Network Approach to the Analytic Continuation
Problem,” Phys. Rev. Lett., vol. 124, no. 5, p. 56401, 2020.
[9] M. H. Hassoun, Fundamentals of artificial neural networks. MIT press, 1995.
[10] M. Van Gerven and S. Bohte, “Artificial neural networks as models of neural information processing,” Front.
Comput. Neurosci., vol. 11, p. 114, 2017.
[11] J. J. Moré, “The Levenberg-Marquardt algorithm: implementation and theory,” in Numerical analysis, Springer,
1978, pp. 105–116.

478
 ECO-TOXIC METAL POLLUTION IN SURFACE WATER AND
ITS SUITABILITY FOR IRRIGATION AROUND A LIGNITE-
FIRED THERMAL POWER PLANT IN TURKEY

Şeref Turhan
Kastamonu University, Arts& Sciences Faculty, Physics Department, Kastamonu/Turkey, [email protected],
ORCID: 0000-0000-5303-3680

Aybaba Hançerliogulları
Kastamonu University, Arts& Sciences Faculty, Physics Department, Kastamonu/Turkey [email protected],
ORCID: 0000-0000-1700-8480

Aslı Kurnaz
Kastamonu University, Arts& Sciences Faculty, Physics Department, Kastamonu/Turkey [email protected],
ORCID: 0000-0002-7910-3461

Cellattin Duran
Kastamonu University, Arts& Sciences Faculty, Geography Department, Kastamonu/Turkey
[email protected], ORCID: 0000-0002-6864-5564

Onur Metin
Kastamonu University, Taşköprü Vocational School, Taşköprü, Kastomunu University, Kastamonu,/Turkey
[email protected], ORCID: 0000-0001-6711-6452

Turhan et al., Eco-toxic metal pollution in surface water and its suitability for irrigation
Cite this paper as: around a lignite-fired thermal power plant in Turkey, 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul.

Abstract: Lignite coal is very important domestic fuel for electricity generation in Turkey. However, toxic
metals found in by-products generated by coal-fired thermal power plants can be extremely
harmful to human health and aquatic life even in small doses because of their non-degradable
properties. Also coal combustion flue gases such as SOx and NOx contribute to acidification of
streams, lakes, oceans and soil. This study executed to investigate the acidification and toxic metal
(Fe, Mn, Cu, Ni, Zn, Cr and Pb) pollution of Tohma stream flowing near Kangal lignite-fired TPP
in Kangal (Sivas, Turkey). All water samples were screened for pH to assess the acidification of
Tohma stream. The average concentrations of Fe, Mn, Cu, Ni, Zn, Cr and Pb in water samples
from Tohma stream were determined as 13.78, 2.27, 1.98, 1.24, 0.32, 0.94 and 0.54 mg/L using
EDXRF spectroscopy, respectively. Metal pollution index (MPI) and metal evaluation index
(MEI) estimated to assess Tohma water pollution with potentially toxic metals. The values of MPI
and MEI varied from 312 to 9715 with an average of 4713 and 181 to 317 with an average of 226,
respectively.
Keywords: Lignite-burning thermal power plant, metal evaluation index, metal pollution index, surface water,
Tohma stream
© 2020 Published by ECRES

1. INTRODUCTION

Lignite-fired thermal power plant (LFTPP) has an indispensable importance in Turkey's electricity production
[1]. However, these LFTPPs produce a huge amount of fly ash, suspended particulate matter, acid gases (SOx
and NOx), and other pollutants from coal combustion and wastewaters [2]. Every day, CFTPPs discharge fly
ashes and wastewaters loaded with toxic metals such as As, Hg, Fe, Mn, Cu, Ni, Zn, Cr and Pb, etc. into

479
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

environment (air, soil and water). The pollution of air, soil, ground, and surface (rivers, lakes, streams, and bays)
waters with toxic metals and other pollutants is an increasing risk to human being and marine life even in small
doses [3].
Some surface waters in Turkey are a very convenient water source for drinking water, irrigation, and aquatic life
like fish farming. Unfortunately, the unplanned urbanization and industrial activities (iron and steel factories,
cement plants, thermal power plants, chemical industry, fertilizer industry etc.) in Turkey have caused pollution
of such water resources to be difficult to restore. Toxic metal pollution has harmful impacts on the water quality
and sediment as well as another marine fauna [4]. Therefore, pollution of the potentially toxic metal is an
essential problem in Turkey as in numerous growing countries.
The knowledge of toxic metal concentrations in environmental samples is very important to evaluate the
potential impacts of the anthropogenic and industrial activities on the environment and human health, and
surface water quality. Recent surveys have focalized the investigation of toxic metal pollution of drinking water,
groundwater, surface water, and sediment [5- 13]. Demirak et al. [14] determined the concentrations of Zn, Pb,
Cd, Cu and Zn in bottom sediment, water, and tissues (gills and muscle) of Leuciscus cephalus from the stream
of Dipsiz in the Yatağan basin in south-western of Turkey, the site of a LFTPP. They showed that the
contaminants from the Yatağan LFTPP can be an origin of toxic metals. Türkmen and Çalışkan [15] investigated
the spatial and seasonal variations of Fe, Zn, Cd, Pb, Co, Mn, Ni, Cu, and Cr in Asi River in the northeast
Mediterranean region of Turkey. Varol and Şen [16] researched the potentially toxic metal pollution of
sediments and surface water collected from the upper Tigris River, which is one of Turkey's most important
rivers, by estimating geo-accumulation index and enrichment factor. They suggested that Zn, Cd, As, Cr, Co,
Mn, Cu and Ni are originated in human activity, especially metallic evacuations of the Ergani Copper Mine
Facility. Kalender and Uçar [17] investigated the metal pollution of sediments from tributaries of the Euphrates
River (between 428120 N latitude and 510760 E longitude) using pollution load index, contamination factor,
and geo-accumulation index. Sari et al. [18] investigated the contamination level of some potentially toxic metals
in surface sediments from Ergene River passing through Turkey's most heavily industrialized regions by
estimating enrichment factor and geo-accumulation index.
They revealed that the studied sediments were moderately polluted with As, Ag, and Cr, moderately to severely
polluted with Ni and Zn and severely polluted with Cu. Dede [19] evaluated the quality of surface waters
sustaining Çamlıdere barrage, Ankara's largest drinking water supply, in accordance with metal pollution using
metal pollution index. Omwene et al. [20] evaluated the contamination levels of toxic metals in surface
sediments from Mustafakemalpaşa creek (Turkey), which is sited in the world's largest borate basin by using
enrichment fact, geo-accumulation index, contamination factor, sediment quality guidelines and multivariate
statistical technics. They found that the potentially toxic metals (Cu, Pb, and Zn) in the sediments were originated
from fly ashes released from CFTPPs, waste leachate of the urban, and weathering of sulphide ore minerals.
Koç and Yılmaz [21] studied the metal pollution (Cu, Zn, Mn, Pb, Cd and Fe) of water samples from Büyük
Menderes River in Aydın (Turkey). They observed that Büyük Menderes River is not under the threat of
pollution. However, there is no detailed study on the investigation of toxic metal pollution of Tohma stream
flowing very close to the Kangal LFTPP in the literature according to our seeking. The aim of the study is (1)
to research the acidification of the Tohma stream due to the Kangal LFTPP, (2) to analyse the concentrations of
Fe, Mn, Cu, Ni, Zn, Cr and Pb in surface water samples collected from Tohma stream using EDXRF
spectroscopy, and (3) to assess the contamination (or pollution) of these potentially toxic metals in surface water
samples and the water quality of the Tohma stream with respect to drinking and agricultural irrigation aims by
estimating ecological indexes such as metal evaluation index and metal pollution index.

2. MATERIAL AND METHODS

Study area: Kangal Basin on the Taurus orogenic line is surrounded by the Tecer-Gürlevik, Felhan, Yaycı, Çatal
and Yılanlı, Gürün and Delidağ, and the Bozbel mountains, and the Uzunyayla Basin, with the threshold field
corresponding to the basin base in the west. The Kangal Basin with these boundaries morphologically
corresponds to a complete basin. The Kangal Basin is an inter-mountainous basin because it is surrounded by
high mountainous areas from the north and south. The Kangal Basin, formed between the north and south thrusts,
is a geologically Piggy-back basin. Kangal basin corresponds to an Upper Miocene-Pliocene age since it is a
continuous sedimentation area during the Upper Miocene-Pliocene. The Kangal Basin also includes Tohma
stream Upper Basin, which is one of the important tributaries of the Euphrates River [22]. Tohma consists of
the combination of streams of Kazikli and Mancinik. Tohma is formed by the combination of the Kazıklı and

480
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Mancınık streams. However, new water sources join in different directions along the route of Tohma stream.
Thus, the valley of the Tohma stream expands and the flow increases. It has the same direction of flow along
the southern slope plateau. It also played a role in the fragmentation of the plateau surface. The Tohma stream
passes close to the Kangal LTPP. It is poured into the Karakaya Dam Lake on the Euphrates River after passing
through several provinces and counties. The climate of the region is continental climate. The snowmelt that
starts in early spring causes the current to rise. The summer drought also markedly declines. The Kangal LFTPP
with a 457 MW total installed capacity is sited 25 km southwest of Kangal country of Sivas province [1]. Kangal
LFTPP consumes approximately 6,832,889 tons of lignite coal annually. At the same time, it produces
approximately 1.420.000 tons of fly ash [1].
Sampling and analysis: Nineteen water samples were gathered from different points on Tohma stream flowing
near the Kangal LFTPP using polyethylene sample containers. The sample points are shown on the map given
in Fig. 1. The sample containers were properly coded for identification of places and transported to the
measurement laboratory. Firstly pH value of the samples, which is an important indicator of water quality and
pollution level in the water ecosystem, was determined by using a pH meter (LaMotte 5 series). Seven ecotoxic
metals (Cu, Zn, Mn, Pb, Cd and Fe) in the water samples were analysed by using an EDXRF spectroscopy
(Spectro Xepos) having a thick binary Pd/Co alloy anode of X-ray tube with 50 W and 60 kV. It has an
autosampler for up to 12 items and software modules. The target changer having up to 8 polarization and
secondary targets presents a lot of various excitation conditions, proving the optimum specification of all
elements between potassium and uranium [23].

Figure 1. Location map of the investigated area showing the sampling site on Tohma stream

Evaluation of pollution water with toxic metals: In recent years, metal pollution index (MPI) and metal
evaluation index (MEI) have been utilized preferentially to assess the degree of toxic metal pollution of
underground and surface waters [4,9,24-25].

481
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

MPI is a powerful index for the assessment of quality of water based on metal concentration and it has been
developed and formulated as follows [9,24,26].

𝑈𝑖 𝑃𝑖
∑𝑁
𝑖=1
𝑈 (1)
𝑀𝑃𝐼 = 𝑁 𝑖
∑𝑖=1 𝑈𝑖

𝑁
|𝐶𝑖 − 𝐼𝑉𝑖 |
𝑃𝑖 = ∑ (2)
𝑆𝑉𝑖 − 𝐼𝑉𝑖
𝑖=1

where N is the number of toxic metals considered; Ui is the unit weightage of the ith parameter; Pi is the subindex
of the ith parameter; Ci is the analysed metal concentration of the ith parameter; IVi is the metal ideal value of
the ith parameter and SVi is the standard value of metal of the ith parameter. MEI is the general index that
considers the probable additive impact of potentially toxic metals about human health. MEI is estimated as
follows [9,27]:
𝑁
𝐶
𝑀𝐸𝐼 = ∑ (3)
𝐶𝑀𝐴𝑋𝑖
𝑖=1

where Ci is the ith metal concentration in a water sample and (CMAX)i is the maximum concentration for the
ith metal. The degree of pollution or water quality according to the HMPI and HMEI values are given in Table
1 [28-29].

Table I. The degree of pollution or water quality classification

Index Water type Value Degree of pollution
MPI <15 Low
Drinking water 15 ≤ MPI <30 Medium
MPI ≥ 30 High
Metal pollution index (MPI)
MPI < 300 Low
Surface water 300 ≤ MPI < 600 Medium
MPI ≥ 600 High
Drinking water MEI < 1.24 Low
1.24 ≤ MEI < 2.48 Medium
MEI ≥ 2.48 High
Metal Evaluation Index (MEI)
Surface water MEI < 150 Low
150 ≤ MEI < 300 Medium
MEI ≥ 300 High

3. RESULTS AND DISCUSSION

The pH values measured for the water samples are presented in the second column of Table 2. The average pH
value was found as 8.36 (range: 8.08 to 8.70). The pH values show that the investigated waters are alkaline and
within the allowable limit range (6.5–9.5) recommended by Turkish standard (TS) [30].The toxic metal
concentrations analysed in the water samples are given in Table 2.
Chromium in the form of Cr (III) and Cr (VI) in the soil, air, and water arises from natural processes
anthropogenic activities. Although the concentration of Cr in drinking water is usually low, the polluted well or
surface water may include the dangerous Cr (IV).

482
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table II. pH values and toxic metal concentrations of the Tohma water samples
Sample code pH Toxic metal concentration (mg/L)
Fe Mn Cu Ni Zn Cr Pb
W-1 8.45 13.30 2.30 2.10 1.10 0.30 0.90 0.40
W-2 8.38 18.20 2.50 1.30 1.20 0.50 0.80 0.30
W-3 8.36 12.30 2.30 2.90 1.50 0.20 1.10 0.40
W-4 8.42 10.60 2.10 2.00 0.80 0.30 0.90 0.50
W-5 8.45 12.20 2.10 2.10 0.90 0.20 0.80 0.40
W-6 8.36 11.20 2.20 1.40 1.10 0.40 1.00 0.50
W-7 8.08 12.80 2.20 1.30 1.30 0.30 1.00 0.60
W8 8.34 12.94 2.24 1.87 1.13 0.31 0.93 0.44
W-9 8.70 11.70 1.80 1.80 1.00 0.10 0.90 0.70
W-10 8.60 19.50 2.60 2.50 1.90 0.50 0.70 0.90
W-11 8.18 13.47 2.23 1.93 1.19 0.31 0.90 0.51
W-12 8.32 14.00 2.00 1.50 1.40 0.40 1.00 0.60
W-13 8.11 13.52 2.21 1.89 1.21 0.32 0.91 0.52
W-14 8.24 18.50 2.80 2.30 1.80 0.40 1.20 0.70
W-15 8.38 12.20 2.30 2.60 1.00 0.20 1.10 0.30
W-16 8.10 12.20 2.20 1.20 0.70 0.20 1.10 1.00
W-17 8.20 19.70 2.90 2.60 1.80 0.80 0.90 0.70
W-18 8.60 10.70 1.80 2.00 1.10 0.20 0.90 0.50
W-19 8.70 12.30 2.40 2.40 1.30 0.20 0.80 0.10
Average 8.36 13.78 2.27 1.98 1.24 0.32 0.94 0.54
Median 8.36 12.80 2.23 2.00 1.19 0.30 0.90 0.50
Standard deviation 0.19 2.93 0.28 0.49 0.33 0.16 0.13 0.21
Standard error 0.04 0.67 0.07 0.11 0.08 0.04 0.03 0.05
Min 8.08 10.60 1.80 1.20 0.70 0.10 0.70 0.10
Max 8.70 19.70 2.90 2.90 1.90 0.80 1.20 1.00

The average concentration of Cr analysed in the investigated water samples was found as 0.94 mg/L (range:
0.70 (W-10) to 1.20 (W-14) mg/L). The guideline value or allowable limit for Cr in drinking waters is determined
as 0.05 mg/L [30-31]. All Cr concentrations in the Tohma stream waters are significantly higher than the
guideline value. The average concentration of Mn analysed in the investigated water samples was found as 2.27
mg/L (range: 1.80 (W-18) to 2.90 (W-17) mg/L). The guideline value or allowable limit for Mn in drinking
water is determined 0.02 mg/L [30]. All Mn concentrations in the Tohma stream waters are significantly higher
than the guideline value. The average concentration of Fe analysed in the investigated water samples was found
as 13.78 mg/L (range: 13.78 (W-4) to 19.70 (W-17) mg/L). The guideline value or allowable limit for Fe in
drinking water is determined 0.05 mg/L [30]. All Fe concentrations in the Tohma stream waters are significantly
higher than the guideline value. The average concentration of Ni analysed in the investigated water samples was
found as 1.24 mg/L (range: 0.70 (W-16) to 1.90 (W-10) mg/L). The guideline value or allowable limit for Ni in
drinking water is determined 0.02 mg/L [30]. All Ni concentrations in the Tohma stream waters are significantly
higher than the guideline value. The average concentration of Cu analysed in the investigated water samples was
found as 1.98 mg/L (range: 1.20 (W-16) to 2.90 (W-3) mg/L). The guideline value or allowable limit for Cu in
drinking water is determined 2 mg/L [31]. The average concentration of Cu in the Tohma stream waters is lower
than the guideline value. The average concentration of Zn analysed in the investigated water samples was found

483
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

as 0.32 mg/L (range: 0.10 (W-9) to 0.80 (W-17) mg/L). The guideline value or allowable limit for Zn in drinking
water is determined 5 mg/L [32]. All Zn concentrations in the Tohma stream waters are significantly lower than
the guideline value. The average concentration of Pb analysed in the investigated water samples was found as
0.54 mg/L (range: 0.10 (W-19) to 1.00 (W-16) mg/L). The guideline value or allowable limit for Pb in drinking
water is determined 0.01 mg/L [31]. All Pb concentrations in the Tohma stream waters are significantly higher
than the guideline value.
The correlations among metals indicate some facts regarding the origin and migration of these metals [28].
Pearson’s correlation coefficient (PCC) is the most commonly used correlation coefficient. Table 3 presents The
PCC matrix estimated for toxic metals analysed in the Tohma water samples. In accordance with the values of
PCC, meaningful positive correlations (for p0.01) existed among the investigated toxic metals.

Table III. Pearson correlation matrix between the analysed toxic metals
Cr Mn Fe Ni Cu Zn Pb
Cr 1
Mn 0.04236 1
Fe -0.1721 0.83054 1
Ni -0.0359 0.70102 0.78937 1
Cu -0.0207 0.40328 0.21672 0.47644 1
Zn -0.1873 0.72826 0.80534 0.64421 0.07328 1
Pb 0.18045 0.09796 0.31977 0.20311 -0.2225 0.24266 1

It can be seen that Fe was significantly correlated with Ni (0.8) and Zn (0.8). Ni was significantly correlated
with Cu (0.5) and Zn (0.6). Mn was significantly correlated with Fe (0.8), Ni (0.7), Cu (0.4) and Zn (0.7). The
high correlation of Mn with Fe, Ni, Cu and Zn indicates these potentially toxic metals were originated from
same sources and further moving together.
The data related to the estimation of MPI and MEI is given in Table 4 [30-32]. The values of MPI and MEI
estimated for the Tohma water samples are presented in Table 5. The MPI values varied from 312 (W-19) to
9715 (W-16) with an average of 4713. All MPI values are greater than the critical value for drinking waters (30)
and the critical value for surface waters (600), except for the W-19 sample. The MPI results indicated that the
investigated water samples were seriously contaminated with respect to Fe, Mn, Cu, Ni, Zn, Cr and Pb. The MEI
values varied from 181 (W-19) to 317 (W-10) with an average of 226. All MEI values are below than the critical
value of 2.48 for drinking water. The MEI values estimated for the W-10, W-14 and W-17 sample are above the
critical limit of 300 for surface water thus these water samples were seriously polluted with respect to Fe, Mn,
Cu, Ni, Zn, Cr and Pb. The other water samples (84% of the total water samples) are in moderately polluted with
the toxic metals.

Table III. Data for the estimation of MPI and MEI

Toxic metal CMAX SV IV
Cr 0.050 0.050 0.100
Mn 0.050 0.400 0.050
Fe 0.300 0.200 0.300
Ni 0.020 0.020 0.100
Cu 2.000 1.500 1.000
Zn 5.000 - 5.000
Pb 0.010 0.010 0.005

484
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table IV. The estimated values of MPI and MEI

Sample code Metal pollution index (MPI) Metal evaluation index (MEI)
W-1 3428 204
W-2 2307 217
W-3 3275 225
W-4 4596 186
W-5 3526 185
W-6 4471 207
W-7 5410 232
W-8 3859 208
W-9 6565 214
W-10 8268 317
W-11 4570 220
W-12 5351 237
W-13 4635 221
W-14 6158 303
W-15 2407 190
W-16 9715 242
W-17 6203 303
W-18 4488 196
W-19 312 181
Average 4713 226
Median 4570 217
Standard deviation 2144 41
Standard error 492 9
Min 312 181
Max 9715 317

4. CONCLUSIONS

The alkalinity of the water samples showed that the Tohma stream in the studied area is not affected by the acid
rain generated by the flue gases released from the Kangal LFTPP. The concentrations of potentially toxic metals
were the highest for Fe followed by Mn > Cu > Ni > Cr > Pb in the Tohma surface water samples. Fe, Ni, Cr,
Pb and Mn concentrations are above the allowable limits recommended by the TS and WHO. The average Cu
concentration is within its limit while the Zn concentration is significantly lower than the allowable limit. Inter-
relationship between some metals analysed in the Tohma water samples indicated that the pollution of the Tohma
stream water was originated from the natural (geologic) and anthropogenic sources. The MPI and MEI values
index estimated for Tohma stream water samples were found above the critical index values. Consequently, the
results reveal that the Tohma stream in the studied area was significantly polluted with the potentially toxic
metals. While this is the case, it is warned that surface water from polluted locations should not be utilized for
irrigation, drinking, and aquatic life water purposes without proper treatment.

REFERENCES

[1] Gören, E., Turhan, Ş., Kurnaz, A., Garad, A.M.K., Duran, C., Uğur, F.A., Yeğingil, Z. 2017. Environmental evaluation of natural
radioactivity in soil near a lignite-burning power plant in Turkey. Applied Radiation and Isotopes 129, 13–18.

485
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[2] Singh, R.K., Gupta, N.C., Guha, B.K. 2014. Assessment of ground water contamination for heavy metals in the proximity of ash
ponds. Elixir Pollution 75, 28016–28019.
[3] Verma, C., Madan, S., Hussain, A. 2016. Heavy metal contamination of groundwater due to fly ash disposal of coal-fired thermal
power plant, Parichha, Jhansi, India. Cogent Engineering 3, 1–8.
[4] Ali, M.M., Ali, M.L., Islam, M.S., Rahman, M.Z. 2016. Preliminary assessment of heavy metals in water and sediment of Karnaphuli
River, Bangladesh. Environmental Nanotechnology, Monitoring and Management 5 27–35.
[5] Khan, K., Lu, Y., Khan, H., Zakir, S., Ihsanullah, Khan, S., Khan, A.A., Wei, L., Wang, T. 2013. Health risks associated with heavy
metals in the drinking water of Swat, northern Pakistan. Journal of Environmental Sciences 25, 2003–2013.
[6] Nazeer, S., Hashmi, M.Z., Malik, R.M. 2014. Heavy metals distribution, risk assessment and water quality characterization by water
quality index of the River Soan, Pakistan. Ecological Indicators 43, 262–270.
[7] Sonawane, S.M. 2015. Effects of waste water and ash of thermal power station on hydrological features of River Tapi at Bhusawal,
District Jalgaon Maharashtra. IOSR Journal of Environmental Science, Toxicology, and Food Technology 9, 19–24
[8] Noli, F., Tsamos, P. 2016. Concentration of heavy metals and trace elements in soils, waters and vegetables and assessment of health
risk in the vicinity of a lignite-fired power plant. Science of The Total Environment 563, 377–385.
[9] Balakrishnan, A., Ramu, A. 2016. Evaluation of heavy metal pollution index (HPI) of ground water in and around the coastal area of
Gulf of Mannar Biosphere and Palk Strait. Journal of Advanced Chemical Sciences 2, 331–333.
[10] Liu, Y., Yu, H., Sun, Y., Chen, J. 2017. Novel assessment method of heavy metal pollution in surface water: A case study of
Yangping River in Lingbao City, China. Environmental Engineering Research 22, 31–39.
[11] Saha, M., Rahman, M.S., Ahmed, M.B., Zhou, J.L,, Ngo, H.H., Guo, W. 2017. Industrial metal pollution in water and probabilistic
assessment of human health risk. Journal of Environmental Management 185, 70–78.
[12] Ahamed, K.R., Manikandan, S., Ananthi Malini, V.C. 2017. Impact of heavy metals on ground water pollution in some towns of
Vellore district, Tamil Nadu, India. International Journal of Science, Engineering and Technology Research 6 1210–1218.
[13] Chaturvedi, A., Bhattacharjee, S., Singh, A.K., Kumar, V. 2018. A new approach for indexing groundwater heavy metal pollution.
Ecological Indicators 87, 323–331.
[14] Demirak, A., Yilmaz, F., Tuna, A.L., Ozdemir, N. 2006. Heavy metals in water, sediment and tissues of Leuciscus cephalus from
a stream in southwestern Turkey. Chemosphere 63, 1451–1458.
[15] Türkmen, M., Çalışkan, E. 2011. Seasonal and Spatial Distributions of the Metals in the Water from the River Asi in Southern East
Mediterranean Area of Turkey. The Black Sea Journal of Sciences 1, 35-42.
[16] Varol, M., Şen, B. 2012. Assessment of nutrient and heavy metal contamination in surface water and sediments of the upper Tigris
River, Turkey. Catena 92, 1–10.
[17] Kalender, L., Uçar, S. 2013. Assessment of metal contamination in sediments in the tributaries of the Euphrates River, using pollution
indices and the determination of the pollution source, Turkey. Journal of Geochemical Exploration 134, 73–84
[18] Sari, E.S., Cukrov, N.C., Franciskovic-Bilinski, S., Kurt, M.A. Halli, M. 2016. Contamination assessment of ecotoxic metals in
recent sediments from the Ergene River, Turkey. Environmental Earth Sciences 75, 1051.
[19] Dede, Ö.T. 2016. Application of the Heavy Metal Pollution Index for Surface Waters: A Case Study for Çamlıdere. Hacettepe
Journal of Biology and Chemistry 44, 499–504.
[20] Omwene, P.I., Öncel, M.S., Çelen, M., Kobya, M. 2018. Heavy metal pollution and spatial distribution in surface sediments of
Mustafakemalpasa¸ a stream located in the world's largest borate basin (Turkey). Chemosphere 208, 782-792.
[21] Koç, Ç., Yılmaz, E. 2019. A Study About of Heavy Metal Pollution in the Büyük Menderes River in Turkey. European Journal of
Science and Technology 15, 571-578.
[22] Sunkar, M. 2006. Geomorphology of the Kangal basin (Sivas). Doctoral Thesis, Institute of Social Science, Fırat University, Elazığ
(in Turkish).
[23] Turhan, Ş., Gören, E., Garad, A.M.K., Altıkulaç, A., Kurnaz, A., Duran, C., Hançerlioğulları, A,. Altunal, V., Güçkan, V., Özdemir,
A. 2018. Radiometric measurement of lignite coal and its by-products and assessment of the usability of fly ash as raw materials in
Turkey. Radiochimica Acta 106, 611–621.
[24] Nalawade, P.M., Bholay, A.D., Mule, M.B. 2012. Assessment of groundwater and Ssrface water qualityiIndices for heavy metals
nearby area of Parli Thermal Power Plant. Universal Journal of Environmental Research & Technology 2, 47–51.
[25] Sirajudeen, J., Arulmanikandan, S., Manivel, V. 2014. Heavy metal pollution index of ground water of Fathima Nagar area near
Uyyakondan Channel Tiruchirappalli District, Tamilnadu, India. World Journal of Pharmacy and Pharmaceutical Sciences 4, 967–
975.
[26] Mohan, S.V., Nithila, P., Reddy, S.J. 1996. Estimation of heavy metal in drinking water and development of heavy metal pollution
index. Journal of Environmental Science and Health A 31, 283–289.
[27] Tamasi, G., Cini, R. 2004. Heavy metals in drinking waters from Mount Amiata (Tuscany, Italy) Possible risks from arsenic for
public health in the province of Siena. Science of The Total Environment 327, 41–51.
[28] Khoshnam, Z., Sarikhani, R., Dehnavi, A.G., Ahmadnejad, Z. 2017. Evaluation of water quality using heavy metal index and
multivariate statistical analysis in Lorestan Province, Iran. Journal of Advances in Environmental Health Research 5, 29-37.
[29] Biswas, P.K., Uddin, N., Alam, S., Sakib, T., Sultana, S., Ahmed. T. 2017. Evaluation of heavy metal pollution indices in irrigation
and drinking water systems of Barapukuria coal mine area, Bangladesh. American Journal of Water Resources 5, 146-151.
[30] TS266 2005. Turkish Standard prepared by Turkish Standard Institution (TSE). Water intended for human consumption. Ankara (in
Turkish).
[31] WHO 2011. Guidelines for Drinking-water Quality, 4th edition, WHO Library Cataloguing-in-Publication Data NLM classification:
WA 675, Geneva.
[32] Güler, C. 2007. Evaluation of maximum contaminant levels in Turkish bottled drinking waters utilizing parameters reported on
manufacturer’s labeling and government-issued production licenses. Journal of Food Composition and Analysis 20, 262–272.

486
 METAMODELING THE INFLUENCE OF FORM AND SHADING
ON THE HEATING AND COOLING LOADS OF A RESIDENTIAL
ZONE IN THE MEDITERRANEAN

Aristotelis Vartholomaios
School of Architecture, Aristotle University of Thessaloniki, Thessaloniki, Greece, [email protected],
ORCID: 0000-0001-6177-1130

Angeliki Chatzidimitriou
School of Architecture, Aristotle University of Thessaloniki, Thessaloniki, Greece, [email protected],
ORCID: 0000-0001-7966-4034

Konstantinos Ioannidis
School of Architecture, Aristotle University of Thessaloniki, Thessaloniki, Greece, [email protected],
ORCID: 0000-0001-8381-7013

Vartholomaios, A, Chatzidimitriou, A, Ioannidis, K, Metamodeling the influence of form and

Cite this paper as: shading on the heating and cooling loads of a residential zone in the Mediterranean. 8th Eur.
Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: The study presents a method of generating simple building energy metamodels for single zone
dwellings using machine learning techniques. The novelty of this research is the parametric
generation of a training dataset through Monte Carlo energy simulations of a single thermal zone.
The study examines the influence of the geometrical aspects that affect the heating and cooling
loads of residential zones such as the compactness and shape of the zone and its relation to
neighboring building masses or other shading volumes. This is achieved by examining the
potential of geometric form indicators to interpret various aspects of the thermal zone energy
balance, while keeping other variables, such as construction and activity, constant. The study
focuses on the Mediterranean climate of Thessaloniki, Greece and streamlines the use of available
simulation and data analysis tools such as Ladybug Tools, EnergyPlus and Scikit-learn under a
unitary workflow. The trained Random Forest models demonstrate a good fit with a less than 5%
Mean Absolute Percentage Error for predicting heating and cooling loads when using the best
combination of geometric indicators. These positive results, although preliminary, are
encouraging for further investigation and model validation against complex real-world urban
locations.
Keywords: Simulation, Heating Loads, Cooling Loads, Residential, Machine Learning
© 2020 Published by ECRES

1. INTRODUCTION

Urban climate resilience and energy sustainability is a common theme across several UN Sustainable
Development Goals [1]. A major hurdle that has yet to be overcome is the need to inform the early, but crucial,
urban design decisions with quick and accurate environmental feedback. Running detailed building energy
simulations at the urban scale requires a significant depth and breadth of user expertise, as well as significant
time to fine-tune and simulate complex models. Researchers have proposed different solutions to this challenge,
from developing simplified urban energy simulation software [2], to estimating building energy consumption
through statistical modelling [3], spreadsheets [4] or more recently through Machine Learning (ML) [5–7].
This paper presents the preliminary results of a research project that seeks to develop ML metamodels for
estimating the influence of urban form on heating and cooling loads of dwellings in the Mediterranean.
Metamodeling is the creation of simpler generalizable models from more complex and well-validated models.

487
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In this case, ML is an ideal tool to model the complexities of a dynamic building energy simulation engine, such
as EnergyPlus [8]. Such metamodels could provide near real-time feedback to the early-stage urban design
process with minimal inputs from the designer. Since urban designers are preoccupied with the potential form
of a neighborhood the presented study seeks to discover a combination of morphological indicators that can be
used to accurately predict the heating and cooling loads of a single thermal zone, while keeping the rest of the
simulation parameters (construction and occupancy) constant. If such combination exists for a single zone, then
it is likely that it can be used for more complex models that are built from several zones, such as an urban block
or a neighborhood.
Previous studies have already studied the relation between morphological indicators and energy consumption at
the building, urban block and neighborhood scales [9–14]. Commonly cited indicators such as the Surface-to-
Volume (S/V) ratio focus on the overall compactness of form while others, such as Solar Admittance (SA) and
the Sky View Factor (SVF) deal with the radiant heat fluxes between the buildings and the sky vault. ML has
also been employed by a number of studies in order to estimate or map urban energy, using a combination of
morphological, demographical, construction and land use indicators [5,7]. The present research, however,
focuses exclusively on the influence of geometrical parameters on building heat balance which are influenced
by early design decisions.

2. SIMULATION METHODOLOGY

Simulation parameters
The study is carried out for the Mediterranean climate (Köppen CSa class) of Thessaloniki, Greece. Thessaloniki
is generally characterized by dry-hot summers and mild-wet winters and belongs to the ‘C’ climatic zone of the
Greek Building Energy Code (KEnAK), were heating needs are greater than cooling needs. According to
KEnAK [15], the heating period for zone C is between 15/10 - 30/04, while the cooling period is between 01/06
– 31/08. The climate of Thessaloniki offers the potential for both passive solar heating in winter and passive
cooling during summer through natural ventilation, due to the nocturnal drop in air temperature.
The dataset used in the ML process is based on a single rectangular residential thermal zone (Fig. 1) that is
generated parametrically using Monte Carlo to randomize the parametric combinations. Each randomly
generated thermal zone is passed to the EnergyPlus engine for simulation while a set of morphometric indicators
are calculated. These indicators, along with the simulated heating and cooling loads, are then written to an output
file and the process is repeated for the next randomly generated zone. Monte Carlo ensures that the examined
parametric space will be evenly sampled regardless of the possible number of potential combinations. The
present study focuses only on parameters of geometric nature that are mostly influenced by early but crucial
design decisions at the urban block scale. Construction, occupancy and internal loads are held constant and in
accordance to the revised 2017 KEnAK specifications for the ‘reference building’ (Table 2). To further simplify
the analysis the height of the zone is also set at 3.0m.

The randomized parameters (Fig. 1) are:

 Zone length (3m – 30m): Corresponds to the front façade of the rectangular residential zone.
 Zone width (3m – 15m): Geometries that exceed this width should be split into two zones.
 Zone rotation (0° – 360°): The zone is randomly rotated to get different sun exposure.
 Floor contact with ground (true – false): Determines whether the zone is in contact with the ground.
 Window-to-Wall-Ratio (WWR) (0.2 – 0.8): Larger openings increase solar gains as well as conductive and convective
gains or losses depending on the season. They also increase natural ventilation potential and infiltration.
 Percentage (%) of exposed back/left/right walls: With the exception of the front façade which is always fully exposed,
the exposure of the other three walls to the outside air varies by a random percentage. It is assumed that the non-exposed
portions of the walls are in contact with another thermal zone of the exact same construction and function and are treated
as adiabatic surfaces.
 Percentage (%) of roof exposed: Roof exposure is treated in same manner as the walls exposure.
 Wall/Roof shader transmittance (0.000 – 1.000): The generated zone is enclosed in a slightly larger envelope (offset by
0.05 m) of variable transparency to simulate the reduction of zone surface irradiation by obstacles. Different random
transmittances are assigned to the ‘roof’ and the ‘walls’ of the shader envelope in order to generate a more dynamic
shading regime. This simplified representation of shading also allows for an easy calculation of SVF for walls and roofs.
Roof SVF is equal to roof shader transmittance. Wall SVF is equal to half the wall shader transmittance since an
unobstructed vertical wall can only “see” half of the sky vault.

488
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. Visualization of the single-zone model explaining the examined parameters and four examples of randomly
generated zones. Adiabatic surfaces are shown in red.

Table 1. Simulation parameters for the single thermal zone.

Construction: Occupancy and operation:
 Concrete structure (20%) /19mm Perforated bricks  5 people/100m2, operation for 18 h/day
(80%). Total zone heat capacity: 280 kJ/(m²K)  Min. fresh air: 15m3/h/person
(post-calculated).  Equipment / Lighting load per area: 2W/m² / 2.5W/m²
 External insulation: expanded polystyrene (variable  Heating/cooling set point: 20°C / 26°C
widths to meet maximum allowed U values):  Natural ventilation:
 Wall: 0.40 W/(m2K)  min/max outdoor temp.: 22°C / 26°C,
 Roof: 0.35 W/(m2K)  min indoor temp.: 24°C
 Ground floor: 0.65 W/(m2K)  operable glazing fraction: 80%
 Glazed opening (20% frame): 2.4 W/(m²K)  HVAC: ideal air loads
 External finishing: plaster (albedo: 0.35).
 Glazing g-value (20% frame): 0.55 Simulation parameters:
 Infiltration: 5.5m³/h per m² of operable glazing  6 timesteps / h
 Solar calculations: full exterior, update shading calcs.
every 15 days.

Morphological indicators
The calculated morphological indicators for each generated zone are separated into three broad families:
 Basic indicators: (i) zone area (m²), (ii) envelope area (m²), (iii) wall area (m²), (iv) exposed wall area (m²) and (v) Zone
volume (m³).
 Compactness indicators: (i) total S/V ratio (m-1), (ii) exposed wall S/V ratio (m-1) and (iii) exposed roof S/V ratio (m-
1
).
 Envelope indicators: (i) average WWR, (ii) average wall SVF, (iii) roof SVF, (iv) floor contacts ground (Boolean
True/False), (v) winter SA, (vi) summer SA.

The family of basic indicators is likely to have little interpretive power with respect to the zone energy behavior
but it is useful for experiment control and for calculating more advanced indicators. The family of compactness
indicators attempt to describe the relation between thermal zone shape and heat gains/losses through the zone
envelope. The simplest indicator is the S/V which can then further be refined into an exposed S/V ratio for the
walls and the roof. Walls and roofs have different construction details and play a different role in the zone energy
balance, which is further differentiated by the exposure of these surfaces to outside air. Finally, the family of
envelope indicators describes the interaction of the zone with the solar radiation fluxes and the ground. Average
WWR determines the percentage of vertical surfaces that are glazed openings. Average wall SVF and roof SVF
express the exposure of these surfaces to the radiant heat fluxes between the zone and the sky vault. However,
SVF does not contain any information on the actual surface irradiation in relation to its orientation and the
seasonal apparent sun trajectory. For this reason, winter/summer SA is used. SA is an advanced solar irradiation
metric that is presented in more detail in the dedicated subsection below.

Simulation results
A total of 10,000 randomly generated thermal zones were simulated using the validated EnergyPlus 8.6 building
energy simulation engine [8]. Ladybug Tools [16] and Grasshopper for Rhino were used for automating the

489
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

process. Climatic data for Thessaloniki were provided by Meteonorm 7.1 [17] for the 1991-2010 period. The
simulations were run at six timesteps per hour, updating shading calculations every 15 days. The required SVFs
for each surface were calculated at a resolution of 0.5m. SAs were calculated using custom Python coding. The
simulated heating and cooling loads ranged from 8.5 to 119kWh/m² per year and from 1.5 to 39.5kWh/m² per
year respectively. The results for each examined parameter are summarized in the histograms below (Fig. 2).
The floor contact parameter is evenly distributed between the two potential options (true-false) and is not shown
below.

Figure 2. Histograms of the parametric simulation results.

Solar Admittance
SA was first proposed by Luc Adolphe [11] as a dimensionless measure that expresses the relative potential of
the vertical surfaces of a given urban fabric to capture solar radiation. SA was calculated as the weighted average
of exterior wall areas with surface contiguity, orientation, shading and reflectance being the individual weights
for each wall surface. While this calculation method was appropriate for assessing the SA of an entire
neighborhood or city, the present study adapts it so it can work for a single residential thermal zone. In this
context, SA is calculated only for exposed surfaces, so the contiguity weight is not required. Instead, WWR can
be used to express the relative amount of surface irradiation that is converted to passive solar gains. Orientation
is assessed in the same way as Adolphe proposed, by using the ‘South Area Equivalent Factor’ (SAEF) of a
building. For any given vertical and unobstructed surface, SAEF is the ratio of its cumulative solar irradiation
to the cumulative solar irradiation of the same surface when oriented to the South. This can be calculated for
any time period, either annual or seasonal. For a southern wall this is equal to 1.0 and for a northern wall it is
approximately 0.2. This roughly means that a northern wall receives only 20% of the annual solar radiation
received by the southern wall. SAEF here is calculated separately for the heating and cooling periods for the
climate of Thessaloniki, since walls facing at the same direction will receive different amounts of solar
irradiation during winter and summer. Consequently, two SAs must be separately calculated, one for the heating
period (SAwinter) and one for the cooling period (SAsummer). The SA used in this study is calculated using Eq. [1]:

𝑆𝐴 = ∑ (𝐴𝑖 𝑆𝑉𝐹𝑖 𝑊𝑊𝑅𝑖 𝑆𝐴𝐸𝐹𝑖 ) / ∑ 𝐴𝑖 (1)

exposed 𝑒𝑥𝑝𝑜𝑠𝑒𝑑
𝑤𝑎𝑙𝑙𝑠 𝑖 𝑤𝑎𝑙𝑙𝑠 𝑖

Where Ai is the area of each exposed wall surface (i) of the thermal zone.

3. MACHINE LEARNING METHODS AND RESULTS

Since this study is preliminary the ML methodology presented here is simplified and its goal is to test the
potential of the proposed indicators as predictors of heating and cooling loads for the single zone model
described above. There are several available ML algorithms, ranging from the ‘traditional’ multi-linear
regression to more sophisticated techniques like deep neural networks [18]. The study uses Random Forests
(RF), an easily interpretable and fairy robust to overfitting, collinearity and outliers algorithm [19]. The Scikit-
learn python library is used to train the RF algorithm using the simulation results to eventually generate two
metamodels, one for the cooling and one for the heating loads. The fine-tuning of RF hyperparameters and the

490
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

testing of metamodel fitness is done with the k-folds cross-validation technique, using 5 folds. ‘Mean Squared
Error’ (MSE) is used as a loss function to train the algorithms. The study compares two different scenarios:
 Scenario 1 uses more generic indicators: (i) total S/V ratio, (ii) average WWR, (iii) floor contacts ground, (iv) average
wall SVF and (v) roof SVF.
 Scenario 2 uses more advanced indicators: (i) exposed wall S/V ratio, (ii) exposed roof S/V ratio, (iii) average WWR,
(iv) floor contacts ground, (v) winter SA (only for heating loads metamodel), (vi) summer SA (only for cooling loads
metamodel).
RF is trained separately for heating and cooling loads by setting the tree population to a constant of 300 and then
conducting a grid-search optimization for the rest of hyperparameters: (i) tree depth (3-15) and (ii) min. samples
per leaf (3-10). For both heating and cooling metamodels and in both scenarios the best results are found when
using a tree depth of 15 and 3 min. samples per leaf respectively. To keep results comparable, the Mean Absolute
Percentage Error (MAPE) metric is used to report model accuracy. Scenario 2 outperforms Scenario 1 in
predicting both heating and cooling loads with smaller errors (Table 1). Partial dependency plots verify the role
of each indicator in the metamodels’ overall behavior (Fig. 2). It can be seen, for example, that an increase in
SA generally increases cooling loads and decreases heating loads (Fig. 2), which is logical and expected. The
heating loads metamodel is also very sensitive to the exposed wall S/V ratio, while exposed roof S/V ratio and
ground contact seem to play a minor role in cooling load estimation.

Table 2. Performance of the trained RF models for the two examined scenarios
Scenario 1 metamodel performance Scenario 2 metamodel performance
Heating loads MAPE: 17.2% (±0.7%) Heating loads MAPE: 3.8% (±0.3%)
Cooing loads MAPE: 19.7% (±0.9%) Cooing loads MAPE: 4.8% (±0.9%)

Figure 3. Partial dependency plots for the examined morphological indicators (Scenario 2).

4. CONCLUSIONS

The study presented the preliminary results of a research project that aims to develop ML metamodels for
estimating the influence of urban form on heating and cooling loads of dwellings in the Mediterranean. Here, a
number of morphological indicators were tested on a single-zone residential energy model proving that: (i) There
is a combination of morphological indicators (seasonal SA, WWR, exposed wall/roof S/V ratio and contact with
ground), that can be used to estimate the heating and cooling loads for any simplified rectangular zone geometry
for the examined climate with reasonably good accuracy (MAPE < 5%), assuming that climatic, construction
and occupancy characteristics remain constant. (ii) Seasonal SA is a good predictor of the influence of shading
on building loads, yet it requires the calculation of SVF which can be a performance bottleneck. This ultimately
depends on the method and the spatial resolution of calculation. (iii) Random Forests, one of the simpler and
easiest to interpret ML algorithms is appropriate for this task. (iv) The datasets for training and testing the ML
algorithms can be generated through Monte Carlo parametric building energy simulations. However, the models,
in their current form are both limited in their practical applicability and require extensive testing in both real-
world and hypothetical urban locations. The presented methodology also includes a number of assumptions on
the actual energy behavior of the examined thermal zone with respect to boundary conditions, shading and zone
geometry. These assumptions may prove to be sources of additional systematic errors when the models are
applied to complex urban geometries consisting of several zones. Nevertheless, the results are promising and
demonstrate the potential of ML applications in rapid urban energy estimation during early design stages by
using a combination of morphological indicators as predicting features.

491
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

ACKNOWLEDGMENT
This research is co-financed by Greece and the European Union (European Social Fund-ESF) through the
Operational Programme “Human Resources Development, Education and Lifelong Learning 2014-2020” in the
context of the project “Estimation of heating and cooling loads of residential urban blocks by using urban
morphological indicators” (MIS 5047833).
The authors would also like to thank Kleo Axarli, Prof. Emeritus of AUTh, for providing valuable feedback
during the project.

REFERENCES

[1] UN, Sustainable Development Goals Knowledge Platform, (2020). sustainabledevelopment.un.org/ (ημερομηνία πρόσβασης
21 Μάιος 2020).
[2] D. Perez, J. Kämpf, U. Wilke, CITYSIM simulation: the case study of Alt-Wiedikon, a neighbourhood of Zürich City, Proc.
CISBAT 2011 - CleanTech Sustain. Build. (2011) 937–940. http://infoscience.epfl.ch/record/174435.
[3] M. Braulio-Gonzalo, M.D.M.D. Bovea, M.J. Ru??, P. Juan, M.J. Ruá, P. Juan, A methodology for predicting the energy
performance and indoor thermal comfort of residential stocks on the neighbourhood and city scales. A case study in Spain,
J. Clean. Prod. 139 (2016) 646–665. doi:10.1016/j.jclepro.2016.08.059.
[4] N. Baker, K. Steemers, LT Method 3.0 - A strategic energy-design tool for Southern Europe, Energy Build. 23 (1996) 251–
256. doi:10.1016/0378-7788(95)00950-7.
[5] J. Ma, J.C.P. Cheng, Estimation of the building energy use intensity in the urban scale by integrating GIS and big data
technology, Appl. Energy. 183 (2016) 182–192. doi:10.1016/j.apenergy.2016.08.079.
[6] V.K. Ojha, D. Griego, S. Kuliga, M. Bielik, P. Buš, C. Schaeben, L. Treyer, M. Standfest, S. Schneider, R. König, D.
Donath, G. Schmitt, Machine learning approaches to understand the influence of urban environments on human’s
physiological response, Inf. Sci. (Ny). 474 (2019) 154–169. doi:10.1016/j.ins.2018.09.061.
[7] A. Nutkiewicz, Z. Yang, R.K. Jain, Data-driven Urban Energy Simulation (DUE-S): A framework for integrating
engineering simulation and machine learning methods in a multi-scale urban energy modeling workflow, Appl. Energy. 225
(2018) 1176–1189. doi:10.1016/j.apenergy.2018.05.023.
[8] D.B. Crawley, F.C. Winkelmann, L.K. Lawrie, C.O. Pedersen, Energyplus: new capabilities in a whole-building εnergy
simulation program, στο: Proc. Seventh Int. IBPSA Conf., International Building Performance Simulation Association, Rio
de Janeiro, 2001: σσ 51–58. http://www.ibpsa.org/proceedings/BS2001/BS01_0051_58.pdf.
[9] A. Vartholomaios, A parametric sensitivity analysis of the influence of urban form on domestic energy consumption for
heating and cooling in a Mediterranean city, Sustain. Cities Soc. 28 (2017) 135–145. doi:10.1016/j.scs.2016.09.006.
[10] J. Natanian, O. Aleksandrowicz, T. Auer, A parametric approach to optimizing urban form, energy balance and
environmental quality: The case of Mediterranean districts, Appl. Energy. 254 (2019) 113637.
doi:10.1016/j.apenergy.2019.113637.
[11] L. Adolphe, A simplified model of urban morphology: application to an analysis of the environmental performance of
cities, Environ. Plan. B Plan. Des. 28 (2001) 183–200. doi:10.1068/b2631.
[12] B. D’Amico, F. Pomponi, A compactness measure of sustainable building forms, R. Soc. Open Sci. 6 (2019).
doi:10.1098/rsos.181265.
[13] A. Tsanas, A. Xifara, Accurate quantitative estimation of energy performance of residential buildings using statistical
machine learning tools, Energy Build. 49 (2012) 560–567. doi:10.1016/j.enbuild.2012.03.003.
[14] E. Nault, P. Moonen, E. Rey, M. Andersen, Predictive models for assessing the passive solar and daylight potential of
neighborhood designs: A comparative proof-of-concept study, Build. Environ. 116 (2017) 1–16.
doi:10.1016/j.buildenv.2017.01.018.
[15] TEE, Climatic Data of Greek Areas, Athens, 2012.
[16] M.S. Roudsari, M. Pak, A. Smith, Ladybug: a Parametric Environmental Plugin for Grasshopper To Help Designers Create
an Environmentally-Conscious Design, στο: 13th Conf. Int. Build. Perform. Simul. Assoc., Lyon, 2013: σσ 3129–3135.
http://www.ibpsa.org/proceedings/BS2013/p_2499.pdf.
[17] METEOTEST, Meteonorm, (2018).
[18] F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V.
Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, E. Duchesnay, Scikit-learn: Machine Learning
in Python, J. Mach. Learn. Res. 12 (2011) 2825–2830.
[19] L. Breiman, Random forests, Mach. Learn. 45 (2001) 5–32.

492
 A FEASIBILITY ASSESSMENT OF OFFSHORE WIND POWER
POTENTIAL BASED ON ENVIRONMENTAL, SOCIAL AND
ECONOMIC IMPACTS: A CASE STUDY ON THE
KIRKLARELI COASTLINE IN TURKEY

Fatih KARİPOĞLU
Izmir Institute of Technology, Department of Energy Systems Engineering, Izmir Institute of Technology, Izmir,
Turkey, 35430, [email protected], ORCID: 0000-0002-3314-134X

Mustafa Serdar GENÇ

Wind and Aerodynamic Research Center, Department of Energy Systems Engineering, Erciyes University, 38039,
Kayseri, Turkey, [email protected], ORCID: 0000-0002-6540-620X

Kemal KOCA
Wind and Aerodynamic Research Center, Department of Energy Systems Engineering, Erciyes University, 38039,
Kayseri, Turkey, [email protected],ORCID: 0000-0003-2464-6466

Karipoglu, F, Genc, M,S, Koca, K.. A feasibility assessment of offshore wind power potential
Cite this paper as: based on environmental, social and economic impacts: A case study on the Kırklareli Coastline
in Turkey. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Energy demand and consumption are increasing due to increased population and
developments in technological devices day by day. Offshore wind energy is seen as
the most effective renewable energy source to provide increased energy demand.
Notwithstanding high wind potential, Turkey does not have any electricity
production from offshore wind farm. This assessment study presents suitability of
the Kırklareli coastal region for offshore wind power installation based on a
comprehensive technical, environmental, social, and economic analysis. For this
selected region, environmental data were obtained from Emodnet, wind potential
assessment data was obtained from Global Wind Atlas. The results showed that the
Kırklareli region had high wind potential and low energy costs in terms of offshore
wind farm installation. Hence, the results indicated that investment of in this region
for installation seemed to be feasible and reasonable.

Keywords: Offshore wind farm; Emodnet; Geographic Information System (GIS); Feasible site selection;
Kırklareli
© 2020 Published by ECRES
Nomenclature
TSMS Turkish State Meteorological Service
GWA Global Wind Atlas
Emodnet European Marine Observation and Data Network
CAPEX Capital expenditures
OPEX Operation expenditures
LCOE Levelized cost of energy
GIS Geographic Information System
Da Array density of turbines
Ca Availability coefficient of wind turbines
Lw Wake loss rate
Nwt Total number of turbines

493
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION

Energy has become the most important need for societies due to the increase in technological devices and
population. Therefore, energy consumption increases on 4% - 5% each year in the world [1] and this rate might
be described as a high rate. Renewable energies such as wind energy play a vital role in electricity production
because of the reduction of the reserves of fossil fuels resources and countries’ greenhouses gas emissions
reduction targets. Particularly, wind energy has grown faster than other renewable energy sources due to the
recently expanding wind energy sector. In this expanding and growing wind energy sector, offshore wind energy
has been also growing rapidly for 10 years. Although offshore wind power plants have high total cost and are
difficult to finance, it has more benefit in the long term due to its higher energy production capacity than onshore
wind energy plants [2]. Due to the using restrictions of land areas, offshore wind energy may soon become
indispensable in the growing wind energy market [3]. Annual cumulative offshore wind power capacity
increased almost triple fold from 2011 to 2016 and reached the 14.384 GW [4].

Table 1 showed overview of grid-connected offshore wind power projects at the end of 2019 [5]. As seen in
table, UK played dominant role in country rank in conjunction with cumulative capacity of 9945 MW and net
capacity connected 1760 MW. Moreover, the rank of Denmark could be surprisingly evaluated because this
country was precursor in advance. But it was clearly seen that Denmark lag behind to Germany in terms of
cumulative capacity and net capacity connected.
Table I. Overview of grid-connected offshore wind power projects at the end of 2019 [5].
NO. OF WIND CUMULATIVE NO. OF WIND NET NO. OF WIND
COUNTRY FARMS CAPACITY FARMS CAPACITY FARMS
CONNECTED1 (MW) CONNECTED CONNECTED CONNECTED
IN 2019 (MW) IN 2019
UK 40 9,945 2,225 1,760 252
Germany 28 7,445 1,469 1,111 160
Denmark 14 1,703 559 374 45
Belgium 8 1,556 318 370 44
Netherlands 6 1,118 365 0 0
Sweden 5 192 80 0 0
Finland 3 70.7 19 0 0
Ireland 1 25.2 7 0 0
Spain 2 5 2 0 0
Portugal 1 8.4 1 8 1
Norway 1 2.3 1 0 0
France 1 2 1 0 0
Total 110 22,072 5,047 3,623 502

Turkey can easily generate electricity from renewables due to its geographical position. According to report
identified in Ref. [6], Turkey’s average annual sunshine duration was 2640 hours and the average annual solar
radiation was1,311 kWh/m2*year. Based on study in Ref. [7], Turkey has the potential of onshore and offshore
48000 MW and 17393.2 MW, respectively. According to specified values, it was seen that Turkey was in good
condition in terms of solar and wind energy.

Electricity demand of Turkey continuously increases even though Turkey had geographical advantages. Turkey
was in top 6 countries in Europe and 20 countries in the world in terms of total installed capacity [8]. Turkey’s
total installed capacity and electricity production were 73 GW and 261.783 GW in 2016, respectively [9]. The
most popular way to generate electricity in Turkey was natural gas and Turkey imported huge amount of natural
gas (nearly 1.7 billion cubic meters) in 2014 [10]. This caused Turkey to be dependent the foreign countries
such as Russia and Azerbaijan. Most amounts of Turkey's energy demand were supplied from coal, natural gas,
and hydropower. 6% of energy demand was obtained from geothermal, wind, solar and other resources [11]. On
wind energy development in Turkey, established in 1988, it began with 7.2 MW wind power plant [12]. The
information received from these wind farms were evaluated and helped to create a Turkey’s potential atlas which
was created at 50 meters high in certain regions.

494
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Turkey did not have any offshore wind farm although it surrounded with three seas. Inherently, there was not
any law and regulation for offshore wind farm installation. Different researchers have studied the offshore wind
energy assessment so far. But nobody investigated Kırklareli coastal. The objective of the present study was to
analyze the multi principles of coastal regions of Kırklareli for offshore wind energy assessment in detail.
Investigated topics consisted of environmental, social, and wind potential issues in which are documented as an
obstacle in terms of installation. In addition to these principles, the current legislation with regards to wind energy
was discussed.

2. Assessment of Wind Potential and Environmental, Social Impacts for Selected Region
To assessment of desired offshore wind power plan for Kırklareli coastline, multi criteria such as wind potential,
environmental, social impacts and study arrangements were performed as follows:

2.1. Selected Region

The selected region located at Kırklareli coastline in the northwest of Turkey. For the depth waters, the perimeter
length of the region seen in Fig. 1 was average 100 km. Within the scope of the literature studies, offshore wind
potential studies were found in many regions such as Gökçeada, Bozcaada, etc. However, no previous study was
performed for this region as mentioned before.

Figure 1. Selected region in Kırklareli coastal

2.2. WIND POTENTIAL

The wind potential in the selected region should be sufficient. The highest, the lowest and average wind speeds
of the region should be provided from the authorized institutions. In Turkey, the Turkish State Meteorological
Service (TSMS) provided wind data information only for onshore locations. In this study, the Global Wind Atlas
(GWA) [13] was used for the necessary data. This atlas ensured wind power density and wind speed data. The
average wind speed of the selected region was 8 m/s as shown in Fig. 2. According to these data, it was
determined that the wind speed was quite high, and the wind potential of the selected region was between 45-50
MW.

495
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. The average wind speed from GWA

2.3. ENVIRONMENTAL AND SOCIAL IMPACTS

One of the most constraints of offshore wind farms was environmental impacts. Because environmental
assessment, such as technical evaluations, was also very important for the establishment of offshore wind power
plant. There were many variables that affect the determination of eligibility to determine appropriate marine
areas. The area for desired offshore wind energy installation should be suitable for military regions, territorial
waters, civil and military aviation, marine ship density, social and environmental concerns, underground
pipelines and communication cables [14]. Emodnet (European Marine Observation and Data Network) [15] was
employed for data acquisition to unlock fragmented, hidden marine data resources and to facilitate investment
in sustainable coastal and offshore activities. According to obtained data from Emodnet, offshore power plant
installation was not a negative impact. Firstly, Turkey was in the territorial waters of the selected region. The
selected region was rectangular, and the corner lengths were 10 and 40 km, respectively. The selected region
had not tourism activities and accommodation. Therefore, there would be no operation and maintenance
problems. There was not military unit and airport in the selected region.

Additionally, there were several sea routes for trade and transportation in the region. However, these roads were
rarely used. According to the result obtained from Emodnet, the sea routes and their density were illustrated in
Fig. 3. Sea vehicles were density on İstanbul sides. Hence, it did not occur any negative impact on the offshore
wind power plant to be installed in the region. Furthermore, there was not private sheltered seafood production
in the region in terms of an environmental and social impact.

Figure 3. Sea routes and ship density from Emodnet.

2.4. BATHYMETRY AND WIND TURBINE CONFIGURATION

Distance of offshore wind farm areas to land and the selected foundation type were the two most important
factors affecting the total cost. As the water depth increased, foundation costs increased, and economically lower
efficiency was obtained. For offshore wind turbines, marine depth maps were very important. In this paper, the
selected region’s bathymetry map was shown in Fig. 4. According to the scale of the bathymetry map, the sea

496
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

depth of the selected region was between 20-50 m. This sea depth was perfectly suitable for offshore investment.
In this article, SG 3.4-134 [16] offshore wind turbines were preferred.

Figure 4. Bathymetry map for the selected region from Emodnet.

It was desired to minimize the effect of tip vortex occurring aerobically in turbine placement. In the turbines
configuration process, the turbines placement rules were applied for tower spacing 4-7 times rotor diameter
(D) spacing between towers within a row and 5D-9D times between row according to the recommendation by
Dvorak et al [17]. Due to these effects and wind potential, 15 SG 3.4-132 wind turbines could be installed in the
selected region. After installation, power of the facility to be established would be 51 MW. The general view of
the turbines installed in the region was shown in Fig. 5.

Figure 5. Location of turbines in the selected region from adopted Google Earth.

3. Economic Analysis for Offshore Wind Power Plant

The wind resource was free, whilst the wind power systems equipment was very expensive [18]. The cost model
consisted of capital expenditures (CAPEX) and operation expenditures (OPEX). The CAPEX of an offshore
wind farm was investigated in four different parts [19] as follows: (i) the wind turbines cost, (ii) the electrical
system cost, (iii) the support system and (iv) installation cost. Additionally, project development and insurance
costs were also added to total cost [8]. The OPEX of an offshore wind farm meant the operational and
maintenance costs of the energy power plants. Considering the study conducted by Kim et al. [20], the OPEX
was estimated to be 1.9% of the annual CAPEX over the 20-years lifetime of the offshore wind energy project.
For the determination of offshore wind farms with 15 wind turbines were analyzed the economic value of OPEX
and CAPEX. After these values were determined, the levelized cost of energy (LCOE) should be obtained. The
LCOE was very important for offshore wind power plants and it was the major investment indicators that helped
the power plant project developers make their investment decisions. This indicator also evaluated integrally
before taking the final investment decision [21]. Apart from CAPEX and OPEX costs, wind potential affected

497
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the economic value [22]. This study was determined Siemens Gamesa SG 3.4-132 as wind turbines types
because of the wind speed potential of the study area. According to Table 4, power capacity of turbines was
3.465 MW level.

Eq. [1] was used for determination the annual wind energy potential of the study area [14]:

𝐷𝑎 ∗(1−𝐿𝑤 )∗𝐶𝑎 ∗𝑁𝑤𝑡 ∗𝑃𝑔𝑒𝑛 ∗8760

𝐸𝑝 = (1)
𝑃𝑤𝑡

where Da was the array density of turbines, Ca was the availability co-efficient of wind turbines, Lw was the
wake loss rate, Nwt was total number of turbines and Pwt was wind turbine capacity. According to Siemens
Gamesa SG 3.4-132 properties [16], Ca and Lw were determined as 0.95 and 0.1, respectively. Furthermore, Da
was determined as 0.3 MW/km2.

Regions with high potential in wind energy investment were more economically effective regions than other low
potential regions. After the wind potential assessment, the CAPEX was analyzed with Eq. [2] suggested by Hong
and Muller [14]:

𝐶𝐴𝑃𝐸𝑋 = 𝐶𝑤𝑡 + 𝐶𝑠 + 𝐶𝑒 + 𝐶𝑖𝑛𝑠 + 𝐶𝑜𝑡ℎ𝑒𝑟 (2)

where Cwt was wind turbine cost, Cs was support cost, Ce was electrical systems cost, Cins was wind turbine
installation cost, and Cother was other costs.

Other costs were considered as 10% of CAPEX in an economic analysis study. After expressed to CAPEX, the
distance to the nearest harbor was the parameter which affected operation and maintenance costs [23]. The
OPEX function was based on in Ref. [12] and was obtained information on the least coast distance to the harbor
from GIS. LCOE was expressed as the unit cost of the energy produced. Generation cost included all cost
parameters like CAPEX and OPEX. LCOE value was very important for offshore wind farm investment. To
obtain more income, this value was desired to be low. LCOE function was defined in Eq. [3] in E/MWh
depended on CAPEX, OPEX:

𝐶 𝐶𝑜𝑚
𝐿𝐶𝑂𝐸 = (𝑎∗𝐸𝑡 + 𝐸𝑝
(3)
𝑝)

where a was annuity factor which depended upon the life span of the wind power plant and the discount rate
[20]. LCOE for offshore wind farms was estimated between100-200 E/MWh [12].

4. Results and Discussion

Offshore power plant investment should be made for the selected region at least 1 year without investing in
offshore power plants. The average wind speed of the selected region according to the data from GWA was 7-
8.5 m/s. The selected wind turbine generated an average 3 MW per hour. Since the selected region would be
placement 15 wind turbines, 45 MW could be produced per hour. This high-power generation could easily meet

498
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

most of the region’s electricity demand. Additionally, according to Eq. (1), the annual wind energy potential
(Ep) was shown in Table 4. The capacity factors were calculated 40% for the study area.

Table II. The results of economic analysis studies

Region Kırklareli Coastline
Turbine Power 3,465 MW
Number of Total Turbines 15
Total Power 75 MW
Distance to Shore 1,7 km
CAPEX (Euro/MW) 15510219,73
OPEX (Euro/MW) 760286,895
Annuity Factor 0,117459
LCOE 104,47 E/MWh

10 regions of Turkey (Çanakkale, Aliağa, Gemlik, etc.) could be installed offshore wind farms when it
considered as a high wind potential [24]. But offshore wind farm installation would not be correct due to
environmental and social constraints even though these regions had high wind potential. For example, Çanakkale
and Aliağa had different military training zones. Similarly, Gemlik was not suitable region due to the heavily
used port for trade.

In this study, the previously uninvestigated Kırklareli coastline was discussed in terms of offshore wind potential
assessment. According to the analysis, the selected region seemed to be suitable for offshore wind farm
installation according to technical, environmental, and social impacts. Economic analysis studies also indicated
that the study area was in good condition because LCOE was a low rate and energy generation was high level,
the establishment of the offshore power plant would be the right decision for investors.

5. Conclusion, Suggestions and The Future Works

Offshore wind power plant installations had been extensively assessed by means of Kırklareli coastline of
Turkey in this comprehensive survey. According to the most important hallmark for offshore wind energy, wind
speed data based on smooth sea surface were calculated with physical oceanography equations. The average
wind speeds of 5 m/s and more were taken into consideration for suitable site selection. Moreover, the sea depth
of sites was assigned as the 20 m-50 m for the shorter infrastructure and fewer installation expenses. Following
critical conclusions based on comprehensive analysis and assessment were obtained as follows:

- The selected region had high wind speed and high wind power density therefore, electrical power generation
would be high.
- According to environmental and social impacts, the Kırklareli coastline was suitable for offshore wind
generation.
- According to the assessment of economic value studies, the study area was more appropriate condition than
other coastline regions.
- Regarding the future works: Acts and legislation should be promulgated because the legislative framework
lacks guidelines or guidance related to the installation, design, and operation of offshore wind power plants.
- After offshore site selection, the suitable offshore wind turbine could be selected and R&D (Research and
Development) could be carried out to extract maximum power generation.
- Turkish government should present the public finance and fiscal encouragement to promote offshore wind
power projects.
Consequently, Turkey where its huge quantity (three quarters) is surrounded by seas, should be disposed to
offshore renewable energy sources if dependency on other countries wants to be decreased. Therefore, this
comprehensive study will be useful for the development and expansion of offshore wind energy in Turkey.

499
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES
[1] Aydın, İ. Balıkesir’de Rüzgâr Enerjisi. Eastern Geographical Review 2013; 18(19): 29-50. <
https://dergipark.org.tr/tr/pub/ataunidcd/issue/2457/31305>
[2] UK Offshore Wind: Opportunity, Costs & Financing. Climate Change Investment Research. 2011.
[3] IRENA, Renewable Energy Technologies: Cost Analysis Series. International Renewable Agency Working Paper 2012.
(https://www.irena.org/documentdownloads/publications/re_technologies_cost_analysis-csp.pdf)
[4] Esteban, M, Leary, D. Current developments and future prospects of offshore wind ocean energy. Applied Energy 2012;
90 (1): 128-136. < https://doi.org/10.1016/j.apenergy.2011.06.011>
[5] Wind Europe Annual Statics, 2019. (https://windeurope.org/wp-content/uploads/files/about-
wind/statistics/WindEurope-Annual-Offshore-Statistics-2019)
[6] Şen, Z. Türkiye’nin Temiz Enerji İmkânları. Mimar ve Mühendis Dergisi 2004; 33: 6-12.
[7] Altuntaşoğlu, Z, T. Türkiye’de Rüzgâr Enerjisi, Mevcut Durum, Sorunlar. Mühendis ve Makina Dergisi 2012; 52(617):
56-63.
[8] Cali, U, Erdogan, N, Kucuksari, S, Argin, M. TECHNO-ECONOMİC analysis of high potential offshore wind farm
locations in Turkey. Energy Strategy Reviews 2018; 22: 325-336. < https://doi.org/10.1016/j.esr.2018.10.007>
[9] TEIAS, 2015. Turkish Government Electricity Transmission Company. Annual Report 2015.
(http://www.teias.gov.tr/FaaliyetRaporlari/faaliyet2015/)
[10] International Energy Agency (IEA). Energy supply security emergency response of iea countries. (www.iea.org 2014)
[11] TEIAS, 2018. Turkish Government Electricity Transmission Company. Annual Report 2018.
(http://www.teias.gov.tr/FaaliyetRaporlari/faaliyet2018/)
[12] Ilkilic, C, Aydın, H. Wind power potential and usage in the coastal regions of Turkey. Renewable and Sustainable
Energy Reviews 2015; 44: 78-86. < https://doi.org/10.1016/j.rser.2014.12.010>
[13] Global Wind Atlas Wind Potential Resources. 2020. (https://globalwindatlas.com.tr)
[14] Hong, L, Moller, B. Offshore wind energy potential in China: under technical, € spatial and economic constraints.
Energy 2011; 36(44): 82-91. < https://doi.org/10.1016/j.energy.2011.03.071>
[15] Emodnet (European Marine Observation and Data Network) 2020. (https://www.emodnet.eu/human-activities)
[16] Specification of Siemens Gamesa SG 3.4-132 Wind Turbine. 2020. (https://www.siemensgamesa.com/products-and-
services/onshore/wind-turbine-sg-3-4-132)
[17] Dvorak, M, J, Archer, C, L, Jacobson, M, Z. California offshore wind energy potential. Renewable Energy 2010;
35(6): 1244-1254. < https://doi.org/10.1016/j.renene.2009.11.022>
[18] Genç, M, S. Economic analysis of large-scale wind energy conversion systems in central Anatolian Turkey. In: Clean
Energy Systems and Experiences (Eds: Kei. Eguchi). ISBN 978-953-307-147-3, Intech-Sciyo. 131-154. 2010.
[19] Gonzalez, A, G. Review of offshore wind farm cost components. Energy Sustainable Development 2017; 37: 10-19.
https://doi.org/10.1016/j.esd.2016.12.001
[20] Kim, J-Y, Oh, K-Y, Kang, K-S, Lee, J-S. Site selection of offshore wind farms around the Korean peninsula through
economic evaluation. Renewable Energy 2013; 54: 189-195. < https://doi.org/10.1016/j.renene.2012.08.026>
[21] Bhaba, R, S. Minimizing maintenance cost for offshore wind turbines following multi-level opportunistic preventive
strategy. Renewable Energy 2016; 85:104-113. < https://doi.org/10.1016/j.renene.2015.06.030>
[22] Energy Next. India forms offshore wind energy panel. 2012.
[23] Singh, R, Kumar, A. Estimation of offshore wind power potential and cost optimization of wind farm in Indian coastal
region by using GAMS. IEEE International Conference on Current Trends toward Converging Technologies. 1-3
March 2018, Coinbatore, India. 2018 International Conference on Current Trends towards Converging Technologies
(ICCTCT) < 10.1109/ICCTCT.2018.8550900>
[24] Argin, M, Yerci, V. The assessment of offshore wind power potential of Turkey. IEEE Electrical and Electronics
Engineering Conference (ELECO). November 2015. Bursa, Turkey. 2015 9th International Conference on Electrical
and Electronics Engineering (ELECO). < 10.1109/ELECO.2015.7394519>

500
 DETAILED FEASIBILITY STUDY ON ENERGY AND
ECONOMIC APPLICABILITY OF OFFSHORE HYBRID
ENERGY SYSTEMS: A CASE STUDY SAROS BAY

Fatih Karipoğlu
Department of Energy Systems Engineering, Izmir Institute of Technology, Izmir, Turkey, 35430,
[email protected] ORCID: 0000-0002-3314-134X

Mustafa Serdar Genç

Wind and Aerodynamic Research Center, Department of Energy Systems Engineering, Erciyes University, 38039,
Kayseri, Turkey, [email protected] ORCID: 0000-0002-6540-620X

Kemal Koca
Wind and Aerodynamic Research Center, Department of Energy Systems Engineering, Erciyes University, 38039,
Kayseri, Turkey, [email protected] ORCID: 0000-0003-2464-6466

Mert Gezici
Department of Architecture, Izmir Institute of Technology, Izmir, Turkey, 35430, [email protected]
ORCID: 0000-0002-7836-9241

Karipoğlu, F, Genç, MS, Koca, K. Detailed feasibility study on energy and economic
Cite this paper as: applicability of offshore hybrid energy systems: A case study Saros Bay. 8th Eur. Conf. Ren.
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In this paper, energy and economic applicability analysis of the offshore solar and wind hybrid
energy systems was presented. Due to the increased growth industry, economy and population,
energy demand was constantly increased. Offshore energy systems were important energy
production resources for countries with seas. The wind and solar energy markets were rapidly
grown in Turkey. Also, Turkish government wanted to meet most of energy demand from
Renewables. Turkey had a lot of onshore wind and solar energy systems. Although Turkey had
bordered the 4 seas and high offshore wind and solar energy potential, there were not offshore
wind or solar hybrid energy systems. This comprehensive analysis study presented energy and
economic applicability of the Saros Bay for offshore hybrid energy systems under environmental,
social, potential and technical constraints. To determine the suitable site, Geographical Information
Systems and Multi Criteria Decision Method were used. Emodnet and Global Wind Atlas were
also used as a necessary data resource. For the analysis of shadow effect, the Revit Software and
analysis of economic value the RETScreen were utilized. The results showed that Saros Bay for
offshore hybrid energy systems installation seemed to be feasible and reasonable.
Keywords: Hybrid Systems; GIS; MCDM; RETScreen; Shadow effect; Saros Bay
© 2020 Published by ECRES

1. INTRODUCTION

Energy was one of the most important needs of current life. With the increase the number of people and the
development of technological devices, the per capita energy need was increased constantly. In addition, carbon
dioxide (CO2) emissions for energy production was increased with unexpected economic growth [1]. For many
years, most of the energy demand was met from fossil fuels such as coal, natural gas and so on. However,
governments directed to alternative energy systems due to the reduction of the reserves of fossil fuel resources

501
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

and environmental hazards [2]. Over the 20 years, the trend towards renewable energy systems has increased
and clean energy production has been achieved [3]. Renewable Energy Systems (RES) were power plants such
as solar, wind, biogas and so on. Countries showed the great interest in environmentally renewable energy
systems with free energy resources. With the use of the renewable energy systems, carbon emission was
decreased and the domestic job creation was achieved around the word [4]. Wind energy systems were the
highest installed power systems (excluding hydropower) around the renewable energy systems [5]. Wind energy
systems were the most preferred type of energy systems in recent years due to high energy production and low
payback period. Therefore, the global wind energy capacity reached 539 GW from 17.4 GW in 2000 to 2017
[6]. Most of the wind installed power covered the onshore applications worldwide. However, the offshore wind
power capacity increased 3 times in the last 5 years and reached approximately 15 GW level [7]. There were
198 onshore wind power plants in Turkey and wind energy capacity had reach to 10 GW [8]. Although Turkey
was surrounded by the seas, there were not offshore wind power plants until now. So, the studies of the offshore
wind energy were new and important for the energy policy in Turkey. Offshore wind energy potential of Turkey
was identified as 69 GW (with the fixed wind turbines 12 GW and floating wind turbines 57 GW) [10].

In the installation of the renewable energy power plants, the multiple energy systems could be combined to
obtain hybrid energy systems. Hybrid energy systems had many applications areas. They were extensively used
to meet the energy demand and for different applications. Continuously generating energy was provided by
using the solar-wind hybrid energy systems. In the uncertain weather conditions, the high energy production
could be provided from the solar-wind hybrid energy systems thanks to low voltage wave [11]. Also, with the
production of the unlicensed electricity, the installed capacity of solar energy systems reached approximately
3600 MW in 2018 [12]. However, the offshore solar energy systems provided 15% more energy production than
onshore applications due to minimal shadow effects on the seas [13]. Also, offshore wind energy systems
provided more energy production than onshore wind power plants due to steady and intensive wind mass.

In the study in Ref. [14], wind energy was higher in the sea and compared to wind turbines of the same capacity,
the wind turbine installed on the sea would produce 50% more energy. In another study in Ref. [15], Determined
10 different suitability criteria were such factors for Turkey’s seas: underground cables, military zones, water
depth and so on. The most importance of these various criteria was water depth. In the study in Ref. [16],
Geographical Information Systems (GIS)-based method of offshore wind power plants, it was the study
describing the use of GIS and Multi-criteria Decision-Making Method (MCDM) for southeast of Poland. Within
the scope of study, environmental, social and technical 12 different restrictions were handled and analyzed by
GIS. In the study of the effect of shadow and dust effect on the solar cells on PV panels [17], the method of
working solar cells was mentioned. Dusts and shadows negatively affected the performance of the solar cells. It
was conducted the techno-economic analysis of land-based hybrid energy systems in Indonesia [18]. In this
analysis study, wind turbines and solar panels with 1 kW power capacity were performed.

Turkey should ensure from offshore energy resources. Until now, there has not been comprehensive study
related to offshore wind-solar hybrid energy systems. In this study, an innovation hybrid energy system was
examined environmental, social, technical and potential aspects. The suitable region was selected by using GIS
and MDCM combination in the Saros Bay. In this way, the suitable region visualization based on potential,
environmental and social restrictions analysis were obtained. Due to wind turbine’s shadow effect on the solar
panels, power generation analysis from solar panels were done by using Revit 2019. Shade sizes and directions
continuously changed according to the length of the sun duration, location and dominant wind direction. In this
study, the economical applicability of the hybrid energy systems for the selected region were examined. Thus,
a comprehensive analysis study was put forward by considering the potential, environmental, social and
economic analysis result of wind-solar hybrid energy system. The output of this comprehensive study provides
information to investors and policymakers for the offshore wind-solar hybrid energy systems, which will be a
new energy production system in Turkey.

502
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2. METHOD AND MATERIALS

Selected of suitable regions was very important for wind energy investments. For choosing the suitable region,
comprehensive studies should be performed in terms of potential, technical, environmental, social and economic
perspective. In Fig. 1, the method of this study was illustrated. The method consisted of 7 steps and a lot of sub
steps. According to method of study, a suitable region was determined by using GIS and MCDM in Saros Bay.
MCDM consisted of environmental, technical and social impacts such as territorial waters, ship density, military
regions and so on. After the determined suitable region, the configuration of wind turbines were made according
to placement rules and vortex effect. The shadow effect caused by the wind turbines on the solar panels were
analyzed by using Revit 2019. The shadow effect, length and direction constantly changed during the day.
According to the shadow effect on the solar panels, the energy production from solar power plants were
determined by months. In the sixth step, the wind-solar hybrid energy systems were combined and obtained the
total installed power and energy production. According to total installed power solar and wind energy
components, the economic analysis was made by using RETScreen. In this way, the economic applicability of
the solar-wind hybrid energy systems were analyzed according to shadow effect and energy production. The
obtained results of analysis study were interpreted and created the results section.

In order to visualization of analysis studies in GIS, update and accurate database were necessary. In first step of
this study, Emodnet (European Marine Observation and Data Net) [19] was preferred for the environmental and
social restrictions. Also, potential and technical data were obtained from Global Wind Atlas (GWA) and
Bathymeter [20, 21]. Generally, analysis of the shadow effect of objects were used the Revit Software. This
software could make analysis of output power to solar panels based on shading. Therefore, it was used in the
energy project to examine shadow effects. RETScreen software was an important program used in the economic
analysis of energy projects [22]. In the seventh step, the RETScreen was used for the economic applicability and
payback period of offshore hybrid systems. Obtained results were also compared to only offshore wind farms
applications in terms of payback period.

Figure 1. The Method Of Study

3. THE SELECTED OF SUITABLE REGION DEPENDENCE ON POTENTIAL, TECHNICAL,

ENVIRONMENTAL AND SOCIAL IMPACTS
3.1. Study Area: Saros Bay
In this study, the Saros Bay was determined as a case study. Saros Bay was located on the northwest part of
Turkey and near to Marmara Sea. In Fig. 2 showed the study area for to create of offshore hybrid energy systems.
There was a tunnel effect in Saros Bay due to geographical position. So, the wind speed was high in the study
area and the study area was preferable for the offshore wind farms investments.

503
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Study Area

Figure 2. The study area (Saros Bay) adopted by Google Maps

3.2. Potential and Technical Assessment

Prior analysis studies should be performed related to wind speed in the wind energy potential assessment since
wind speed was one of the most important parameters for wind energy investments. More stable flow and high
wind speeds occurred on the seas due to there were not obstacles to disrupting the winds [23]. Therefore, the
offshore wind energy had higher potential than onshore wind farms. The wind speed of the study area were
analyzed by using GIS according to obtained data from GWA for 100 m elevation. In Fig. 3 showed the
disruption of wind speed with the legend in the Saros Bay. The average wind speed of Saros Bay was between
7-10 m/s and this speed value was suitable to established to wind farms.

Figure 3. The disruption of wind speed at 100 m elevation

1.3. Assessment of Environmental and Social Impacts

After the analysis of wind speed, environmental and social impacts should be determined for the offshore wind
farms. These were very important to determine suitable regions like to wind speed analysis. For the determined
environmental and social impacts, MCMD were used in the energy projects. During the last decades, MCDM
method has gained great interest to support complex energy decision-making. Environmental, social impacts
and border zones were performed as follows

 Territorial waters: Selecting study region must be inside territorial waters [17]. According to study in
Ref. [3] of the United Nations Treaty signed in 1982, the seas width was determined as 12 nautical miles
distances.
 Military Regions: The use of military regions for different purposes was dangerous and prohibit. The
offshore wind farms should be installed far away 5 kilometers from military regions due to it might
negatively affect military operations [14].

504
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

 Main Ports, Ship Routes and Ship Density: The seas in Turkey was used by native or foreign people.
The analysis of main ports, ship routes and ship density were very important to determine study area.
The offshore power plants might be dangerous and hazard in the stormy weather. Therefore, the offshore
wind turbines should be located far away 3 km from intensive ship routes and main ports [24].
 Fishery Zones and Fault Lines: The offshore wind turbines could negatively affect the nature sea
habitats. Therefore, 3 kilometers far away should be preferred from fishery zones [25]. Also, Turkey
had a lot of fault lines and Turkey was an earthquake country. Desired offshore wind farms should be
far away 150 meters from active fault lines [26].
 Bird Migration Paths: The wind turbines could kill these migratory birds. Therefore, the offshore wind
farms should be far away 3 kilometers from bird migration birds [27].
Necessary data were obtained for the environmental and social impacts mentioned above. Data was then put in
GIS and was created the border zones. Finally, according to the potential, technical, environmental and social
impacts, the result of these impacts was obtained in Fig. 4. With the result of analysis studies, the corners
coordinates of suitable regions were determined.

Figure 4. The analysis of environmental and social impacts by using GIS

3.4.Position of Offshore Wind Turbines

Considering the real flow conditions around a wind turbine, aerodynamic friction and tip vortices losses due to
turbine blades could not be neglected [28]. In this study, 4*D-7*D towers distance were preferred. The Siemens
Gamesa (SG 2.2.) offshore wind turbine were preferred for this study. The rotor diameter of this wind turbines
was 122 meters, energy capacity was 2.2 MW level [29]. Fig. 7 showed the corner points and wind turbines placement
according to corner point of suitable region.

Figure 5. The corner points and wind turbines locations (adopted by Google Maps)

4. SHADOW ANALYSIS
Shadow negatively affects energy production from solar panels. Its effect caused reduction of energy output.
Therefore, shadow effect should be ignored for maximum energy production when solar energy systems were

505
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

installed [30]. In addition, dust and tilt of ground affected energy production importantly [20]. 22 offshore wind
turbines were placed in the study area and the solar panels area, which was region of between wind turbines,
was calculated about 46.475 m2 by using GIS and Revit. . Revit 2019 calculated the energy production from
solar panels according to days, hours, sun duration, location, season and so on. In this study, daily energy
production was calculated for the study area by months. Solar panels with 250 W (15480 pieces) power capacity
was used and energy efficient of solar panels was also calculated in %16 by the software. In Fig. 6 and Fig. 7
showed the shadow effect in the study area.

Fig 6. The shadowing in the study area in 21 June; Fig 7.) The top view of 21 December at 12:30 pm.

The change of energy production was based on sunshine, day duration, temperature, shading. Annual energy
production was calculated as 42785,422 kWh. During the year, day of the highest energy production was 21
June with the 209.5 kWh and day of lowest energy production was 21 December with the 13.5 kWh.

5. ECONOMIC ANALYSIS AND APPLICABILITY OF OFFSHORE HYBRID SYSTEMS

Economic analysis of energy projects was very important for the applicability and sustainability. So, detailed
equipment selection and economic analysis should be taken into consideration. The costs of energy power plants
constantly changed due to associated risks and payback period perceived by investors [31]. If the cost of
foundation decreased, the development of offshore wind energy would increase. In this study, the monopile
offshore wind turbines foundation type was connected the solar panels foundations. Therefore, only foundations
costs were occurred in the study area except for foundation of solar panels. In addition, for the hybrid energy
systems, hybrid inverter with the lower cost was used according to total power of hybrid energy systems in the
study area. Thus, the payback period of offshore hybrid energy systems was prompted shorter than only offshore
wind farms. The total annual energy production was about 233.518 kWh in the hybrid energy system. The initial
cost of hybrid system was calculated in 125.000.000 $ and the annual maintenance cost of system was calculated
in 2.274.000 $ by using analysis software. According to these data, the payback of offshore hybrid systems was
calculated about 8.7 years. Therefore, the payback period of offshore wind farms was lowered from 10.5 years
to 8.7 years. Thus, the offshore wind and solar hybrid systems was preferable in terms of economic value in
Saros Bay.

Figure 8. The results of economic analysis by using RETScreen

506
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

6. CONCLUSIONS
This comprehensive study investigated applicability of hybrid energy systems (solar and wind energy) in terms
of economic and energy in the study area. Due to the high foundation and other costs, the offshore wind farms
were not available economically. In this study, offshore hybrid systems was examined to connect at the same
foundation and inverter. Decreasing of high costs were targeted to occur the only foundation costs and electrical
cost in the study area. GIS and MCDM combination were performed to determine the suitable site selection of
desired hybrid energy system. Different data resources were also used in terms of environmental, potential,
technical and social impacts.

With the identified restrictions and border zones, the analysis of study area was made in GIS and the visualization
was obtained in map layers. According to placement rules, the 22 offshore wind turbines were placed in the
selected area and the study area was calculated about 46,475 m2. After these process, 15480 pieces with 250 W
polycrystal solar panels were placed between the wind turbines in the Revit software. Monthly energy production
from solar panels were then determined based on shadow effect. The energy production of solar panels changed
constantly due to location, sunshine, sun duration and shadow effects. Considering the all these impacts, the
output of energy was obtained from Revit. It was observed that month of the highest energy production was June
while the lowest energy production was December. In addition, days of the highest energy production and the
lowest energy production were 21 June and 21 December, respectively. Due to, these days were the longest and
the shortest sunshine in a year. Thus, the sunshine importantly effected energy production of solar panels. In the
same study area, the annual total energy production was obtained about 233,518 kWh instead of 191,792 kWh.
Thus, annual energy production was increased and costs of some equipment was decreased. Finally, the payback
period decreased from 10.5 years to 8.7 years. Thus, offshore solar and wind hybrid systems could be an
economically viable investment. This study was proved that offshore solar and wind hybrid energy systems were
useful and advantageous. For the 2023 energy strategies of Turkey, offshore hybrid systems should be given
importance to ensure high energy production with the low investment costs.

With regulation of the hybrid energy systems in Turkey, the important of hybrid energy systems were thought
to increase day to day. For the policymakers and investors, this study contained important information in terms
of restrictions, to determine the suitable region, economic viability, analysis of shadow effects. This study can
be further extended with third energy system such as wave energy system. Also, the electrical design and subsea
energy transmission will be considered in the further study.
REFERENCES

[1] Kazim, A. Strategy for a sustainable development in the UAE through hydrogen energy. Renewable Energy 2010; 35:
2257-2269. < https://doi.org/10.1016/j.renene.2010.03.013>
[2] El Chaar, L, Lamont, L.Global solar radiation: Multiple on-site assessments in Abu Dhabi, UAE. Renewable Energy
2010; 35:1596-1601. < https://doi.org/10.1016/j.renene.2009.10.007>
[3] Zervos, A, Adib, R. Renewables 2018 global status report,” Paris: REN21 Secretariat, 2018.
[4] Değirmenci, S, Bingöl, F, Sofuoğlu, S. MCDM Analysis of wind energy in Turkey: decision making based on
environmental impacts. Environmental Science and Pollution Research 2018; 25: 19753-19766. <
https://doi.org/10.1007/s11356-018-2004-4>
[5] Saidur, R, Islam, M, Rahim, N, Solangi, K. A review on global wind energy policy. Renewable and Sustainable. Energy
Reviews 2010; 14 (7): 1744–1762. < https://doi.org/10.1016/j.rser.2010.03.007>
[6] Global wind energy council (GWEC), Global wind statistics. (http://www.gwec.net).
[7] Sun, X, Huang, D, Wu, G. The current state of offshore wind energy technology development. Energy 2012; 41 (1):
298–312. < https://doi.org/10.1016/j.energy.2012.02.054>
[8] TÜREB 2020, Türkiye Rüzgâr Enerjisi Birliği (TÜREB), 2020. Türkiye rüzgâr enerjisi birliği istatistik ocak ayı raporu.
(http://www.tureb.com.tr/turebsayfa/duyurular/turkiye-ruzgar-enerjisi-istatistik-raporu-ocak-2020).
[9] Enerji Piyasaları İşletim Anonim Şirketi, EPİAŞ, 2020. (https://www.epias.com.tr/).
[10] ESMAP GOING GLOBAL-Expanding offshore wind to emerging markets, October 2019 Report.
[11] Li-qun, L, Wang, Z. The development and application practice of wind-solar energy hybrid generation systems in
China. Renewable and Sustainable Energy Reviews 2009; 13: 1504-1512. < doi: 10.1016/j.rser.2008.09.021>
[12] Türkiye Bilimler Akademisi, TUBA, Güneş Enerji Teknolojileri Raporu, Ankara, 2018.
(http://www.tuba.gov.tr/tr/yayinlar/suresiz-yayinlar/raporlar/tuba-gunes-enerjisi-teknolojileri-raporu).
[13] Bajpai, P, Dash, V. Hybrid renewable energy systems for power generation in stand-alone applications. Renewable
and Sustainable Energy Reviews 2012; 16(5): 2926-2939. < https://doi.org/10.1016/j.rser.2012.02.009>

507
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[14] Deveci, M, Özcan, E, John, R. Offshore Wind Farms: A fuzzy approach to site selection in a Black Sea region. In:
IEEE Conference, College Station, TX, USA, 6-7 February 2020, 2020 IEEE Texas Power and Energy Conference
(TPEC). < 10.1109/TPEC48276.2020.9042530>
[15] Emeksiz, C, Demirci, B. The determination of offshore wind energy potential of Turkey by using novelty hybrid site
selection method. Sustainable Energy Technologies and Assessments 2019; 36: 100562. <
https://doi.org/10.1016/j.seta.2019.100562>
[16] Szurek, M, Blackhowski, J, Nowacka, A. GIS-based method for wind farm location multi-criteria analysis. Mining
Science 2104; 21: 65-81. < doi: 10.5277/ms142106>
[17] Ibrahim, A. Effect of Shadow and Dust on the Performance of the Silicon Solar Cell. Journal of Basic and Applied
Scientific Research 2011; 1(3): 222-230. < ISSN 2090-424X>
[18] Hiendro, A, Kurnianto, R, Rajagukguk, M, Simanjuntak, J. Techno-economic analysis of photovoltaic/ wind hybrid
energy systems for onshore/remote area in Indonesia. Energy 2013; 59: 652-657. <
http://dx.doi.org/10.1016/j.energy.2013.06.005>
[19] Emodnet (European Marine Observation and Data Net), 2020. (https://www.emodnet.eu/human-activities).
[20] Global Wind Atlas, Wind Speed Database, 2020. (https://globalwindatlas.info/).
[21] Emodnet / Bathymeter Data Map, 2020. (https://portal.emodnet-bathymetry.eu/).
[22] Satir, M, Murhpy, F, McDonnel, K. Feasibility study of an offshore wind farm in the Aegean Sea, Turkey. Renewable
and Sustainable Energy Reviews 2018; 81: 2552-2562. < http://dx.doi.org/10.1016/j.rser.2017.06.063>
[23] Bilgili, M, Yasar, A, Simsek E. Offshore wind power development in Europe and its comparison with onshore
counterpart. Renewable and Sustainable Energy Reviews 2011;15: 905–1005.
<http://dx.doi.org/10.1016/j.rser.2010.11.006>
[24] Hong, L, Möller, B. Offshore wind energy potential in China: under technical, spatial and economic constraints. Energy
2011; 36: 4482-4491. < doi:10.1016/j.energy.2011.03.071>
[25] Mahdy, M, Bahaj, A. Multi criteria decision analysis for offshore wind energy potential in Egypt. Renewable Energy
2018; 118: 278-289. < https://doi.org/10.1016/j.renene.2017.11.021>
[26] Demirtaş, R. Kentsel planlamada diri faylar etrafında tampon bölge oluşturma esasları. Ankara University Press,
Ankara, 2005.
[27] Zimmerling, R., Pomeroy, C., d’Entremont, M. V., Francis, C. M. Canadian estimate of bird mortality due to collisions
and direct habitat loss associated with wind turbine developments. Avian Conservation Ecology 2013; 8(2):10. <
10.5751/ACE-00609-080210>
[28] İlhan, A, Bilgili, M, Sahin, B. Analysis of aerodynamic chacteristics of 2 MW horizontal axis large wind turbine. Wind
and Structures 2018; 27(3):187-197. < https://doi.org/10.12989/was.2018.27.3.187>
[29] Specification of Siemens Gamesa SG-2.2 Wind Turbine, 2020. (https://www.siemensgamesa.com/products-and-
services).
[30] Ribeiro, AED, Arouca, MC, Coelho, DM. Electric energy generation from small scale solar and wind power in Brazil,
The influence of location, area and shape. Renewable Energy 2016; 85: 554-563. <
http://dx.doi.org/10.1016/j.renene.2015.06.071>
[31] Dvorak, MJ, Corcoran, BA, Ten Hoeve, JE, McIntyre, NG, Jacobson, MZ. US East Coast offshore wind energy
resources and their relationship to peak-time electricity demand. Wind Energy 2012; 16: 977-997. <
https://doi.org/10.1002/we.1524>

508
 THERMAL-FSI OF MINICHANNEL HEAT SINK FOR
CALCULATING THE OPTIMUM FLOW RATE: A NUMERICAL
APPROACH

Shoukat Ahmad Adnan

Department of Mechanical Engineering, Institute of Space Technology, Islamabad, Pakistan,
ORCID: 0000-0002-0587-8774

Anwar Muhammad
Department of Mechanical Engineering, Institute of Space Technology, Islamabad, Pakistan,
ORCID: 0000-0001-8899-852X

Koten Hasan
Istanbul Medeniyet University, Mechanical Engineering Department, Istanbul, 34700, Turkey,
ORCID: 0000-0002-1907-9420

Shoukat A, Koten, H. Thermal-FSI of Minichannel Heat Sink for Calculating the Optimum
Cite this paper as: Flow Rate: A Numerical Approach . 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: In the present work, numerical investigations were carried out on the thermal fluid-structure
interaction (Thermal-FSI) of minichannel heat sinks. Based on the different meshing
configurations, the optimum flow rate for the microchannel heat sink was evaluated. Results in
terms of equivalent stress, thermal resistance, base temperature, and overall heat transfer
coefficient were calculated and compared for different meshing heat sink configurations. An in-
depth comparison was made between the mesh configuration and the relative simulation time. The
effect of flow rate and the simultaneous effect of temperature and flow rate on the produced
stresses was also observed. The optimum flow rate of 0.7 LPM was measured for an average
meshing size of 0.85mm at the threshold temperature of 103.05○C. Moreover, the produced
stresses were calculated to be 124.2 MPa. In addition, heat sink geometry with the best thermal
performance was also simulated for higher heat fluxes within the same operating limits. Moreover,
results were validated using available correlations and previously published literature. Flow
maldistribution, pressure drop and local heat transfer coefficient in the shape configuration are also
presented in the current study.
Keywords: FSI, Thermal Stresses, Microchannels, CFD, Heat sink
© 2020 Published by ECRES

1. INTRODUCTION

Heat transfer enhancement in minichannel heat sinks is an imperative technique for efficient thermal
management of microelectronics systems, nowadays. Previously, many researchers have contributed to the
optimization of geometrical parameters to enhance the overall heat transfer rate. For instance, Muhammad et.al
[1] investigated the header design of minichannel heat sinks and suggested a mathematical and analytical model
improve the header distributor and collector by the influence of pressure drop. It was also reported that the
channel geometry is sensitive to flow rate and the pressure drop, thermal resistance and flow maldistribution
could be reduced up to 74% based on their proposed/modified header design geometry. In addition, S. Kim et.al
[2] presented three numerical and two analytical optimization models for the thermal performance of
minichannel heat sinks. Based on the five optimized models; a comparison was drawn between the different
aspect ratios of minichannel heat sinks with numerical simulation and it was suggested that the porous optimized

509
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

method was the best optimization method. Saeed et.al [3] investigated the channel configuration and flow rate
effect on thermal resistance and heat transfer enhancement using Al 2O3-H2O nanofluids, when the fluid was
used as a mixture. It was observed that the thermal performance was improved using nanofluid and changing
the fin spacing of the minichannel heat sink. Moreover, two-phase mixture models gave a better prediction than
the single-phase model, when the results were compared with the experimental data. Shoukat et.al [4],
minimized the base temperature of the minichannel heat sink by using stable Nano-fluids with different
volumetric concentrations and compared the experimental results with numerical simulation. The results
revealed that the AgO-H2O nano-fluid with 0.001% volumetric concentration showed 7.12% better thermal
performance on pin-fin minichannel as compared to Al2O3-EG nano-fluids. Different geometric parameters were
optimized at constant volumetric flow rates and different geometric configurations were also investigated. The
pumping power is a very important factor in these confined flows through minichannel heat sinks. It was
observed that the sandwich and trapezoidal porous geometric designs give more effective heat transfer but
sandwich porous design configuration showed the lowest pressure drop [5,6–12]. Some researchers numerically
optimized the design of the minichannel heat sink by analyzing uniform temperature and flow distribution using
different geometric configurations in different flow regimes. The thermal and hydraulic performances of
different flow structures were discussed in terms of the surface temperature and pressure drop, which gave the
optimized Ψ-shaped heat sink [13,14]. Moreover, few analytical models were reported to minimize the thermal
resistance of minichannel heat sinks and those models were also validated through CFD approaches [15].

The geometry of the heat sink and computational modeling

Pin-fin mini channel heat sink was selected to analyze the thermal performance parameters and the effect of
thermal stresses. The geometry with dimensions of the heat sink is shown in Fig. 1 (a) and (b). Steady-state
forms of continuity, momentum and energy equations (Eq.1 to 3) were solved using CFD software to simulate
the conjugate heat transfer problem for the heat sink and then one-way FSI procedure was adopted to compute
the thermal stresses. The temperature was transferred from the CFD solver to the structural part of the software
or CSD solver and the effect of imported temperature on the heat sink was observed. The CSD solver calculated
the thermal stress by using the Eq.4.
(1)
=- (2)
( =k (3)
(4)
σ = αEΔT –

(a) (b)
Figure 1: Pin-Fin Heat Sink

The flow rates of 0.5 LPM to 1.5 LPM were used in the current study. The Reynold’s number can be calculated
from Eq. 5. The flow in the pin fin microchannels was modeled as turbulent flow, so the SST turbulence model
was chosen to simulate the heat sink. Moreover, the fluid and solid domains of the sink geometry were
discretized with tetrahedral control volumes. To ensure node to node connections at the interface between fluid
and solid domains, both domains meshed with the same topology, where the relevance was 100.

510
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Re = (5)

The major purpose of transferring the temperature loads from CFD solver to CSD Solver was to build the mesh
to mesh model so that the loads can be applied more smoothly on each node. ANSYS ® was used for both of
these purposes to evaluate the design of the heat sink. The commercial code of CFX was used as CFD solver
and ANSYS Mechanical was used as CSD solver. The variation of temperature distribution overtime was the
main purpose of this work, which produces thermal stresses. It was pertinent to find the threshold range of
temperature at which the minimum number of thermal stresses (because more the thermal stresses, more chances
to failure of the heat sink) produced at different flow rates. A comparison is drawn between the different mesh
sizes, temperature drop, thermal stresses at different flow rates. The mesh size of 0.5mm, 0.75mm and 1 mm
was used to find out the temperature distribution and the thermal stresses were measured at the flow rates of 0.5
LPM to 1.5 LPM. The dimensions of minichannel heat sink used in the current study are illustrated in Table.1.

Table 1 Heat Sink Dimensions

Sr.No Parameters Dimensions (mm)
1 Active Width of Heat Sink 66
2 Total Width of Heat Sink 70
3 Active length of Heat Sink 101.6
4 Total length of Heat Sink 147
5 Pin-Fin Width 1
6 Pin-Fin Height 1
7 Heat Sink’s Base Height 5
8 No. of Pin-Fins 43 x 34

Mass flow and pressure boundary conditions were imposed on inlet and outlet and inlet temperature and outlet
pressure for all simulations were set to be 25oC and 1atm, respectively. Surface roughness was not included in
the model, hence no-slip condition was implemented on the walls and the interface (an interface was created
between the solid and fluid domain) for Fluid-Solid Interaction (FSI). There was no loss of heat flux except for
interface and all other walls were set to no-slip at the adiabatic condition. Distilled water was utilized as a
working fluid at a different flow rate of 0.5 LPM to 1.5 LPM and a heat source of 230 W was applied, which
was simulated by a protrusion on the bottom.
The abovementioned assumption was taken because the momentum and the energy equations were solved
separately but when the loads were transformed from the fluid domain to a solid domain, the solver solved both
equations simultaneously.
The coupling energy balance equations for Thermal-FSI (fluid and solid contact layer) are shown below
+ =0 (6)

The heat flux through the contact layer between the fluid and the solid can be formulated as follow.
= T (7)
= T (8)
Where ƛ is the function of Reynold’s number, wall y+ and strain rate.
The basic understanding of thermal-FSI is illustrated by the Fig.2. The Thermal Stresses were observed during
the Thermal-Fluid-Structure Interaction of the minichannel heat sink and water in its flow condition.

511
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2: Illustration of Thermal FSI

RESULTS
In the present study, the effect of flow rate on the temperature drop and thermal stresses was observed on a
minichannel heat sink, which is used in a microprocessor. The thermal FSI analysis includes the heat transfer
between the separated mediums for the different flow rates. The heat flux is defined as

(9)
= cp (Tout-Tin)

The other parameters like overall heat transfer coefficient (U), Interface Area (Ai) log mean temperature
difference (LMTD) and Thermal resistance (R th) can be calculated using the following equations.

(10)
LMTD =

Ai=2(a.d)–4*0.5*[0.5*(a-c)*0.5*(d-b)]+8.89(2c+2b+4edges)+N(lpf.wpf) (11)

(12)
U= cp

(13)
Rth =

MESH INDEPENDENCE
The mesh independence study was completed for three different meshes M-1, M-2 and M-3 for identical
geometry, as shown in Table 2. All the meshes were generated with tetrahedral elements having the same
topology because of low computational cost, although it took almost 10 hours to complete one case for each
flow rate. It is better to use Hexahedral elements because of the less computational time but the computational
cost will be high. In the current mesh independence study, the base temperature (T b) and thermal stresses were
compared with each mesh size.

Table 2 Mesh details

Mesh M-1 M-2 M-3
Element Size maximum [mm] 0.5 0.75 1
Total Number of Nodes 706964 565115 518614
Computation Time (per 10 900 500 80
Allocated Memory
iterations) [s] [MB] 2197.48 1738.58 500

512
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

DISCUSSION
These numerical simulations were validated by published literature related to thermal stresses in a minichannel
heat sink. Accordingly, it was assumed in the current study that the shear stresses produced at an increased flow
rate dominated the stresses produced by the temperature gradient in the minichannel heat sink. Hence, this
assumption paved the way to find out the optimum flow rate at which overall produced stresses, and the
temperature was minimal so that the material failure could be avoided. In the current model, a complete
temperature profile on the surface was considered. Fig.3 illustrates one example of 0.5mm mesh element size
at1.5 LPM flow rate. The area-weighted average on the surface was computed using the Eq. 14, as shown below:

Tb = ʃ T. = (14)

Figure 3: Temperature Distribution

The area in the above equation is the surface area which is in contact with the microprocessor directly.
‘i’ is an element on the surface of that area and ‘n’ represents the total number of elements on the
surface. Accordingly, maximum thermal stresses can be evaluated from the maximum drop in base
temperature. The other parameters like, heat transfer coefficient (U), Interface Area (Ai) log mean
temperature difference (LMTD), Thermal resistance (Rth) and Nusselt number (Nu) were also calculated
and compared with the thermal stresses.

CONCLUSIONS

It is evident when the mesh size was changed from higher to lower i.e. 1 mm to 0.5mm, the base
temperature decreased but the equivalent stress was increased. For 1mm and 0.75mm mesh size, the
temperature and stress curve coincided when the Reynold’s number was 4000. By taking the average
value of both mesh size (~0.85mm), the optimum flow rate was calculated to be 0.8LPM. This work
was conducted to find out the optimum flow rate for various structural parameters that show the best
thermal performance parameters in which generated stresses were studied together by applying the
Thermal FSI technique. Pin-fin Mini channel heart sink was analyzed for different mesh sizes at
varying flow rates from 0.5 LPM to 1.5 LPM with an increment of 0.25 LPM. It was observed that a
decrease in mesh size lead to an increase in generated stresses and a higher pressure drop was also
noticed for increased flow rates.
REFERENCES

[1] S. Muhammad and M. Kim, “Header design approaches for mini-channel heatsinks using analytical and numerical
methods and numerical methods,” Applied Thermal Engineering, vol. 110, no. September, pp. 1500–1510, 2016.

513
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[2] S. Kim, “Methods for Thermal Optimization of Microchannel Heat Sinks,” Heat Transfer Engineering, vol. 7632, no.
25:1, pp. 37–49, 2010.
[3] M. Saeed and M. Kim, “Heat transfer enhancement using nanofluids ( Al2O3 -H2O ) in mini-channel heatsinks,”
International Journal of Heat and Mass Transfer, vol. 120, pp. 671–682, 2018.
[4] A. A. Shoukat, M. Shaban, A. Israr, O. R. Shah, M. Z. Khan, and M. Anwar, “Stability of nano-fluids and their use for
thermal management of a microprocessor: an experimental and numerical study,” Heat and Mass Transfer, pp. 1–12,
Mar. 2018.
[5] X. D. Wang, Bin An, and J. L. Xu, “Optimal geometric structure for nanofluid-cooled microchannel heat sink under
various constraint conditions,” Energy Conversion and Management, vol. 65, pp. 528–538, 2013.
[6] P. Naphon and S. Wongwises, “Investigation on the jet liquid impingement heat transfer for the central processing unit
of personal computers ☆,” International Communications in Heat and Mass Transfer, vol. 37, no. 7, pp. 822–826,
2010.
[7] K. Azha, M. Annuar, and F. S. Ismail, “Optimal Pin Fin Arrangement of Heat Sink Design and Thermal Analysis for
Central Processing Unit,” 2014.
[8] X. L. Xie and W. Q. Tao, “Numerical Study of Turbulent Heat Transfer and Pressure Drop Characteristics in a Water-
Cooled Minichannel Heat Sink,” Journal of Electronic Packaging, vol. 129, no. September, pp. 247–255, 2007.
[9] S. Mukherjee, “Preparation and Stability of Nanofluids-A Review,” IOSR Journal of Mechanical and Civil Engineering,
vol. 9, no. 2, pp. 63–69, 2013.
[10] D. B. Tuckerman and R. F. W. Pease, “High-Performance Heat Sinking for VLSI,” IEEE ELECTRON DEVICE
LETTERS, VOL. EDL-2, NO. 5, MAY 1981 High-Performance, no. 5, pp. 126–129, 1981.
[11] N. Goldberg, “Narrow Channel Forced Air Heat Sink,” IEEE TRANSACTIONS ON COMPONENTS, HYBRIDS, AND
MANUFACTURING TECHNOLOGY, VOL. CHMT-7. NO. 1. MARCH 1984, pp. 154–159, 1984.
[12] S. U. S. Choi, Z. G. Zhang, W. Yu, F. E. Lockwood, and E. A. Grulke, “Anomalous thermal conductivity enhancement
in nanotube suspensions,” vol. 2252, no. 2001, pp. 14–17, 2013.
[13] B. Ramos-alvarado, P. Li, H. Liu, and A. Hernandez-guerrero, “CFD study of liquid-cooled heat sinks with
microchannel flow field configurations for electronics , fuel cells , and concentrated solar cells,” Applied Thermal
Engineering, vol. 31, no. 14–15, pp. 2494–2507, 2011.
[14] C. A. Rubio-jimenez, A. Hernandez-guerrero, J. G. Cervantes, D. Lorenzini-gutierrez, and C. U. Gonzalez-valle, “CFD
study of constructal microchannel networks for liquid-cooling of electronic devices,” vol. 95, pp. 374–381, 2016.
[15] S. V Garimella, “Analysis and Optimization of the Thermal Performance of Microchannel Heat Sinks,” CTRC
Research Publications, 2005.

514
 THERMODYNAMIC ANALYSIS OF A REFRIGERATION
CYCLE USING NATURAL REFRIGERANTS WITH LOW
GLOBAL WARMING POTENTIAL

Touaibi Rabah
Laboratory of Industrial Fluids, Measurements and Applications (FIMA), Djilali Bounaama University, Ain Defla,
Algeria,
ORCID: 0000-0003-1236-0516

Koten Hasan
Istanbul Medeniyet University, Mechanical Engineering Department, Istanbul, 34700, Turkey,
ORCID: 0000-0002-1907-9420

Touaibi, R, Köten, H. Thermodynamic analysis of a refrigeration cycle using natural

Cite this paper as: refrigerants with low global warming potential. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
2020, Istanbul, Turkey

Abstract: This work presents an energy analysis of a vapor system refrigeration cycle using natural coolants
with low GWP including fluids R290, R717 and R 170 using fluid R134a as a reference. The study
was carried out to avoid the problem of the ozone layer and the conservation of the environment;
to do this, a calculation code was developed using the EES software in order to determine all the
thermodynamic characteristics and the performances of the cycle, in particular the coefficient of
performance. This detailed study between the coolants was carried out to determine the best fluid,
which gives a good coefficient of performance. The parameter taken into account for the
comparative study is the isentropic compressor efficiency. Results show that R717 fluid gives a
good coefficient of performance compared to the others; the latter has a low GWP gives a
coefficient of performance higher than that of the R134a fluid.
Keywords: Thermodynamic Analysis- refrigeration cycle- natural refrigerants global- warming potential
© 2020 Published by ECRES

Nomenclature
h specific enthalpy, kJ/kg
𝑚̇ Masse flow rate, kg/s
𝑄̇ Heat rate, kW
𝑇 Temperature, °C
𝑊̇ Mechanical Power, kW
COP Coefficient of performance
ODP Ozone depletion potential
GWP Global warming potential

1. INTRODUCTION

In last decades, the demand for cold has increased due to multiple applications, such as domestic cold,
commercial cold and industrial cold. The international market is constantly looking for more environmentally
friendly refrigerants with even less impact on the environment. Coolant systems still in use are mechanical vapor
compression systems. These systems use refrigerants as the working fluid, the majority of which are considered
air pollutants. According to the Montreal Protocol in 1987 and Kyoto in 1979 necessitate the search for new
fluids. Some harmful materials kinds of ODP were forbidden under the Montreal Protocol because of their
harmful effects on the ozone layer which protects from harmful lights [1-4]. The environmental impact is stated

515
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

due to the ODP and the GWP. These parameters become a more efficient factor in the selection of refrigerants
[5-8]. Many works are currently in search of new ecological fluids which can have less impact on the ozone
layer thus also on the global warming of our planet. The objective of this work is to make an energy analysis of
a vapor compression coolant cycle by natural coolants with low GWP including fluids R290, R717 and R 170
using fluid R134a as a reference.

2. MATERIALS AND METHODS

The vapor compression refrigeration cycle is generally used refrigeration system for air conditioning in buildings
today.

Description of system:
The coolant system consists of four different devices: compressor, condenser, expansion valve and evaporator.
The coolant comes to the compressor at low temperature and low pressure. The refrigerant leaves the compressor
and enters the condenser; the condenser is essentially a heat exchanger; heat is transferred from the refrigerant
to the outside environment. As the refrigerant comes to the expansion valve, it expands and releases pressure.
Therefore, the temperature drops at this point. As a result of these changes, the refrigerant leaves the expansion
valve as a mixture of liquid vapor. Throttle valves play two crucial roles in the vapor compression cycle. In the
evaporator, the refrigerant is at a temperature lower than that of its environment. Consequently, it evaporates
and absorbs the latent heat of vaporization. Heat detraction occurs at low pressure and temperature (Figure 1).

Condenser
3 2

Valve Comp

4 1
Evaporator

Figure 1. Diagram of vapor compression refrigeration cycle.

Thermodynamic analysis
In this section, a thermodynamic modeling study is presented and the mass balance is defined as follows Eq [1]
[9-10].

∑ 𝑚̇𝑖𝑛 = ∑ 𝑚̇𝑜𝑢𝑡 (1)

Where, 𝑚̇ is the mass flow rate (kg/s),

According to first law, the energy equation as follows Eq [2],:

∑ 𝑚̇𝑖𝑛 ℎ𝑖𝑛 + ∑ 𝑄̇𝑖𝑛 + ∑ 𝑊̇𝑖𝑛 = ∑ 𝑚̇𝑜𝑢𝑡 ℎ𝑜𝑢𝑡 + ∑ 𝑄̇𝑜𝑢𝑡 + ∑ 𝑊̇𝑜𝑢𝑡 (2)

Where; 𝑄̇, 𝑊̇ and ℎ are the heat transfer rate, work rate and specific enthalpy.
The energy analysis of the Vapor Compression Refrigeration Cycle requires the application of the first law of
thermodynamics. The general equations corresponding to this principle are presented below Eqs [3-6]:

𝑚̇1 ℎ1 + 𝑊̇𝐶𝑜𝑚𝑝 = 𝑚̇2 ℎ2 (3)

𝑚̇2 ℎ2 = 𝑚̇3 ℎ3 + 𝑄̇𝐶𝑜𝑛𝑑 (4)

𝑚̇3 ℎ3 = 𝑚̇4 ℎ4 (5)

𝑚̇4 ℎ4 + 𝑄̇𝐸𝑣𝑎𝑝 = 𝑚̇1 ℎ1 (6)

516
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Performance evaluation
Efficiency of the Coolant Cycle is determined by the following equations Eq [7]:

𝑄̇𝐸𝑣𝑎𝑝
𝐶𝑂𝑃 = (7)
𝑊̇𝐶
3. RESULTS AND DISCUSSION

In this part we study the performance of a vapor refrigeration cycle using different natural refrigerants with low
global warming potential in order to see the performance of some refrigerants. Natural coolants are substances
that occur normally in the space. With zero ODP and also so little or no GWP, they are evaluated the definitive
method to the environmental damage caused by synthetic refrigerants. The main characteristics of the most
commonly used natural refrigerants are listed in the Table 1:

Table 1. Thermodynamic properties of the refrigerants

Refrigerant Safety group GWP ODP Critical temperature ( °C) Critical pressure (bar)
R717 B2L 0 0 132.25 113.33
R290 A3 3 0 96.74 42.512
R170 A3 6 0 32.172 48.722
R134a A1 1430 0 101.06 40.593

This part presents a comparative study based on a mechanical compression refrigeration cycle, using natural
refrigerants. A thermodynamic model was developed to determine the cop of the coolant cycle and to see the
variation of a certain quantity depending on the compressor performance, in particular the mechanical power
consumed by the compressor and the cycle performance coefficient. We set certain parameters and vary the
evaporation temperature. Figure 2 shows the variation of the mechanical power consumed by the compressor as
a function of the efficiency of the compressor using certain natural refrigerants (R717, R170 and R290) taking
into account the R 134a fluid as reference refrigerant. The results show that for each fluid, the mechanical power
consumed by the compressor decreases with the incease in the efficiency of the compressor over the interval
[0.6; 1], in this interval the mechanical power consumed by the compressor for R 7171 decreases from 4.85 to
2.91. It is also noted that in the case of the fluid R170 the compressor requires a large mechanical consumption
to supply the compressor, on the other hand the fluid R717 requires less like the fluid R134a.

18
16
R134a R290
14
R717 R170
12
W_C, kW

10
8
6
4
2
0
0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Eta_C

Figure 2. Variation of the mechanical power consumed by the compressor as a function of the compressor efficiency

Figure 3 shows the variation of the coefficient of performance of the mechanical power consumed by the
compressor as a function of the efficiency of the compressor using certain natural refrigerants (R717, R170 and
R290) taking into account the R 134a fluid as the reference refrigerant. The results show that for each fluid, the
coefficient of performance increases considerably with the increase in the efficiency of the compressor over the
interval [0.6; 1] for all fluids, in this interval the coefficient of performance increases from 2.05 to 3.43. We

517
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

notice that the R717 refrigerant gives a good coefficient of performance compared to other refrigerants. its value
exceeds the value of R134a refrigerant.

4
3.5 R134a R290
3 R717 R170
2.5

COP, -
2
1.5
1
0.5
0
0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Eta_C

Figure 3. Variation of the coefficient of performance as a function of the compressor efficiency

4. CONCLUSION

In this work, a comparative study of the performances of certain refrigerants to see the performances of tails of
natural refrigerants with low GWP taking into account the fluid R 134a as a reference. Different refrigerants
were analyzed using numerical software and discussed their performances in terms of the cooling system
equipment. In literature, refrigerants play critical importance for the system efficiency. The work was followed
by parametric study to see the effect of compressor performance on the performance of the refrigeration cycle.
This led us to the following conclusions.
- The R717 gives a coefficient of performance, which is close to the R134a fluid.
- The mechanical power consumed by the compressor decreases with the increase in the isentropic efficiency of
the compressor.
- The coefficient of performance of the vapor compression refrigeration cycle increases with the increase in the
isentropic efficiency of the compressor.

ACKNOWLEDGMENT
The authors thank the University Djilali Bounaama Khemis Miliana (Algeria) for funding this study.
REFERENCES
[1] Abas, N., Kalair, A. R., Khan, N., Haider, A., Saleem, Z., & Saleem, M. S. Natural and synthetic refrigerants, global warming: A
review. Renewable and Sustainable Energy Reviews,2018; 90: 557-569.
[2] Sarbu, I. A review on substitution strategy of non-ecological refrigerants from vapour compression-based refrigeration, air-
conditioning and heat pump systems. International Journal of Refrigeration, 2014; 46: 123-141.
[3] Andersen, S. O., Sherman, N. J., Carvalho, S., & Gonzalez, M. The global search and commercialization of alternatives and substitutes
for ozone-depleting substances. Comptes Rendus Geoscience, 2018 ; 350(7) : 410-424.
[4] Botticella, F., De Rossi, F., Mauro, A. W., Vanoli, G. P., & Viscito, L. Multi-criteria (thermodynamic, economic and environmental)
analysis of possible design options for residential heating split systems working with low GWP refrigerants. International Journal
of Refrigeration, 2018; 87: 131-153.
[5] Ciconkov, R. Refrigerants: There is still no vision for sustainable solutions. International Journal of Refrigeration, 2018; 86: 441-
448.
[6] Kim, B., Lee, S. H., Lee, D., & Kim, Y. Performance comparison of heat pumps using low global warming potential refrigerants with
optimized heat exchanger designs. Applied Thermal Engineering, 2020; 114990.
[7] Makhnatch, P., Mota-Babiloni, A., Rogstam, J., & Khodabandeh, R. Retrofit of lower GWP alternative R449A into an existing R404A
indirect supermarket refrigeration system. International Journal of Refrigeration, 2017; 76:184-192.
[8] Vaitkus, L., & Dagilis, V. Analysis of alternatives to high GWP refrigerants for eutectic refrigerating systems. International Journal
of Refrigeration, 2017; 76:160-169.
[9] Touaibi R, Köten H, Feidt M, Boydak O. Investigation of three Organic Fluids Effects on Exergy Analysis of a Combined Cycle:
Organic Rankine Cycle/Vapor Compression Refrigeration. Journal of Advanced Research in Fluid Mechanics and Thermal Sciences,
2018; 52(2): 232-245.
[10] Touaibi R, Feidt M, Vasilescu E. E, Abbes M. T. Parametric study and exergy analysis of solar water-lithium bromide absorption
cooling system. International Journal of Exergy, 2013; 13(3): 409-429.

518
 EXPERIMENTAL STUDIES OF NOVEL HYDROGEN
PRODUCTION SYSTEM INTEGRATED WITH NANOFLUID
BASED PVT SYSTEM
S.Senthilraja
SRM Institute of Science and Technology, Kattankulathur, India – 603203, [email protected]
ORCID: 0000-0002-3472-4467

R.Gangadevi
SRM Institute of Science and Technology, Kattankulathur, India – 603203, [email protected]
ORCID: 0000-0003-2327-9836

Hasan KÖTEN
Istanbul Medeniyet University, Istanbul 34700, TURKEY, [email protected],
ORCID: 0000-0002-1907-9420
P.Ravichandran
Department of Mechatronics Engineering, Kongu Engineering College, Erode, [email protected]
ORCID: 0000-0002-9896-219X

Senthilraja s, Gangadevi R, Hasan Koten, Ravichandran P. Experimental studies of novel

Cite this paper as: hydrogen production system integrated with nanofluid based PVT system. 8th Eur. Conf. Ren.
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The performance of a newly developed solar powered hydrogen production system with
CuO/water nanofluid was studied and the results are discussed in this article. In this study, the test
was carried out during the month of May from 8.00 to 16.00 in Chennai, India and the impact of
volume fraction on the PVT performance and hydrogen yield rate have been analyzed. Results
indicated that the nanofluid cooled PVT solar collector outperforms the water and air cooled
collectors having remarkable performance enhancement and hydrogen yield rate. Moreover, the
results showed that PVT solar collector with 0.2% volume fraction extracts more heat from the PV
module and produce more electrical energy. In comparison to PV module, the enhanced hydrogen
yield rate was 41.08%, 46.12% and 56.44% on air, water and 0.2% volume fraction of nanofluid
respectively.
Keywords: Hydrogen Yield rate, PVT solar collector, Electrolysis and Nanofluid
© 2020 Published by ECRES

Nomenclature
F Faraday’s Constant (96485˚C/mol)
ṁ mass flow rate of fluid‘F’ -.
Cp Specific heat of fluid
Toutand Tin temperature of fluid at inlet and outlet side of thermal collector respectively
G solar radiation
AC Area of the Thermal collector
VOC Open circuit coltage
ISC Short Circuit Current
AP Area of the PV module
R Gas Constant
Iel Input current to the electrolyzer
Ta Atmosphere Temperature
P Atmosphere Pressure
t Time duration
Z Excess Electrons

519
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION
Over the past few decades, our society is mainly depends on fossil fuel for energy production. As per the data,
around 81% of energy demand is satisfied by the fossil fuels [1]. However, due to the increase of vehicle
population and modernization of industries, the energy demand is increasing every day. On the other side,
depletion rate of fossil fuels and the emissions of greenhouse gases such as carbon dioxide and Nitrogen oxide
are increased rapidly [2,3]. To overcome these issues, many developed and developing countries have
accelerated the use of renewable energy sources in different applications. This leads the researchers to start their
research to use the hydrogen as secondary fuel in different applications [4 – 8].
Generally, hydrogen can be produced from water by using electrolysis process. Because of simple and cheap,
this is one of the most widely used method to produce hydrogen. In this method, the external power source is
required. Because of clean and huge available in nature, solar powered electrolysis method gained more attention
among the researchers [9 -12]. Generally, temperature is one of the prime factor that affects the power output of
PV module. In order to reduce the temperature of PV module, thermal collector with traditional heat transfer
fluid is attached with the rear side the PV module. Naturally, the traditional heat transfer fluids possess less heat
transfer characteristics than solids [13]. In 1995, SUS Choi, has developed a new kind of heat transfer fluid with
high heat transfer characteristics by suspending nanometer sized metal particle in base heat transfer fluid. This
special fluid is named as “Nanofluid” [14]. After that, researchers using different techniques studied the thermal
and physical properties of nanofluids. Example Vajjiha and Das [15], Prsakash et al. [16], Sahooli and Sabbaghi
[17], Babar and Ali [18], Sajid and Ali [19, 20], Babar et al. [21] studied the properties of different nanofluids.
From these studies they observed that the thermal conductivity of all nanofluids is directly affected by the
temperature and preparation time.
From the results of earlier studies, it is observed that the PVT system with nanofluid produces more power
output than the PV module. Also the hydrogen production rate increases with increase of power input. Till now,
few research articles have been published related to solar powered hydrogen production system. But none of the
researchers have produced the hydrogen with nanoparticle based PVT system. In this work, water based CuO
nanofluids with 0.05, 0.1 and 0.2% volume fractions are prepared and the thermophysical properties have
studied. Then the influence of preparation time, volume concentration and mass flow rate of the nanofluids on
the effectiveness of PVT system and hydrogen yield rate are also analysed.

2. EXPERIMENTAL PROCEDURE AND APPARATUS

2.1 Nanofluid Preparation
In order to prepare huge quantity of cuO/water nanofluid, two step method is used in this research. To prepare
nanofluid, CuO nanoparticles (daverage = 27nm, density (ρ) = 6310 kg/m3 and specific surface area (A) 29
m2/gm) is purchased from Sigma Aldrich Corporation, India. The required quantity of CuO particles to disperse
in base fluid is calculated by using the following equation. Then the required quantity of nanoparticles dispersed
in the water. In order to avoid the quick sedimentation of particles, the particle dispersed fluid was sonicated for
4hrs using ultrasonicatior. Then it was stirred for about 1 hour. Finally, the prepared nnaofluids were placed
without any disturbance for about 7 days, and then it was observed that very few particles were sediment at the
bottom of the beaker and the uniform dispersion of particle is also observed.
w cuo
 cuo
% Volume concentrat ion ()  100
w cuo w water

 cuo  water (1)
2.2 Experimental setup
An experimental setup with necessary elements as shown in Fig. 1 was fabricated in this research. This system
consists of one photovoltaic module and thermal collector module. The PV panel with the maximum power,
open circuit voltage and short circuit current of 50W, 21.5V and 3.25A respectively used in this research. The
thermal collector consists of thermal collector, fluid storage tank, pump, flow meter, flow regulating valve and
heat rejection unit. A copper tube with an outside diameter of 12.7mm and wall thickness of 1.22mm was used
to make the thermal collector. This copper tube is bent into spiral shaped and attached with the rear side of the
PV system. To avoid the heat loss, the thermal collector is covered with Rockwool insulation material. The
water/nanofluid storage tank with the capacity of 10 liter is used in this study. Because of high cost of
nanoparticle, the nanofluid is passed through the heat transfer unit and recirculated into the thermal collector.
The air is supplied into the system by using compressor. The Resistance Temperature Detector (RTD) sensors
measures the panel and fluid temperature and sends the values to the data storage system though data acquisition

520
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

module The solar radiation values are measured by using pyrometer. To start the electrolysis process, the
electrical output of the PV module is connected with the Hoffman’s Voltameter setup. The chemical reaction
for the electrolysis process is as follows [22].
1
H 2O  2 F  H 2  O2
2 (2)

Fig.1 Block diagram of experimental

3. DATA REDUCTION
The thermal, electrical and total efficiencies of the developed hybrid system can be computed by using the
equation 3, 4 and 5 respectively.

m C p Tout  Tin 
the 
G  Ac (3)
V I
ele  OC SC
G  Ap
(4)
total  ele  the (5)
The volume flow rate of the hydrogen can be calculated as
 R  I el  Ta  t 
VH 2  F  P  Z   ml 
QH 2    7.445I el  
t t  min  (6)

4. RESULT AND DISCUSSIONS

In this study, a novel PVT based hydrogen production system with different heat transfer fluids was developed.
The role of fluid flow rate, volume concentration of nanoparticles on the hybrid solar system performance and
the hydrogen production rate is investigated. This study was carried out in three phases. In first phase, the
different volume concentration of nanofluids were prepared then the thermal and physical properties were
analyzed. In the second phase, the performance of hybrid solar system with different fluids was studied. The
impact of nanofluids and fluid flow rate on the hydrogen output was observed in the third phase. The results
obtained in all these phases are discussed in this section.
Plotted in Fig. 2(a) is the thermal efficiency of PVT system as a function of volume fraction and flow rate.
According to this figure, thermal efficiency starts to increase from 8.00 and reaches the peak value at 12.00 then
it decreases till the day end (i.e.16.00). Furthermore, at all times, the lowest thermal efficiency values of are
recorded for air based PVT system with 20LPH flow rate. Similarly, the highest values are recorded for water
based PVT system with 0.2% volume fraction at 40LPH. AT 12.00, the thermal efficiency of PVT system
operated with air and water are noticed as 27.8 % & 30.3% for 20LPH and 31.1% & 33.8% for 40LPH
respectively. The addition of 0.05%, 0.1% and 0.2% (v/v) nanoparticles, improved the thermal efficiency to
57%, 68% & 70% for 20LPH and 62%, 71% & 79% for 40LPH respectively.
When the fluid contains less quantity of nanoparticles and circulated less fluids in the thermal collector, the
surface contact between the fluid and PV module surface is drastically reduced. On the other hand, at low volume

521
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

fraction, less number of particle chains are formed in the base fluid. Because of these reasons, less heat transfer
takes place between the fluid and the PV surface. This facts leads to increase the thermal efficiency with
increasing flow rate and volume fraction. Kumar et al. [23] also observed the similar kind of result in their
research.

Figure
2(a). Hourly thermal efficiency 2(b) Electrical efficiency variation for PVT system with different heat transfer fluids

Fig. 2(b) depicts the correlation between the electrical outputs produced by the PV module cooled with different
fluids. It can be realized form this figure, the peak electrical output is achieved at 12.00 local time then it
gradually decreased till the day end. The similar trend of results are obtained for all fluids at all times. The
obtained electrical output was found least on air cooled PV system at 20LPH and highest on 0.2% volume
fraction nanofluid cooled PV system with 40 LPH at all times. At 12.00 and 20LPH, the electrical output of the
air and water cooled PV system is 7.6% and 7.8% respectively. While the electrical output of the PV system
cooled with 0.05%, 0.1% and 0.2% nanofluids were 10%, 11.9% and 12.6% respectively. In the case of 40LPH,
the values are drastically increased to 8%, 8.5%, 11%, 12.6%, & 13.5% for air, water and .05%, 0.1% and 0.2%
nanofluids respectively
The variation of total efficiency with nanofluid volume fraction is depicted in Fig. 3(a). As expected from the
previous figure (Fig. 2(a) and Fig. 2(b)), the peak total efficiency of the PVT system cooled with 0.2%
concentration of nanofluid was observed 92.5% at 12.00 and the corresponding flow rate was 40LPH. The
maximum total efficiency observed in the case of 0.05% and 0.1% concentration of nanofluids at 40LPH were
found as 73% and 84% respectively. The observed maximum total efficiency of the air and water cooled PVT
system were as 39.1% and 42.3% respectively.

Fig. 3(a). Hourly Total efficiency 3(b) Hydrogen Yield rate variation for PVT system with different heat transfer fluids.

Fig. 3(b) shows the hourly changes of hydrogen generation rate for PVT system with different heat transfer
fluids at constant solar radiation. As seen in this figure, the hydrogen generation rate is gradually increase as the
time increased till Noon (i.e. 12.00) then it gradually decreased till the day end (i.e. 16.00). On the other hand,

522
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the rate of water splitting is increasing as the electrical efficiency of PVT system increased. Hence more amount
of hydrogen generation was obtained for PVT system cooled with 0.2% concentration of CuO nanofluids at
40LPH. At 12.00, the minimum hydrogen generation rate was found as 8.19 ml/min while using PV module
without thermal collector. However, the hydrogen generation rate was slightly increased and obtained as 13.9
ml/min and 15.2 ml/min while using air and water as heat transfer fluid in PVT solar collector at 40 LPH.
Obviously, the intensity of solar radiation is very less at the beginning of the experiment (i.e. 8.00), but at 12.00,
more heat is absorbed by the PV module and produce more electrical output. At the same time more solar
radiation is wasted as heat and accumulated in PV module. Hence the PV module efficiency is reduced
drastically. On the other hand, while circulating fluid with 0.2% volume CuO nanoparticles at 40LPH, more
fluid and solid particle interaction is increasing and decreasing the PV module temperature noticeably. These
are the possible reasons to supply more electrical power to the electrolyzer and achieve more hydrogen
generation rate.
5. Conclusions
The impact of nanofluids on a solar powered hydrogen production system was analyzed experimentally at
constant solar radiation. The tests were conducted continuously for three days by circulating different heat
transfer fluids. The important outcomes of this research have been drawn
1. The observed thermal conductivity ratio (Knf/Kf) due to the addition of 0.05, 0.1 and 0.2% volume
fraction of CuO nanoparticles into water had been 1.03, 1.037 and 1.044 at 4 Hrs sonication time
respectively.
2. Similarly, the obtained viscosity (μnf) for the same nanofluids at 4 Hrs preparation time had been 0.75,
0.78 and 0.8 mPa-S respectively.
3. With the increase in temperature and flow rate, the solid and fluid interaction increased, and the
convection heat transfer also increased. Thus the peak thermal and electrical output was obtained as
79% and 13.5% respectively for PVT system with 0.2% volume fraction of nanofluids prepared with
4Hrs preparation time.
4. Hydrogen generation rate of about 10.3%, 12.1% and 16.5% were obtained for solar powered hydrogen
production system with air, water and 0.2% volume fraction of nanofluid at 20LPH respectively. While
using the fluid at 40LPH, the yield rate increased to 13.9%, 15.2% and 18.8% respectively.
Thus the naofluid enhance the performance of PVT system and hydrogen yield rate significantly. The hydrogen
yield rate can be further improved by using different nanoparticles with different heat transfer fluids.

REFERENCES

[1]. Bank World. Fossil fuel energy consumption. Data, energy Foss fuels Consum. 2016. Available from: http://data.
worldbank.org/indicator/EG.USE.COMM.FO.ZS.
[2]. Muradov, NZ, Veziroğlu, TN, Green path from fossil-based to hydrogen economy: an overview of carbon-neutral
technologies. International journal of hydrogen energy 2008. 33(23): 6804-6839.
https://doi.org/10.1016/j.ijhydene.2008.08.054
[3]. Jacobson, MZ. Review of solutions to global warming, air pollution, and energy security. Energy & Environmental
Science 2009. 2(2): 148-173. DOI: 10.1039/B809990C
[4]. Karagöz, Y, Balcı, Ö, Köten, H. Investigation of hydrogen usage on combustion characteristics and emissions of a
spark ignition engine. International Journal of Hydrogen Energy 2019. 44(27): 14243-14256.
https://doi.org/10.1016/j.ijhydene.2019.01.147
[5]. Demirci, A, Koten, H, Gumus, M. The effects of small amount of hydrogen addition on performance and emissions of
a direct injection compression ignition engine. Thermal science 2018. 22(3): 1395-1404.
https://doi.org/10.2298/TSCI170802004D
[6]. Wang, J, Huang, Z, Tang, C, Miao, H, Wang, X, Numerical study of the effect of hydrogen addition on methane–air
mixtures combustion. International journal of hydrogen energy 2009. 34(2): 1084-1096.
https://doi.org/10.1016/j.ijhydene.2008.11.010
[7]. Vancoillie, J, Demuynck, J, Sileghem, L, Van De Ginste, M, Verhelst, S, Comparison of the renewable transportation
fuels, hydrogen and methanol formed from hydrogen, with gasoline–Engine efficiency study. International journal of
hydrogen energy 2012. 37(12): 9914-9924. https://doi.org/10.1016/j.ijhydene.2012.03.145
[8]. Ji, C, Zhang, B, Wang, S, Enhancing the performance of a spark-ignition methanol engine with hydrogen
addition. International Journal of hydrogen energy 2013. 38(18): 7490-7498.
https://doi.org/10.1016/j.ijhydene.2013.04.001
[9]. Calderón, M, Calderón, AJ, Ramiro, A, González, J.F, González, I. Evaluation of a hybrid photovoltaic-wind system
with hydrogen storage performance using exergy analysis. International journal of hydrogen energy 2011. 36(10):
5751-5762. https://doi.org/10.1016/j.ijhydene.2011.02.055

523
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[10]. Akyuz, E, Oktay, Z, Dincer, I, Performance investigation of hydrogen production from a hybrid wind-PV
system. International journal of hydrogen energy 2012. 37(21): 16623-16630.
https://doi.org/10.1016/j.ijhydene.2012.02.149
[11]. Abanades, S, Flamant, G, Thermochemical hydrogen production from a two-step solar-driven water-splitting cycle
based on cerium oxides. Solar energy 2006. 80(12): 1611-1623. https://doi.org/10.1016/j.solener.2005.12.005
[12]. Pregger, T, Graf, D, Krewitt, W, Sattler, C, Roeb, M, Möller, S, Prospects of solar thermal hydrogen production
processes. International journal of hydrogen energy 2009. 34(10): 4256-4267.
https://doi.org/10.1016/j.ijhydene.2009.03.025
[13]. Senthilraja, S, Karthikeyan, M, Gangadevi, R, Nanofluid applications in future automobiles: comprehensive review
of existing data. Nano-Micro Letters 2010. 2(4): 306-310. doi:10.3786/nml.v2i4.p306-310
[14]. Choi, SU, Eastman, JA, Enhancing thermal conductivity of fluids with nanoparticles (No. ANL/MSD/CP-84938;
CONF-951135-29). Argonne National Lab., IL (United States) 1995.
[15]. Vajjha, RS, Das, DK, Experimental determination of thermal conductivity of three nanofluids and development of
new correlations. International Journal of Heat and Mass Transfer 2009. 52(21-22): 4675-4682.
https://doi.org/10.1016/j.ijheatmasstransfer.2009.06.027
[16]. Prakash, V, Diwan, RK, Characterization of synthesized copper oxide nanopowders and their use in nanofluids for
enhancement of thermal conductivity. Indian Journal of Pure & Applied Physics (IJPAP) 2015. 53(11): 753-758.
[17]. Sahooli, M, Sabbaghi, S, Investigation of thermal properties of CuO nanoparticles on the ethylene glycol–water
mixture. Materials Letters 2013. 93: 254-257. https://doi.org/10.1016/j.matlet.2012.11.097
[18]. Babar, H, Ali, HM, Towards hybrid nanofluids: preparation, thermophysical properties, applications, and
challenges. Journal of Molecular Liquids 2019. 281: 598-633. https://doi.org/10.1016/j.molliq.2019.02.102
[19]. Sajid, MU, Ali, HM, Recent advances in application of nanofluids in heat transfer devices: a critical
review. Renewable and Sustainable Energy Reviews 2019. 103: 556-592. https://doi.org/10.1016/j.rser.2018.12.057
[20]. Sajid, MU, Ali HM. Thermal conductivity of hybrid nanofluids: a critical review. International Journal of Heat and
Mass Transfer 2018. 126: 211-234. https://doi.org/10.1016/j.ijheatmasstransfer.2018.05.021
[21]. Babar, H, Sajid, MU, Ali, HM. Viscosity of hybrid nanofluids: a critical review. Thermal Science 2019. 23: 1713-
1754. https://doi.org/10.2298/TSCI181128015B
[22]. Bianchini, C. and Barbaro, P. eds., 2009. Catalysis for sustainable energy production. Wiley-Vch.
[23]. Kumar, A, Kumar, S, Yadav, A. Thermal performance analysis of evacuated tubes solar air collector in Indian climate
conditions. International Journal of Ambient Energy 2016. 37(2): 162-171.
https://doi.org/10.1080/01430750.2014.915884

524
 A WIND EMULATOR ASSOCIATED WITH DOUBLY-FED
INDUCTION GENERATOR

Hallouz Mohamed*
[email protected],ORCID: 0000-0002-0815-8553

Kabeche Nadir*
[email protected], ORCID: 0000-0002-3505-8873

Moulahoum Samir*
[email protected], ORCID: 0000-0002-7482-0621

*Laboratory of Electrical Engineering and Automatic*University of Medea, Algeria

Hallouz, Mohamed, Kabache, Nadir, Moulahoum Samir. A wind emulator associated with
Cite this paper as: doubly-fed induction generator. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,
Turkey

Abstract: While analyzing a wind energy conversion system (WECS), the accurate information about the
wind behavior and speed is the important key to achieve an appropriate control diagram for the
whole system. In this purpose, the present paper involves the development and the use of an
emulator for a wind turbine to simulate its real behavior. The emulator associated with a DFIG is
analyzed and tested by simulation and experimentation for different operating modes. .
Keywords: Field Oriented Control, MPPT, Wind Emulator
© 2020 Published by ECRES

1. INTRODUCTION

Wind energy considered as a free and sustainable source of clean energy, it develops at high speed over time
using different control techniques and different generators. In recent years, the wind energy conversion systems
(WECS) deployed are systems based on variable speed operation in order to capture the maximum energy wind,
the generator adapted to this operating mode is the doubly fed induction generator (DFIG) given its operation in
a wide speed variation, unlike asynchronous cage machines or synchronous machines which are forced to
operate at the speed of synchronism or in its vicinity. The interest of speed variation is to extract the maximum
power [1]–[5].
In this paper, we have made a study on an experimental test interpreting a chain of the WECS which is composed
of a wind emulator based on a Dc motor, and a model of DFIG executed in simulation. To ensure a proper
operation at variable speed and to extract the maximum power, the MPPT control strategy with speed control is
introduced with the emulator to monitor the point of maximum power. The mechanical speed can be controlled
with different regulators. The famous field oriented control (FOC) is then used to control the active and reactive
powers of the stator, we can use another strategies that can achieve the independent control between the active
and reactive powers, like the DPC or DTC controls [2], [6].

2. WECS MATHEMATICAL MODEL

The topology of the WECS for performing variable speed operation is shown in figure (1). For the next step, we
will develop the different parts of the system through mathematical equations as well the applied control
strategies.
The first part is the wind turbine that allows the conversion of wind aerodynamic power into mechanical power
represented by the rotary movement of the mechanical shaft [6].

525
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In this paper, an emulator based on a DC motor is used to reproduce the aerodynamic of the true wind turbine
and thus drive the generator (Fig.2).

Figure 1. Global schem of WECS based on DFIG.

For the model of the emulator (DC motor), the electric behavior is given as follows:
𝑑𝑖
𝑈 = 𝑅. 𝑖 + 𝐿 +𝐸 (1)
𝑑𝑡

For the mechanical part, the behavior is given by:

𝑑Ω
𝐶𝑒𝑚 − 𝐶𝑟 = 𝑗 𝑑𝑡 + 𝑓Ω (2)

With:

𝐶 = 𝑘𝑒 . 𝑖
{ 𝑒𝑚 (3)
𝐸 = 𝑘𝑒 . Ω

Figure 2. Wind emulator model.

The operation of the wind emulator occurs by capturing the mechanical speed to generate an aerodynamic torque
which then represents the image of the reference current of the Dc motor; this process allows us to emulate the
real behavior of a wind turbine.

526
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Block diagram of the DFIG model.

The second part is the DFIG that can be considered as a highly non-linear system. According to the rotating
frame its simplified model for active and reactive powers is illustrated on the figure (3) (for more details see
[7]).

3. FIELD ORIENTED CONTROLOF THE DFIG

The DFIG active and reactive powers are controlled by using a FOC diagram. The decoupling between the
electromagnetic torque and the flux allows controlling independently the active power (Ps) and the reactive
power (Qs). The controlled variables can be expressed then as fellow:
𝑀𝑠𝑟
𝑃𝑠 = −𝑉𝑠 𝑖𝑞𝑟
𝐿𝑠
𝑉𝑠2 𝑉𝑠 𝑀𝑠𝑟
𝑄𝑠 = ꞷ − 𝑖𝑑𝑟 (4)
𝑠 𝐿𝑠 𝐿𝑠
𝑀𝑠𝑟
{ 𝐶𝑒𝑚 = −𝑃 𝐿𝑠
𝜙𝑠 𝑖𝑞𝑟

4. EXPERIMENTAL TEST SETUP

The experimental test is carried out by a 1.5kw DC motor as emulator and a DFIG as being a generator. The
MPPT control strategy is included in the emulator control diagram to obtain an optimal speed ration and
thereafter the extraction of the maximum of the power (Fig.4). After applying a wind profile and capturing the
mechanical speed, the MPPT block with speed control allows us to generate an electromagnetic torque
representing a reference current for the DC motor-based emulator. The emulator is controlled a current standard
PI via a dSPACE-1104 device. The device sends the control signals to the IGBT switches constituting the arms
of the Dc-Dc converter which supplies the Dc motor allowing it the to operate in its overall four quadrants.

5.RESULTS INTERPRETATION

The study is carried out on the WECS by simulation and experimentation. The obtained results show a good
according between simulation results (Fig. 5-11) and experimentation results (Fig. 12-17).
Figures (6 and 12) represent the speed of the mechanical shaft of the system, we can see that the values captured
follow perfectly their references generated by the MPPT strategy. By controlling the speed of rotation, we obtain
an electromagnetic torque that is the image of the DC motor reference current (Fig. 7, 13) which is followed by
the measured current. The speed ratio (Fig. 9, 15) and the power coefficient (Fig. 8, 14) take constant values for
any variation of wind speed allowing power maximization. For the DFIG performances, the FOC (with stator
flux orientation) that was applied for the control of stator active and reactive powers, the controlled variables
fellow their references in both simulation (Fig. 10, 11) and experimental test (Fig. 16, 17).

527
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 4. Global model of wind energy converter system.

Figure 5. Wind profil.

Figure 6. Mechanical speed. Figure 7. DC motor current

528
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 8. Power coefficient. Figure 9. Optimal speed ratio.

Figure 10. Stator active power. Figure 11. Stator reactive power

.
Figure 12. Mechanical speed. Figure 13. DC motor current.

Figure 14. Power coefficient. Figure 15. Optimal speed ratio.

529
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 16. Stator active power. Figure 17. Stator reactive power.

6. CONCLUSION

In this paper, a wind energy conversion systems (WECS) was investigated. The system consisted of a DC motor
driven by DC-DC converter via an 1104-dSPACE device card which allows emulating the turbine in reality, in
association with a mathematical model of DFIG through MATLAB/SIMULINK software operating as a
generator. An MPPT control strategy with speed control is applied to the emulator to extract the maximum
power whatever the wind speed. The FOC was used for the monitoring of The DFIG active and reactive powers.
The results obtained throw both simulation and experimental test are satisfactory for the emulator as for the
DIFG. Therefore, the perspectives can be extended toward the implementation of the whole WECS control
scheme.

REFERENCES

[1] T. K. Roy, M. A. Mahmud, et A. M. T. Oo, « Nonlinear backstepping controller design for improving fault ride through
capabilities of DFIG-based wind farms », in 2018 IEEE Power & Energy Society General Meeting (PESGM), 2018, p.
1–5.
[2] A. G. Abo-Khalil et al., « Design of State Feedback Current Controller for Fast Synchronization of DFIG in Wind Power
Generation Systems », Energies, vol. 12, no 12, p. 2427, 2019.
[3] M. A. Kashkooli, S. M. Madani, et R. Sadeghi, « Improved direct torque control of DFIG with reduced torque and flux
ripples at constant switching frequency », in 2016 7th Power Electronics and Drive Systems Technologies Conference
(PEDSTC), 2016, p. 58–63.
[4] A. Mohammedi, N. Kabache, S. Moulahoum, et H. Houassine, « FPGA hardware in the loop validation of direct torque
control for induction motor », in Methods and Models in Automation and Robotics (MMAR), 2015 20th International
Conference on, 2015, p. 812–816.
[5] M. Hallouz,S. Moulahoum, N. Kabache, et S. Kouadria, « Neural Network Based Field Oriented Control for Doubly-
Fed Induction Generator », Int. J. Smart Grid-IjSmartGrid, vol. 2, no 3, p. 183–187, 2018.
[6] M. Hallouz, N. Kabache, et S. Moulahoum, « DTC Backstepping Based Control of Double Fed Induction Generator »,
in 2019 Progress in Applied Electrical Engineering (PAEE), 2019, p. 1–6.
[7] M. DOUMI, A. G. AISSAOUI, A. TAHOUR, M. ABID, et K. TAHIR, « Nonlinear integral backstepping control of
wind energy conversion system based on a Double-Fed Induction Generator », Przegląd Elektrotechniczny, vol. 92, no
3, p. 130–135, 2016.
[8] C. Saudemont, L. Leclercq, B. Robyns, G. Cimuca, et M. M. Radulescu, « Développement d’un émulateur temps réel
d’un système de génération éolienne associé à un stockage inertiel d’énergie », Rev. REE, no 11, p. 49–59, 2004.

530
 STUDY OF THE EFFECT OF PREPARATION PARAMETERS ON
THERMAL CONDUCTIVITY OF METAL OXIDE NANOFLUIDS
USING TAGUCHI METHOD

Nikhil S. Mane
Department of Mechanical Engineering, BITS Pilani K K Birla Goa Campus, Sancoale, Goa, India,
[email protected], ORCID: 0000-0002-9938-6841

Vadiraj Hemadri
Department of Mechanical Engineering, BITS Pilani K K Birla Goa Campus, Sancoale, Goa, India,
[email protected], ORCID: 0000-0003-1353-2746

Mane, SN, Hemadri, VA. Study of the effect of preparation parameters on thermal
Cite this paper as: conductivity of metal oxide nanofluids using Taguchi method. 8th Eur. Conf. Ren. Energy Sys.
24-25 August 2020, Istanbul, Turkey

Abstract: In thermal applications, the nanofluids are used because they provide higher heat transfer than
their base fluid. The nanofluid preparation employs different processes and is usually
complemented with the addition of surfactants in different quantities to maximize stability. In
this work, the effect of varying the surfactant material, surfactant weight and ultrasonication time
is studied on CuO-water, Fe3O4-water, and CuO+ Fe3O4-water nanofluids. Taguchi L9
orthogonal array is used for the design of the experiment and 9 samples are prepared using this
array. The thermal conductivity of each sample is measured and the effect of each level of the
process parameter on thermal conductivity is analyzed by calculating Signal to Noise Ratio
(SNR). The Analysis of Variance (ANOVA) is also conducted in this study to explore the
contribution of each parameter in the results of thermal conductivity. The results explore the
effect of each process parameter on the thermal conductivity and optimum levels of these
parameters are identified. The ANOVA shows that stability plays a crucial role in the thermal
conductivity and is as important as nanoparticle material for delivering high thermal conductivity
nanofluids.
Keywords: Nanofluids, Thermal conductivity, ANOVA, Taguchi analysis
© 2020 Published by ECRES

Nomenclature
ANOVA Analysis of Variance
CA Citric Acid
CNT Carbon nanotubes
CTAB Cetyl Trimethyl Ammonium Bromide
OA Orthogonal array
SDS Sodium dodecyl sulfate
SNR Signal Noise ratio

1. INTRODUCTION

The nanofluids offer many thermal, chemical and practical advantages in the field of medicine, engineering and
thermal applications. In thermal applications, users are generally interested in the thermal advantages offered by
the nanofluids in the form of properties like thermal conductivity, viscosity, wettability, and specific heats. The
nanofluids used in thermal applications provide higher thermal conductivity than their base fluids [1]. In thermal
applications, CNT, Cu, CuO, Al2O3, Fe3O4 are the most commonly used nanoparticles used in the aqueous
base fluid. The higher thermal conductivity and specific heat of the nanofluids provide higher heat transfer in
applications like heat pipes [2], heat exchangers [3] and IC engines [4,5]. Many researchers have ascribed the

531
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

improvement in thermal conductivity to the presence of solid particles in the nanofluid in the form of
nanoparticles [6], though most researchers agree on this; some of them argue that the behavior of the nanofluid
is rather complex and multiple mechanisms are involved. The thermal conductivity mechanism of nanofluids
can be categorized into two components viz. static thermal conductivity and Brownian thermal conductivity.
The increase in static thermal conductivity can be attributed to various mechanisms as discussed by different
researchers. One of the less discussed mechanisms is liquid layering of particle surface also known as interfacial
nano-layer. The liquid molecules form a very thin but organized layer around nanoparticles. Henderson et al. [7]
conducted a study on fluid and wall interface based on which the authors suggested that there is a thin layers of
fluid on the surface of nanoparticle, the first layer has thickness less than half the diameter of the nanoparticle.
This liquid layer has a higher thermal conductivity because of the organized molecules around the solid particles.
The thermal conductivity of the layer decreases along the thickness of the layer away from the particles surface
[8]. When these successive layers are formed around the particles in the nanofluid, a collective high thermal
conductivity volume is formed in the nanofluid which increases the effective thermal conductivity. The increase
in static thermal conductivity is also sometimes attributed to thermal diffusion of heat in close particles and
clusters of nanoparticles. The Brownian motion of the nanoparticles in the liquid results in collision, which in
turn results into solid to solid contact. The solid to solid contact between the nanoparticles allows heat transfer
through conduction which leads to an increase in the overall thermal conductivity of the nanofluid.
The stability is another important aspect of the nanofluids. The stability of the nanofluid is characterized by zeta
potential, hydrodynamic diameter and visual observations. The nanoparticles in any nanofluid tend to
agglomerate and settle down in the fluid over a period of time; this is called sedimentation of the nanofluids.
Due to this, the nanofluids lose their advantage and show thermal properties similar to that of the base fluid [9].
Hence in order to improve the stability of nanofluids and to avoid sedimentation, surfactants are added to the
nanofluid. The surfactant molecules consist of a hydrophobic tail and a hydrophilic head. When the surfactant
is mixed in nanofluid the hydrophobic tail attaches itself to the nanoparticle, while the hydrophilic head is in
contact with water. This makes a cluster of nanoparticle and surfactant molecules which is hydrophilic [10]. The
addition of surfactant also increases the surface charge which leads to an increase in repulsive forces between
the nanoparticles, which collectively results in the enhancement of the stability of the nanofluid [11]. Many
studies have shown that the surfactant, in addition to increasing the stability, also affects the thermal properties
of nanofluids. Xuan et. al. [12] performed an experimental investigation to study the effect of SDBS surfactant
on the heat transferability of the Cu-water nanofluid. The findings of this study show that the presence of SDBS
reduces the heat transfer coefficient in jet impingement heat transfer. The heat transfer further reduces with
increase in the mass concentration of the surfactant. The ultrasonication of nanoparticles is also employed to
improve the stability of the nanofluids [13, 14]. It works by giving physical shocks to agglomerates formed by
nanoparticles and physically breaks them into smaller particles.
As discussed earlier, the rise in thermal conductivity is outcome of different mechanisms. As the Brownian
thermal conductivity and liquid solid interface layering mechanism are crucial for stable nanofluids to function
effectively; the thermal conductivity is also indirectly dependent on the stability of nanofluids. In this paper, the
evaluation of the process parameters and their effect on thermal conductivity of nanofluids is conducted using
the Taguchi method. This paper presents the effect of stability contributors (surfactant material, surfactant
weight and ultrasonication time) and nanoparticle material (CuO, Fe3O4, and CuO+Fe3O4) on the thermal
conductivity of nanofluids. This paper also highlights the role of stability in the thermal conductivity of
nanofluids.

2. NANOFLUID PREPARATION

The different surfactant material, weight and sonication time are used for the preparation of water based CuO,
Fe3O4 and hybrid CuO+Fe3O4 nanofluids. The weight of all the three nanoparticles is maintained at 0.1% of
base fluid weight. Three surfactants; Citric Acid (CA), Sodium dodecyl sulfate (SDS) and Cetyl Trimethyl
Ammonium Bromide (CTAB), three quantities of these surfactants; 0.05%, 0.1% and 0.5% of base fluid weight
and three ultrasonication times; 15 minutes, 30 minutes and 60 minutes, are selected as preparation parameters
of the nanofluids. Based on L9 Taguchi Orthogonal Array (OA) 9 samples are prepared using these preparation
parameters. The nanofluid is prepared by mixing water with nanoparticles using a stirrer with a speed of 500
rpm and ultrasonication. The ultrasonication frequency is maintained at 20 kHz with 400-Watt power.

532
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. STABILITY OF NANOFLUIDS

The stability of nanofluids is studied both by measuring its turbidity over a period and by visual observations.
Table 1. Stability of Nanofluid samples
Sample Composition Turbidity Visual Observations
No. 0 minutes 360 minutes
1 CuO + CTAB (0.05%) + 15 64
after 801
after Stable for more than 24 hours
2 Fe3O4 + CTAB
minutes (0.1%) + 30
sonication 241
preparation 343
preparation Stable for more than 12 hours
3 (CuO +minutes
Fe3O4)sonication
+ CTAB (0.1%) 38 599 Stable for more than 24 hours
4 (CuO+ +60Fe3O4)
minutes+ sonication
CA (0.05%) + 71 26  Sedimentation initiates within the 2
5 CuO 30
+ CA (0.1%)
minutes + 60 minutes
sonication 97 50  Sedimentation initiates within the 45
hours after preparation
6 Fe3O4 +sonication
CA (0.5%) + 15 134 71  Sedimentation
minutes
Completeaftersettling
preparation
initiates
of within the 2
nanoparticles
7 Fe3O4minutes
+ SDSsonication
(0.05%) + 60 204 213  hours
Stable
takesfor more
after
Complete
place than
settling
in 9 hours24 of
preparation hours
nanoparticles
8 CuO + minutes
SDS (0.1%) + 15 minutes
sonication 46.89 768  Complete
Stable
takesfor more
place than
settling
in 6 hours24 of
hours
nanoparticles
9 (CuO + Fe3O4) + SDS (0.5%) +
sonication 37.55 781 Stable
takes place in 9 hours hours
for more than 24
30 minutes sonication
It can be observed from Table 1, that Sample 1, 2, 3, 7, 8 and 9 shows higher turbidity than Sample 4, 5 and 6.
Visual observations indicate that the samples having low turbidity at 360 minutes after the preparation are
settling down rapidly and visibly show high rate of sedimentation.

5. MEASUREMENT OF THERMAL CONDUCTIVITY

6.
The jacketed beaker and constant temperature bath are used to maintain the temperature of the nanofluid at
250C. Table 2 shows the specifications of the components used in this experimental setup. The KS-1 sensor is
dipped into the nanofluid and readings are taken. In order to increase the accuracy of measurement, the multiple
readings are taken and 15 minutes of the period is maintained between two readings.
Table 2. Specifications of Components of Experimental Setup
Components Variant/Model Make Accuracy
Thermal Property Analyzer KD2 Pro with KS-1 Sensor Decagon Devices ±5%
Constant Temperature Bath F25 Julabo 0.010C
Jacketed beaker - Optech -

5. TAGUCHI ANALYSIS

Taguchi analysis allows the extraction of maximum data from minimum number of experiments. Whenever a
large number of experiments are required to be performed to study some complex phenomenon, the Taguchi
method is often utilized. It is a statistical tool which provides a sophisticated method to for the design of
experiments. In Taguchi analysis, experiments are designed using Orthogonal Arrays (OA). These orthogonal
arrays give a specific arrangement of process parameters. This arrangement is such that it gives the minimum
number of experiments to be performed and while ensuring that all the necessary combinations of process
parameters are covered in the design. The output of the experiments cannot be interpreted directly as a common
baseline to compare each level of process parameter is generally not available. Hence a signal to noise ratio
(SNR) is calculated to interpret the results of these experiments [15]. The performance of each level of operating
parameter is evaluated using SNR and a higher value of SNR shows a higher impact on the output of the
experiment. In this work, surfactant material, surfactant weight, ultrasonication time and nanoparticle material
are selected as the process parameters and three levels of each process parameters are taken. Based on the number
of parameters and levels, L9 orthogonal array is selected for the design of experiments. As higher values of
thermal conductivity are desired from the nanofluids, the ‘Larger the Better’ category is selected for the analysis.
The calculated SNR for each process parameters is shown in Figure 2.

533
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) Surfactant Material (b) Surfactant Weight

(c) Sonication Time (d) Nanoparticle Material

Figure 1. Effect of process parameters on Thermal Conductivity

Figure 1 (a) shows that CA shows the lowest SNR for thermal conductivity, and nanofluid samples with CTAB
as surfactant shows highest SNR. The average value of thermal conductivity for nanofluids with CTAB and SDS
is respectively 3.28% and 2.9% higher than nanofluids with CA. This indicates that for high thermal conductivity
CTAB is best performing surfactant material in this study. The nanofluid samples with CA shows lowest thermal
conductivity and shows merely 0.5% enhancement in thermal conductivity than that of water. The nanofluid
samples prepared using CA are also least stable nanofluids and hence the lower thermal conductivity is the result
of agglomeration and rapid settlement of nanoparticles in the base fluid.
The effect of surfactant quantity on the thermal conductivity can be observed from the mean value of SNR in
Figure 1(b). No considerable change is observed in the SNR for thermal conductivity with surfactant quantity.
Figure 1(b) shows that the increase in the quantity of surfactant weight does not guarantee specific enhancement
in the thermal properties. As the chemistry of the nanofluids is much more complex, the mere quantity of
surfactant is not absolutely important in the nanofluid composition. The optimum quantity of surfactant is
dependent on the surfactant material, nanoparticle shape and size, and the base fluid.
Figure 1(c) shows that, with increase in ultrasonication time the mean value of SNR decreases. This implies that
with the increase in ultrasonication time the thermal conductivity of the nanofluid decreases. This is because the
increase in ultrasonication time exposes nanoparticles to excessive ultrasonic vibrations which lead to the
removal of the surfactant layer from the surface of the nanoparticles. The absence of surfactant layer on the
nanoparticle surface increases the agglomeration and reduces the Brownian motion of the nanoparticles. This
collectively leads to a decrease in the thermal conductivity of the nanofluid. It can be observed that the thermal
conductivity of nanofluid is a direct product of the thermal conductivity of the nanoparticle material. Figure 1(d)
shows that the CuO-water nanofluid has the highest SNR for thermal conductivity as CuO material has higher
thermal conductivity than Fe3O4 and a mixture of CuO+Fe3O4 material. The CuO-water nanofluids have
highest thermal conductivity and Fe3O4-water nanofluids shows lowest thermal conductivity in this study. It is

534
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

found that, the average value of thermal conductivity of CuO-water, Fe3O4-water and CuO+Fe3O4-water
nanofluid is higher than that of the base fluid by 3.65%, 1.19% and 2.62%, respectively.

6. ANALYSIS OF VARIANCE (ANOVA)

From Taguchi analysis and Analysis of variance (ANOVA), the best performing levels of each parameters and
their effect on the thermal conductivity of the nanofluid is found out and listed in the Table 3.
Table 3. Optimum Conditions
Process Parameter Optimum Level Effect on Thermal Conductivity

Surfactant Material CTAB Dominant

Surfactant Weight (%) 0.5 Significant
Sonication Time (minutes) 15 Dominant
Nanoparticle Material CuO Dominant

Figure 2. Effect of process parameters on thermal Figure 3. Comparison of impact of stability

conductivity contributors and nanoparticle material on thermal
conductivity

Analysis of variance (ANOVA) is conducted to investigate the individual contribution of surfactant material,
surfactant weight, ultrasonication time and nanoparticle material on the thermal conductivity of the nanofluid.
Figure 2 show that the surfactant material has maximum impact on the thermal conductivity which is almost
twice more than the material of nanoparticles.
Even though apparently it looks very self-contradictory that the surfactant material has a higher contribution to
the thermal conductivity than the material; this it can be easily explained from a stability point of view.
Considering the complex behavior of nanofluids, it is not correct to conclude that the surfactant material,
surfactant weight and ultrasonication time of nanofluid preparation have a direct impact on the thermal
conductivity of the nanofluid. The surfactant material, surfactant weight and ultrasonication time have a more
direct impact on the stability of the nanofluids and hence the dominant effect of these factors on the thermal
conductivity is a product of the stability of nanofluids.
The optimum levels of the surfactant material, surfactant weight and ultrasonication time provide stable
nanofluids and thus provide higher thermal conductivity of the nanofluid sample compared to a less stable
sample of the same nanofluid material. The visual observations of the nanofluid stability also suggest that the
stable nanofluid provides considerably higher thermal conductivity than less stable nanofluids of same material.
From all these observations and results it can be concluded that the stability of nanofluid has a high impact on
the thermal conductivity. As the surfactant material, surfactant weight and ultrasonication time plays direct role
in the stability of the nanofluid, these parameters can be considered as stability parameters. From Figure 3 it can
be observed that the stability is more influential on the thermal conductivity of the nanofluid than the material
of the nanofluid.

535
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

CONCLUSIONS

In this paper, Taguchi analysis of thermal conductivity of nanofluids is conducted using surfactant material,
surfactant weight, ultrasonication time and nanoparticles materials as process parameters. This approach
explores the role of preparation parameters and stability in the thermal conductivity of nanofluids. It is found
that CuO nanoparticle with CTAB surfactant and preparation parameters of 0.5% surfactant weight and 15
minutes ultrasonication has the best thermal conductivity. Taguchi analysis shows that the CuO-water nanofluid
delivers the highest thermal conductivity compare to Fe3O4-water and CuO+Fe3O4-water nanofluids. The
nanofluid samples with CTAB and SDS surfactant show higher turbidity than the nanofluids with CA surfactant,
it is also observed that the SNR for CA is considerably lower than the CTAB and SDS. The stable nanofluids
with CTAB and SDS surfactant provide higher thermal conductivity than the unstable nanofluids with CA
surfactants.

The results of ANOVA shows that the nanoparticle material is not solely responsible for the thermal conductivity
of the nanofluids, but surfactant material, quantity and ultrasonication time collectively play an overwhelmingly
dominant role. This work also highlights that though it is statistically observed that the surfactant material,
quantity and ultrasonication time play a dominant role in the thermal conductivity of nanofluid, they do not
directly influence the thermal conductivity. Their effect is actually manifested as an improvement in stability
which in turn results in the overall enhancement of the thermal conductivity of the nanofluid.

ACKNOWLEDGMENT

We acknowledge the funding provided under Additional Competitive Research Grant by BITS Pilani K.K.
Birla Goa Campus for this work.

REFERENCES:
[1] Kleinstreuer C, Feng Y. Experimental and theoretical studies of nanofluid thermal conductivity enhancement: a review.
Nanoscale Res Lett 6. 2011;229.
[2] Sureshkumar R, Tharves Mohideen S, Nethaji N. Heat transfer characteristics of nanofluids in heat pipes: A review.
Renewable and Sustainable Energy Reviews. 2013;20:397-410.
[3] Kulkarni D, Das D, Vajjha R. Application of nanofluids in heating buildings and reducing pollution. Applied Energy.
2009;86:12:2566-2573.
[4] Senthilraja S, Karthikeyan M, Gangadevi R. Nanofluid Applications in Future Automobiles: Comprehensive Review
of Existing Data. Nano-Micro Lett. 2, 306–310.
[5] Sidik N, Yazid M, Mamat R. A review on the application of nanofluids in vehicle engine cooling system, International
Communications in Heat and Mass Transfer. 2015;68:85-90.
[6] Sarviya R, Fuskele V. Review on Thermal Conductivity of Nanofluids, Materials Today: Proceedings. 2017;4:4022-31.
[7] Henderson J.R., Frank S. On the interface between a fluid and a planar wall, Molecular Physics: An International Journal
at the Interface Between Chemistry and Physics. 1984;51:4:991-1010.
[8] Rizvi I., Jain A., Ghosh S., Mukherjee P.S. Mathematical modelling of thermal conductivity for nanofluid considering
interfacial nano-layer. Heat Mass Transfer. 2013;49:595–600.
[9] Kouloulias K., Sergis A., Hardalupas Y. Sedimentation in nanofluids during a natural convection experiment.
International Journal of Heat and Mass Transfer. 2016;101:1193-1203.
[10] Slamet, Redjeki, A. S. Interaction between surfactant and titania in a detergent nanofluid system. 2017.
doi:10.1063/1.5011920.
[11] Choudhary R, Khurana D, Kumar A, Subudhi S. Stability analysis of Al2O3/water nanofluids. Journal of Experimental
Nanoscience. 2017;12-1:140-51.
[12] Xuan Y., Li Q., Tie P. The effect of surfactants on heat transfer feature of nanofluids. Experimental Thermal and Fluid
Science. 2013;46:259–262.
[13] Mahbubul I, Saidur R, Amalina M, Elcioglu E, Okutucu-Ozyurt T. Effective ultrasonication process for better colloidal
dispersion of nanofluid. Ultrasonics Sonochemistry. 2015;26:361-369.
[14] Asadi A, Alarifi I, Ali V, Nguyen H. An experimental investigation on the effects of ultrasonication time on stability
and thermal conductivity of MWCNT-water nanofluid: Finding the optimum ultrasonication time. Ultrasonics –
Sonochemistry. 2019;58:104639.
[15] Taguchi G. Introduction to quality engineering. Whiter Plains, New York: Kraus International Publications. 1986.

536
 CHB CONVERTER WITH BBESS SUPPORT FOR LARGE
SCALE PHOTOVOLTAIC POWER PLANTS

Álvaro Pérez Mayo

CEIT-Basque Research and Technology Alliance (BRTA), Donostia / San Sebastián, Spain, [email protected],
Universidad de Navarra-Tecnun, Donostia / San Sebastián, Spain, ORCID: 0000-0003-2715-0431

Ainhoa Galarza
CEIT-Basque Research and Technology Alliance (BRTA), Donostia / San Sebastián, Spain, [email protected],
Universidad de Navarra-Tecnun, Donostia / San Sebastián, Spain, ORCID: 0000-0002-9584-9482

Fernando Martin
CEIT-Basque Research and Technology Alliance (BRTA), Donostia / San Sebastián, Spain, [email protected],
Universidad de Navarra-Tecnun, Donostia / San Sebastián, Spain, ORCID: 0000-0003-2567-3490

Javier Vadillo
CEIT-Basque Research and Technology Alliance (BRTA), Donostia / San Sebastián, Spain, [email protected],
Universidad de Navarra-Tecnun, Donostia / San Sebastián, Spain, ORCID: 0000-0002-7079-6353

Pérez Mayo, Á, Galarza, A, Martin, F, Vadillo, J. CHB Converter with BBESS Support for
Cite this paper as: Large Scale Photovoltaic Power Plants. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: This paper presents a modified grid tied CHB (Cascaded H-Bridge) converter for large scale
photovoltaic power plants. Apart from the advantages inherent to the CHB topology such as
modularity or highly efficient MPPT (Maximum Power Point Tracking), the proposed system
enables a new control strategy to face power unbalances. This feature is achieved as a result of
including a BBESS (Battery Based Energy Storage System) in a limited area of the converter.
Therefore, there are two different types of cell, one type is in charge of performing the MPPT of
the solar panels whereas the other also regulates the connection to the BBESS. In addition, the
connection that takes place in the BBESS creates a common DC bus through which power can
flow from one phase to another to overcome unbalance conditions. These aspects are taken into
consideration to develop an appropriate control method that is validated via simulations. The
simulations are conducted on a downsized 12 H-bridge model for validation purposes.
Keywords: CHB, Photovoltaic Plants, BBESS, Unbalance
© 2020 Published by ECRES

1. INTRODUCTION

Since the year 2010 the costs of wind and solar energies have dropped dramatically [1] making them more
appealing than other sources of energy. In addition, the costs of battery-based storage systems have also been
reduced heavily [2]. As a result, it is estimated that by the year 2050 almost half of the electric energy consumed
all around the world will come from a renewable source of energy [3], leading to a reduction of pollution.
This situation is going to raise a challenge for wind and solar generation from the technological perspective and
for the grid as a whole, as the effect of renewable source power plants massively connected to the grid has not
been thoroughly analyzed yet. From the technological point of view, one of the main problems that have to be
overcome is the high voltages that power converters will have to face. Furthermore, depending on the power of
the power station, the connection line voltage may be higher than 66 kV [4].
Nowadays, the tendency is that the capacity of photovoltaic plants is increasing up to several MW as it shown
in [5], where a NPC (Neutral Point Clamped) converter handles 40 MVA connected to 3,3 kV and 4,16 kV.

537
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Nevertheless, due to the increasing voltage and power that semiconductor devices have to deal with, the widely
used two level inverter and the NPC type converters result in inefficient or inviable solutions [6].
Another topology that has gained attention lately is the CHB [6]–[13], whose main advantage lies on its modular
capacity that allows devices to operate with lower voltage although the final output may be several times higher.
The modular nature of the CHB topology is also the root of its potential scalability and it also provides a more
efficient MPPT feature than their two-level counterparts [10].
Generally, the voltage oriented control (VOC) is widely used in CHB applications although it is possible to find
other references focused on direct power control [14]. The VOC for grid tied inverters usually consists of a
voltage loop that controls the equivalent DC voltage in the DC side of the H-bridge and combined with dq
current control loops, the reference voltages are generated for the Phase Shifted Pulse Width Modulation (PS-
PWM) [15], [16].
However, in photovoltaic applications the difference of irradiation received by each panel due to shadows casted
by far objects may cause differences in the electric energy produced by different sections of the power plant [6],
[16]. Furthermore, the temperature, which cannot be controlled all over the photovoltaic station, plays a key role
in the energy conversion effectiveness and the tolerances of the passive components, photovoltaic panels and
switching devices contribute to enlarge the differences [17]. These variations are the reason why unbalances
may appear in photovoltaic CHB plants. These unbalances can be classified in two ways: per cell unbalance and
phase unbalance. The first one means that the energy generated by the cells of the same phase is different for
each cell and can cause damages in the converter as a consequence of DC voltage drift. The other type of
unbalance causes disturbances on the currents injected to the grid, which is not allowed by grid codes [18].
Although both unbalances, per cell and per phase, can be compensated with appropriate control strategies such
as zero sequence vector injection [16], [17], [19], the available DC voltage bus limits the capacity of achieving
balanced operation.
As a consequence, there has been diverse hardware modifications proposed to improve the features of
photovoltaic stations. Batteries and other type of energy storage systems have been employed for improving
harmonics [20], supporting the energy production [21], [22], fault tolerance [13] or maximizing economic
benefits [23]. Although in [24] batteries could be used to compensate imbalances, the compensation strategy is
still only focused on zero sequence vector injection. Another approach is to couple magnetically the three phases
so that the excess of power in one phase can flow to the others. This is explained in [25] but this solution has
been criticized for implementation difficulties [26]. As an alternative, in [26], [27] it is shown that stablishing a
common DC voltage bus can remove the imbalance problem. However, these solutions are not optimal and some
resources have to be duplicated.
In the present paper, a new scheme is proposed to optimize the resources needed for compensating the
unbalances generated among phases. In this scheme, with the help of an energy storage system, the common DC
voltage bus is confined to a certain area of the photovoltaic installation and as a result, the number of components
can be reduced while a balanced operation is ensured.
The paper is structured as follows. In section 2 the architecture of the system as a whole is explained whereas in
section 3 the control strategy is analyzed. Subsequently simulation results are shown in section 4 and finally the
conclusions are expressed in section 5.

2. SYSTEM ARCHITECTURE

The proposed scheme is shown in Figure 1 where it can be seen a CHB grid tied inverter in a photovoltaic plant.
There are two different groups of components that may be identified as different type of cell whose functionality
is determined by their structure. The first type of cell is directly connected to photovoltaic panels and they are
located at the bottom of Figure 1, whereas the other type of cell, located at the top of Figure 1, count with a
BBESS.
As it can be seen in Figure 1, the system architecture is CHB converter of 12 H-bridges. The cells 1, 2 and 3 of
each phase are connected to the photovoltaic panels and the fourth cell of each phase, apart from being connected
to the solar panels, they are connected through a bidirectional isolated converter to the same battery. This way,
a common DC bus is created, as it is indicated in Figure 1, and the power injected by each phase can be adjusted.
In addition to the nominal power and phase voltage, the quantity of each type of cell can be modified depending
on the foreseen unbalances that might take place. The function of each cell is explained in the following
subsections.

538
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. Proposed CHB architecture

Photovoltaic Cells

The structure of this type of cell is depicted on Figure 1, cells 1,2 and 3 of phase A. There, it can be seen how
the photovoltaic panels are connected to an isolated unidirectional DC/DC converter whose main function is to
perform the MPPT, apart from providing galvanic isolation. This feature avoids creating a high voltage bus in
the solar panels side, which is limited or regulated in some countries and may result in damages in the panels
due to isolation issues [28]. As the DC/DC is tracking MPP, the group form by the solar panels and the DC/DC
converter can be modelled as a current source [29]. Then, the H-bridge converts the DC power into AC power
while the DC voltage level is kept under control.

Battery Based Energy Storage System Cells

The details of this type of cell is shown in Figure 1, cell number 4 of phase A, and there appear three different
converters. The unidirectional DC/DC converter that is connected to the solar panels performs the same
functions as its counterpart of the previous section, i.e. carrying out MPPT. The presence of the battery-
connected bidirectional isolated DC/DC converter plays a key role in the whole system. It is in charge of
regulating the capacitor’s voltage Vc and it is able to couple the three phases allowing power flow between them
as the battery can be the same for the three phases. The last component is the H-bridge that carries out the same
function as in the other type of cell, however, the behaviour of the cell is completely different.
In the previous type of cell, the power that flows through the unidirectional converter is the same that flows
through the H-bridge if efficiency is high enough. On the contrary, in cells with BBESS the power that flows
through the unidirectional DC/DC can be different from the power that flows through the H-bridge and that
difference may be compensated by the bidirectional DC/DC connected to the battery. Otherwise the capacitor
voltage would drift from its reference. Therefore, the group consisting of solar panels, battery and DC/DC
converters can be modelled as a constant voltage source from the H-bridge point of view.

539
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. CONTROL STRATEGY

The proposed control strategy is mainly based on VOC as it can be seen in Figure 2 but it has been modified for
this particular application. There, it is shown how the MPPT is performed by the DC/DC converters from cell 1
to 3 for the three phases, whereas for the cell 4 in each phase an additional bidirectional converter controls the
DC voltage Vc and enables the connection to the BBESS (Figure 1). Besides, the outer control loop regulates
the average of the DC-link voltage Vc and provides the direct current reference (𝑖𝑑∗ ) for the system to achieve
the necessary active power P while the reactive power Q may be stablished depending on the application, which
leads to a certain quadrature current reference (𝑖𝑞∗ ). Then, with the electrical angle obtained from the phase
voltage measurements through a PLL (Phase Locked Loop), the phase currents are measured and transformed
to the dq reference frame. Subsequently, the current control loops guarantee the reference tracking and, before
the PWM signal generation, the cell and phase imbalance compensation is carried out.

Figure 2. System control scheme

Phase Imbalance Compensation

As a result of inhomogeneous irradiation, temperature differences and solar panel mismatching, each cell will
provide different power levels that have to be taken into account. It should be bored in mind that for the BBESS
cells a grid injected power reference must be provided. These references are a function of the electric power
generated by the solar panels that causes the phase imbalance, the state of charge (SOC) of the BBESS and the
nominal power of the converters. The sum of the power delivered to the grid by each phase must be the same to
achieve balanced operation. The power reference for the BBESS cells is set to compensate the imbalances.
Nevertheless, there are several different solutions that fulfil the balanced operation condition stablished in Eq.
[1].
4 4 4

∑ 𝑃𝑎,𝑖 = ∑ 𝑃𝑏,𝑖 = ∑ 𝑃𝑐,𝑖

(1)
𝑖=1 𝑖=1 𝑖=1

Where 𝑃𝑗,𝑖 is the power provided by each cell. However, as it was stated previously, the power provided by the
cell number four of each phase, i.e. the power reference for those cells, will be given by Eq. [2].

𝑃𝑗,4 = 𝑃𝑃𝑉 + 𝑃𝐵𝐵𝐸𝑆𝑆

(2)

Where 𝑃𝑃𝑉 is the power generated in the photovoltaic panels and 𝑃𝐵𝐵𝐸𝑆𝑆 is the power obtained from the battery.
Hence, it is necessary to consider the SOC to choose one option. The power reference can be higher than the
power generated by the panels to reduce the SOC, or lower if the opposite effect is desired. However, the

540
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

limitations of the power references are influenced by the H-bridge’s nominal power, the bidirectional converter’s
nominal power and the charging profile of the BBESS.

Cell Inner Control

To achieve a balanced operation, it is necessary to take into consideration the nature of each cell. For
photovoltaic cells, the capacitor’s voltage is controlled to a certain value modifying the voltage reference
obtained from the current loop. The whole process is carried out in the imbalance compensation stage and it is
depicted in Figure 3 (a).

Figure 3. (a) Photovoltaic cell inner control (b) BBESS cell inner control

In the same way, for the BBESS cells a similar compensation is conducted but instead of voltage reference a
power reference is used. This reference must be set according to conditions mentioned in the previous section.
The modification can be seen in Figure 3 (b).

4- SIMULATION RESULTS

To verify the proposed control strategy, the architecture of Figure 1 has been tested via the PSIM simulation
tool. The chosen scenario presents per cell and per phase imbalance to show how the system regulates the
capacitor voltage and how the phase unbalance is compensated by BBESS cells as it can be seen in Figure 4 and
Figure 5. Figure 4 shows the phase to phase voltages generated by the CHB converter, the currents, the active
and reactive power injected in the grid and in Figure 5 the values of the capacitor’s voltage and the power
generation of each cell are depicted

Figure 4. (a) injected currents, (b) phase-to-phase voltages, (c) active and reactive power

541
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 5. (a) Phase A cells capacitor voltage, (b) Phase B cells capacitor voltage, (c) Phase C cells capacitor voltage, (d)
Phase A cells power, (e) Phase B cells power and (f) Phase C cells power

As it can be seen in the simulation results, the currents injected to the grid are balanced and the capacitors’
voltages are kept under control while the cells 𝑃𝑎4 , 𝑃𝑏4 and 𝑃𝑐4 compensate the differences in the power
generated in each phase. In this case, the power generated by photovoltaic cells is 256 kW for phase A, 224 kw
for phase B and 240 kW for phase C. So, to avoid unbalanced operation the power references set for BBESS
cells are 84 kW, 116 kW and 100 kW respectively. As a consequence, a balanced behavior is achieved.

5-CONCLUSION

The proposed topology can effectively maintain a balanced operation under unbalanced conditions due to the
implementation of BBESS cells. This type of cells enables the transference of energy between phases and
contribute to reduce the disturbances in the power injected to the grid. There are limitations for this strategy
related to the duration of the imbalance, the nominal power of the converters and the features of the BBESS.
Nevertheless, if the imbalance cannot be overcome employing this technique, the zero-sequence vector injection
can be implemented, although in that case, only currents would be balanced while phase voltage could show
disturbances. Furthermore, in comparison with other solutions, the proposed architecture confines the common
DC bus to a region of the converter, which can improve the efficiency of power exchange between phases apart
from providing a new strategy to endure imbalances.

ACKNOWLEDGMENT

The authors acknowledge funding by the Basque Government in the frame of ELKARTEK 2019 ELPIVE
project.

542
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] A. International Renewable Energy Agency Abu Dhabi, Renewable generation costs 2018. IRENA, 2019.
[2] G. Albright, J. Edie AllCell Technologies, P. Crossley, and A. Vassallo, “IRENA Battery Storage Report 2015,” 2015.
[3] M. Dent and C. Martin, “The World Will Get Half Its Power From Wind, Solar by 2050 - Bloomberg.” [Online].
Available: https://www.bloomberg.com/news/articles/2019-06-18/the-world-will-get-half-its-power-from-wind-and-
solar-by-2050. [Accessed: 10-Sep-2019].
[4] M. Ding, Z. Xu, W. Wang, X. Wang, Y. Song, and D. Chen, “A review on China‫׳‬s large-scale PV integration: Progress,
challenges and recommendations,” Renew. Sustain. Energy Rev., vol. 53, pp. 639–652, Jan. 2016.
[5] S. Kouro et al., “Recent Advances and Industrial Applications of Multilevel Converters,” IEEE Trans. Ind. Electron.,
vol. 57, no. 8, pp. 2553–2580, Aug. 2010.
[6] S. Rivera, S. Kouro, B. Wu, J. I. Leon, J. Rodriguez, and L. G. Franquelo, “Cascaded H-bridge multilevel converter
multistring topology for large scale photovoltaic systems,” in 2011 IEEE International Symposium on Industrial
Electronics, 2011, pp. 1837–1844.
[7] K. Wang, X. Wu, F. Liu, Y. Wang, and R. Zhu, “Cascaded H-bridge multilevel converter topology for large-scale
photovoltaic system with balanced operation,” in IECON 2017 - 43rd Annual Conference of the IEEE Industrial
Electronics Society, 2017, pp. 6452–6457.
[8] B. Xiao, L. Hang, J. Mei, C. Riley, L. M. Tolbert, and B. Ozpineci, “Modular Cascaded H-Bridge Multilevel PV Inverter
With Distributed MPPT for Grid-Connected Applications,” IEEE Trans. Ind. Appl., vol. 51, no. 2, pp. 1722–1731,
Mar. 2015.
[9] N. D. Marks, T. J. Summers, and R. E. Betz, “Reactive power requirements for cascaded H-Bridge photovoltaic
systems,” in IECON 2014 - 40th Annual Conference of the IEEE Industrial Electronics Society, 2014, pp. 2219–2225.
[10] T. H. de Abreu Mateus, J. A. Pomilio, R. B. Godoy, and J. O. P. Pinto, “Distributed MPPT scheme for grid connected
operation of photovoltaic system using cascaded H-bridge multilevel converter under partial shading,” in 2017 IEEE
Southern Power Electronics Conference (SPEC), 2017, pp. 1–6.
[11] M. M. Haji-Esmaeili, M. Naseri, H. Khoun-Jahan, and M. Abapour, “Fault-tolerant structure for cascaded H-bridge
multilevel inverter and reliability evaluation,” IET Power Electron., vol. 10, no. 1, pp. 59–70, 2017.
[12] C. Wang, K. Zhang, J. Xiong, Y. Xue, and W. Liu, “An Efficient Modulation Strategy for Cascaded Photovoltaic
Systems Suffering From Module Mismatch,” IEEE J. Emerg. Sel. Top. Power Electron., vol. 6, no. 2, pp. 941–954,
Jun. 2018.
[13] A. A. Stonier and B. Lehman, “An Intelligent-Based Fault-Tolerant System for Solar-Fed Cascaded Multilevel
Inverters,” IEEE Trans. Energy Convers., vol. 33, no. 3, pp. 1047–1057, Sep. 2018.
[14] M. Malinowski, M. P. Kazmierkowski, and A. M. Trzynadlowski, “A comparative study of control techniques for
PWM rectifiers in AC adjustable speed drives,” IEEE Trans. Power Electron., vol. 18, no. 6, pp. 1390–1396, Nov.
2003.
[15] B. Li, R. Yang, D. Xu, G. Wang, W. Wang, and D. Xu, “Analysis of the Phase-Shifted Carrier Modulation for Modular
Multilevel Converters,” IEEE Trans. Power Electron., vol. 30, no. 1, pp. 297–310, Jan. 2015.
[16] S. Kouro, B. Wu, A. Moya, E. Villanueva, P. Correa, and J. Rodriguez, “Control of a cascaded H-bridge multilevel
converter for grid connection of photovoltaic systems,” in 2009 35th Annual Conference of IEEE Industrial Electronics,
2009, pp. 3976–3982.
[17] H. Akagi, S. Inoue, and T. Yoshii, “Control and Performance of a Transformerless Cascade PWM STATCOM With
Star Configuration,” IEEE Trans. Ind. Appl., vol. 43, no. 4, pp. 1041–1049, 2007.
[18] P. Sochor, N. M. L. Tan, and H. Akagi, “Low-Voltage-Ride-Through Control of a Modular Multilevel Single-Delta
Bridge-Cell (SDBC) Inverter for Utility-Scale Photovoltaic Systems,” IEEE Trans. Ind. Appl., vol. 54, no. 5, pp. 4739–
4751, Sep. 2018.
[19] Y. Yu, G. Konstantinou, B. Hredzak, and V. G. Agelidis, “Operation of Cascaded H-Bridge Multilevel Converters for
Large-Scale Photovoltaic Power Plants Under Bridge Failures,” IEEE Trans. Ind. Electron., vol. 62, no. 11, pp. 7228–
7236, Nov. 2015.
[20] A. Marquez, J. I. Leon, S. Vazquez, L. G. Franquelo, and S. Kouro, “Operation of an hybrid PV-battery system with
improved harmonic performance,” in IECON 2017 - 43rd Annual Conference of the IEEE Industrial Electronics
Society, 2017, pp. 4272–4277.
[21] S. Vavilapalli, U. Subramaniam, S. Padmanaban, and V. K. Ramachandaramurthy, “Design and Real-Time Simulation
of an AC Voltage Regulator Based Battery Charger for Large-Scale PV-Grid Energy Storage Systems,” IEEE Access,
vol. 5, pp. 25158–25170, 2017.
[22] P. Wu, W. Huang, and N. Tai, “Novel grid connection interface for utility-scale PV power plants based on MMC,” J.
Eng., vol. 2019, no. 16, pp. 2683–2686, Mar. 2019.
[23] A. Saez-de-Ibarra, A. Milo, H. Gaztanaga, V. Debusschere, and S. Bacha, “Co-Optimization of Storage System Sizing
and Control Strategy for Intelligent Photovoltaic Power Plants Market Integration,” IEEE Trans. Sustain. Energy, vol.
7, no. 4, pp. 1749–1761, Oct. 2016.
[24] H. Hasabelrasul, X. Yan, and A. S. Gadalla, “Power conditioning system control strategy for cascaded H-bridge
converter battery energy storage system,” J. Eng., vol. 2019, no. 16, pp. 663–667, Mar. 2019.

543
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[25] S. Essakiappan, H. S. Krishnamoorthy, P. Enjeti, R. S. Balog, and S. Ahmed, “Multilevel Medium-Frequency Link
Inverter for Utility Scale Photovoltaic Integration,” IEEE Trans. Power Electron., vol. 30, no. 7, pp. 3674–3684, Jul.
2015.
[26] K. Wang, R. Zhu, C. Wei, F. Liu, X. Wu, and M. Liserre, “Cascaded Multilevel Converter Topology for Large-Scale
Photovoltaic System With Balanced Operation,” IEEE Trans. Ind. Electron., vol. 66, no. 10, pp. 7694–7705, Oct. 2019.
[27] C. D. Fuentes, C. A. Rojas, H. Renaudineau, S. Kouro, M. A. Perez, and T. Meynard, “Experimental Validation of a
Single DC Bus Cascaded H-Bridge Multilevel Inverter for Multistring Photovoltaic Systems,” IEEE Trans. Ind.
Electron., vol. 64, no. 2, pp. 930–934, Feb. 2017.
[28] C. Buccella, C. Cecati, H. Latafat, and K. Razi, “A grid-connected PV system with LLC resonant DC-DC converter,”
in 2013 International Conference on Clean Electrical Power (ICCEP), 2013, pp. 777–782.
[29] P.-H. Wu, Y.-C. Su, J.-L. Shie, and P.-T. Cheng, “A Distributed Control Technique for the Multilevel Cascaded
Converter,” IEEE Trans. Ind. Appl., vol. 55, no. 2, pp. 1649–1657, Mar. 2019.

544
 PHOTODEPOSITION OF PD ON MOS2/TIO2
NANOCOMPOSITES FOR ENHANCED PHOTOCATALYTIC
HYDROGEN PRODUCTION UNDER VISIBLE LIGHT

Güy Nuray
1 Sakarya University, Department of Chemistry, Science & Arts Faculty, 54050 Sakarya, Turkey
2 Sakarya University, Biomedical, Magnetic, Semiconductor Materials Application and Research Center
(BIMAS-RC), 54050 Sakarya, Turkey, [email protected], ORCID: 0000-0002-1126-789X

Güy, N. Photodeposition of Pd on MoS2/TiO2 Nanocomposites for Enhanced Photocatalytic

Cite this paper as: Hydrogen Production Under Visible Light. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
2020, Istanbul, Turkey

Abstract: TiO2 is commonly utilized as a semiconductor for hydrogen production since it has superior
photocatalytic efficiency and chemical stability. But, its spread usage is importantly restricted by
wide band gap and rapid recombination of photocarriers. To eliminate these problems, TiO2 can
be combined with other narrow band gap semiconductors and metal species can be deposited on
TiO2. In recent years, MoS2 as a promising co-catalyst has drawn attention for photocatalytic H2
evaolution. Thus, we prepared Pd/MoS2/TiO2 nanocomposite the photocatalytic water splitting
efficiency of it under visible light was analyzed. The photocatalytic efficiency of Pd/MoS2/TiO2
was compared with the efficiency of MoS2, TiO2, MoS2/TiO2. Pd/MoS2/TiO2 indicated more
superior photocatalytic water splitting efficiency than that of MoS2, TiO2, MoS2/TiO2. This study
depicts that the H2 evolution performance enhances by the doping Pd and heterojunction.
Keywords: H2 production, MoS2, Pd doping, TiO2
© 2020 Published by ECRES

1. INTRODUCTION

Photocatalytic water splitting is one of the most remarkable technologies to convert solar energy to hydrogen to
resolve the increasing energy crisis and environmental issues [1]. Different photocatalysts have been improved
for gerenration of hydrogen by the water splitting, a clean and environmentally friendly renewable fuel [2-3].
Among various of photocatalysts for H2 evolution, TiO2 stands out such as its being cheap, nontoxic, stability
and great photoelectrochemical features [3]. But, the wide band gap of TiO2 restricts its applications under
visible light, which makes up approximately 43% of the solar spectrum [4]. To solve this problem, TiO 2 can be
combined with a narrow band gap energy semiconductor, modified with noble metals and co-catalyts [4]. As a
two-dimensional (2D) layered semiconductor, Molybdenum disulfide (MoS 2) has a band gap energy of ~1.29
eV is a favorable material for H2 production, due to its high-efficiency conductivity and great performance [3,5].
Furthermore, with its excellent electrochemical structure, it is ideal for to the migration of charge carriers [3,5].
So, it is expected that the combination of TiO2 with MoS2 may advance the photocatalytic efficiency of TiO2
[3]. The modifiying of metal nanoparticles via TiO2 or MoS2 is also exhibited as a reasonable viewpoint for the
improvement of their photocatalytic activities [3,4,6].
In this paper, Pd metallic nanoparticles as a co-catalyst was deposited into MoS2/TiO2 nanocomposites by a
photo-deposition method and the efficiency of the MoS 2, TiO2, MoS2/TiO2 and Pd/MoS2/TiO2 for photocatalytic
hydrogen evolution were examined under visible light.

2. EXPERIMENTAL

Chemicals
Hydrofluoric acid (HF), tetraisopropyl titanate (C12H28O4Ti), sodium hydroxide (NaOH), sodium molybdate
(Na2MoO4.2H2O), thiourea, ethanol (C2H5OH), palladium chloride (PdCl2) were procured from Sigma-Aldrich.

545
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Synthesis of Products
TiO2 nanopowders were prepared by microwave-assisted hydrothermal method based on approach of Soner et
al [7]. For synthesis of MoS2/TiO2, 30 mg TiO2 powder was homogeneously dispersed in 60 mL of distilled
water for 30 min. Then 90 mg of Na2MoO4.2H2O and 180 mg of thiourea were added to suspension. The mixture
was stirred and put to a Teflon-lined autoclave (100 mL) and heated at 200 °C for 10 h. The autoclave was then
cooled down. MoS2 was prepared by the same method without TiO2. The MoS2/TiO2 was dispersed in 50 mL of
distilled water and they mixed with a given amount of PdCl 2 solution to deposition 5.0 wt% Pd in the
Pd/MoS2/TiO2. Finally, the mixture was exposed to irradiation by using 300 W Xe lamp for 20 min.

Characterization
The structure of all products was perused by using a PAN-alytical X-ray diffractometer (XRD, Netherlands) and
diffuse reflectance measurements were performed by a Shimadzu UV-2600 UV–Vis spectrophotometer.

Photocatalytic Hydrogen Production

The H2 generation efficiencies of the samples were detected in a in a closed glass system. 50 mg of the product
was added into100 mL aqueous solution including 0.5 M Na 2S and 0.5 M Na2SO3 as a sacrificial agent. 300 W
Xe lamp equipped with a 420 nm cut-off filter was employed as a light source. The H2 concentration was perused
by an online gas chromatograph (GC Agilent-6890, TCD, with N2 as carrier).

3. RESULTS AND DISCUSSION

The XRD patterns of TiO2, MoS2, MoS2/TiO2 and Pd/MoS2/TiO2 are represented in Figure 1a. As demonstrated
in Figure 1a, the XRD peaks of TiO2 at 25.3°, 37.8°, 48.0°, 53.9°, 55.1° and 62.7° index to the (101), (004),
(200), (105), (211) and (204) reflections of anatase phase of TiO2 (ICDS: 98-015-4601). The observed
diffraction peaks of 2θ around 14.1°, 33.2°, 39.4° and 58.6° can be assigned to the (002), (103), and (105)
reflections of MoS2, respectively. For Pd/MoS2/TiO2 nanocomposite, the observed diffraction peaks are well
matched to (101), (220), (311), (004), (331), (200), (422), (511) and (215) crystal planes of anatase phase of
TiO2 and MoS2, respectively. The diffraction peaks marked with “#” correspond to the (111), and (200)
crystalline planes of metallic Pd. These results prove that the metallic Pd nanoparticles have been modified on
the surface of MoS 2/TiO2.

Optic absorption features and band gap energies of the obtained prodcuts were examined with UV–Vis diffuse
reflectance spectroscopy (DRS), as depicted in Figure 1b. From Figure 1b, it is observed that only TiO 2 has a
vigorous absorption in the UV region owing to its large band gap. However, MoS 2, MoS2/TiO2 and
Pd/MoS2/TiO2 indicate absorption in the visible region since MoS 2 has the narrow band gap energy. The
calculated band gap energies of TiO2, MoS2, MoS2/TiO2 and Pd/MoS2/TiO2 are 3.20, 1.64, 2.94 and 1.73 eV,
respectively.

H2 evolution efficiencies of products were analyzed using visible light. The H2 production rates of the as-
obtained TiO2, MoS2, MoS2/TiO2 and Pd/MoS2/TiO2 are exhibited in Figure 2. The H2 production rates of TiO2,
MoS2, MoS2/TiO2 and Pd/MoS2/TiO2 are 2.5, 4.5, 318 and 408 μmol/g, respectively. The MoS2/TiO2
nanocomposites exhibit increased photocatalytic performance because the layered MoS 2 as an efficient
cocatalyst can accelerate the charge carriers separation for H2 production than TiO2. In the presence of a Pd
nanoparticles in the nanocomposite, the photocatalytic H2 generation rate of Pd/MoS2/TiO2 is further enhanced
to 90 μmol/g compared to MoS2/TiO2. This great activity can be primarly assigned to surface plasmon resonance
of Pd, and the synergistic interaction between MoS 2 and TiO2. Pd nanoparticles capture photinduced electrons
and retard the photocarriers, thus advancing the photocatalytic efficiency.

546
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)
Figure 1. (a) XRD spectra, (b) the band gap energies of all products.

Figure 2. H2 generation by photocatalytic water splitting of the products.

4. CONCLUSIONS

In summary, Pd/MoS2/TiO2 nanocomposites were successfully constructed and photocatalytic H2 evolution

efficiency examined. Pd/MoS2/TiO2 displays much increased photocatalytic hydrogen activity than TiO2, MoS2,
MoS2/TiO2, which is primarly owing to the synergistic interest between TiO2 and MoS2 in the MoS2/TiO2 and
surface plasmon resonance effectof Pd. This study supplies a new perspective in the preparation and
improvement of photocatalysts containing of TiO2 for production of hydrogen evolution applications.

ACKNOWLEDGMENT
The author would like to thank to Prof. Mahmut Özacar and Associate Professor Dr. Keziban Atacan for their
valuable contributions.

547
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] Wang B, He S, Zhang L, Huang X, Gao F, Feng W, Liu P. CdS nanorods decorated with inexpensive NiCd bimetallic
nanoparticles as efficient photocatalysts for visible-light-driven photocatalytic hydrogen evolution. Applied Catalysis
B: Environmental 2019; 243:229-235 <https://doi.org/10.1016/j.apcatb.2018.10.065>
[2] Ou W, Pan J, Liu Y, Li S, Li H, Zhao W, Wang J, Song C, Zheng Y, Li C. Two-dimensional ultrathin MoS2-modified
black Ti3+ –TiO2 nanotubes for enhanced photocatalytic water splitting hydrogen production. Journal of Energy
Chemistry 2020; 43:188–194 < https://doi.org/10.1016/j.jechem.2019.08.020>
[3] Liu Y, Xu C, Xie Y, Yang L, Ling Y, Chen L. Au-Cu nanoalloy/TiO2/MoS2 ternary hybrid with enhanced photocatalytic
hydrogen production. Journal of Alloys and Compounds 2020; 820:153440
<https://doi.org/10.1016/j.jallcom.2019.153440>
[4] Guo L, Yang B, Marcus K, Li Z, Luo B, Zhou L, Wang X, Du Y and Yang Y. MoS2/TiO2 heterostructures as nonmetal
plasmonic photocatalysts for highly efficient hydrogen evolution. Energy Environ. Sci. 2018; 11:106
<https://doi.org/10.1039/C7EE02464A>
[5] Liu Y, Li Y, Peng F, Lin Y, Yang S, Zhang S, Wang H, Cao Y, Yu H. 2H- and 1T- mixed phase few-layer MoS2 as a
superior to Pt co-catalyst coated on TiO2 nanorod arrays for photocatalytic hydrogen evolution. Applied Catalysis B:
Environmental 2019; 241:236–245 <https://doi.org/10.1016/j.apcatb.2018.09.040>
[6] Tao j, Chen S, Guan L, Chen G, Yu C, Chen L, Cheng X, Zhang H, Xie X. Well-patterned Au nanodots on MoS2/TiO2
hybrids for enhanced hydrogen evolution activity. Electrochimica Acta 2018; 283:419-427
<https://doi.org/10.1016/j.electacta.2018.06.194>
[7] Ünlü B, Çakar S, Özacar M. The effects of metal doped TiO2 and dithizone-metal complexes on DSSCs performance.
Solar Energy 2018; 166: 441-449 < https://doi.org/10.1016/j.solener.2018.03.064>

548
 DESIGN OF A VISUAL TESTING SYSTEM IN A ROBOTISED
WELDING WORKSHOP OF JACKET NODES BY MEANS OF
DES AND ERGONOMIC SOFTWARE

Adolfo Lamas Rodríguez

Navantia, Ferrol, Spain, [email protected], ORCID: 0000-0002-2405-2513

Javier Pernas Álvarez

UMI Navantia-UDC, Ferrol, Spain, [email protected], ORCID: 0000-0001-9185-3583

Inés Taracido López

UMI Navantia-UDC, Ferrol, Spain, [email protected], ORCID: 0000-0002-3078-4411

Lamas Rodríguez, A, Pernas Álvarez, J, Taracido López, I. Design of a Visual Testing System
Cite this paper as: in a Robotised Welding Workshop of Jacket Nodes by means of DES and Ergonomic
Software. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Offshore wind energy is believed to be the key to decarbonizing the power sector yet nowadays
faces some degree of uncertainty with respect to its eventual cost competitiveness. Regarding
jacket foundations, the leap from manual to robotized nodes welding may signify a 30% reduction
in manufacturing cost. In this context, we present a case study where the 3D discrete-events
simulator FlexSim coupled with the ergonomic package Tecnomatix Jack have been used to design
and optimized the implementation of a new system for visual testing of jacket’s nodes. This system
is to be incorporated into a future workshop considered by the international Spanish shipbuilder
NAVANTIA as a key investment in its medium-term roadmap. Therefore, we first formulated
several scenarios in FlexSim to analyze the new system and developed detailed 3D animations of
worker’s performance. These animations posed a need for in-depth analysis of operator’s
execution from an ergonomics viewpoint. Then, we used Tecnomatix Jack to undertake RULA
analyses of the task. As a result, we proposed an ergonomic alternative. Eventually, we
incorporated these changes into the 3D animation in FlexSim thus obtaining an ultimate integrated
model.
Keywords: Offshore wind, Industry 4.0, Discrete Event Simulation, Ergonomic analysis, Jackets nodes
welding, Improved Fatigue Life of Welded Jacket Connections (JaCo)
© 2020 Published by ECRES

1. INTRODUCTION

Offshore wind energy holds nowadays very high expectations regarding both installed capacity and planned
investments in the medium and long-term. Whereas 150 new offshore wind projects are scheduled over the next
5 years, Europe deserves special mention as to be leading the global market, accounting for the 80% of global
installed offshore wind capacity in 2018 [1]. In line with this, offshore wind energy is posed to be the key to
decarbonizing the power sector and driving energy transitions [2], yet there still exists some degree of
uncertainty when it comes to its eventual attractiveness and cost competitiveness. Long-term vision regulations,
efficient supply chains and remarkable technology improvements will be essential for the success of offshore
wind energy.
In close proximity, Digital Manufacturing (DM) is becoming essential for current manufacturing processes, in
total alignment with Industry 4.0. As part of the DM, Digital Simulation allows forecasting a precise result or
scenario, especially when certain variables are blurry or even unknown. It is thought to basically rely on three
fields: Discrete Event Simulation (DES), Virtual ergonomics and Virtual commissioning, here concerning the
first two.

549
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In this context, this paper presents a case study where 3D discrete-event simulation (DES) coupled with a human
simulation package have been used to design and optimized the implementation of a new system for Visual
Testing (VT) of jacket’s nodes in a future workshop. Previously, we had also used 3D DES to entirely design
from scratch the workshop for robotized welding of nodes, which is planned to be set up by the international
renowned Spanish shipbuilder NAVANTIA S.A.
Overall, we present an innovative approach in line with the noticed challenges offshore wind must tackle. The
reduction of costs will undoubtedly depend in part on offshore foundations which currently evidence a
significant room for improvement regarding technology and automatization. Concretely, jacket foundations are
15% of capital expenditures in an offshore wind farm [3] and the leap from manual to robotized nodes welding
may signify both 60% and 30% reductions in manufacturing time and cost respectively [4]. Due to its
complexity, nodes weld turns out to be a cost driver within jacket manufacturing and thus becomes crucial for
the right performance of the whole project.
As for DES, this well-aged simulation paradigm currently used by many companies around the world allows
handling process’ variability and testing changes in stochastic systems in an efficient way without committing
resources and thus incurring costs. Such is partly the case of [5] where several improvements are proposed by
means of DES to reduce flowtime per jacket and a minimize the probability of missing due dates. Likewise, [6]
uses DES as a tool to perform a quantitative assessment of risk exposure and associated cost impacts in a jackets
manufacturing process.
All in all, the recent outstanding improvements in 3D graphics applied to DES has enhanced its use in the design
of new systems and facilities, and so we have done here. However, there is an apparent lack of literature when
it comes to design a process from the grounds up, probably due in part to a certain reticence to shift from
traditional methods based on personnel’s know-how and benchmarks. This is also pointed by [7] who used DES
as a decision-making tool in the first stages of a hospital construction project.
Finally, when it comes to ergonomics, it must be noted that nowadays many studies on the subject are undertaken
as post analyses, to study a certain real processed prone to cause any kind of ergonomic issue, as stated in [8].
They propose a design methodology to promote process sustainability by virtual simulations from a viewpoint
similar to the one given here, since our approach is based on taking actions before coming down to real world.
Although [9] turned back to the post-study methodology, they used Tecnomatix JACK from Siemens PLM to
put forward possible risks on workers during their daily activities, a close instance to the one given here either
on the software and the sort of study carried out.
Prediction of the ergonomic performance can represent a fundamental step in preventing ergonomic issues since
the early design phase of a production process [10]. Not only have we analyzed the impact of a node’s rotator
in terms of productivity and cost, but also by adopting a human-centered perspective and caring about the
wellness of the worker which should be that of the company as well.

2. CASE DESCRIPTION

NAVANTIA, the fifth-largest shipbuilder in Europe, has deemed a new workshop for robotized nodes’ welding
to be profitable and of great importance to its medium-term roadmap. It has diversified its production over the
last decade by undertaking construction projects of offshore wind foundations and sees the possible investment
as a big step forward in his business plan. In addition, the company counts already with a robotized welding
robot from a still ongoing R&D project named JaCo (Improved Fatigue Life of Welded Jacket Connections). Its
objective is to reduce the cost of offshore wind by optimizing the design of jacket foundations through improved
fatigue standards and validation of faster test and fabrication methods: that is, robotized welding. The robot at
issue is already full operational and could be further adapted to different typologies of nodes.
So far, the production of nodes has been subcontracted to local companies and, as a result, the actual costs
incurred and process’ stages cycle times come to be unclear. Thus, the company have required further basis on
flowtime, necessary expenditures and return on investment to proceed with the investment. Therefore, it has
thought of 3D DES as the framework against which define tasks’ times, personnel, and spatial and technology
requirements which would be eventually determine by company’s pundits. The current flow time which is to be
attained is 1 jacket per week.

Jacket’s Nodes
Jackets are can be coarsely divided into four parts: the stabbings; the Jacket Lower Block (JLB) and Jacket
Upper Block (JUB) which make up the main body of the structure; and the Transition Piece (TP) between the

550
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

jacket body and the wind turbine tower. The JUB and JLB are in turn made up of braces and legs which, for
their part, consist of tubular elements connected by transition nodes, namely, jacket’s nodes. So, as showed in
Figure 1, these latter are steel tubular steel joints made up of at least two stubs and a central chord. It’s worth
noting that the jackets in question present three different types of nodes depending on their shape and role: X,
Y and K-shaped nodes, although in principle K nodes have been discarded in the design of the workshop since
they are currently unmanageable by the robot.

Figure 1. Main components of a jacket.

Manual Welding and Visual Testing

Although the manual welding process comprised several phases, we can roughly ascertain 3 stages sequentially
performed: root pass, rest of filling and capping passes and subsequent Non-Destructive Test (NDT) and
Dimensional Control (DC) processes. Upon this, we note two main points:
 The limited accessibility and the stubs preheating temperature make the operation conditions
particularly difficult from the inside and require welders with high manual skills.
 NDT process is composed of several phases, one of which is VT. Once the inspector has carried out the
VT, he fulfills the proper form template to further record it into the project database. This two-
sequential-step procedure takes several hours to have the data available within the enterprise resource
planning (ERP) software. However, since the manual welding is slower, there has not been so far need
for time reduction.

Robotized welding and automated NDT

On top of all the aforementioned, the robotized welding may increase the deposition rate up to 7 times the manual
one while lowering the defect rate and thwarting reworks from taking place.
When it comes to automation of NDT, Navantia has thought of a node’s adjustable rotator to perform the whole
process and a portable, non-contact weld and joint inspection system for VT. The process must be executed by
an operator, but we can count at least four advantages over the traditional method:
 Quicker inspection as the system automatically interprets the information being read.
 Wi-Fi connection automatically records information in the company’s ERP system.
 No need for external means to reposition the node during the process.
 Significant ergonomic improvement.

3. METHODOLOGY

Data collection
Personnel of the Production and Engineering Division (PED) together with information from previous projects
have served as data sources either during the model conceptualization, data collection and validation phases.

Discrete-Event Simulation Models

Due to the uncertainty presented in the case and the need for detailed analysis and fluent discussion between
expert modelers and non-expert personnel, 3D DES has been considered as an adequate approach to carry out

551
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the design of the new workshop. The selected software is FlexSim, mainly due to modelers’ experience with it
and its powerful 3D orientation, which strongly boosts the communication of results.
We first integrated the current layout of the predefined company facility for the workshop in FlexSim, including
the CAD plane and existing resources such as overhead travelling cranes. Afterwards, we designed and modeled
the different stages that will make up the workshop: stubs cutting and beveling, DC phases, root pass beads,
JaCo robot and final NDT. All this was done by always respecting the shipyard flow time of 1 jacket weekly.
FlexSim’s animations module allowed us to develop realist animations of every operation and incorporate VT
changes with a high degree of accuracy. Every step was examined with Navantia experts, which also provided
us with cycle times distributions, reductions in defect rates respect to manual welding and information about
upcoming workload. They also participated in the validation phase thanks to 3D graphics which highly facilitate
the understanding of the model’s functioning.

Tecnomatix JACK models

Once the need for ergonomic analysis on VT was identified, we proposed the use of Tecnomatix JACK software
for Siemens to undertake the study. The choice was mostly based on its extended use in the literature review
which acknowledge its intuitive interface usability.
We centered the study on the evaluation of postural load, one of the risk factors most commonly associated with
the appearance of musculoskeletal-type disorders, and ways to reduce it. As evaluating method, we chose RULA
(Rapid Upper Limb Assessment) method, one of the most widespread. We evaluated the VT of both X- and Y-
nodes and divided the movement into 2 postures.

4. RESULTS AND DISCUSSION

Figure 2 shows the proposed process flow over the 3D layout in FlexSim for the robotized welding workshop
(Figure 5). This layout satisfies the requested flow time of all X and Y-nodes of a jacket per week. In the last
span of the shop there is the welding robot at the bottom and the NDT as last phase (in parallel with final node
DC) on the top. As we can see, the layout only defines a zone dedicated to NDT without determining an accurate
number of stations. To our purposes, this layout corresponds to a base scenario with traditional NDT procedure,
which in turn includes VT and is entirely performed in the same location. NDT cycle time is defined by an
exponential distribution with 7h as location and 4 as scale, due to some present degree of uncertainty.
The model is fully parametrized in order to carry out sensitivity analyses of numbers of resources (operators,
stations) and machine cycle times. Its development also entailed a model optimization to achieve a quasi-optimal
configuration of the layout in terms of resources needed.
Once the global layout was set up, we performed the detailed analysis of NDT. Table 1 summarize the 5
scenarios contemplated during the study:
 First, we examined the number of slots that would be necessary to fulfill the required flow time with
traditional NDT and VT procedures. Scenario 1 with 2 stations yielded a flow time of 9.54 days to
produce all the nodes of a jacket, very far from the objective requested. Scenario 2 adds an extra slot yet
requested flow time is still not achieved.
 Scenario 3 with 4 NDT stations ensures a flow time of 7days/jacket. The main issue here is the need for
more room in the span, which is not available given the proposed layout.
 Scenario 4 incorporates one real scale node’s rotator along with the new VT system. New NDT cycle
times are modelled by a triangular distribution with 6h as mean and 4 to 8 as lower and upper bounds
respectively in accordance with PED, who also foresees a reduction in the variability of the process.
Besides the fact the node’s rotator should be able to adjust both for X- and Y-nodes, which would make
it much more expensive, the model shows that this option does not reach the requested flow time.
 Scenario 5 shows the final approach, with 2 node’s rotators, one per type of node. This approach solves
3 questions simultaneously: it ensures the required flow time with certain floating time; it is less
expensive than designing a multi-node rotator; and it fulfills the spatial requirements.

552
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1. Summary of DES scenarios and flow rates.

NDT avg.
NDT Weeks/1 Jacket's
Scenario Description cycle time
Slots nodes
(h/node)
1 Traditional method. 1 station per type of node. 2 9 0.73
Traditional method. 2 stations for X-Nodes, 1 for Y-
2 3 9 0.96
nodes
3 Traditional method. 2 stations per type of node. 4 9 1.01
4 1 rotator and new inspection system 1 6 0.89
5 1 rotator per node type and new inspection system 2 6 1.02

Thus, scenario 5 was the most suitable approach in the design of the VT stage. For this scenario we also
developed 3D animations of the operator performing the task (Figure 3) and hence we used FlexSim to model
the movements performed by the operator in charge of the VT. By doing so we identified the need for
undertaking an ergonomic analysis especially because of the operator's difficulty in accessing certain parts of
the node due to its height.
Final Robotised Workshop Proposed Layout
Material
(Chords)
Material (Chords and Output
Process Stubs) Input
Flow
Direction

Chords Entry
Buffer NDT and DC
Zone

Stubs Entry
Buffer Assembly and
DC1 Stations

Root Pass
Welding
Cutting and and DC 2
Beveled Stubs
beveling Stations
Buffer Welding
Rework
Station
Station Buffer

Welding Robot

Figure 2. Final proposed layout of the robotized welding Figure 3. Node’s VT 3D animation in FlexSim without
workshop. platform.

Ergonomic Analysis
Once we implemented the 3D models in Tecnomatix Jack, we examined 3 scenarios summarized in Table 2:
 Scenario 1: The most unfavorable scenario in which the operator performs the inspection from the
ground with predefined node’s rotations during the VT.
 Scenario 2: An intermediate scenario in which we incorporated a platform to raise the operator and kept
the standard rotation scheme.
 Scenario 3: The most ergonomic scenario. It includes the platform and we also improved the rotation
scheme with more precise and ergonomic rotations from the perspective of facilitating operator’s access.

Table 2: Summary of results obtained in ergonomic analysis.

Score RULA Score RULA Score RULA Score RULA
N Scenario Description Posture 1-Node Posture 2- Posture 1- Posture 2-
X NodeX NodeY NodeY
1 VT from the ground 6 6 7 7
2 VT from the platform 3 3 3 5
VT from the platform
3 and new rotation 2 2 2 2
scheme

553
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 4. Screenshots of different scenarios examined in Tecnomatix Jack.

In Scenario 3 RULA analysis indicated level 1 for all positions, that is, the postures are acceptable if not
maintained or repeated for long periods of time. Finally, we set up the new proposed node’s rotation scheme
back again in FlexSim so that we obtained a final 3D animation of the noticed operation integrated with the
process simulation (Figure 6).

Figure 5. Screenshot of the final robotized workshop Figure 6. Node’s inspection animation after the ergonomic
DES model. study.

5. CONCLUSIONS

In this paper we present an innovative approach which consists in the use of the 3D DES software FlexSim
coupled with the simulation and ergonomic software Tecnomatix Jack to carry out the study and design the VT
process of a new jacket’s nodes robotized workshop both from production and ergonomics point of view.
Once we developed the 3D model of the new robotized workshop, we studied the NDT process and brought in
a new solution for VT in accordance with Navantia personnel based on a node’s adjustable rotator and a new
inspection system. This allowed us reducing the number of stations needed as well as task time and variability
with respect to traditional NDT. The implementation of detailed 3D animations in FlexSim posed a need for in-
depth analysis of operator’s task performance from an ergonomics standpoint. Then, we used Tecnomatix Jack
to undertake RULA analyses and we proposed a much more ergonomic alternative made up of a platform to
raise the operator along with a redefinition of the node’s rotation trajectories. Eventually, we incorporated these
changes into the 3D animation in FlexSim thus obtaining a ultimate integrated model.
This case study mainly puts forward the potential use of 3D DES to assess the performance of tasks by humans
through detailed 3D animations in the design stages of a process. However, it also places the need for an external
human-simulation software to evaluate them from an ergonomics viewpoint so that afterwards the results may
be reflected again into the 3D DES software. Therefore, this iterative approach allows optimizing a task or
process by adopting a human-centered perspective at the same time the productivity analysis is considered.

REFERENCES

[1] International Energy Agency, “Offshore Wind Outlook 2019 – Analysis - IEA,” 2019.
[2] EWEA, “Reviving wind markets and delivering on our 2030 objetives,” Wind Eur., 2017.
[3] P. Dvorak, “Global Wind Network updates state of U.S. wind-turbine manufacturing for onshore and off,” 2014.
[Online]. Available: https://www.windpowerengineering.com/global-wind-network-updates-state-u-s-wind-turbine-

554
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

manufacturing-onshore/. [Accessed: 23-Mar-2020].

[4] S. M. L. P. GmbH, “Sustainability in jacket structures for offshore wind energy plants | Automated node welding,
maintenance-free corrosion protection,” Mar. 2020.
[5] A. Lamas-Rodríguez, D. Crespo-Pereira, and R. Sánchez-Tutor, “Discrete Events Simulation to improve
manufacturing process of jackets offshore structures,” in 15th International Conference on Modeling and Applied
Simulation, MAS 2016, 2016, pp. 85–94.
[6] A. Lamas-Rodríguez, D. Chas-Álvarez, and A. Muiña-Dono, RISK MANAGEMENT IN JACKETS
MANUFACTURING PROJECTS USING DISCRETE EVENTS SIMULATION. 2017.
[7] I. W. Gibson, “An approach to hospital planning and design using Discrete Event Simulation,” Proc. - Winter Simul.
Conf., pp. 1501–1509, 2007.
[8] M. Peruzzini, S. Carassai, and M. Pellicciari, “The Benefits of Human-centred Design in Industrial Practices: Re-
design of Workstations in Pipe Industry,” Procedia Manuf., vol. 11, pp. 1247–1254, 2017.
[9] D. Cortes et al., “A model for plant digitalisation, simulation and improvement: A case study in the automotive tier
one supplier,” in Proceedings - 2019 IEEE International Conference on Engineering, Technology and Innovation,
ICE/ITMC 2019, 2019.
[10] F. Caputo, A. Greco, M. Fera, and R. Macchiaroli, “Digital twins to enhance the integration of ergonomics in the
workplace design,” Int. J. Ind. Ergon., vol. 71, pp. 20–31, 2019.

555
 ACCELEROMETER MASS LOADING STUDY BASED ON A
DAMAGE IDENTIFICATION METHOD USING FUNDAMENTAL
LAWS IN CLOSED SYSTEMS

Tarık Tufan
Istanbul Medeniyet University, Istanbul, Turkiye, [email protected], ORCID: 0000-0001-9324-2401

Hasan Köten
Istanbul Medeniyet University, Istanbul, Turkiye, [email protected], ORCID: 0000-0002-1907-9420

Tufan, T, Köten, H. Accelerometer Mass Loading Study Based on a Damage Identification

Cite this paper as: Method using Fundamental Laws in Closed Systems. 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: For a limited time, civil engineering structures like bridges can be model as a closed system in
which can energy exchange exists but not mass change with its surroundings. Besides, the
damage process can be represented as a thermodynamic process. In this study, it is aimed to
show that the damage to the closed systems can be successfully estimated by using a damage
indicator. In the analytical study, a 100-element 200 degree of freedom fixed-free beam is
investigated and the assessment of the damage position is verified numerically. Except for the
piece at the free end of the beam; It has been determined that the damage indicator calculated on
the damaged elements is 60 times larger than the value of undamaged elements. In the elements
at the free end of the beam; it was observed that this ratio is between 6 and 40 depending on the
mass of the accelerometer.
Keywords: Modal Excitation Energy,, Modal Plot, Damage Detection, Operational Modal Analysis,
Acceleration Mass Loading, Structural Material.
© 2020 Published by ECRES

Nomenclature
SHM Structural Health Monitoring
DOF Degree of Freedom

1. INTRODUCTION

During the operation of the critical structures, the total mass of the structure may change. However, it can be
assumed that the mass in the bridge is preserved within a certain period but the position of mass in the system
due to live load may change. Hence, such a system can be considered as a closed system where energy transfer
is possible and the mass in the system is displaced, provided that it is preserved. For the "closed" term, which
is taken from thermodynamics in the system definition, to be meaningful, the damage in the system must also
be considered from a thermodynamic perspective.

Vibration-based structural health monitoring aims to detect the instantaneous state of the system during the
system operation. Since the external forces can apply to the white noise assumption where the frequency
components in the target range excited ‘uniformly’ (Yang, 1975, Peeters, 2000, Ren et al., 2004), system
properties of a system can be identified under daily (operational) loads. However, as stated above, in order to
be operational, real structures like bridges should be considered as closed systems.

556
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In the literature, damage detection using modal variables and their functions has been one of the popular issues
since the early days (Doebling et al. 1996, Tufan et al., 2019). Besides, researchers have started to use learning
algorithms such as artificial neural networks (Riva et al., 1992) and deep learning (Guo et al., 2014) since
1992 and 2014 respectively. A drawback of these studies is that they are not appropriate for generic structures
but rather they have to be adopted for a specific model or a structure. Even if advanced methods are used,
selecting a damage parameter that is sensitive to damage directly affects the success of the method. Due to the
urgency of the issue, these advanced methods are used in the monitoring of real structures. However, it is seen
in the literature that damage assessment studies have been carried out for beams and similar simple systems
(Krishnanunni et al., 2019, Vashisht et al. 2020). There are successful studies in this regard, though not
completely. Since complete success has not been achieved yet, further studies are needed in damage detection
of beams and similar systems. In addition, there is no study to investigate the weight of the accelerometer
depending on damage identification method in Structural Health Monitoring (SHM) literature.

In this study, an enhanced and valid damage indicator and detection method are proposed for closed systems
that can successfully detect damage in any of the elements. It was observed that there is a difference in the
damage indicator according to the accelerometer mass at the boundary conditions of the system examined. A
criterion is developed for the utilization of the damage indicator according to the accelerometer mass. To the
best of our knowledge, there has been no such technique in SHM literature that the selection of the
accelerometer mass according to a damage detection procedure and therefore it represents a novelty for the
literature. The validity of the damage identification method for the closed systems is numerically verified and
supported with an experimental study. Moreover, modal plots, which is a tool to monitor modal excitation
energy, are also constructed by utilizing frequency response functions in this study.

2. DAMAGE DETECTION

Inelastic deformation processes of solid bodies represent thermo-mechanical processes due to the involved
dissipation of mechanical work which is also true in such cases when the process runs isothermally (Lehmann,
1989). The Inelastic deformations in steel, many ceramics and aggregate stem from dislocations along slip
planes are usually accompanied by certain damage processes due to microdefects. These damage processes
may start at a certain stage of the deformation process and result in the development of macrodefects in the
materials (Brünig, 2004). In many cases larger inelastic deformations are accompanied by certain irreversible
damage processes which also have to be considered as thermo-mechanical processes (Lehmann, 1989).

𝑤̇𝑡 = 𝑤𝑒̇ + 𝑤̇ 𝑖 (1)

where 𝑤̇𝑡 denotes inelastic work rate, 𝑤𝑒̇ shows the dissipated part of the 𝑤̇𝑡 , and 𝑤̇ 𝑖 represent the non-
dissipated part of 𝑤̇𝑡 which corresponds to thermodynamical processes.

For the structural health monitoring concerns, the input forces acting on the system provide the energy to
excite the global and local modes of the system, but these loads also drive different types of damage
mechanisms, which is thermo-mechanical processes as mentioned, in the material according to a certain
threshold value. It is generally agreed that the damage identification methods can be systematically
categorized into four levels [see, e.g., (Rytter, 1993; Fassosi et al., 2009)]:

Level 0: (Fault Detection) Determining whether damage exists in the structure.

Level 1: (Fault Identification) Detection of the geometric location of the damage.

Level 2: (Magnitude Estimation) Quantification of the severity of the damage.

Level 3: Prediction of the remaining service life of the structure.

557
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In addition to the aforementioned levels, some researchers (Ayres et al., 1998) emphasize that the type of
damage is also important. Some of the examples of the damage entities are cracking, rusting in the
reinforcement bars, material deformation, loosening of the bolts.

Although there are many sensitivity-based studies in the literature (Doebling et al. 1996), the damage indicator
developed by Yuen (1985) regarding levels 1 and 2 has been studied. In the study, the damage in the cantilever
beam finite element model was represented by the reduction of the modulus of elasticity, and the position of
the damage is successfully identified. In Eq. [2], the eigenparameters were proposed as a damage indicator to
reflect well both the location and the size of the crack in the cantilever for the first mode only:

Φ𝑖𝑑 Φ𝑖
Φ𝑖∗ = 𝑑 2
− 2 (2a)
(𝜔1 ) (𝜔 ) 1
𝜃𝑖𝑑 𝜃𝑖
𝜃𝑖∗ = 𝑑 2
− 2 (2b)
(𝜔1 ) (𝜔1 )

In these expressions, Φ and θ are the mass normalized translational and rotational mode shapes, respectively;
ω is the natural frequency; subindex 1, i represents the first mode and the Degree of Freedom (DOF) number,
respectively, superscript d is a mark for the properties of the damaged system. The method proposed by Yuen
(1985) is applied only to a cantilever beam. The method is tested for the simply supported beam by Dong
(1994). The results show that due to symmetry in the system, the location of the damage cannot be determined
adequately.

In this study, the eigen-parameters in Eq. [2] and the damage detection method is enhanced in which (i) the
mode shape is not normalized by mass matrix but by the reference DOF (ii) instead of the eigenvalue in the
denominator, the natural-frequency is used and (iii) the difference between the DOF of the mode shape is used
instead of the mode shape. With this simplification, it is not necessary to know or calculate the mass matrix
and damage identification method can be utilized for different types of structures and damage can be identified
with other modal information. The following steps were followed for the proposed damage identification
method:

i. The absolute values of the mode shape differences are summed (to utilize the damage indicator for other
modes)

ΔΦ𝑗,𝑖 = |Φ𝑗,𝑖 − Φ𝑗,𝑖−1 | (3)

𝑖

𝜙̃𝑗,𝑖 = ∑ ΔΦ𝑗,𝑘 (4)

𝑘=1
ii. The mode-shape like 𝜙̃𝑗 vector is normalized by its value 𝜙̃𝑗,𝑅 at the reference DOF and the the differences
between consecutive values of 𝜙̃𝑗 vector are calculated.

𝜙̃𝑗,𝑖
𝜙̌𝑗,𝑖 = (5)
𝜙̃𝑗,𝑅
Δ𝜙̌𝑗,𝑖 = 𝜙̌𝑗,𝑖 − 𝜙̌𝑗,𝑖−1 (6)

iii. The proposed damage indicator can be calculated as:

∗
Δ𝜙̌𝑗,𝑖
𝑑
Δ𝜙̌𝑗,𝑖
𝜙𝑗,𝑖 = − (7)
𝜔𝑗𝑑 𝜔𝑗
iv) The damage parameter can also be calculated as the sum of several modes:

558
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑛𝑚
Δ𝜙̌𝑗,𝑖
𝑑
Δ𝜙̌𝑗,𝑖
𝜙𝑖∗ = ∑( 𝑑 − ) (8)
𝜔𝑗 𝜔𝑗
𝑗=1
v) (Optional) Both damage indicators may be normalized by their maximum values as
∗
𝜙𝑗,𝑖 𝜙𝑖∗
𝜙̅𝑗,𝑖 = ̅
∗ or 𝜙𝑖 = (9)
max 𝜙𝑗,𝑖 max 𝜙𝑖∗
𝑖 𝑖
It should be noted that by the normalization step in Eq. [9], information regarding the extent of damage is lost.
In Case Study 1, Eq. [9] is required to provide summary graphs regarding the location of the damage. Another
important issue is that if rotations are explicitly defined as DOFs, a rotational modal shape may be used
instead of translational mode shapes.

Since the proposed damage indicators are associated with how the mode shape changes across nodes, it is
possible to say that the probable damage is between the detected measurement point and the previous
measurement location. In case studies, both translational and rotational damage indicators are examined and it
is found that the rotational damage indicator was more suitable for optimization.

3 CASE STUDIES

Case study I: Numerical Proof of Damage Indicator in Closed Systems

A case of the known damage states and the known system where each variable can be accurately calculated, is
considered. The main aim is to examine the effectiveness of the proposed damage detection method under a
number of damage conditions of a finite element of a cantilever beam model. The numerical model is a 200
DOF system with 100 beam elements in series as shown in Figure 1. For every element, rotational and
translational DOFs are defined at nodes. Section properties and modulus of elasticity of the beam elements are
identical with system used in the previous case study. In the console systems, the numbering is done from the
fixed end to the free end. In the numerical proof, it is desired that all modal parameters are accurate and that
possible errors due to the system identification step do not play a role. Therefore, natural-frequency and mode
shapes have been found by solving the eigenvalue problem of the related systems. Firstly, modulus of
elasticity of each element is reduced one by one and the damage position was determined in the case of 5%
stiffness reduction (damage). The fact that the magnitude of the value obtained in the damage position is
several times greater than the value obtained in the undamaged positions, enables the damage location to be
easily identified.

Figure 1.the beam model and the section properties

559
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The damage indicator value of the damaged element calculated using Eq. [9] is always 1. In cases where each
element is damaged separately, the largest and the smallest values of the damage indicator taken other than the
damaged position are plotted in Figure 2. In this way, the horizontal axis indicates the position of the damage;
the vertical axis shows the largest and smallest values of the damage indicator in undamaged positions. It is
understood that the value of the damage indicator in the damaged position in each case is approximately 60-
100 times greater than values in the undamaged locations omitting the element at the free end and 6 to 20
times greater than values in the undamaged locations at the element at the free end.

It is observed that the damage indicator interval for the undamaged elements varies according to accelerometer
mass. As the accelerometer mass increases near the free end, it is observed that this range expands and the
positive values get closer to the highest value. However, it is seen that the element at the free end receives the
greatest damage indicator value at the value with the lowest accelerometer mass.

The first mode damage indicator, the first two modes and the first three modes damage indicators are
calculated for all damage cases using Eqs. [7-9]. In the elements between 11/20 and 16/20 part of the beam,
the damage indicator interval is higher in the first mode damage indicator, while it is close and lower in the
other two damage indicators as can be seen in Figure 3. However, the first mode damage marker appears to be
more reliable between the 16/20 of the beam and the free end. Between the fixed end and the 11/20 of the
beam, all three damage indicators are calculated very close to the values.

Figure 2. Damage Indicator Range for Undamaged Elements for Varying Mass of Accelerometers

Figure 3. Damage Indicator Range for Undamaged Elements for Varying Accelerometers Mass 0.06 of the Beam Mass for
the Different Damage Indicators.

560
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Case study II: Accelerometer Mass Effect

In this section, damage detection based on different damage parameters and accelerometer mass is investigated
for the element at free end. In the previous section, in case the element at the free end is damaged, in contrast
to what is expected, the largest undamaged element damage parameter value (error) for the lowest mass
accelerometer was seen. For this reason, at the free end, the error is compared according to the mass of the
accelerometer according to the damage indicator values in Figure 4. In determining the damage in the last
element by utilizing the first mode damage indicator, the error increases as the accelerometer weight
decreases. However, in cases where the accelerometer mass is 0.08 times heavier or more than the mass of the
beam, the first mode and the first two modes seem to behave similarly and they are better damage indicator
than the third damage indicator. While the accelerometer weight is 0.08 times smaller than the weight of the
beam, the first two and first three modes damage indicators appear to have fewer errors than the first mode
damage indicator.

Figure 4. Accelerometer Loading Effect.

4. CONCLUSIONS

This study has three important objectives: (i) to propose a damage indicator applies to closed systems, (ii) to
examine the effect of the accelerometer mass used on the damage parameter.

In the damage detection section, for the 100-element and 200-freedom finite element model of a cantilever
beam with translational and rotational modes, it has been shown that it is possible to detect the damage in any
position by using only rotation modes. Moreover, the effect of the mass of the accelerometer on the damage
identification is investigated. Three different damage parameters (Eq. [8]) are used: the first mode, the first
two modes and the first three modes damage indicators. For all sections of the beam, the least error is observed
by utilizing the first two mode damage indicator. It is observed that the accelerometer mass does not have a
serious effect if the damage is between the middle element and the fixed part. In the other half, if the ratio of
the accelerometer mass to the beam mass is 0.4, it is observed that the damage determination is adversely
affected. It is suggested that if the damage is in the free end element, the accelerometer mass can be more than
0.08 times the mass of the beam.

REFERENCES

[1] Ayres, J.W., Lalande, F., Chaudhry, Z. and Rogers, C.A. Qualitative impedance-based health monitoring of civil
infrastructures. Smart Mater. Struct. 7(5), 599, 1998. https://doi.org/10.1088/0964-1726/7/5/004.
[2] Brünig, M. An Anisotropic Continuum damage Model: Theory and Numerical Analyses. Latin American Journal of
Solids and Structures 1, no. 2: 185-218, 2004.

561
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[3] Doebling, S.W., Farrar, C.R., Prime, M.B. and Shevitz, D.W. Damage identification and health monitoring of structural
and mechanical systems from changes in their vibration characteristics: a literature review. Research Report No. LA-
13070-MS; Los Alamos National Lab,, NM, United States, 1996.
[4] Dong, C., Zhang, P.Q., Feng, W.Q. and Huang, T.C. The sensitivity study of the modal parameters of a cracked beam.
Proceedings of the 12th International Modal Analysis, Schenectady, New York, USA, January,1994.
[5] Fassois, S. D. and Sakellariou, J. S. Statistical Time Series Methods for SHM. In Encyclopedia of Structural Health
Monitoring. (eds Boller, C., Chang, F. and Fujino, Y.), 2009. https://doi.org/10.1002/9780470061626.shm044.
[6] Guo, J., Xie, X., Bie, R., Sun, L. Structural health monitoring by using a sparse coding-based deep learning algorithm
with wireless sensor networks. Personal and ubiquitous computing. Dec 1;18(8):1977-87, 2014.
[7] Krishnanunni, C.G., Raj, R.S., Nandan, D., Midhun, C.K., Sajith, A.S., Ameen, M. Sensitivity-based damage detection
algorithm for structures using vibration data. Journal of Civil Structural Health Monitoring. 13;9(1):137-51, 2019.
[8] Lehmann, T. Some thermodynamical considerations on inelastic deformations including damage processes. Acta
mechanica 79, no. 1-2: 1-24, 1989.
[9] Peeters B. System identification and damage detection in civil engeneering. Ph.D. Dissertation, Katholieke Universiteit
Leuven, 2000.
[10] Ren, W.X. and Zong, Z.H. Output-only modal parameter identification of civil engineering structures. Structural
Engineering and Mechanics. 17(3-4), 429-444, 2004. http://dx.doi.org/10.12989/sem.2004.17.3_4.429.
[11] Riva, A., Giorcelli, E. Dynamic System Identification by Means of Neural Networks. Proceedings of the 10th
International Modal Analysis Conference, pp. 928-933, 1992.
[12] Rytter, A.Vibrational based inspection of civil engineering structures. Ph.D. Dissertation, Aalborg University, 1993.
[13] Tufan, T., Akalp, S. Modal plot—System identification and fault detection. Structuctural Control and Health
Monitoring, e2347, 2019.
[14] Vashisht, R., Viji, H., Sundararajan, T., Mohankumar, D., Sumitra S. Structural Health Monitoring of Cantilever
Beam, a Case Study—Using Bayesian Neural Network and Deep Learning. InStructural Integrity Assessment (pp.
749-761). Springer, Singapore, 2020.
[15] Yang, J.N.: Application of optimal control theory to civil engineering structures. J. Eng. Mech. Div.ASCE, 101(6),
819-838, 1975.
[16] Yuen, M.M.F.A Numerical Study of the Eigenparameters of a Damaged Cantilever. J Sound Vib, 103, 301–310, 1985.
https://doi.org/10.1016/0022-460X(85)90423-7.

562
 FORECASTING THE POWER PRODUCTION FROM A MICRO
GRID PV SYSTEM

Penka V. Georgieva
Burgas Free University, Burgas, Bulgaria, [email protected], ORCID: 0000-0001-9598-5997

Georgieva, P. V. Forecasting the Power Production from a Micro Grid PV System. 8th Eur.
Cite this paper as:
Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In this paper, an AI approach for short-term predicting the electricity power production of a micro
grid PV system is proposed. In the presented forecasting model, the input variables are three: solar
radiation, wind speed and air temperature; and the output variable is one: the quantity of electrical
power produced from the installed PV panels in the micro system. An artificial neural network has
been used in the implementation of the model. This research is part of a project for optimizing the
energy consumption of a building with the use of independent alternative renewable energy
sources. The model is implemented and tested on real data, collected in 5 minutes from a
specifically designed meteorological station which is installed on the building of Burgas Free
University.
Keywords: artificial neural network, micro grid PV system, smart building management
© 2020 Published by ECRES

Nomenclature
PV (system) photovoltaic (system)
ANN artificial neural network
BFU Burgas Free University

1. INTRODUCTION

Photovoltaic (PV) systems are designed to convert solar radiation into electricity through photovoltaic elements.
PV systems’ capacity ranges from several kilowatts to hundreds of megawatts. Currently, the commercial
photovoltaic systems are connected to the electricity grid, while stand-alone systems make up a small portion of
the market share. [1]
Generally, a micro grid PV system consists of three modules: (1) a power source including the photovoltaic
panels and the rechargeable batteries; (2) a controller and an AC inverter; (3) consumers - the standard AC
consumers and the DC consumers. Loads are divided into two main groups –consumers receiving energy directly
from the batteries and consumers receiving energy inverted into AC voltage. [2]
For the correct operation of the PV panels and their proper coordination with the consumers, the systems are
provided with a controller for monitoring the proper charging/discharging of the batteries, the operation of the
photovoltaic panels and detecting emergencies. The controller is the most important element of the system and
the proper choice and appropriate software are key for the smooth operation of the system. [3]
The operational characteristics of the electricity produced by the PV system are out of the scope of this paper.
In [1], two fuzzy inference systems for supporting the process of managing micro grid PV systems are proposed.
The amount of power produced by PV panels is an important parameter in each of these management systems
and its successful prediction increases significantly the systems’ efficiency.
Predicting is a key element of the process of decision making. For example, different methods used for solar PV
power forecasts are outlined in [4].
The development of computer technologies leads to the emergence of new methods of predicting, e.g. fuzzy
systems. Unlike in traditional computing (hard computing), the keystone in modern artificial intelligence (AI)
is the understanding that accuracy and security have a high price and tolerance for imprecision and uncertainty

563
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

in calculations, reasoning and decision making should be admitted (when and where possible). The current
computing paradigms of AI are Fuzzy Logic, Neural Networks, approximate conclusions and non-differential
optimization methods such as Genetic Algorithms and Simulated Annealing. [5]
А detailed overview of applications of AI techniques as “alternate approaches to conventional techniques or as
components of integrated systems” can be found in [6].
Nowadays AI techniques are widely used in managing various stand-alone PV plants, off the grid PV systems
and different hybrid systems for generating electricity by using photovoltaics. [7-11]
This study is focused on the predicting the amount of electric power produced by photovoltaic panels installed
on a single building and connected with the grid. The purpose is to use this forecast to effectively manage the
energy flows of the building. The forecasting model is a neural network, trained on real data collected from the
micro energy system, installed on the building of Burgas Free University.

2. THE MICRO GRID PV SYSTEM

The predicting model, proposed in this paper, aims to support a more efficient management system of a micro
grid PV system. The research is part of a project for optimizing the energy consumption of a building with the
use of independent alternative renewable energy sources. The input data for solar radiation, wind speed and air
temperature are received from BFU METEO - a meteorology station specially designed and installed on the
building of Burgas Free University for the needs of this project. BFU METEO consists of sensors for monitoring
the input variables, Fronius Sensor Box, and FroniusDatalogger Web. Data is collected in a 5-minutes interval
and stored in a database in a server.
The micro grid PV system consists of 20 polycrystalline PV panels, with max power approx. 235Wp each, an
inventor SolarMax 13MT2 and 10 monocrystalline PV panels, with max power approx. 150Wp each; an invert
or IBC Serve Master 3300 MV. Some of the components of the micro grid PV system are shown in Fig. 1.

Fig. 1. Components of the micro grid PV system

The data are collected from the BFU METEO sensors every 5 minutes and then automatically stored in the
database. The data for the quantity of the electric power produced by two types of installed PV panels
(polycrystalline and monocrystalline) is also collected and stored in the database at 5-minute intervals. (Fig. 2)
[12]

Fig. 2. Data flow in managing the micro grid PV system

564
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. THE FORECASTING MODEL

Machine Learning
Machine learning (ML) is a branch in computer science for implementing algorithms capable of “learning” how
to perform a task without being explicitly programmed for it. ML is a broad term for adaptive algorithms able
to “learn” from examples and find patterns in data. They are used in the process of decision making, as well as
for predictions in various areas: medical diagnosis, computer vision, face recognition, stock trading, credit
scoring, energy performance forecasting, computational biology, natural language processing, etc. ML is
preferred when: 1) difficulties in formal descriptions of knowledge exist; 2) the future states are unpredictable
and 3) the objects of interest are dynamically changing, and it is difficult or impossible to predict all the changes
over time.
Two are the main techniques used in ML: supervised and unsupervised learning. In the supervised learning, the
training is carried out on known input and output data so future outputs from new data are predicted, while in
unsupervised learning hidden patterns are sought and eventually found in input data. (Fig. 3)
Supervised ML aims at building models based on evidence under the conditions of some source of uncertainty.
A supervised learning algorithm uses a known set of input data and known output data and “teaches” the model
to generate feasible results from new data. Supervised learning uses regression techniques and classification to
develop predictive models.
Regression techniques are used to predict continuous responses (e.g. changes in temperature or in electricity
demand).
Classification techniques are mainly used for predicting future values of discrete variables by classifying the
input data into categories (e.g. whether in a picture there is an animal or plant or human).
Unsupervised ML finds hidden patterns or intrinsic structures in data. It is used to draw inferences from datasets
consisting of input data without labeled responses. Clustering is the most common unsupervised learning
technique. It is used for exploratory data analysis to find hidden patterns or groupings in data.
Applications for clustering include gene sequence analysis, market research, and object recognition.
The selection of a machine learning algorithm is a process of trial and error, and is subject to finding a balance
between the algorithms characteristics: speed; memory usage; predictive precision; interpretability.

Fig. 3. Main techniques in machine learning

Commonly Used Regression Algorithms
Logistic Regression is a predictive analysis tool and is used to explain dependencies between one dependent
binary variable and one or more independent variables. Logistic regression is often used as a starting point for
binary classification problems.
Linear Regression is a statistical modeling technique used to describe a continuous response variable as a linear
function of one or more predictor variables. Because linear regression models are simple to interpret and easy
to train, they are often the first model to be fitted to a new dataset. Linear regressions are easy to interpret and
fast to fit and thus can be used as first stage of building more complex regression models.
Nonlinear Regression is a statistical modeling technique that is used for describing nonlinear relationships in
experimental data. Nonlinear regression models are assumed to be parametric, where the fit function is a
nonlinear function of the parameters. This type of regressions is commonly used when data have nonlinear trends
and cannot be easily transformed into a linear space.

565
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Artificial Neural Networks (ANNs) are inspired by the human brain, and thus consist of highly connected
networks of neurons. The neurons and the connections are taught to map the input data onto the desired outputs
by iteratively modifying the weights of the connections. ANNs are best used for modeling highly nonlinear
systems, when there are unexpected changes in the input data, when model interpretability is not a major
requirement.
The biological model of a neural network is interesting for data processing because it operates in a similar to the
human brain way. Humans are very good at recognizing objects and faces, recognizing the environment, solving
various situational problems, actions and reactions because of the organization and functioning of the brain.
From birth, the brain can develop its own rules based on previous experience. These rules are made by
strengthening the old and creating new connections between neurons. The human brain has billions of neurons
that are interconnected and communicate through electrochemical signals.
The performance of ANNs is closely resembling that of the human brain and is designed to solve complex
problems in machine learning. ANNs training is not done for each neuron individually but as a system globally.
All neurons learn together to build a network that can solve problems with a high degree of difficulty. One of
the most common types of neural networks is the back-propagation neural network. The reverse propagation of
the error is implemented after each iteration. The activation function for each neuron is determined by the value
of the output from the previous layer of neurons and the weights of the connections between them. Firstly, error
in the output (the discrepancy between the expected and the actual result) is calculated (forward propagation).
Once the error is known, its minimization is achieved by backpropagation. This means that the algorithm
propagates backwards, from the output layer to the input layer, and along the way lies an error for each of the
weights between the neurons. These values must be modified to minimize the overall error [13].
The ANN Forecasting Model
The Deep Learning Toolbox™ (formerly Neural Network Toolbox™) provides a framework for designing and
implementing deep neural networks with algorithms, pretrained models, and apps. Convolutional neural
networks and long short-term memory networks are used for classification and regression on images, time-
series, and text data. Advanced network architectures can be developed by using custom training loops, shared
weights, and automatic differentiation. Apps and plots visualize activations and thus support editing and
analyzing network architectures and monitor training progress. Models can be exchanged in TensorFlow™ and
PyTorch through the ONNX format and imported from TensorFlow-Keras and Caffe. The toolbox supports
transfer learning with a library of pretrained models. The training can be accelerated on a single- or multiple-
GPU workstation (with Parallel Computing Toolbox™), or scale up to clusters and clouds, including
NVIDIA®GPU Cloud and Amazon EC2®GPU instances. [14]
The selected for this research MatLab-implemented function for fitting a neural network with a specified hidden
layer size and a predefined training function is:

fitnet(hiddenSizes,trainFcn).
In adjusting the architecture of the ANN the next procedure is followed: 1) data cleaning; 2) launching the neural
network; 3) obtaining results; 4) comparing and adjusting the neural network (number of layers and/or number
of neurons); 5) going to 2).
A general scheme of the ANN forecasting model is shown in Fig. 4.

input ouput

solar radiation

forecast
for
wind speed ANN
electic power
production
air
temperature

Fig. 4 A general scheme of the forecasting model

566
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The first question that must be answered is on what data the neural network will be trained on. As can be seen
from Fig. 5, the 5-min observations show significant fluctuations during day and night. However, although there
are many zero values, they cannot be omitted because without them the ANN loses accuracy. After trials on
different number of days, the best results in training the neural net are obtained when 1-day sequence of 5-min
observations starting at 00:00 am is used (288 observations).
After the implementation of the described model of an ANN in MatLab, experimental tests have been carried
out. The tests aim at studying the behavior of the ANN while changing the values of the parameters of its
architecture – the number of hidden layers and the number of neurons in each layer.

Fig. 5. Visualization of 5-min observations for 6 consecutive days: Solar radiation (left axis); Wind speed (right axis); Air
temperature (left axis); Power production (right axis)
After the conducted numerous tests with different architectures of the ANN, best results are obtained from a
neural network with the following parameters:
fitnet([10,5,3,10],trainlm),
which is an ANN with four hidden layers with respectively 10, 5, 3, and 10 neurons and Levenberg-Marquardt
backpropagation algorithm (Fig. 6). Details about this algorithm can be found in [15].

Fig. 6. The architecture and the achieved performance of the ANN

567
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The best validation performance achieved after 13 iterations (epochs) is 13.246, the corresponding gradient is
10,2775.
The error histogram and fluctuations after training, validations and testing the ANN is shown in Fig. 7.

Fig. 7. Error histogram and fluctuations after training, validations and testing the ANN

4. CONCLUSION
In this article a predicting model for supporting the process of managing micro grid PV systems is proposed.
The model is based on machine learning. An artificial neural network has been used in the implementation of
the model for predicting the electrical power production of a micro grid PV system. The trained ANN shows
satisfactory results that can be used in managing the energy flows of the micro PV system.
The training and the tests are conducted in MatLab. This research is part of a project for optimizing the energy
consumption of a building with the use of independent alternative renewable energy sources. The input test data
for the optimization are the amount of used energy for lighting, heating, computers power supply and other
needs of one particular building (the building of Burgas Free University).
A useful development of the presented forecasting model is to create an autonomous software for managing
photovoltaic systems. Such a program will be widely used given the pace of development of this technology
around the world.
Optimizing PV systems with respect to their cost is another priority in the development of the proposed models.

ACKNOWLEDGMENT
This research is funded by National Research Fund of Bulgaria under Contract No. KP-06-COST-8/ 06.08.2019
for providing national co-financing for the participation of Bulgarian teams in approved actions under the
European program for cooperation in the field of research and technology COST under the project
"Characteristics prediction and optimization of a photovoltaic system with artificial intelligence methods”.

REFERENCES

[1] Georgieva, P. V., Dolchinkov, R. Fuzzy Models for Managing a Micro Grid PV System. In: 42nd International
Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO), 2019, pp.
1066-1071
[2] Skoplaki, E., Palyvos, J. A. Operating temperature of photovoltaic modules: A survey of pertinent correlations.
Renewable Energy, Vol. 34 № 1, 2009, pp. 23-29
[3] Dolchinkov, R., Georgieva P., Effectiveness of solar tracking systems. Yearbook of Burgas Free University, Vol.
XXVII, 2012 (in Bulgarian)
[4] Tuohy, A. and coauthors, Solar forecasting: Methods, challenges, and performance. IEEE Power and Energy Magazine
vol. 13 (6), 2015, pp. 50–59
[5] Georgieva, P. V., Fuzzy rule-based systems for decision-making. Engineering Sciences, Bulgarian Academy of
Sciences, vol. 53 (1), pp. 5–16, 2016
[6] Mellit, A., Kalogiroub, S. Artificial intelligence techniques for photovoltaic applications: A review. Progress in Energy
and Combustion Science vol. 34, 2008, pp. 574–632
[7] Kardakos, E., Alexiadis, M., and coauthors, Application of time series and artificial neural network models in short-
term forecasting of PV power generation. 48th International Universities' Power Engineering Conference (UPEC),
2013

568
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[8] Welch, R., Ruffing, S., Venayagamoorthy, G. Comparison of feedforward and feedback neural network architectures
for short-term wind speed prediction. Proceedings of international joint conference on neural networks, 2009
[9] Huang, Y., Lu, J. and coauthors, Comparative study of power forecasting methods for PV stations. International
Conference on Power System Technology (POWERCON), Hangzhou, China, pp. 1 – 6, 2010
[10] Wang T., Su-Yuan Tsai, Solar Panel Supplier Selection for the Photovoltaic System Design by Using Fuzzy Multi-
Criteria Decision-Making Approaches. Energies vol. 11(8), 2018
[11] Lee, A., H.-Y. Kang, Lin, C.-Y., Shen, K.-C. An integrated decision-making model for the location of a PV solar plant.
Sustainability, vol. 7(10), 2015, pp. 13522–13541
[12] Georgieva, P., Popchev, I. P., Stoyanov. S. N. Cybernetics and Information Technologies, Vol. 15, No 3, 2015, pp.
41-51.
[13] M. A. Nielsen, Neural Network and Deep learning, Determination Press, 2015.
[14] https://www.mathworks.com/products/deep-learning.html
[15] https://se.mathworks.com/help/deeplearning/ref/trainlm.html;jsessionid=b8f21ff67e433e3e6fe7f697887f

569
 PMSM PARAMETER ESTIMATION USING A KIEFER-
WOLFOWITZ BASED SEARCH ALGORITHM

Artun SEL
TOBB University of Economics and Technology, Ankara, Turkey, [email protected], ORCID: 0000-0002-3142-
3650

Cosku KASNAKOGLU
TOBB University of Economics and Technology, Ankara, Turkey, [email protected],ORCID: 0000-0002-9928-
727X

Sel, A, Kasnakoglu, C. PMSM Parameter Estimation Using a Kiefer-Wolfowitz Based Search

Cite this paper as:
Algorithm. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In this paper, a PMSM parameter estimation procedure is presented. The parameter estimation is
performed using the Kiefer-Wolfowitz Algorithm which is a gradient-based stochastic
approximation method. For the parameter estimation, an assumed nonlinear PMSM model is
considered which contains nonlinearity. For the aforementioned algorithm to be applicable, a
couple of manipulations are carried out. After the problem is posed in a way that is suitable to
implement gradient-based algorithm, parameters estimation can be done. The manipulations that
are done on the system and the implementation of the algorithms are given in detail. Furthermore,
the noise tolerance of the algorithm is also studied, and the results are interpreted. The algorithm
is tested on a numerical computation environment, MATLAB and the efficacy of the parameter
estimation and other scenarios where the algorithm can be applied is discussed.
Keywords: PMSM, Kiefer-Wolfowitz Algorithm, Nonlinear Systems, System Identification
© 2020 Published by ECRES

1. INTRODUCTION

PMSM(permanent magnet synchronous motor) is a type of AC motor that is widely used in the industry due to
the decreasing cost of the magnetic materials and its ability to generate greater torque in comparison to its
counterparts in same volume and weight. In addition, since PMSM does not possess slip speed between the
magnetic and the electrical fields, this simplifies the control algorithms that are needed [1]. These factors
amounts to the fact that PMSM is highly preferred in the renewable energy applications such as wind farms and
in the EV(electrical vehicle) industry [2]. For the speed or rotor position control of PMSM, often sensorless
control algorithms are preferred. Sensorless control indicates that the control algorithm does not require an Hall
effect sensor for the measurement of the motor magnetic field and only operates under the knowledge of the
stator current signals [3]. FOC(field oriented control) [4] and DTC(direct torque control) are two prominent
sensorless controls that can be named [5].

For the control algorithms that require some dependence on the plant model, and for the reasons of maintenance
of a large-scale system, it is desirable to keep track of the plant parameters using a reliable system identification
method [8]. In addition to the model-based control algorithms and a maintenance needs, system identification
methods are also employed in fault detection schemes [6,7]. However, model-based control design stands out as
the prominent reason of the utilization of a system identification [9].

Most linear control algorithms require a model of the system which results in a system identification method to
be implemented on an experimental data, and an estimated system model is used in the generation of the control
parameters. This type of control assumes that the plant in question maintains its parameters or neglects the
change in time-varying parameters. There are robust control algorithms in the literature that are designed to

570
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

maintain the closed-loop operation in spite of the variation of those parameters, which is achieved by modeling
this variations as a disturbance and designing a control law that is able to suppress the maximum assumed
disturbance due to the parameter variations [10-14]. However, for the systems where a reliable linear or
nonlinear state dynamics are known and a presence of a set of parameters that vary with time in the dynamics,
adaptive control algorithms are also preferable. Adaptive control algorithms which contain model update rules
in their structure have the capacity to mitigate the effects of the time-varying parameters [15]. Model update
rules are designed based on parameters estimation techniques, if a parameters set is known to be varying with
time. Depending on the speed of the variation of a parameter that is to be estimated in the model update rule part
of an adaptive control, the estimation can be performed sequentially or in a batch process. Beside parameter
varying plants, there are systems whose dynamics change structurally due to an external effects rather than a
parameter variation, and this change in dynamics is complicated to model that depends on a parameter, which
requires the adaptive control algorithm to account for that structural change rather than a parameter dependent
change using a specified system identification method [16]. ERA(eigensystem realization algorithm) is one of
the algorithms that estimates the whole dynamics rather than a parameter estimation by computing the state
space system matrices by using the input and output signal measurements. ERA produces system matrices
however it is not directly preferred in the parameter estimation [17].

In this paper, a parameter estimation algorithm which is based on the Kiefer-Wolfowitz Algorithm for PMSM
is discussed [18]. For the parameter estimation, an assumed simplified nonlinear model is considered, however
this method can be extended to the cases where an assumed nonlinear model is not readily available. For the
Kiefer-Wolfowitz Algorithm to be implementable to the problem, using the dynamics of the plant, a couple of
manipulations are required which is to be detailed in the next section. The performance of the algorithm is tested
in a numerical computation environment, MATLAB and the robustness against the measurement noise is
discussed.

2. THE PMSM NONLINEAR MODEL

In this section a simplified nonlinear model of PMSM is given [19]. For the sake of brevity, a simplified
version of the rotating reference frame model is studied, although the parameter estimation algorithm can be
applied to the problem in a more general setting. The dynamics of the four state of the plant is given as,

𝑑 𝑅 𝜆 1 (1)
𝑖𝑎 = (− ) 𝑖𝑎 + ( ) 𝜔 sin(𝜃) + ( ) 𝑢𝑎
𝑑𝑡 𝐿 𝐿 𝐿
𝑑 𝑅 𝜆 1 (2)
𝑖𝑏 = (− ) 𝑖𝑏 + (− ) 𝜔 cos(𝜃) + ( ) 𝑢𝑏
𝑑𝑡 𝐿 𝐿 𝐿
𝑑 3𝜆 3𝜆 𝐹 (3)
𝜔 = (− ) 𝑖𝑎 sin(𝜃) + ( ) 𝑖𝑏 cos(𝜃) + (− ) 𝜔
𝑑𝑡 2𝐽 2𝐽 𝐽
𝑑 (4)
𝜃=𝜔
𝑑𝑡
In the study, the plant is considered as a system which has 5 parameters to be determined. There are cases
where a part of the parameters comprising the parameter set is known to a sufficient certainty.

Table I. The plant parameters and definitions

Parameter Unit Definition
𝑅 Ω Winding Resistance
𝐿 𝐻 Winding Inductance
𝜆 - Magnetic flux constant of
𝐽 𝑘𝑔 𝑚2 Moment
the motorof inertia of the
𝐹 - Viscous
rotor andfriction
the loadconstant
of the rotor

571
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

There are 5 parameters given in Table I, all of whom are to be estimated in the problem setting.

The states and their definitions are given in Table II,

Table II. The plant states and definitions

State Unit Name
𝑖𝑎 𝐴 Direct-axis stator winding
𝑖𝑏 𝐴 current
Quadrature-axis stator
𝜔 𝑟𝑎𝑑/𝑠 Rotor mechanical
winding current rotation
𝜃 𝑟𝑎𝑑 Rotor
speed mechanical position

For the concerns of tractability, the parameters and states are denoted with 𝑝 and 𝑥 with their respective
subscripts. The system model is then given by,
𝑑 𝑝1 𝑝3 1 (5)
𝑥 = (− ) 𝑥1 + ( ) 𝑥3 sin(𝑥4 ) + ( ) 𝑢1
𝑑𝑡 1 𝑝2 𝑝2 𝑝2
𝑑 𝑝1 𝑝3 1 (6)
𝑥2 = (− ) 𝑥2 + (− ) 𝑥3 cos(𝑥4 ) + ( ) 𝑢2
𝑑𝑡 𝑝2 𝑝2 𝑝2
𝑑 3𝑝3 3𝑝3 𝑝5 (7)
𝑥3 = (− ) 𝑥1 sin(𝑥4 ) + ( ) 𝑥2 cos(𝑥4 ) + (− ) 𝑥3
𝑑𝑡 2𝑝4 2𝑝4 𝑝4
𝑑 (8)
𝑥 = 𝑥3
𝑑𝑡 4
Using forward Euler discretization, the discrete state space dynamics are obtained as,

𝑝1 𝑝3 1 (9)
𝑥1 (𝑘 + 1) = (− 𝑇𝑠 + 1) 𝑥1 (𝑘) + ( 𝑇𝑠 ) 𝑥3 (𝑘) sin(𝑥4 (𝑘)) + ( 𝑇𝑠 ) 𝑢1 (𝑘)
𝑝2 𝑝2 𝑝2
𝑝1 𝑝3 1 (10)
𝑥2 (𝑘 + 1) = (− 𝑇𝑠 + 1) 𝑥2 (𝑘) + (− 𝑇𝑠 ) 𝑥3 (𝑘) cos(𝑥4 (𝑘)) + ( 𝑇𝑠 ) 𝑢2 (𝑘)
𝑝2 𝑝2 𝑝2
3𝑝3 3𝑝3 𝑝5 (11)
𝑥3 (𝑘 + 1) = (− 𝑇𝑠 ) 𝑥1 (𝑘) sin(𝑥4 (𝑘)) + ( 𝑇𝑠 ) 𝑥2 (𝑘) cos(𝑥4 (𝑘)) + (− 𝑇𝑠 + 1) 𝑥3 (𝑘)
2𝑝4 2𝑝4 𝑝4
𝑥4 (𝑘 + 1) = 𝑇𝑠 𝑥3 (𝑘) + 𝑥4 (𝑘) (12)
Using the parameter set defined as,

𝑝1 𝑝3 1 (13)
[𝓅1 𝓅2 𝓅3 ] ≔ [(−
𝑇𝑠 + 1) ( 𝑇𝑠 ) ( 𝑇𝑠 )]
𝑝2 𝑝2 𝑝2
𝑝1 𝑝3 1 (14)
[𝓅4 𝓅5 𝓅6 ] ≔ [(− 𝑇𝑠 + 1) (− 𝑇𝑠 ) ( 𝑇𝑠 )]
𝑝2 𝑝2 𝑝2
3𝑝3 3𝑝3 𝑝5 (15)
[𝓅7 𝓅8 𝓅9 ] ≔ [(− 𝑇) ( 𝑇 ) (− 𝑇𝑠 + 1)]
2𝑝4 𝑠 2𝑝4 𝑠 𝑝4
[𝓅10 𝓅11 ] ≔ [𝑇𝑠 1] (16)

Finally, the state dynamics are obtained, where the parameters {𝑝1 , 𝑝2 , 𝑝3 , 𝑝4 , 𝑝5 } are to be estimated. The
estimation procedure requires a pre-processing of the dynamics given in Eqs [9-12].

3. THE PARAMETER ESTIMATION PROCEDURE

In this section, for the parameter estimation algorithm to be applicable to the problem, a series of manipulations
are conducted to simplify the estimation process. The dynamics are given in Eqs [9-12]., and the dynamics are
nonlinear for the 4 states. For the ease of manipulation, 4 new states are defined as,

𝑥5 (𝑘) ≔ 𝑥1 (𝑘) sin(𝑥4 (𝑘)) (17)

𝑥6 (𝑘) ≔ 𝑥2 (𝑘) cos(𝑥4 (𝑘)) (18)

572
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑥7 (𝑘) ≔ 𝑥3 (𝑘) sin(𝑥4 (𝑘)) (19)

𝑥8 (𝑘) ≔ 𝑥3 (𝑘) cos(𝑥4 (𝑘)) (20)
At this point a new function is defined, to simplify the derivations ahead,

𝑥1 (𝑘) 𝑥1 (𝑘) (21)

𝑥2 (𝑘) 𝑥2 (𝑘)
𝑥3 (𝑘) 𝑥3 (𝑘)
𝑥4 (𝑘) 𝑥4 (𝑘)
𝑓𝑖𝑛 − − − ≔ −−−
𝑥5 (𝑘) 𝑥1 (𝑘) sin(𝑥4 (𝑘))
𝑥6 (𝑘) 𝑥2 (𝑘) cos(𝑥4 (𝑘))
𝑥7 (𝑘) 𝑥3 (𝑘) sin(𝑥4 (𝑘))
([ 𝑥8 (𝑘) ]) [𝑥3 (𝑘) cos(𝑥4 (𝑘))]
Using this inner-nonlinearity function, a new system is defined as,

𝑥1 (𝑘 + 1) 𝑥1 (𝑘) (22)
𝑥2 (𝑘 + 1) 𝑥2 (𝑘)
(
𝑥3 𝑘 + 1 ) 𝑥3 (𝑘)
𝑥4 (𝑘 + 1) 𝑥4 (𝑘)
𝑢 (𝑘)
− − − = [𝐴𝑀 ]𝑓𝑖𝑛 − − − + [𝐵𝑀 ] [ 1 ]
𝑢2 (𝑘)
𝑥5 (𝑘 + 1) 𝑥5 (𝑘)
𝑥6 (𝑘 + 1) 𝑥6 (𝑘)
(
𝑥7 𝑘 + 1 ) 𝑥7 (𝑘)
[𝑥8 (𝑘 + 1)] ([ 𝑥8 (𝑘) ])
Where, state matrices 𝐴𝑀 and 𝐵𝑀 are given as,

𝓅1 0 0 0 0 0 𝓅2 0 𝓅3 0 (23)
0 𝓅4 0 0 0 0 0 𝓅5 0 𝓅6
[𝐴𝑀 ] = 0 0 𝓅9 0 𝓅7 𝓅8 0 0 , [𝐵 ] = − − − −
0 0 𝓅10 𝓅11 0 0 0 0
𝑀 𝟎𝟐×𝟐
−−−−−−−−−−−−−−− − −−−
[ 𝟎𝟒×𝟒 ] [ 𝟎𝟒×𝟐 ]
For the output of the system, 𝐶𝑀 and 𝐷𝑀 matrices are also defined so that the output of the system is the first
four states. Using this change of variables and state extension, the plant dynamics are illustrated by the block
diagram given in Figure 1 and Figure 2.

Figure 1. Block diagram of the plant dynamics given in Eqs [9-12].

573
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. Block diagram of the new defined dynamics given in Eq [22]

States [𝑥1 (𝑘) 𝑥2 (𝑘) 𝑥3 (𝑘) 𝑥4 (𝑘)]𝑇 are considered to be measurable where the measurement noise
𝑣~𝒩 (04×1 , 𝜎 2 𝐼4 ) is present. In the estimation, first, the state vector is measured when the input signal is applied
for a sufficient amount of time and measured state signals are recorded. Using the measured state vector
[𝑥1 (𝑘) 𝑥2 (𝑘) 𝑥3 (𝑘) 𝑥4 (𝑘)]𝑇 , the extended states [𝑥5 (𝑘) 𝑥6 (𝑘) 𝑥7 (𝑘) 𝑥8 (𝑘)]𝑇 are generated using
the rule defined in Eq [21]. This generated extended state variables are denoted with barred notation not to
confuse these states with the states given in Eq [22]. At this stage, the parameter estimation of the system whose
dynamics are given in Eqs [9-12] is equivalent to the estimation of the 𝐴𝑀 and 𝐵𝑀 matrices for the state space
representation of a new defined virtual system, given as

𝑥1 (𝑘 + 1) 𝑥1 (𝑘) (24)
(
𝑥2 𝑘 + 1 ) 𝑥2 (𝑘)
𝑥3 (𝑘 + 1) 𝑥3 (𝑘)
(
𝑥4 𝑘 + 1 ) 𝑥4 (𝑘)
𝑢 (𝑘)
− − − = [𝐴𝑀 ] − − − + [𝐵𝑀 ] [ 1 ]
𝑢2 (𝑘)
𝑥̅ 5 (𝑘 + 1) 𝑥̅ 5 (𝑘)
𝑥̅ 6 (𝑘 + 1) 𝑥̅ 6 (𝑘)
𝑥̅ 7 (𝑘 + 1) 𝑥̅ 7 (𝑘)
(
[𝑥̅ 8 𝑘 + 1 ]) [ 𝑥̅ 8 (𝑘) ]
Since state vector information is available for this system, Eq [24] is arranged as,

𝐴𝑇𝑀 (25)
[𝑥1 (𝑘 + 1) … 𝑥8 (𝑘 + 1)] = [𝑥1 (𝑘) … 𝑥8 (𝑘) │ 𝑢1 (𝑘) 𝑢2 (𝑘)] [ − ]
𝑇
𝐵𝑀
Now that the Eq [25] is in a linear form, the measured terms are stacked on top of each other given as,

𝑥1 (𝑘 = 2) … 𝑥8 (𝑘 = 2) (26)
[ ⋮ ⋮ ⋮ ]
⏟𝑥1 (𝑘 = 𝑁) … 𝑥8 (𝑘 = 𝑁)
𝕏𝑘+1
𝑥1 (𝑘 = 1) … 𝑥8 (𝑘 = 1) │ 𝑢1 (𝑘 = 1) 𝑢2 (𝑘 = 1) 𝐴𝑇𝑀
=[ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ][ − ]
𝑇
⏟𝑥1 (𝑘 = 𝑁 − 1) … 𝑥8 (𝑘 = 𝑁 − 1) │ 𝑢1 (𝑘 = 𝑁 − 1) 𝑢2 (𝑘 = 𝑁 − 1) ⏟𝐵𝑀
𝕏𝑘 𝜃
Where, 𝑁 is the sample point number that is measured. At this stage, the problem is in a form that is suitable for
the Kiefer-Wolfowitz Algorithm. The problem is restated as,

1 (27)
min. 𝐽 (𝜃) = ‖𝕏𝑘+1 − 𝕏𝑘 𝜃‖2𝐹
𝜃 2
𝑠. 𝑡. 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟𝑠 𝑐𝑜𝑛𝑠𝑡𝑟𝑎𝑖𝑛𝑡𝑠
The parameter constrains are the limitation that are imposed by the a priori knowledge on the parameters, it can
be specified for each of the parameters so that each parameter lies in an interval. The solution of this problem is
given by the algorithm who is given in a diagram in Figure 3. Where, in the given algorithm, 𝛼𝑖 denotes the

574
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝜕𝐽(𝜃)
learning rate and the computation of | can be done numerically. After the search algorithm given in Figure
𝜕𝜃 𝜃𝑖
3, a computed 𝜃 vector is used to compute the parameter sets. In the next section, numerical simulations
conducted in Matlab are examined.

Figure 3. Block diagram of the iterative search algorithm

4. THE SIMULATION RESULTS

In this section, a simulation that is conducted to test the parameter estimation for PMSM plant algorithm is
described and the results are interpreted. First, the parameters to be estimated and their values are given in Table
III, where the parameters are chosen arbitrarily.

Table III. The plant parameters and definitions

Parameter Value
𝑅 0.1
𝐿 0.1
𝜆 1
𝐽 1
𝐹 1

The PMSM model is simulated and the noise corrupted measurements are recorded until 𝑁 data point is obtained
for each of the states. Using these measurements the extended state signals are produced in accordance with the
rule that is stated in Eq [21]. In the next step, the data matrices, 𝕏𝑘+1 and 𝕏𝑘 are produced as described in Eq
[26]. Using the cost function that is given in Eq [27] and the search algorithm to solve that, given in Figure 3,
the parameter estimations are obtained as soon as the convergence is achieved.

The system is discretized at sample period of 𝑇𝑠 = 10−3 and the parameter numerical values are given in
TableIII. In the Figure 4, parameter estimation accuracy with respect to the measurement noise variance 𝜎 2 is
illustrated. It can be noted that there might be additional noise eliminating algorithms that can be employed

575
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

before the search algorithm is started to mitigate the effects of the measurement noise. It can be seen that the
resistance and inductance related terms are most affected by the presence of noise.

Figure 4. The parameter estimation accuracy with respect 𝜎 2

Using the estimated parameters, the real and model states are obtained for the case where the variance of noise
is 10−3 , and is given in Figure 5.

(a) (b)

(c) (d)
Figure 5. Real states and model states for (a) state 𝑥1, (b) state 𝑥2 ,(c) state 𝑥3(d) state 𝑥4

To excite the system an input signal must be produced so that the inner dynamics are sufficiently excited which
in turn results in the increase in the accuracy of the estimation procedure.

5. CONCLUSION AND FUTURE WORKS

In the study, Kiefer-Wolfowitz based parameter estimation algorithm is studied and the required procedure to
implement is detailed. The measurement noise aspect of the estimation is addressed in the light of the conducted
numerical simulations. It can be noted that the parameter estimation is achieved to a sufficient degree which is
impeded by the power of measurement noise whose affects can be suppressed using a noise elimination pre-

576
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

processing before the estimation. In this work, it can also be discerned that the complete state estimation
assumption is present which might be reasonable for a class of industrial systems. For the cases where only a
part of the state measurable, ERA based algorithms can be considered in conjunction with the described
algorithm. ERA is applied for the dynamic system identification but there are number of cases where using a
priori knowledge on the internal state dynamics, ERA can be employed to estimate the system parameters.

REFERENCES

[1] R. Mocanu and A. Onea, "Passivity Based Torque Control of PMSM used in electrical vehicles," 2015 19th International
Conference on System Theory, Control and Computing (ICSTCC), Cheile Gradistei, 2015, pp. 803-810.
[2] R. Mocanu and A. Onea, "Passivity Based Torque Control of PMSM used in electrical vehicles," 2015 19th International
Conference on System Theory, Control and Computing (ICSTCC), Cheile Gradistei, 2015, pp. 803-810.
[3] Teiar, Hakim et al. “Almost parameter-free sensorless control of PMSM.” IECON 2015 - 41st Annual Conference of
the IEEE Industrial Electronics Society (2015)
[4] Tino Jerčić, Šandor Ileš, Damir Žarko & Jadranko Matuško (2017) Constrained field-oriented control of permanent
magnet synchronous machine with field-weakening utilizing a reference governor, Automatika, 58:4, 439-449.
[5] Ali Ahmed Adam, Kayhan Gulez & Halit Pastaci (2009) A direct torque control algorithm for permanent magnet
synchronous motors with minimum torque pulsation and reduced electromagnetic interference noise, International
Journal of Electronics, 96:11, 1191-1196.
[6] Zdeněk Čeřovský & Miroslav Lev (2015) Permanent Magnet Synchronous Machine Parameters Identification for Load
Characteristics Calculation, Automatika, 56:2, 217-225.
[7] Andraž Kontarček, Asst. Prof. Mitja Nemec, Primož Bajec & Prof. Vanja Ambrožič (2014) Single Open-phase Fault
Detection with Fault-Tolerant Control of an Inverter-fed Permanent Magnet Synchronous Machine, Automatika, 55:4,
474-486.
[8] M. Schreier, "Modeling and adaptive control of a quadrotor," 2012 IEEE International Conference on Mechatronics and
Automation, Chengdu, 2012, pp. 383-390.
[9] J. Nathan Kutz, Steven L. Brunton, Bingni W. Brunton, and Joshua L. Proctor. 2016. Dynamic Mode Decomposition:
Data-Driven Modeling of Complex Systems. SIAM-Society for Industrial and Applied Mathematics, Philadelphia, PA,
USA.
[10] C. Foias, Hitay Ozbay, Allen R. Tannenbaum, and M. Thoma. 1996. Robust Control of Infinite Dimensional Systems:
Frequency Domain Methods. Springer-Verlag, Berlin, Heidelberg.
[11] Kurkcu, B., Kasnakoglu, C., (2018). Robust Temperature Control of a Thermoelectric Cooler via µ-Synthesis, Journal
of Electronic Materials, 47(8), 4421–4429.
[12] I. Aliskan, K. Gulez, G. Tuna, M. Kizilkaya, "Lyapunov function-based control of active power filters to reduce the
adverse effects of harmonics on induction motors," 2013 3rd International Conference on Electric Power and Energy
Conversion Systems, 2013, pp. 1-5.
[13] Kurkcu, B., Kasnakoglu, C., Efe, Ö., (2018). Disturbance/Uncertainty Estimator Based Robust Control of
NonminimumPhase Systems, IEEE/ASME Transactions on Mechatronics, 23(4), 1941–1951.
[14] Aliskan, I. (2019) "Adaptive Model Predictive Control for Wiener Nonlinear Systems" Iranian Journal of Science and
Technology, Transactions of Electrical Engineering.
[15] E. Lavretsky and K. A. Wise,Robust and Adaptive Control with Aerospace Applications. Springer-Verlag London,
2013, pp. 161–176.
[16] Demir, R., Barut, M. (2018). Novel hybrid estimator based on model reference adaptive system and extended Kalman
filter for speed-sensorless induction motor control. Transactions of the Institute of Measurement and Control, 40(13),
3884–3898.
[17] Juang, J.N. & Pappa, Richard. (1985). An Eigensystem Realization Algorithm for Modal Parameter Identification and
Model Reduction. Journal of Guidance Control and Dynamics. 8. 10.2514/3.20031.
[18] Isermann, Rolf and Marco Münchhof. “Identification of Dynamic Systems: An Introduction with Applications.”
(2010).
[19] Simon, Daniel. (2006). Optimal State Estimation: Kalman, H∞, and Nonlinear Approaches.

577
 MULTICRITERIA OUTRANKING METHODS AND THE
HYDROPOWER EXPANSION PLANNING IN BRAZIL

Raupp Igor
Cepel, Rio de Janeiro, Brazil, [email protected], ORCID: 0000-0003-0375-7239

Clímaco João
Institute for Systems Engineering and Computers, INESC-Coimbra, Portugal, [email protected], ORCID: 0000-0001-
6655-8590

Costa Fernanda
Rio de Janeiro State University, Rio de Janeiro, Brazil, [email protected], ORCID: 0000-0002-4414-7755

Raupp I, Clímaco J, Costa F. Outranking methods and the Hydropower Expansion Planning
Cite this paper as:
in Brazil. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The hydropower expansion planning in Brazil is done through a serie of studies. One of the first
steps is the River Basin Inventory Studies and its main objective is to design and analyze all
possible alternatives of dimensioning and location arrangement of hydropower plants to be built
in a river basin to select the best one, which will determine the hydropower potential to be exploited
in the basin. The decision-making to select the best alternative uses the weighted sum method and
it considers three criteria: the cost-energy benefit ratio, the negative socio-environmental impact
and the positive socio-environmental impact. This article presents a proposal to change the
multicriteria method so that it is no longer necessary to inform relative weights (trade-offs)
between the criteria. Thus, the ELECTRE III and IV methods were tested in an inventory study
based on the Inventory of the Middle Tocantins in Brazil, to verify the applicability of the
proposals.
Keywords: River Basin Inventory Studies, Multicriteria Analysis, Socio-environmental Impacts, Hydropower
© 2020 Published by ECRES

1. INTRODUCTION

The Brazilian electric power matrix is strongly based on renewable sources and a huge part of it is due to the
use of the great hydropower potential of the Brazilian river basins. The hydropower source represents 66% of
the Brazilian installed capacity [1], enabling a generation of electricity using a clean and renewable source of
low-cost power generation. The Brazilian hydropower potential is estimated in 250,000 MW and it has been
progressively explored, and, nowadays, Brazil has exploited 43% of this potential [2] and half of the potential
still available to be exploited is concentrated in the north region of Brazil, characterized by environmental
relevant areas, rich biodiversity and presence of traditional and indigenous people. Thus, the development of
new hydropower plants must be done carefully dealing with the permanent paradox of society, that increases its
demand for energy for its continuous development but, on the other hand, it is increasingly aware of
environmental impacts and excessive use of natural resources.
In Brazil, the development of new hydropower projects considers different stages [3], from the hydropower
expansion planning to the operation of the plant, with different details and successive approaches. The Inventory
Studies [4], focus of this paper, is one of the first studies of the hydropower’s plant development stages and it is
a strategic study, as at this stage, technical and financial resources have not yet been committed to the
implementation of any future hydropower plant that will compose the river basin head division. Therefore, in
this study, all possible head division alternatives should be raised and studied in order to select the one that
presents the “best” energy and socio-environmental efficiency. As the alternatives comprise the whole river
basin, it is an opportunity to evaluate the synergies and cumulative effects of the set of projects in the same
alternative concerning energy benefits and socio-environmental impacts. The projects of the selected alternative

578
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

constitute the hydropower potential to be harnessed over the years in the basin and they will later be forwarded
individually to the respective Feasibility Studies.
The procedures and methodologies of the Inventory Studies are consolidated in the Inventory Manual [5][6] and
it has been updated over time, especially regarding the decision making to select the most preferred alternative.
This paper presents the current methodology for selecting the most preferred alternative and a proposal to change
the multicriteria decision support method regarding this selection, in order to avoid the need to define specific
relative criteria weights, reflecting trade-offs among criteria. The paper also presents the application of this
proposal in a case study based on a real Inventory Study in Brazil.

2. DECISION-MAKING INVENTORY STUDIES

Until 1997, the decision to select the most preferred alternative only considered economic-energy efficiency.
The third version of the Inventory Manual published in 1997, presented great advances in the environmental
area and the selection of the most preferred alternative became a multicriteria analysis through two criteria:
economic-energy efficiency and negative socio-environmental impacts. In 2007, a new revision of the Inventory
Manual [5][6] was published, being the current version of the Manual, and it included one more criterion in
decision making: the positive socio-environment impact. So, the basic criterion of the current Inventory Studies
decision-making is to maximize economic-energy efficiency, while minimizing negative socio-environmental
impacts, and also maximizing positive socio-environmental impacts provided by the implementation of
hydropower plants in the river basin. Therefore, in the decision-making process three criteria are considered,
each one associated with an objective of the basic criterion.

Criterion 1: Economic-energy efficiency (a minimization criteria intending to maximize efficiency)

The economic-energy efficiency is assessed by the cost/energy benefit index (ICB), calculated using the ratio of
cost (construction, operation and maintenance cost of the alternative’ set of projects) to their energy contribution,
calculated by simulating the operation of the set of projects considering the worst hydrological period of the
historical inflows, respecting the other water uses in the basin. The higher value of this index indicates a more
expensive alternative (higher energy cost), and thus worsening the economic-energy efficiency of the alternative.

Criterion 2: Negative socio-environmental impact (minimization)

The analysis of negative socio-environmental impacts is necessary to ensure that such impacts are minimized in
the river basin and it is expressed by the negative socio-environmental impact index (IAN). This index should
express the magnitude of the total negative impact brought about by the set of projects of the alternatives on the
study area and the higher value of this index indicates a greater impact, resulting in a poorer socio-environmental
performance. The evaluation and quantification of this impact is based on 6 themes: Aquatic Ecosystems,
Terrestrial Ecosystems, Ways of Life, Territorial Organization, Regional Economy, Indigenous
Peoples/Traditional Communities.
The IAN is calculated in two steps: (i) alternative impact quantification considering each theme; and (ii) IAN
calculation per alternative considering the whole socio-environmental system, obtained by the weighted sum of
the themes’ impacts quantified on the previous stage. Both impacts (per theme and per alternative) range from
zero (no impact) to one (total devastation). Given the interdisciplinary nature of this activity and the great margin
for subjectivity involved, it is necessary to systematize the procedures for weighting the synthesis components
using specific methods and techniques. These methods should make it possible to represent the subjective
assessments in a system of weights, as well as combining the opinions of different decision agents, such as
Analytic Hierarchy Process (AHP) [7].

Criterion 3: Positive socio-environmental impact (maximization)

The main benefit of the alternatives is the electricity generation which is achieved by maximizing the economic-
energy efficiency of the river basin. However, the analysis of potential positive impacts can also include the
social and economic regional development induced by the implementation of hydropower plants. Thus, the
evaluation and quantification of this impact index is based on four benefits: Increased municipal revenues,
dynamization of the local labor market, improvement of road infrastructure and opportunity for the efficient use
of water resources.
This criterion is represented by the positive socio-environmental impact index (IAP), that is calculated per
alternative in two steps. The first one is the quantification of the positive impact related to each benefit and then,
the IAP obtained by the weighted sum of the benefits’ impact indexes for the alternative prior quantified.

579
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Multicriteria Analysis based on an Additive Model

According to the current Inventory Manual, the selection of the most preferred alternative uses the weighted
sum method, an additive model, classified as a full compensatory method, to aggregate the three criteria into a
single index and the alternative with the “best” value (lowest value) is selected. The aggregation is performed
in two steps. The first step calculates the preference index by the weighted sum of the most important criteria:
the economic-energy efficiency maximization, represented by the ICB, and the minimization of the negative
socio-environmental impact, represented by the IAN. So, it must be defined the weight regarding ICB (p cb), and
consequently obtaining the IAN weight (1-pcb). In the second step, the Modified Preference Index (I') is
calculated by the weighted sum of the pre-calculated Preference Index (I) and the positive socio-environmental
impact maximization criterion, represented by the IAP. Note that, as it is intended to maximize IAP, it is used
the corresponding minimization of (1- IAP). In this step, it must be defined the weight regarding the
"minimization (1-IAP)" (pap), obtaining the I weight (1-pap).
The weights pcb and pap reflect a subjective evaluation by the decision maker concerning the trade-offs between
pairs of criteria values. The multidisciplinary team directly involved in the Inventory Study must define them,
reflecting the opinion of the society and taking into account the context in which the analysis is inserted and the
time the studies are carried out. The Inventory Manual recommends that p ap must be lower 0.25, because this
criterion has less importance than the others. As suggested in the Inventory Manual, the selection of the most
preferred alternative must identify the most robust alternative, however, by the current method, this search is
performed varying the weights without a formal methodology, which can become quite complex.

3. A NEW MULTICRITERIA INVENTORY APPROACH BASED ON THE USE OF ELECTRE

METHODS

A criticism that is made in Inventory Studies is the definition of the criteria weights in the decision-making,
required by the weighted sum method. These weights reflect/estimate the trade-offs or the marginal substitution
rates among criteria, and according to the Inventory Manual, they must be chosen by a multidisciplinary team
in order to reflect society's wishes. However, such a choice becomes a subjective task and each Inventory Study
may choose different ways to define them, resulting in different trade-offs/weights among criteria by different
decision makers. The Inventory Manual does not propose a procedure to consider these uncertainties, to take
subjectivity into account. Of course, the next step is the identification and test of appropriate multicriteria
methods to overcome, or, at least mitigate this difficulty. In [3] it was proposed the use of VIP-Analysis, keeping
the simplicity of the additive models based on weighted sums, but incorporating explicitly the incompleteness
of information related to these weights.
In this paper a second experience is proposed. The use of, in some sense, more flexible methods is proposed. In
this case, ELECTRE III and ELECTRE IV are considered. Before an outline introduction of these methods we
emphasize the two following remarks. First, although the non-compensatory approaches, namely those based on
outranking relations as ELECTRE methods, may be more adequate - to the type of real world problems we are
dealing with - than methods based on weighted sums, the results are less conclusive in many situations and the
number of parameters it is necessary to fix is large; secondly ELECTRE III and IV are dedicated to the ranking
problematic, and the inventory problem we are dealing with is a choice problem. However, the use of ELECTRE
III/IV besides the identification of the most preferred alternative (the first in the ranking), enables to identify
possible incomparabilities among alternatives and, as we shall see, they incorporate “distillation” procedures
leading to the identification and quantification of the number of the alternatives that are outranked by each
alternative along the distillation procedures. This type of information can be useful, namely suggesting further
experience changing some parameters of the model, in order to better assist the decision maker in the
identification of the most satisfactory alternative for exploring the hydropower potential of the basin under study.
In ELECTRE III and IV partial relations concerning pairwise comparison of alternatives are built (the, so called,
outranking relations). Then an exploitation procedure is used in order to rank the alternatives (although in many
situations it is not possible to get a complete preorder).
We say that an alternative a outranks b when “a is at least as good as b”, and, as we shall see, in the case of
ELECTRE III they are fuzzy. The outranking relations are based on global concordance and discordance indices.
Let us start by defining these indices. First, we define Partial Concordance Index, for each pseudo-criterion (j).
A pseudo-criterion establishes the following structure of preferences, considering two thresholds, an indifference
threshold (q) and a strict preference threshold (p): (i) Indifference: When the difference in the performance of
two alternatives in the given criterion is lower than “q”; (ii) Weak Preference: When the difference in the

580
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

performance of two alternatives in the given criterion is greater than “q” and lower than “p”; and (iii) Strict
Preference: When the difference in the performance of two alternatives in the given criterion is greater than "p".
It must be remarked that the pseudo-criterion structure of preferences is more flexible and realistic when
compared with the sharp criteria structure.
The partial concordance index for a pseudo-criterion expresses the intensity of concordance with the statement
“alternative a is at least as good as alternative b”, for that criterion. The global concordance index Cab is the sum
of the partial concordance indices for all pseudo-criteria, weighted by the weights assigned to each pseudo-
criterion. A remark must be added at this point: ELECTRE III requires the user being able to fix weights for
criteria, however the meaning of these weights is not the same as in the additive model. In this case, they do not
express trade-offs but only the relative importance of the criteria, and so, elicit them is much less demanding in
cognitive terms.
The discordance index, concerning each criterion j, and considering the performances of a and b, measures how
far we are from the assertion “a is at least as good as b”. They are related, to veto thresholds.
Now, we are ready to define the associated “outranking relations” for pairs of alternatives using credibility
indices based on the global concordance and the discordance. If discordance is not considered, as in the case
presented in next section, the credibility index coincides with C ab. Otherwise, the discordance may weaken the
credibility index by a factor between zero and one. Finally, the second phase, i.e. the exploitation of the
outranking relation matrix (from the Credibility Indices for all pairs of alternatives, a credibility matrix is
assembled), in order to build a procedure to look for a final ranking of alternatives, although in many situations
just enabling to get a partial preorder.
Defining distillation thresholds and cutoff levels two distillation procedures are performed, based on the concept
of qualification (difference between the number of alternatives outranked by it and the number of alternatives
outranking it). The alternatives are ranked resulting in two preorders, one calculated from the best alternative to
the worst (descending distillation) and the other from the worst to the best (ascending distillation). The final
ranking is obtained by the intersection of these two preorders, but it may include incomparabilities among
alternatives.
ELECTRE IV is a method close to the ELECTRE III, however in this case it is admitted that the user is not able
or doesn’t feel comfortable fixing values for the weighs, so it is not possible to build the Global Concordance
indices as in ELECTRE III. In ELECTRE IV a nested sequence of outranking relations is considered. The
conditions to establish five types of dominance, in the pairwise comparison of pairs of alternatives, for instance
a and b, are:
 Quasi-dominance: if a is preferred or indifferent to b (note that in the indifference a can be better or worse than b) in
all criteria; and if the number of criteria for which b is better than a is lower than the number of criteria for which a is
better than b. In this case, the index of credibility for a quasi-dominates b is 1.0.
 Four other types of dominance, progressively less rich, are built. Predefined conditions, as for the quasi-dominance
case, are used. Remember that they are modeled such that the sequence of dominances is nested. In short, they are the
canonic dominance, with a credibility degree 0.8; pseudo-dominance, with credibility 0.6; sub-dominance, with
credibility degree 0.4; and veto dominance with credibility 0.2. Of course, if there no dominance at al., the credibility
is 0.0.
The reminder of the method is conceptually similar to ELECTRE III. For more details on ELECTRE III and IV
see [8-10]. It is interesting to compare the application of these two methods in the Hydropower Inventory
Problem dealt with in this paper. On the one hand, it is clear that fixing the weights is, in any case, one of the
critical points on the use in any method, but, on the other hand the Inventory Manual requiring the explicit
consideration of several criteria, also recommends the aggregation of the results using a weighted sum model
[5][6].

4. APPLICATION

In this paper it is proposed a new multicriteria approach to evaluate the replacement of the weighted sum method
by ELECTRE III or IV. To verify their effectiveness and performance, they were applied to a case study based
on an Actual Inventory Study of the Middle Tocantins Basin in Brazil [11]. The Tocantins river basin has a
drainage area of 767,000Km² and represents 7.5% of the Brazilian territory. The Inventory Study identified 23
dam sites (places where it is possible to construct a hydropower project) and 26 hydropower projects
(considering that, in some cases, it is possible to design two different layout options of hydropower projects for
the same dam site). These 26 projects were aggregated in 6 alternatives (sets of projects), to exploit the
hydropower potential of the river basin. Table 1 presents the performance of the alternatives for the three criteria,

581
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

highlighting the best alternative for each criterion in green and the worst in red. From Table 2, 7A and 3R are
the best and the worst alternatives concerning the energy efficiency (ICB), respectively; 9A and 7B are the best
and the worst alternatives regarding the negative socio-environmental impact (IAN); and 9B and 3R are the best
and the worst concerning the positive socio-environmental impact (IAP).

Weighted sum method

This application followed the current methodology according to the Inventory Manual, using the weighted sum
approach as the multicriteria method. The weight values defined in the Inventory Study were used in this case
study. Namely: pcb = 0.5 and pap = 0.17. The results are shown in Table 2 and the most preferred alternative, that
obtained the lowest value for the Modified Preference Index (I’) is alternative 7A, the alternative values are
close, in special 7B, 9B and 1R. Alternative 3R is the worst.

Table 2. Alternatives ranking.

Alternatives Number of Reservoirs total Average Energy I’ Ranking
ICB IAN IAP
projects area (km²) (MWmed)
7A 14 8100 2612 0,535 0,612 0,530 0,556 1
9A 14 8060 2473 0,566 0,606 0,545 0,564 2
1R 12 8070 2716 0,543 0,667 0,540 0,579 3
7B 14 6980 2650 0,570 0,616 0,490 0,580 4
9B 13 8090 2581 0,573 0,654 0,560 0,584 5
3R 14 5390 2510 0,611 0,608 0,470 0,596 6

An additional analysis was performed changing the weight values of ICB, IAN and IAP. The results indicated
that alternative 7A is a robust choice, but it is not very common in Inventory Studies. On the other hand, there
are changes in the ranking of alternatives 1R, 7B and 9B, justified by the fact that they present modified
preference index (I’) values very close to each other (0.579, 0.580 and 0.584, respectively). So, they can be
considered as tied.

ELECTRE III
The information required by ELECTRE III, from the decision maker, regards the preference and indifference
thresholds for each criterion and the weights reflecting the relative importance of criteria. Note that, in this case
study, it was considered inadequate to consider “discordance”. In order to define the values of the required
thresholds, so that the same type of metric could be adopted for all criteria and taking into account that the
performance of the alternatives may present different levels of variability according to the criterion under review,
the standard deviation (SD) was adopted, because it is a measure of the performance’s dispersion. Six different
cases (Table 4) were considered, varying the thresholds and the relative importance of IAP. Figure 3 presents
the outputs of ELECTRE III.
As expected, it can be concluded that as the values of the thresholds decrease, tendentially the ranking of the
alternatives includes a smaller number of ties, that is, most of the alternatives can be more differentiate from
each other. This can be confirmed, for instance, observing the ranking corresponding to 3D, 3E and 3F (using
as the preference and indifference parameters 2SD and 1SD, respectively). In all cases alternative 7A appears
as the best alternative. Considering less restrictive thresholds, 7A is indifferent to 9A (case 3A) or 9B (case 3C),
and considering more restrictive thresholds, the alternatives 7A and 9A are tied on the top for all sets of weights.
In terms of the worst alternative, it is always 3R, and, in some cases, it is tied with alternative 7B.

582
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020
7A 9A 7B 9B 1R 3R

7A 9A 1R 9B 3R 7B
Table 4. Data used in ELECTRE III application.
3A

3B 7A 1R 9A 7B 9B 3R Criteria Weights Parameters

Case
3C 7A 9B 1R 9A 7B 3R ICB IAN IAP Indifference Preference
3A 1 1 0,25 2 SD 3 SD
3D 7A 9A 1R 9B 3R 7B 3B 1 1 0,50 2 SD 3 SD
3C 1 1 1 2 SD 3 SD
3E 7A 9A 1R 9B 3R 7B
3D 1 1 0,25 1 SD 2 SD
1R 3E 1 1 0,50 1 SD 2 SD
3F 7A 9A 9B 3R 3F 1 1 1 1 SD 2 SD
7B SD 0,03 0,02 0,03
Figure 3. Alternatives ranking, according to ELECTRE III.

ELECTRE IV
As referred to previously, ELECTRE IV also uses preference and indifference thresholds for each criterion
provided by the decision maker, but it is not necessary to fix weights regarding the relative importance of the
criteria. In this approach, three cases were considered, differentiated by the threshold pairs (Table 5). Figure 3
presents the outputs of ELECTRE IV. In case 4A, there is incomparability between alternatives 7A and 9A and,
also, between 9B and 1R. Despite the incomparability, both alternatives 7A and 9A are better than alternatives
9B, 1R, 7B and 3R. In case 4B, it is obtained a total ranking of the alternatives, where 9A is on the top, without
ties and eliminating the previous incomparabilities. A third experience (case 4C) led to tie among alternatives
1R, 9B and 7B, maintaining 9A as the best alternative (followed by 7A) and 3R as the worst alternative.
In short, ELECTRE III and ELECTRE IV aim at ranking the alternatives, however these methods do not
guarantee to obtain a complete preorder, because in many situations they are not able prevent incomparabilities
among alternatives. Furthermore, an adequate consideration of several values for the preference and indifference
thresholds may play an important role to clarify which are the alternative(s) that are the most satisfactory.

Table 5. ELECTRE IV execution data.

Criteria Parameters
Case
ICB IAN IAP Indifference Preference
4A - - - 1 SD 2 SD
4B - - - 2 SD 3 SD
4C - - - 3 SD 4 SD
SD 0,03 0,02 0,03
Figure 4. Alternatives ranking, according to ELECTRE
IV.

5. CONCLUSION

This paper aimed at presenting potential improvements in the decision aiding of the Hydropower Inventory
Studies through the proposal of new multicriteria methods for selection of the preferable head division
alternative, avoiding the need of specifying weights that reflect marginal rates of substitution (trade-offs) among
the criteria. In this paper some experiences using ELECTRE III and ELECTRE IV methods were presented,
applied to a real Inventory Study. Three different pairs for the indifference and preference thresholds were
considered to test how much they contribute to vary the outputs of ELECTRE IV. Concerning ELECTRE III
two different pairs of indifference and preference thresholds were considered, and it was also tested three
different weight values for IAP, fixing equal values to the most important criteria (ICB and IAN), assessing its
influence on ELECTRE III output. The results obtained using ELECTE III are congruent with those obtained
by the additive model current in use. However, we believe that ELECTRE III is more flexible because, instead
of sharp criteria, it uses pseudo-criteria considering indifference and preference thresholds; and the requirement
of fixing weights is much less exigent in cognitive terms than fixing weights reflecting trade-offs as it is required
by the additive model. In short, the additive model is a full compensation approach where the weights represent
trade-offs between criteria values, and in ELECTRE III they just represent the relative importance attributed to
the criteria. So, it is much easier to elicit these ones. As it was discussed above the results obtained using
ELECTRE IV are a bit different from those obtained using ELECTRE III but not very contradictory. In any

583
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

case, although the ELECTRE IV is less demanding for the user, since no weights are required, the Decision
Maker felt not very comfortable due to lack of intervention on the relative valuation of criteria. So, ELECTRE
IV doesn’t seem adequate to study the inventory problem dealt with in this paper.

REFERENCES

[1] ONS. Capacidade Instalada no SIN - 2020/ 2024. Available at: < http://www.ons.org.br/paginas/sobre-o-sin/o-sistema-
em-numeros>. Acessed: May 2020.
[2] Eletrobras. SIPOT - Potencial Hidrelétrico Brasileiro 2015. Available at:
<http://www.eletrobras.com/elb/data/Pages/LUMIS21D128D3PTBRIE.htm>. Acessed: May/2020.
[3] Raupp, I, Climaco, J, Costa, F, Miguez, M. Multicriteria Methods and the Hydropower Plants Planning in Brazil. In
Multiple Criteria Decision Making - Beyond the Information Age: Selected works from MCDM 2019 Istanbul,
Istanbul: Springer, 2020.
[4] Costa, F, Damazio, J, Raupp, I, Pires, SH, Matos, D, Paz, L, Mollica, A, Menezes, PC. Hydropower inventory studies
of river basins in Brazil. International Journal on Hydropower & Dams 2011, v. 2/2011, p. 31-36.
[5] MME. Manual de Inventário Hidroelétrico de Bacias Hidrográficas. Rio de Janeiro: E-papers, 2007.
[6] MME. River Basin Hydropower Inventory Manual. Rio de Janeiro: E-papers, 2010.
[7] Saaty, T. Método de Análise Hierárquica. São Paulo: McGraw-Hill, Makron,1991.
[8] Dias, J, Figueira, J, Roy, B. The Software ELECTRE III and IV- Methodology and user manual (version 3.X). Centro
de Estudos de Gestão IST, Working Paper nº4/2006.
[9] Roy, B, Bouyssou, D. Aide Multicritère à la Décision: Méthodes et Cas, Paris: Economica, 1993.
[10] Figueira, J, Mousseau, V, Roy, B. ELECTRE Methods, in Multiple Criteria Decision Analysis:State of the Art Surveys,
pp 133-153. New York: Springer, 2005.
[11] Themag Engenharia Ltda. Bacia do Medio Tocantins. Estudos Finais de Inventário – Apêndice H. Relatório Final.
1987.

584
 SATELLITE CARRIER STRUCTURE ANALYSIS AND
AERODYNAMIC CHARACTERISTICS FOR LOW ALTITUDE
DELIVERY SYSTEM

Koten Hasan
Istanbul Medeniyet University, Istanbul, Turkiye, [email protected], ORCID: 0000-0002-1907-9420

Tufan Tarık
Istanbul Medeniyet University, Istanbul, Turkiye, [email protected], ORCID: 0000-0001-9324-2401

Demir Enes
Marmara University, Istanbul, Turkiye, [email protected], ORCID: 0000-0002-3454-6882

Köten, H. Tufan, T, Demir E., Satellite Carrier Structure Analysis and Aerodynamic
Cite this paper as: Characteristics for Low Altitude Delivery System. 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: This study investigates analysis of low altitude satellite carrier system structure and aerodynamic
characteristic properties to find the optimum parameters of low altitude delivery system.
Computational Fluid Dynamics (CFD) tools are used to solve the governing equations of the
system and to illustrate the airflow around the carrier system. Computer Aided Design (CAD)
structure of the satellite carrier system is generated using Space-Claim and exported to ANSYS
solver. Commercial CFD software, ANSYS Fluent code is used to compute airflow velocities,
pressure, the drag force and the drag coefficient and reported. In this study, the main goal is to
obtain the drag force and the drag coefficient in the velocity range of 0.6 and 4 Mach number.
Also aerodynamic analysis of the system velocity and pressure contours are investigated and
reported in different conditions.
Keywords: Satellite carrier, aerodynamics, drag force, drag coefficient, rocket-target.
© 2020 Published by ECRES

1. INTRODUCTION

The Satellite carrier model characteristics and design parameters as well as the exterior fluid structures are
analyzed and investigated using CFD code. Generally these engineering designs are estimated at different
carrier’s velocities in literature. The current model was inspired by the demand to develop a carrier for the low
altitude satellite system which is about 4 meters in length and 0.3 meters in diameter. The satellite carrier’s
maximum velocity is intended to be less than desired value. According to literature one of optimal nose cones
for this carrier situation is parabolic one [1, 2]. As known, fluid structure around the satellite carrier system
and aerodynamics design characteristics such as velocity, airflow, pressure, lift and drag force and others have
a big impact on the exterior lunch design of the satellite carrier system. The lift-drag force and the lift-drag
coefficient are the main important design parameters for the satellite carrier system. Because necessary thrust
properties of the system are directly related to exterior design parameters. CFD tools are widely preferred to
investigate the equipment structures as the conceptual and preliminary design steps to decrease the production
time and the costs in aeronautical applications before the experimental studies and production [3-5]. This study
is related with the aerodynamic analysis of the satellite carrier system to be produced and the first aim is to

585
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

obtain the drag and lift forces and the drag and lift coefficients in different velocity range of 0.6-4 Mach
number.

Figure 1 Mesh display of the satellite carrier.

2. CFD CASES FOR THE SATELLITE CARRIER SYSTEM

In Figure 1 computational mesh structure is given around the system generated in ANSYS mesh module [6].
The velocity contours for different Mach numbers, around the carrier system with nose and cones and around a
carrier frame without a nozzle cone were given in Figure 2. The carrier diameter is 0.3 m. and the length of
this system without nose is 4 m. The satellite carrier system length with the nose totally is 4.2 m and the nose
diameter in the medium is 0.15 m.

586
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

@ mach 4 @ mach 3

@ mach 2 @ mach 1

@ mach 0.9 @ mach 0.8

@ mach 0.7 @ mach 0.6

Figure 2 Velocity contours in different mach numbers.

Frontal area and drag coefficient calculations:

6.82
𝐹𝑟𝑜𝑛𝑡𝑎𝑙 𝑎𝑟𝑒𝑎 (𝐴) = 4 ∗ 0.2 ∗ 6.8 + 𝜋 ∗ = 120.44 𝑐𝑚 2 = 0.01204 𝑚 2
4
𝐷𝑟𝑎𝑔 𝑓𝑜𝑟𝑐𝑒(𝐹𝑑 )
𝐷𝑟𝑎𝑔 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 (𝐶𝑑) =
0.5 ∗ 𝜌 ∗ 𝑉 2 ∗ 𝐴

587
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3 Drag coefficient versus Mach number.

Table1. Drag coefficients at different mach numbers.

Mach numbers Drag coefficients
0.6 mach 0.342
0.7 mach 0.484
0.8 mach 0.685
0.9 mach 0.775
1 mach 1.02
2 mach 3.54
3 mach 5.63
4 mach 7.65
In figure 3 and Table 1 drag coefficient calculations were given. According to literature drag and lift
coefficients and also drag and lift forces are widely agreement with similar studies [7-8]

3. CONCLUSIONS

In this study, CFD calculations and aerodynamic performance of the satellite carrier system with nose and
flow for the different cases were investigated and the desired aerodynamics properties for simulations of
exterior structure were obtained. As a result, satellite carrier system drag force is significantly lower compared
to the drag force of the system in low Mach numbers. Again, the drag coefficient of the carrier system is 0.342
at M = 0.6 and 7.65 at M = 4 as maximum and minimum value. The satellite carrier system drag coefficient
varies in the range of 0.342- 7.65 for different Mach numbers. This design was developed for the low altitude
satellite carrier systems which will be used in national project. Also this system successfully was tested by
computational techniques. Accordingly, satellite carrier system actual flight trajectory is widely and
qualitatively agreement with the similar experimental and numerical studies in the literature.

REFERENCES

[1] Fedaravičius A., S. Kilikevičius, A. Survila. 2012. “Optimization of the rocket’s nose and nozzle design parameters in
respect to its aerodynamic characteristics”. Journal of Vibroengineering, Vilnius: Vibromechanika 14(3) : 1390-1398.
[2] Department of Defense.1990. Military Design Handbook. Design of Aerodynamically Stabilized Free Rockets, U.S.
Army Missile Command.
[3] Schütte A., G. Einarsson, A. Madrane, B. Schöning, W. Mönnich, W.R. Krüger. 2002. Numerical simulation of
maneuvering aircraft by CFD and flight mechanic coupling, RTO Symposium. Paris.

588
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[4] Kroll N., C.C. Rossow, D. Schwamborn, K. Becker, G. Heller. 2002. MEGAFLOW-a numerical flow simulation tool
for transport aircraft design. In Proceedings of ICAS Congress, 1105.1-1105.20.
[5] Langtry R.B., M. Kuntz, F. Menter. 2005. “Drag prediction of engineairframe interference effects with CFX-5”.
Journal of Aircraft, 42(6) : 1523-1529.
[6] ANSYS FLUENT-Solver Theory Guide, ANSYS, Inc., Southpointe 275 Technology Drive Canonsburg, PA 15317.
[7] Menter F.R., M. Kuntz, R. Langtry. 2003. Ten years of industrial experience with the SST turbulence model. In
Turbulence, Heat and Mass Transfer 4, (ed.: K. Hanjalic, Y. Nagano and M. Tummers), 625-632. Begell House, Inc.
[8] Mills A.F., R.D. Irwin. 1995. Basic Heat and Mass Transfer, University of California, Los Angeles.

589
 MAXIMUM POWER POINT TRACKING UNDER PARTIAL
SHADING CONDITIONS USING NOVEL MINE BLAST
ALOGRITHM

Denis Karabetsky
NAU, Kyiv, Ukraine, [email protected], ORCID: 0000-0002-1325-9279

Victor Sineglazov
NAU, Kyiv, Ukraine, [email protected], ORCID: 0000-0002-3297-9060

Karabetsky, D., Sineglazov, V.. Maximum power point tracking under partial shading
Cite this paper as: conditions using novel mine blast alogrithm. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
2020, Istanbul, Turkey

Abstract: At this paper will be considered problematics of the partial shading conditions and impact of it on
the PV system. Main problem that is introduced by PSC that P-V characteristic could have several
extremums, some of them are local and only one is global. Current conventional methods for
tracking, such as hill climbing, P&O etc., can not find such global maximum point on the curve
and as result efficiency of PV subsystem could be dramatically decreased. For solving such
problem only algorithms that are ready for tracking global maximum should be used. This paper
will introduce more formal problem statement of PSC phenomena, introduce evaluation metrics
for algorithms, MATLAB/Simulink simplified model of standalone PV system and will show
adaptation of novel mine blast algorithm to be used with MPPT to control of DC-DC converter.
The main objective for MPPT is solving optimization problem by evaluating duty cycle that
corresponds to GMPP of attached PV array. At result will be given brief comparison with PSO
algorithm and points for future research or such class of the meta-heuristic optimization.
Keywords: Maximum Power Point Tracker, Partial Shading Conditions, Mine Blast Algorithm
© 2020 Published by ECRES

1. INTRODUCTION

Renewable energy is energy produced by a resource that quickly fills (recovers) as a result of a natural or nonstop
natural process.
From an economic point of view, renewable energy sources (RES) can be considered as an effective means of
stimulating innovation and business activity in national economies, creating additional jobs, creating new
significant sources of income from equipment import.
According to the International Energy Agency (IEA) [1], electricity consumption in the world will reach 26
trillion kW/h by 2025. Additionally, the installed capacity of the power plants will reach 5500 GW. The
corresponding parameters are also estimated to reach 32 trillion kW/h and 5900 GW by 2035. The heads of the
leading states consider RES to play a significant part (about 44%) in achieving the stated parameters since the
traditional methods of generating electricity with limited primary resources, cause some environmental damage
[2].
It has been proved that increasing the volume of RES production and their share in the energy balances
contributes to improving the efficiency of economic activities of different energy consumers and enhances trust
between the countries that have included RES in the list of strategic priorities of their development.
In order to obtain the maximum possible power at the output of PV modules, electric generators of wind turbines,
electric generators with variable rotation speed and torque, electric motors operating in regenerative braking
mode, the maximum power point tracking (MPPT) method is used.
Most often, MPPT devices are used to optimize the maximum power of solar photovoltaic batteries (photo
modules) when using solar energy. This is because the illumination of PV modules during the day varies

590
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

significantly from the position of the sun in the sky, cloudiness, precipitation, which causes a profound change
in the load characteristic of photo modules. To obtain the maximum power point, it is necessary to change the
current taken from the photo modules, while the voltage on the photovoltaic battery changes.

2. PROBLEM STATEMENT

The output power of any solar panel depends on many parameters: environmental conditions, level of solar
irradiation, temperature of solar panels, connected load. For example, with an increase in the level of solar
radiation, the output power will increase, but with a simultaneous increase in the temperature of the solar panels,
the output will decrease. Based on that solar cells have a rather complex relationship between environmental
parameters and output power, thus current-voltage characteristic also changes from these parameters, therefore,
it can be represented as a certain curve that depends on specific instantaneous values of the environment. On the
I-V curve there is a specific point at which certain values of current and voltage, depending on the selected load
(according to Ohm's law), will correspond to the point of maximum power, which is the global maximum on the
P-V curve.
For the operation of solar panels at the point of maximum power, DC-DC converters are most often used, which
are placed between the solar panels and the load (consumers). An additional unit that provides the
implementation of the algorithm for finding the maximum power point controls these DC-DC converters [3].
The maximum power of the panel at the operating point can be represented:
𝑃𝑀𝑃𝑃 = 𝐼𝑀𝑃𝑃 𝑉𝑀𝑃𝑃 (1)
where 𝑉𝑀𝑃𝑃 , 𝐼𝑀𝑃𝑃 is the voltage and current at the maximum power point, respectively.
Due to the fact that MPP can move on the power curve at different moments in time and system operation, due
to the dependence on environment variables, for its search in real time, MPPT algorithms should be applied,
which will be discussed later in this paper.
When it comes to harvesting solar energy in large scales, the PV modules are arranged and connected in series
and parallel combinations to produce the desired voltages and currents. Conventional MPPT techniques perform
well with uniform solar irradiance. However, shades from different objects (e.g. buildings, trees), clouds etc.
can cause changes to operating conditions of PV panels, which leads to partial shading phenomena.
During PSC the output power could be noticeably decrease. In addition, these conditions could make impact
physically on PV cells, such as creating of hotspots on modules, that is why during bundling cells into
modules/panels manufacturers using bypass diodes. These hot spots are begin to appear after cell’s power loss
or degradations in string connected configurations. Because each cell at string should maintains constant current
and part of lost power generated as heat. To remove those affect each cell (ideally) should be reverse biased with
bypass diode, that will turn-off shaded or damaged cell. This is a reason why P-V curve will experience several
types peaks (figure 1): local and global maximum power points [4]. Many conventional MPPT algorithms don’t
have ability to track accurately GMPP.
When developing a search method, it is necessary to solve the following problems:
 improvement of the search speed;
 adaptability to dimming conditions;
 reducing the cost of the system;
 correction of defects in conventional search algorithms.
Accordingly, it is possible to formulate the main criteria for search methods:
 complexity of the algorithms;
 the complexity of the implementation of the algorithm is applicable to solar energetic system;
 global maximum power point search speed;
 accuracy and search efficiency.

591
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. The PV array characteristics under fully illuminated and partially shaded conditions. (a) P-V curve (b) I-V curve.

3. OBJECTIVE FUNCTION

In this section, objective function would be built for SEPIC (Single Ended Primary Inductance Converter).
SEPIC has many benefits such as same polarity from input to output, could convert voltage to any direction:
lower or higher voltage, but the main drawback is that control loop is more complicated. This DC-DC converter
incorporated into simplified standalone PV system topology designed at MATLAB/Simulink software as shown
at figure 2. SEPIC converter used to interface the PV module output to the load and to track the maximum power
point of the PV module [5].

Figure 2. Standalone PV System topology (MATLAB/Simulink).

The main objective for MPPT is solving optimization problem by evaluating duty cycle corresponds to GMPP
of attached PV array.
Objective function itself represents PV array output power:
𝑃𝑃𝑉 = 𝑉𝑃𝑉 𝐼𝑃𝑉 (2)
Where 𝑉𝑃𝑉 and 𝐼𝑃𝑉 – voltage and currency at the outputs of PV array.
Voltage gain of SEPIC:
𝑉𝑜𝑢𝑡 𝐷
= (3)
𝑉𝑖𝑛 1−𝐷
SEPIC current gain could be expressed by using DC-DC power balance statement:
𝑉𝑖𝑛 𝐼𝑖𝑛 = 𝑉𝑜𝑢𝑡 𝐼𝑜𝑢𝑡 (4)
𝐼𝑖𝑛 𝐷
= (5)
𝐼𝑜𝑢𝑡 1 − 𝐷
Complete objective function that represent PV array output power that is expressed through SEPIC’s duty cycle
(where 𝑉𝑖𝑛 = 𝑉𝑃𝑉 and 𝐼𝑖𝑛 = 𝐼𝑃𝑉 ) :
1−𝐷 𝐷
𝑃𝑃𝑉 = 𝑉𝑃𝑉 𝐼𝑃𝑉 = 𝑉𝑜𝑢𝑡 𝐼𝑃𝑉 = 𝑉𝑃𝑉 𝐼 (6)
𝐷 1 − 𝐷 𝑂𝑈𝑇
Duty cycle of SEPIC is a subject to constraint [6]:

592
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝐷𝑚𝑖𝑛 < 𝐷 < 𝐷𝑚𝑎𝑥 (7)

where 𝐷𝑚𝑖𝑛 and 𝐷𝑚𝑎𝑥 are minimum and maximum duty cycle according to calculation at [6].

4. MINE BLAST ALGORITHM

At this paper we will adopt novel Mine Blast algorithm proposed by work [7, 8], this algorithm is used to solve
big variety of constrained problems. Its main task is to reduce computational time, improve accuracy, and require
less parameters.
The idea of the algorithm is expressed in the fact that the concept of a mine explosion in the real world is adopted,
shrapnel that fly in different directions, at different speeds, to different distance and times, which produce a
subsequent explosions within a certain zone. This collision lead us to discover the most explosive mine, which
could be considered as fitness of the objective function at the mine bomb location.
Algorithm begins by defining the starting point where the first mine would be blasted 𝑋0𝑓 , where 𝑓 – is a number
of first shot points. First explosion will define an initial population and this number is equal to number of
shrapnel 𝑁𝑠 .
The first blast point could be calculated:
𝑋0 = 𝐿𝐵 + 𝑟𝑎𝑛𝑑 ∗ (𝑈𝐵 − 𝐿𝐵) (8)
where 𝑋0 – first shot point, LB and UB – lower and upper bound of the problem, 𝑟𝑎𝑛𝑑 is a number produced
by random number generator.
Current location of mine bomb is:
𝑋 = {𝑋𝑚 }, 𝑚 = 1. . 𝑁𝑑 (9)
where 𝑁𝑑 is a search space dimension that is equal to number of independent variables.
Number of shrapnel 𝑋𝑛 produced by bomb explosion causes next mine to explode at 𝑋𝑛+1 :
𝑓
𝑓 𝑓 𝑚𝑛+1
𝑋𝑛+1 = 𝑋𝑒(𝑛+1) + exp (−√ 𝑓
) 𝑋𝑛𝑓 , 𝑛 = 0, 1, 2, 3 … (10)
𝑑𝑛+1
𝑓 𝑓 𝑓
where 𝑋𝑒(𝑛+1) – location of exploding mine bomb, 𝑚𝑛+1 – direction or slope of produced shrapnel, 𝑑𝑛+1 –
distance of produced shrapnel.
𝑓 𝑓
𝑋𝑒(𝑛+1) = 𝑑𝑛 ∗ 𝑟𝑎𝑛𝑑 ∗ cos(𝜃) , 𝑛 = 0, 1, 2, 3 … (11)
360
where 𝜃 – angle of the shrapnel (𝜃 = )
𝑁𝑠
Distance and direction of shrapnel pieces:
𝑓 𝑓 𝑓 2 𝑓 𝑓 2
𝑑𝑛+1 = √(𝑋𝑛+1 − 𝑋𝑛 ) − (𝐹𝑛+1 − 𝐹𝑛 ) , 𝑛 = 0, 1, 2, 3 … (12)
𝑓
𝑓 𝐹𝑛+1 − 𝐹𝑛𝑓
𝑚𝑛+1 = 𝑓
, 𝑛 = 0, 1, 2, 3 … (13)
𝑋𝑛+1 − 𝑋𝑛𝑓
𝑓
𝑑𝑛−1
𝑑𝑛𝑓 = , 𝑛 = 0, 1, 2, 3 … (14)
𝑘
exp(𝛼 )
where F is the function value of the X, k.
In order to conduct exploration of the design space at smaller/larger distances, exploration constant was
introduced (𝜇). IF it is hight then the iteration number the exploration of the solution space is:
𝑓
𝑑𝑛+1 = 𝑑𝑛𝑓 (|𝑟𝑎𝑛𝑑𝑛|)2 , 𝑛 = 0, 1, 2, 3 … (15)
𝑓 𝑓
𝑚𝑛+1 = 𝑑𝑛+1 cos(𝜃) , 𝑛 = 0, 1, 2, 3 … (16)

As a termination criteria could be used some variables like: CPU time, or difference threshould between two
last results or the maximum number of iterations.
Flow chart of the MBA algorithm that could be used as pseudo implementation for MPPT controller shown at
Fig. 3.

593
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Flow chart of the MBA algorithm.

5. RESULTS AND DISCUSSIONS

At this paper were presented simplified scheme of PV standalone system at MATLAB/Simulink which were
used to perform initial testing of several power point tracking algorithms. The main objective of this work was
present adaptation of novel meta-heuristic mine blast algorithms for usage with MPPT device, which could be
included to DC-DC control module. For the mind blast algorithm, at this paper its general definition and
technical flow chart were presented, that requires only one additional step to add by any engineer: computation
of objective function and constraints and definition of convergence criteria. At out study for this needs SEPIC
controller was used and objective function was adopted for it.
During the experiments with MBA by comparison with it PSO algorithm it is shown that it has faster time to
seek GMPP and constantly better efficiency.
For the future study of this research the MBA algorithm and its improved [8] version should be compared to
different set of meta-heuristic algorithms with coverage of different PSC patterns and PV array layouts with a
full specter of defined metrics (that were proposed at section 2 of this paper).

REFERENCES

[1] Naciri, M , Aggour, M , Ait Ahmed, W . Wind energy storage by pumped hydro station. Journal of Energy Systems
2017; 1: 32-42 <http://dergipark.gov.tr/jes/issue/30882/329315>
[1] Sineglazov, V.M. Prospects for the development of the global electric power industry until 2035. Electricity,
transmission and distribution 2011; 2: 103.
[2] Tugcu C., Ozturk I., Alper А., Renewable and non-renewable energy consumption and economic growth relationship
revisited: Evidence from G7 countries. Energy Economics 2012; 34(6); 1942-1950.
<https://doi.org/10.1016/j.eneco.2012.08.021>

594
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[3] Rahul R, Chandel S. S.. Review of Maximum‐Power‐Point Tracking Techniques for Solar‐Photovoltaic Systems.
Energy Technology 1.8 2013; 438-448.
[4] Farzaneh, J., Keypour, R., & Khanesar, M. A New Maximum Power Point Tracking Based on Modified Firefly
Algorithm for PV System Under Partial Shading Conditions. Technology and Economics of Smart Grids and Sustainable
Energy 2018; 3(1); 1-13.
[5] Banaei M. R., Shirinabady M. R., Mirzaey M. MPPT control of Photovoltaic using SEPIC converter to reduce the input
current ripples. Int. Journal of Engineering Research and Applications 2014; 4(1); 160-166.
[6] Ridley R. Analyzing the Sepic converter. Power Systems Design Europe 2006; 14-18.
[7] Sadollah A., Bahreininejad A., Eskandar H., Hamdi M. Mine blast algorithm for optimization of truss structures with
discrete variables. Computers and Structures 2012; 102–103; 49–63.
[8] Sadollah A., Eskandar H., Bahreininejad A.,⇑, Joong Hoon Kim. Water cycle, mine blast and improved mine blast
algorithms for discrete sizing optimization of truss structures. Computers and Structures 2015; 149; 1–16.

595
 HYBRID ENERGY STORAGE SYSTEMS

Victor Sineglazov
NAU, Kyiv, Ukraine, [email protected], ORCID: 0000-0002-3297-9060

Denis Karabetsky
NAU, Kyiv, Ukraine, [email protected], ORCID: 0000-0002-1325-9279

Sineglazov, V., Karabetsky, D. Hybrid energy storage systems. 8th Eur. Conf. Ren. Energy
Cite this paper as:
Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: This paper considers the problems of designing hybrid energy storage systems (HESS). It is shown
that these systems are a complex of energy storage and storage technologies, each of which has its
own advantages and disadvantages. This type of system is highly efficient with proper design,
which in turn will provide high output power stability. The HESS system may include flywheels,
supercapacitors, batteries, fuel cells, SMES, etc. It is analyzed the use of every possible component
as a part of HESS in different operating conditions. It is proposed a hybrid genetic algorithm for
solving multicriteria optimization problems. Its usage permits to determine the structure and
parameters the components of HESS. The use of a genetic algorithm at the first stage of the neural
network optimization will allow to expand the area of possible solutions search and to increase the
approaching chances to the point of the global minimum. At the second stage the use of
optimization algorithm which is based on the method of steepest descent (Adagrad, RMSProp, and
Adam) will permit the guarantee to search the point of global minimum with higher probability.
Keywords: Hybrid Energy Storage Systems; Genetic Algorithm; Multi-objective Optimization;
© 2020 Published by ECRES

1. INTRODUCTION

In the modern world, when applying and designing hybrid power systems, the classic problem of storing,
accumulating and buffering generated energy arises. One approach that is ready to solve this problem is the use
of a hybrid energy storage system (HESS). HESS is a rather promising technology.
The list of tasks that this system can handle includes:
 smoothing the unpredictability of generated energy, its intermittent nature, into a predictable and, most
importantly, uninterrupted energy source, which in turn helps to integrate with the region's energy grid;
 energy arbitrage, for example, energy storage during the period of a cheap tariff (night time) and selling
during an increase (daytime);
 compensation of peak loads
 power grid support during blackout.
Accordingly, when using the HESS, these objectives must be taken into account in order to ensure
adequate quality and satisfy the constraints of the system. Since the content of HESS involves the use of many
different technologies that may be suitable for solving some problems, but not suitable for solving other
problems, which requires at the design stage to combine many positive and negative aspects of each of the
technologies.
It is common when designing HESS such main problems should be solved globally: the choice of storage
technology, the choice of their scale, the choice of the system control algorithm. Having solved all these three
interrelated problems, we can assume that the designed HESS is optimal, which makes the solution of this
problem very difficult.

596
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2. ANALYSIS OF HYBRID ENERGY STORAGE SYSTEMS

Hybrid energy storage system characteristics can be various, such as: energy and power density, charge /
discharge / self-discharge rate, service life, efficiency. If these systems are used with photovoltaic stations, they
get additional work, compensation for highly fluctuating loads, etc. The energy management system is used to
provide increased efficiency. HESS also includes two types of storage devices: HPS (high power storage device
used to smooth out high-frequency oscillations) and HES (high energy storage device used to smooth out low-
frequency oscillations). HPS - often can consist of superconducting magnetic energy storage devices (SMES),
flywheels, supercapacitors, etc., HES - various types and types of storage batteries (PbAc, VRB, LiPo, Li-ion,
etc.).
HESS could consist of following subsystems:
 Superconducting magnetic energy storage devices (SMES) - electromagnetic energy storage, which
could store electric energy in magnetic field, used for damping of energy provision processes, smoothing
power fluctuations, control of power consumption and its quality;
 rechargeable batteries (PbAc, Li-ion etc) - electrochemical storage of energy, used if connected
consumers require to provide additional energy for a long time;
 flywheels - mechanical storage of energy used mainly in industry to stabilize power surges;
 supercapacitors are a more advanced type of capacitor that can store the right amount of energy to
quickly dump it at the right moment;
 fuel cells - electrochemical energy storage, use redox reactions for energy conversion, mostly based on
hydrogen and oxygen.
HESS has a huge number of tasks: determining the size of the subsystem, choosing the optimal storage
systems, optimizing the life cycle, as well as management tasks: coordinating, monitoring and control the status
of the subsystem [1].

3. HYBRID GENETIC ALGORITHM FOR THE DESIGN SOLUTION

It is considered a multi-objective sizing methodology for hybrid energy storage systems. It is proposed a Hybrid
Genetic Algorithm for Solving Multicriteria Optimization Problems [2-3].

Features of the developed algorithm:

 use the concept of Pareto-optimality, dominance and density of solutions to calculate the adaptability of
the solution;
 adaptive determination of the probability of crossing and mutation;
 use of a special type of crossover – neighborhood crossover;
 support of an external multitude of individuals (archive) for the realization of elitism;
 clustering to eliminate clumps of points and increase the representativeness of the solution;

Description of the algorithm

Inputs:
 N (population size),
 NA (archive size),
 T (maximum number of generations),
 Pc (initial value of the probability of crossing),
 Pm (initial value of mutation probability).

Output: А (a plurality of non-dominant individuals).

Step 0. Initialization: Initial population is generated P0, for this:

1) Randomly (according to the uniform distribution law), N of the possible solutions of the problem in the area
of feasible solutions is chosen. If the solution is a certain vector X(x1 ... xk), then each component of this vector
is randomly selected.
2) The resulting N random resolutions are encoded into a binary code. Depending on the constraints on the
minimum and maximum values of the junction (or individual components of a vector), the number of bits

597
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

required to encode that value (or a given component of a vector) is determined. For example, using 10 bits, each
of which corresponds to one bit, you can encode numbers from 0 to 210=1024. The encoded binary representation
(the vector of zeros and ones) formed by the encoding is called a chromosome.
We also create an empty archive А0 =  and set the iteration number t = 0.

Step 1. Determination of fitness: performed based on the use of the concept of Pareto domination, algorithm
for calculating which (fitness function F) for each of the individuals in Рt and Аt has the following form.
1) Suppose we have a plurality of individuals that make up a population Pt and archive At, where each
individualіis assigned a value S(i)[0,1), called "power" (which shows how many decisions it dominates), which
is proportional to the number of members of the population j Рt, for which f(i) ≥ f(j)in the case of multi-criteria
optimization. The proportion is as follows:
𝑛
𝑆(𝑖 ) = (1)
𝑁+1
where N is the population size; n is the number of individuals who dominate the condition f(i) ≥ f(j).
The "power" of each individualіin many individuals of the archive At and populations Pt will be defined as the
sum of the "force" per individual and "force" given the dominance coded by I criterion over criterion encoded
by j:
𝑆(𝑖 ) = |{𝑗|𝑗𝜖𝑃𝑡 + 𝐴𝑡 ∧ 𝑖 ≻ 𝑗}| (2)
where 𝑥𝑘+1 . . . х1 is the dominance coded by i criterion over criterion encoded by j
2) Based on value S(i) is calculated raw the value of the fitness function R(i) traits i, which is calculated by
summing the "forces" of all individuals j, which dominate or weakly dominate the individual and by all criteria
𝑥𝑘+1 … х1 𝑅(𝑖) = ∑𝑗 ∈ 𝑃𝑡 + 𝐴𝑡 , 𝑗 > 𝑖𝑆(𝑗).
where Pt a plurality of individuals of the population; At is the multiple individuals archive
3) The density estimation method is an adaptation of the k-th neighbor method, where density at any point is a
(decreasing) function of the distance to the k-th nearest data point. A function is called a drop-down at some
interval if for any argument value from that interval; a larger argument value corresponds to a smaller function
value. The density inversion is taken to invert the distance to the k-th nearest neighbor. For each individual and,
distances (in the criterion space) to the individuals j in the archive and in the general sample are counted and
stored in the list.
To calculate the value of the fitness function, the value of the density of the individuals is used: for each
individual, the Cartesian distance from it to the other individuals is calculated j in the archive and population.
After sorting the list in ascending order kth the element gives the desired distance to the individual і, which is
indicated 𝜎𝑖𝑘 . That is 𝜎𝑖𝑘 denotes the distance from the individualіto the nearest k-th the neighbor. We use k,
equal to the square root of the sample size.
We calculate the density value D(i) for the individual i:
1
𝐷(𝑖) = 𝑘 (3)
𝜎𝑖 + 2
𝑘 = √(𝑁 + 𝑁𝐴 (4)
where N is the size of the population; NA is the size of the archive; k is rounded to the nearest integer.
A denominator is added to the denominator to ensure that its value is greater than zero and that 𝐷(𝑖 ) < 1
4) Finally, adding 𝐷(𝑖 ) to the original fitness value 𝑅(𝑖 ) individual i gives its fitness 𝐹 (𝑖 ). Therefore, the final
value of the fitness function F(i) for the individualіis defined as 𝐹 (𝑖 ) = 𝑅 (𝑖 ) + 𝐷(𝑖 ).

Step 2. Archive upgrade. Create an interim archive 𝑥𝑖 .

а) copy individuals whose decision vectors are invariant with respect to Рt in Аt*;
b) remove those individuals from Аt*, whose corresponding decision vectors are weakly dominated by Аt*;
c) reduce the number of individuals stored in the archive and place the resulting reduced number of individuals
in the At+1. The following procedure is used to reduce the number of individuals in the archive:
1) All nondominant individuals (values of fitness functions less than 1) from the archive and populations are
copied to the archive of the next iteration:
𝐴𝑡+1 = {𝑖|𝑖 ∈ 𝑃𝑡 + 𝐴𝑡 ∧ 𝐹(𝑖) < 1}, 𝑖=1,N̅̅̅̅̅ (5)
where Pt is the intermediate population; Аt is the intermediate archive; Аt+1 is the reduced archive; F(i) are fitness
features for the individual і; N is the population size.
2) If the size of the created archive corresponds to the desired one (|𝐴𝑡+1 | = 𝑁𝐴 ), then the step is completed,
otherwise two situations are possible.

598
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3) The size of the created archive:

a) less than desired (|𝐴𝑡+1 | < 𝑁𝐴 ). In this case, the best 𝑁𝐴 − |𝐴𝑡+1 | individuals that dominate the individuals
in the previous archive and the population are copied to the new archive. This can be done by sorting the
combined set Р𝑡 + А𝑡 according to the values of their fitness functions and the first 𝑁𝐴 − |𝐴𝑡+1 | individuals
with 𝐹 (𝑖 ) ≥ 1 from the sorted set are copied to the archive А𝑡 + 1;
b) more than desired (|𝐴𝑡+1 | = 𝑁𝐴 ). In this case, the archive size reduction process begins, which iteratively
removes the individuals from А𝑡 + 1, until (|𝐴𝑡+1 | = 𝑁𝐴 ).
 create a blank list D;
 for each individual і its Cartesian distance is calculated d(i, j) to all other individuals j. The
smallest of the distances obtained 𝑑𝑖 = 𝑚𝑖𝑛 𝑑(𝑖, 𝑗)is added to the list 𝐷 = 𝐷і, where
𝑖 = ̅̅̅̅̅
1, 𝑁.
If there are several individuals with a minimum distance, then the problem is solved, taking into account the
second shortest distance, etc.
 the values of the received list D are sorted in ascending order. The person corresponding to the
lowest value in the list D is removed from the archive;
 the process will be repeated until the archive reaches the desired size
(|𝐴𝑡+1 | = 𝑁𝐴 ).

Step 3. Ranking. Individuals in the population are sorted according to descending fitness values (from best
individual to worst).

Step 4. Grouping. The individuals are divided into groups, each consisting of two individuals. These two
individuals are selected from the beginning of the list of sorted individuals.

Step 5. Crossing and mutation. In each of the formed groups there is a crossover and a mutation.
Crossing For any two individuals, the need to perform the crossing operation is determined randomly:
1) A number is randomly generated рс* from the interval [0,1]. The obtained number is compared with the given
probability of crossing. If рс* >рс(t), where рс(t) –probability of crossing on t iterations, the crossing does not
occur and the parent individuals remain unchanged. Otherwise, the process of crossing occurs.
2) This algorithm uses the single-point crossover variant - both selected "parents" are cut at a randomly selected
point, after which their chromosomes exchange their fragments.

Mutation
The need to perform a mutation is determined similarly to a similar operation for crossing:
3) A number is randomly generated рm* from the interval [0,1]. The obtained number is compared with the given
mutation probability. If рm* > рm(t), where рm(t) – the probability of a mutation at t iteration, then the mutation
does not occur and the parent individuals remain unchanged. Otherwise, a mutation process takes place (see
paragraph 4, step 5).
4) A mutation is associated with a random change of one or more genes in the chromosome.
Two parent individuals are formed from two parent individuals. Parents are removed from the group.

Adaptation of probabilities of crossing over mutations

5) Crossing progress values are calculated 𝐵𝐼𝐽 and the progress of the mutation 𝑋 = (∑𝑛𝑖=1 𝑥𝑖 )/𝑛. New values
of crossing and mutation probabilities are calculated based on the calculated values of the progress:
𝑝𝑐(𝑡 + 1) = 𝑝𝑐(𝑡) + 𝛼 · 𝑝𝑐 (𝑡) and 𝑝м (𝑡 + 1) = 𝑝м(𝑡) – 𝛼 · 𝑝м(𝑡), if 𝐶𝑃 ̅̅̅̅ > 𝑀𝑃
̅̅̅̅̅
(6)
𝑝𝑐(𝑡 + 1) = 𝑝𝑐(𝑡) – 𝛼 · 𝑝𝑐(𝑡) and 𝑝м(𝑡 + 1) = 𝑝м(𝑡) + 𝛼 · 𝑝м(𝑡), if 𝐶𝑃 ̅̅̅̅ > 𝑀𝑃
̅̅̅̅̅
where α is the adaptation rate; α belongs [0.10; 0.15].

Step 6. All the children are grouped together into a new population Pt.

Step 7. End. Put Рt  1 Аt and t  t 1 . If t ≥ Тor some other stop criterion is then fulfilled Аt– there is a
multitude of resolutions you want, otherwise go to step 1.
Step 8. Eliminate clots of dots
1) Initialize multiple clusters С. Each individual iА forms a separate cluster Ci.

599
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2) If |С| ≤ NА, go to step 5 of step 9, otherwise go to step 3 of step 9.

3) Calculate the distance between all possible pairs of clusters. Remoteness dc two clusters Ci and CjС is defined
as the average distance between pairs of individuals belonging to these clusters:
1
𝑑𝑐 = ∑ ||𝑐𝑖 − 𝑐𝑗 ||, (7)
|𝐶𝑖 | ⋅ |𝐶𝑗 |
𝑐𝑖 ∈𝐶𝑖,𝑐𝑗 ∈𝐶𝑗
de operator || ⋅ || reflects the Euclidean distance (in the space of goals) between two individuals ci and cj, operator
|| ⋅ || determines the number of elements in the set.
Let 𝑥 = (𝑥1 , 𝑥2 , . . . , 𝑥𝑘 )and 𝑦 = (𝑦1 , 𝑦2 , . . . , 𝑦𝑘 )– decisions that are relevant to the individual X and Y, then
the Euclidean distance between these individuals can be defined as ||𝑋 − 𝑌|| = ||𝑥 − 𝑦|| = √∑𝑘𝑖=1(𝑥𝑖 − 𝑦𝑖 )2
4) Identify two clusters Ci and CjСwith a minimum distance dc. These clusters merge into one larger cluster
𝐶𝑖𝑗 = 𝐶𝑖 ∪ 𝐶𝑗 . Go to Step 2.
5) Select a representative individual for each cluster, and remove all other individuals from the cluster. Thus, a
representative individual is a centroid, a point with a minimum average distance to all other points of the cluster.
Define a reduced archive by combining representative individuals of all clusters.

4. RESULTS OF MULTILEVEL OPTIMIZATION ALGORITHM

The result of the proposed algorithm is the approximation of the Pareto set is representative – the points are
evenly distributed, there is no condensation, there are no dominant points.
The use of a genetic algorithm at the first stage of the neural network optimization will allow to expand the area
of possible solutions search and to increase the approaching chances to the point of the global minimum. At the
second stage the use of optimization algorithm which is based on the method of steepest descent will permit the
guarantee to search the point of global minimum with higher probability.
During the experiments, it was used the following tuning optimizers: for the steepest descent algorithm,
Adagrad, RMSProp, and Adam a training coefficient was 0.01, for an accelerated Nesterov gradient used a
training coefficient of 0.01 and a impulse coefficient of 0.9, for Adadelta optimizer used a training coefficient
of 1. The values of coefficients were chosen experimentally with the purpose of algorithm functioning efficiency
increasment.

5. CONCLUSIONS

In this paper, an algorithm was presented that can be used in design problems. Despite the fact that it solves only
one of the parts of the design problem for optimal HESS, optimal design also needs to take into account the
control algorithm of this system, which in turn requires additional modification of the design and optimization
process.
As for continuing research in this area, focus should be on a more detailed description of the design process with
full stops at all stages, starting from storage technologies, their characteristics, as well as their applicability in
solving various problems, which will help to introduce a qualitative assessment of each of them, as well as it
allows to correctly estimate the size of the system. Also, it is necessary to analyze and show in more detail the
available HESS control algorithms, as well as their general impact on the design problem being solved, which,
as a result, will help to find the optimal control strategy.

REFERENCES

[1] Coello С. A. A comprehensive survey of evolutionary-based multiobjective optimization techniques. Laboratorio

Nacional de Informatica Avanzada, Veracruz, Mexico, 1998.
[2] Fonseca C. M. Multiobjective optimization and multiple constraint handling with evolutionary algorithms. A unified
formulation. Technical report 564; University of Sheffield, Sheffield, UK, January 1995.
[3] Diederik P. Kingma, Jimmy Lei Ba. Adam. A Method for Stochastic Optimization. In: International Conference on
Learning Representations 2015; 30 Jan 2017; pp. 1–15.

600
 ASSESSING S-CO2 MIXTURES FOR POTENTIALLY
IMPROVING SUPERCRITICAL POWER PLANTS
PERFORMANCE

Robert Valencia-Chapi
Universidad Politécnica de Madrid, Madrid, Spain, [email protected],
Universidad Técnica del Norte, Ibarra, Ecuador, [email protected], ORCID: 0000-0003-1977-2118

Luis Coco-Enríquez
Universidad Internacional de la Rioja, La Rioja, Spain, [email protected], ORCID: 0000-0002-3655-2654

Javier Muñoz-Antón
Universidad Politécnica de Madrid, Madrid, Spain, [email protected], ORCID: 0000-0002-1980-0863

Valencia-Chapi, Robert, Coco-Enríquez, Luis, Muñoz-Antón, Javier. Assessing s-CO2

Cite this paper as: Mixtures for Potentially Improving Supercritical Power Plants Performance. 8th Eur. Conf.
Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: This article aims to show that power cycle working fluid chemical composition is a key issue in
new-generation of the Concentrated Solar Power Plant coupled to s-CO2 Brayton Power Cycles
(CSP-sCO2) design. In this work, two Brayton power cycle configurations are considered:
recompression (RCC), and recompression with reheating (RCC-RH). The solar power plant
reference power is 50 MW and the CIT is fixed to 51 ºC. The overall conductance of the low- and
high-temperature recuperators varies from 2.5-15 MW/K, and the supercritical Brayton power
cycle performance parameters are optimized with the BOBYQA algorithm. The thermal efficiency
is improved by adding high-molecular-weight substances (COS, H2S, NH3, R12, R32, R40, and R161)
to pure s-CO2. The RCC-RH configuration achieves the best performance with carbonyl sulfide
(COS = 30% and s-CO2 = 70% mole fractions), with a maximum thermal efficiency 46.2% (for
UA = 30 MW/K). The working fluid selection is closely related with the power plant location and
depends on the local annual ambient temperature profile.
Keywords: Brayton cycle; fluid mixture; solar plant; supercritical CO2.
© 2020 Published by ECRES

Nomenclature
CO2 Carbon Dioxide
COS Carbonyl Sulfide [O=C=S]
CIT Compressor Inlet Temperature
CSP-sCO2 Concentrated Solar Power Plant coupled to s-CO2 Brayton Power Cycles
H2S Hydrogen Sulfide
NH3 Ammonia
NIST National Institute of Standards and Technology
PTC Parabolic Trough Solar Collector
PC Precooler
R12 Dichlorodifluoromethane
R32 Difluoromethane
R40 Methyl Chloride
R161 Fluor ethane
RCC Recompression
RCC-RH Recompression with Reheating
REFPROP Reference Fluid Thermodynamic and Transport Properties Database
s-CO2 Supercritical Carbon Dioxide
UA Heat Exchanger Conductance

601
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION

The next generation of concentrated solar power (CSP) plants are likely to be founded on supercritical carbon
dioxide (s-CO2) Brayton power cycles coupled to concentrated solar collectors. In parallel, the latest research
on supercritical Brayton power cycles has focused on defining the optimum chemical composition for the
supercritical working fluid [1]. The ever-increasing necessity to reduce the environmental impact of industrial
and urban energy conversion processes has led engineers to consider the use of s-CO2 mixtures as working fluid
for thermodynamic power and refrigeration cycles [2], emphasizing the mitigation of air pollution and the
emission of greenhouse gases, and reducing energy production costs [3].
The main purpose of this study is to compare the benefits of various s-CO2 mixtures (s-CO2/COS, s-CO2/H2S,
s-CO2/NH3, s-CO2/R12, s-CO2/R32, s-CO2/R40, and s-CO2/R161) used in s-CO2 Brayton power cycles coupled with
line-focusing solar power plants.

2. ASSUMPTIONS AND METHODS

For plant performance calculations, a turbine inlet temperature of 550 °C has been established. In addition to the
recent publications, other complex Brayton cycle configurations have been studied: recompression (RCC),
recompression with reheating (RCC-RH) (see Figure 1).

(a) (b)
Figure 1. Brayton power cycles layout: (a) recompression (RCC), and (b) recompression with reheating (RCC-RH). MC:
main compressor; RC: recompressor; G: generator; T: turbine; PC: precooler; FS: fluid split; FM: fluid mixture; LTR:
low-temperature recuperator; HTR: high-temperature recuperator; PHX: primary heat exchanger; SF: solar field; RHX:
reheating exchanger.

The efficiency of these cycles compared to their total recuperator conductance, using pure s-CO2 as the working
fluid, is shown in Figure 2(a). Figure 2(b) shows the temperature vs entropy diagrams for the cycles Brayton
using pure s-CO2 as working fluid.

(a) (b)
Figure 2. (a) Cycle efficiency vs UA for RCC, and RCC-RH Brayton cycles using pure s-CO2. (b) Thermodynamic plane
T-S (UAtotal = 30 MW/K) for RCC, and RCC-RH Brayton cycles using pure s-CO2.

602
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The Supercritical Concentrated Solar Power Plant (SCSP) software [4] has been used to simulate the
performance of the Brayton cycle configurations using pure s-CO2 and s-CO2 mixtures. The fluid properties
were obtained from the REFPROP [5] database developed by the National Institute of Standards and Technology
(NIST) in the USA. The computer program SCSP has been developed by members of the Energy Engineering
Department at the Technical University of Madrid and it is based on the software core developed by J. Dyreby
in his doctoral thesis [6]. SCSP sets a constant heat recuperator total conductance for the performance calculation
methodology and uses optimization algorithms (SUBPLEX [7], NEWUOA [8] and BOBYQA [9]) to calculate
the optimal cycle performance.
Figure 3 shows the evolution of the critical pressure and temperature of the mixtures as a function of the mole
fraction. Figure 3(a) shows that the critical temperature distributions of the mixtures follow a monotonous,
almost-linear upward trend. Figure 3(b) indicates that the pressure varies for different binary mixtures based on
s-CO2, displaying a nonlinear relation between the critical pressure and the mole fraction.

(a) (b)
Figure 3. (a) Variation of the mixture’s critical temperature vs the mole fraction. (b) Variation of the mixture’s critical
pressure vs the mole fraction.

The main assumptions in this work are summarized in Table 1.

Table 1. Input parameters.

Nomenclature Value Unit
Net power output W 50 MW
Compressor inlet temperature T1 51 ºC
Compressor inlet pressure P1 optimized MPa
Turbine inlet temperature T6 550 ºC
Turbine inlet pressure P6 25 MPa
Compressor efficiency [10] ηc 0.89 -
Turbine efficiency [10] ηt 0.93 -
Heat exchanger conductance for LTR UALT 2.50  15.00 MW/K
Heat exchanger conductance for the HTR UAHT 2.50  15.00 MW/K
Split fraction γ optimized -
Drop pressure LTR and HTR [11] ΔP/PLTR | ΔP/PHTR 0.015 | 0.01 -
Drop pressure precooler [11] ΔP/PPC 0.02 -
Drop pressure PHX and RHX [11] ΔP/PPHX | ΔP/PRHX 0.015 | 0.015 -

3. RESULTS AND DISCUSSION

The best results for the two configurations studied in this work were obtained with mixtures of s-CO2/COS
(70/30), s-CO2/H2S (60/40), s-CO2/NH3 (81/19), s-CO2/R12 (80/20), s-CO2/R32 (70/30), s-CO2/R40 (91/09), and
s-CO2/R161 (84/16) (molar fractions shown in parentheses). This is because, with these molar fractions, the

603
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

critical temperature of each mixture is just below 51 °C, corresponding to the study temperature. Figure 4(a)
suggests that, in the RCC cycle, the mixtures achieve enhanced efficiency of up to 4% compared with the cycles
using pure s-CO2. In the RCC-RH cycle [Fig. 4(b)], the efficiency improves by slightly more than 4%.

(a) (b)
Figure 4. Cycle efficiency vs UA: (a) RCC using s-CO2 mixtures, and (b) RCC-RH cycle using s-CO2 mixtures.

The use of s-CO2 mixtures has a limit due to the pinch point in the low and high-temperature recuperators
(between 5-10 ºC). The pinch point decreases exponentially when the UA increases (Figure 5).

(a) (b)
Figure 5. (a) LTR Pinch point vs UA for RCC using s-CO2 mixtures. (b) HTR Pinch point vs UA for RCC cycle using s-
CO2 mixtures.

The values of the pressure ratio for the RCC cycle optimal efficiency are shown in Figure 6 where the range for
s-CO2 pure is 2.39 to 2.79 and a 2.71 to 3.3 range for the s-CO2 mixtures. Also was confirmed the beneficial
impact in the power cycle thermal efficiency of increasing the compressor inlet pressure above the working fluid
critical pressure, when the compressor inlet temperature is above the working fluid critical temperature, as
previously studied for pure s-CO2 in J. Dyreby thesis [6].

604
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 6. Pressure ratio vs UA for RCC using s-CO2 mixtures.

4. CONCLUSIONS

The main conclusion of this work is that the s-CO2 mixtures with substances studied (COS, H2S, NH3, R12, R32,
R40, and R161) have a beneficial impact on CSP efficiency, as previously reported by Binotti et al. [11] with TiCl 4
(Titanium Tetrachloride) blend in CSP-sCO2. The line-focusing solar plant with s-CO2 binary-mixture Brayton
power cycles offers an alternative, cost-competitive technical solution to CSP plants with similar power cycles.

This study shows the necessity of researching more complex customized supercritical fluids as the working fluid
for Brayton power cycles and as a heat transfer fluid in the solar fields. The RCC-RH configuration provides the
best plant efficiency and the addition of COS achieves the best performance results (cycle efficiency equal to
46.2% for UA=30MW/K). COS is a linear molecule consisting of a carbonyl group double bonded to a sulfur
atom (O=C=S), with similar structure as CO2 molecule. A double bond in chemistry is a covalent bond between
two atoms involving four bonding electrons instead of the usual two; hence, the COS stability at high
temperature is presumable good.

The adoption of a thermal energy storage system in the solar field, and for the ultimate heat sink, will play an
important role in stabilizing the system’s annual performance, absorbing solar irradiance and ambient
temperature variations, and should be compatible with the working fluid selected.

As future work a CIT sensing analysis will be developed for studying the power cycle behavior under ambient
temperature fluctuations. The target is defining the CSP-sCO2 optimized performance map, considering the
Direct Normal Irradiance and ambient temperature as main variables, and recompression fraction, compressor
inlet pressure and UA as the parameters to be optimized. Also is important to corroborate the added substances
potential chemical decomposition at high-pressure and temperature do not impact negatively on the Brayton
cycle performance.

ACKNOWLEDGMENT

This work has been supported by the Industrial Doctorates program of the Community of Madrid
(IND2018/IND-9952).

REFERENCES

[1] R. Valencia-Chapi, L. Coco-Enríquez, and J. Muñoz-Antón, “Supercritical CO2 Mixtures for Advanced Brayton Power
Cycles in Line-Focusing Solar Power Plants,” 2019.

605
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[2] M.-J. Li and W.-Q. Tao, “Review of Methodologies and Polices for Evaluation of Energy Efficiency in High Energy-
Consuming Industry,” Appl. Energy, vol. 187, pp. 203–215, 2017.
[3] L. Rivera-González, D. Bolonio, L. F. Mazadiego, and R. Valencia-Chapi, “Long-Term Electricity Supply and Demand
Forecast (2018-2040): A LEAP Model Application Towards a Sustainable Power Generation System in Ecuador.,”
Sustain., vol. 11, no. 19, 2019.
[4] L. Coco-Enríquez, “Thesis: Nueva Generación de Centrales Termosolares con Colectores Solares Lineales Acoplados
a Ciclos Supercriticos de Potencia,” Universidad Politécnica de Madrid, 2017.
[5] E. W. Lemmon, I. H. Bell, M. L. Huber, and M. O. McLinden, “NIST Standard Reference Database 23: Reference Fluid
Thermodynamic and Transport Properties-REFPROP, Version 10.0,” Natl. Inst. Stand. Technol. Stand. Ref. Data
Program, Gaithersbg., 2018.
[6] J. J. Dyreby, “Thesis: Modeling the Supercritical Carbon Dioxide Brayton Cycle with Recompression.,” University of
Wisconsin-Madison, 2014.
[7] T. H. Rowan, “Thesis: Functional Stability Analysis of Numerical Algorithms SUBPLEX,” 1990.
[8] M. J. D. Powell, “The NEWUOA Software for Unconstrained Optimization without Derivatives,” pp. 255–297, 2006.
[9] M. Powell, “The BOBYQA Algorithm for Bound Constrained Optimization without Derivatives,” NA Rep.
NA2009/06, p. 39, 2009.
[10] M. Kulhánek and V. Dostál, “Thermodynamic Analysis and Comparison of Supercritical Carbon Dioxide Cycles,” pp.
2–8, 2011.
[11] M. Binotti and G. Manzolini, “Supercritical Carbon Dioxide/Alternative Fluid Blends for Efficiency Upgrade of Solar
Power Plant,” 2019.

606
 NUMERICAL SIMULATION ON MATHEMATICAL MODELING
OF VERTICAL CONTACT MODE TRIBOELECTRIC
GENERATORS

Shimna Shafeek
University of Bolton, United Arab Emirates, s.shafeek @bolton.ac.uk, ORCID: https://orcid.org/0000-0001-7396-6808

Mohammad Asim Riaz

University of Bolton, United Arab Emirates, [email protected], ORCID: https://orcid.org/0000-0003-3122-7219

Shafeek, Shimna, Riaz, Mohammad Asim. Numerical Simulation on Mathematical Modeling

Cite this paper as: of Vertical Contact mode Triboelectric Generators. 8th Eur. Conf. Ren. Energy Sys. 24-25
August 2020, Istanbul, Turkey

Abstract: Triboelectric generators which emerged as an upcoming energy harvesting device works on the
principle of contact electrifications and electrostatic induction. To understand the working
principle various theoretical models have been given by researchers. The models discussed vary
in the principle of explanation from the classical model of TENG based on Gauss –Maxwell
equation, the capacitive nature, the Electric field generation and it's characteristic with the distance,
till the electric charge lumped mode. This study aims to simulate the mathematical model which
can accurately simulate the charge transferred (Q), the current produced (I), the voltage generated
(V), Power produced (P) with respect to time and power generated to varying resistances for all
types of contact mode triboelectric nanogenerators. For this study and simulation, the conductor-
dielectric triboelectric generator was taken as a base reference and a numerical model was made
using MATLAB. The MATLAB simulation study of the mathematical model is applied to
dielectric-dielectric triboelectric nanogenerator and validated computationally using Comsol
which further can be extended to all modes of triboelectric generators for finding the output
characteristics for a set of input data.
.
Keywords: Mathematical model, numerical, simulation, triboelectric generator, energy harnessing
© 2020 Published by ECRES

1. INTRODUCTION

Scavenging waste mechanical energy from the surroundings as a sustainable energy source is emerging as a
trending field in the research area where it focuses on the development of energy source for micro nanosystems.
With the development of IoT systems, an enormous number of batteries are required for powering the sensors
used in it where the maintainability, life span, health and environmental hazards will be unpractical [1]. Sensors
to work as active power sensor is a developing scope of study in the research field area due to these reasons.
Triboelectricity as a sustainable energy harvesting source for powering sensors due to the reason of cost-
effectiveness, environmental friendly[2] and high efficiency for low-frequency low power systems[3].
The triboelectric effect can be observed between nearly any two known materials; be they metals, polymers, or
even wood. However, a material's ability to lose and gain electrons, a fundamental quality of triboelectricity,
depends on its polarity and decides its position on the triboelectric series [4]. The Triboelectric effect revolves
around a phenomenon where when an object becomes electrically charged when contacted with various
materials having different electron affinities through friction. On separation of the materials in contact, a few
associated atoms tend to retain additional electrons, and a few tend to give them away and perhaps produce
electrostatic tribocharges on the contact surfaces [5]. Studies have shown the recent development in this area
where energy harnessing using triboelectric can be done from body motion to sea waves which can contribute a
good amount of energy in microwatts to Mega and Giga values[5].

607
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

TEG which is organic based generator works on the principle of electrostatic induction and contact electrification
[6]. TEG works on four different modes based on the type of contact between the materials as contact mode,
sliding mode, single electrode mode and free-standing triboelectrification. Vertical contact mode applications
are studied here as more applications are coming under this mode[7]. It is one of the simplest forms of
triboelectric generators. It consists of two dielectric layers facing each on one side and other having electrodes
deposited on the other side and the electrodes is connected with a load. So, when the dielectric come in contact
with each other due to mechanical input, triboelectric charges are produced on the surface. As the layers separate
and a small gap is produced between them a potential drop between the electrodes is created. This causes
electrons from the higher potential electrode to flow towards the lower potential electrode to balance the
electrostatic field produced by the triboelectric charges. As the dielectrics come into contact again by external
means the potential difference is reversed and electrons flow back [8].
To explain the working principle thereby predicting the output characteristics of the circuit various theoretical
modeling have been published. The models are defined on different principles where Simiao Niu(2013)
explained the theoretical framework based on Gauss theorem whereas Dharmasena(2017) explained the concept
using Maxwell's equations[9-10].
In this study, the theoretical mathematical model for TEG based on Gauss theorem is investigated for numerical
simulation using MATLAB. The analysis is performed on a Case study of input parameters which is considered
in dielectric- dielectric and conductor- dielectric vertical contact mode. The study also aimed to evaluate the
comparative performance of output characteristics of dielectric-dielectric and conductor- dielectric vertical
contact mode in the current, power output characteristics with varying resistance values.

2. MATHEMATICAL MODEL OF TRIBOELECTRIC GENERATOR

The basic working theory of TEG in vertical contact mode includes two dielectrics separated by a distance of x
and with the external excitation static induced charges are produced as shown in Figure.1. The two tribomaterials
are of thicknesses d1and d2.[11-12]. The contact mode triboelectric generator classified into two types of TEGs:
Conductor-dielectric and dielectric-dielectric both of which are shown in the figure below. In Conductor-
dielectric in contrast to the first kind, there will be only one dielectric which is dielectric2 and metal1 will play
both the role of a dielectric1 and electrode [13].

Figure.1. Working model of a vertical contact mode TEG (a) Dielectric-Dielectric (b) Conductor-Dielectric

The Voltage Discharge Distance (V-Q-x) relationship in TEG parameters are defined as follows [14].

The strength of the electric field inside:

𝑄
𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐 𝐹𝑖𝑒𝑙𝑑 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 , 𝐸1 = − (1)
𝑆𝜀0 𝜀𝑟1

Where εr1 and εr2 are the relative permittivity of dielectric 1 and 2
𝑄
− +𝜎(𝑡)
𝑆
𝐸 𝑎𝑡 𝑡ℎ𝑒 𝑎𝑖𝑟 𝑔𝑎𝑝, 𝐸𝑎𝑖𝑟 = ; . is the surface density on contact surfaces. (2)
𝜀0

𝑄
𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐 𝐹𝑖𝑒𝑙𝑑 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦, 𝐸2 = (3)
𝑆𝜀0 𝜀𝑟2

608
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑉𝑜𝑙𝑡𝑎𝑔𝑒 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑡ℎ𝑒 𝑡𝑤𝑜 𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑑𝑒𝑠, 𝑉 = 𝐸1 𝑑1 + 𝐸2 𝑑2 + 𝐸𝑎𝑖𝑟 𝑥 (4)

The Voltage generated in the conductor-dielectric,

𝑄 𝑑 𝜎𝑥(𝑡)
𝑉 = 𝐸2 𝑑2 + 𝐸𝑎𝑖𝑟 𝑥 = − 𝑆𝜀 (𝜀 2 + x(t)) + 𝜀0
(5)
0 𝑟2

𝑑𝑖
𝐸𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠, 𝑑0 = ∑𝑛𝑖=1 , where 𝜀𝑟𝑖 is the relative permittivity (6)
𝜀𝑟𝑖

Substituting the effective dielectric thickness in Eqn. [5],

𝑄 𝜎𝑥(𝑡)
𝑉 = − 𝑆𝜀 (𝑑0 + x(t)) + 𝜀0
(7)
0

Based on electrodynamics the output characteristics under open circuit and short circuit conditions,
𝜎𝑥(𝑡)
𝑂𝑝𝑒𝑛 𝑐𝑖𝑟𝑐𝑢𝑖𝑡 𝑣𝑜𝑙𝑡𝑎𝑔𝑒, 𝑉𝑂𝐶 = (8)
𝜀0

𝑆𝜎𝑥(𝑡)
𝑆ℎ𝑜𝑟𝑡 𝐶𝑖𝑟𝑐𝑢𝑖𝑡 𝐶ℎ𝑎𝑟𝑔𝑒, 𝑄𝑆𝐶 = (9)
𝑑0 +𝑥(𝑡)

𝜀0 𝑆
𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑎𝑛𝑐𝑒, 𝐶 = 𝑑 (10)
0 +𝑥(𝑡)

Assuming periodic contact to be harmonic motion, where

𝑥𝑚𝑎𝑥 𝑥𝑚𝑎𝑥 𝑣
𝑥(𝑡) = cos (𝑥 )𝑡 (11)
2 2 𝑚𝑎𝑥

Equation (8) by applying the harmonic condition Eqn. [11] gives the below equations [13-15],
1 1 𝑡 1 𝑡 1 𝑡 1
𝑄(𝑡) = 𝑅 𝑒𝑥𝑝 [− 𝑅 ∫0 𝐶(𝑥(𝑡)) 𝑑𝑡 ] ∫0 𝑉𝑂𝐶 (𝑥(𝑡))𝑒𝑥𝑝 [𝑅 ∫0 𝐶(𝑥(𝑡)) 𝑑𝑡 ] (12)

𝑉𝑂𝐶 1 1 𝑡 1 𝑡 1 𝑡 1
𝐼(𝑡) = − 𝑅2𝐶 𝑒𝑥𝑝 [− 𝑅 ∫0 𝐶(𝑥(𝑡)) 𝑑𝑡 ] ∫0 𝑉𝑂𝐶 (𝑥(𝑡))𝑒𝑥𝑝 [𝑅 ∫0 𝐶(𝑥(𝑡)) 𝑑𝑡 ] 𝑑𝑡 (13)
𝑅

1 1 𝑡 1 𝑡 1 𝑡 1
𝑉(𝑡) = 𝑉𝑂𝐶 − 𝑒𝑥𝑝 [− ∫0 𝑑𝑡 ] ∫0 𝑉𝑂𝐶 (𝑥(𝑡))𝑒𝑥𝑝 [ ∫0 𝑑𝑡 ] 𝑑𝑡 (14)
𝑅𝐶 𝑅 𝐶(𝑥(𝑡)) 𝑅 𝐶(𝑥(𝑡))

Governing Equations
Assuming uniform motion under constant velocity, then the equations for generating real time power generation
of this theoretical model can be written as [16]:
𝐹 𝐹
𝑄(𝑡) = 𝜎𝑆 [1 − 𝑒𝑥𝑝(−𝐴𝑡 − 𝐵𝑡 2 ) + √2𝐹 𝑒𝑥𝑝(−𝐴𝑡 − 𝐵𝑡 2 ) × 𝐷𝑎𝑤𝑠𝑜𝑛 ( ) − √2𝐹 × 𝐷𝑎𝑤𝑠𝑜𝑛 ( + √𝑏𝑡)] (15)
√2 √2

𝐹 𝐹
𝐼(𝑡) = 𝜎𝑆 [𝑒𝑥𝑝(−𝐴𝑡 − 𝐵𝑡 2 )(𝐴 + 2𝐵𝑡) − √2𝐹 × 𝑒𝑥𝑝(−𝐴𝑡 − 𝐵𝑡 2 )(𝐴 + 2𝐵𝑡) × 𝐷𝑎𝑤𝑠𝑜𝑛 ( ) − 2𝐴 ( + √𝑏𝑡) ×
√2 √2

𝐹
𝐷𝑎𝑤𝑠𝑜𝑛 ( + √𝑏𝑡)] (16)
√2

𝐹 𝐹
𝑉(𝑡) = 𝜎𝑆𝑅 [𝑒𝑥𝑝(−𝐴𝑡 − 𝐵𝑡 2 )(𝐴 + 2𝐵𝑡) − √2𝐹 × 𝑒𝑥𝑝(−𝐴𝑡 − 𝐵𝑡 2 )(𝐴 + 2𝐵𝑡) × 𝐷𝑎𝑤𝑠𝑜𝑛 ( ) − 𝐴 + 2𝐴 ( +
√2 √2

𝐹
√𝑏𝑡) × 𝐷𝑎𝑤𝑠𝑜𝑛 ( + √𝑏𝑡)] (17)
√2

609
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Where A, B and C are constants which can be defined as:

𝑑0 𝑣 𝑑0 𝑥
𝐴= ; 𝐵= ; 𝐹= ;𝐷𝑎𝑤𝑠𝑜𝑛(𝑥) = 𝑒𝑥𝑝(−𝑥 2 ) ∫0 𝑒𝑥𝑝(𝑦 2 ) 𝑑𝑦
𝑅𝑆𝜀0 2𝑅𝑆𝜀0 √𝑅𝑆𝜀0 𝑣

3. NUMERICAL SIMULATION

The mathematical model discussed in the above study has been numerically investigated using MATLAB for
the vertical contact mode of the conductor-dielectric and dielectric-dielectric model. The materials chosen for
the numerical study is based on the electron affinities from the triboelectric series. PDMS PA6 combination is
used as the triboelectric pair. The surface charge density though it used to decay within the material as shown
in experimental studies, here the value is taken as constant σ =20µC/m2 as the aim is to investigate on the
performance of TEG by numerical analysis on its mathematical model. The input parameters used for the
analysis is given in Table I.
Table I. The Input parameters
Item Input Parameters Dielectric-Dielectric Conductor-Dielectric
no
1 Dielectric 1 PDMS 𝜀𝑟1 = 2.75 Metal

2 Dielectric 2 PA6 𝜀𝑟2 = 3.6 PA6 𝜀𝑟2 = 3.6

3 Thickness d1=100𝜇𝑚 d2=100𝜇𝑚 d1=0 d2=100𝜇𝑚

4 Surface Area of the

Dielectrics, S 4 × 10−4 𝑚2 4 × 10−4 𝑚2
6 Average velocity ,v 0.1𝑚/𝑠 0.1𝑚/𝑠
Moving distance,
5 xmax 0.005𝑚 0.005𝑚
A range of resistive load from 0,1MΩ, 100MΩ, 1GΩ, 10GΩ to Infinity are taken for the analysis.

Figure2.Dielectric-Dielectric Contact mode Output characteristics (a) Charge (b) Voltage Output (c) Current Output (d)
Power Output

610
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure3. Conductor-Dielectric Contact mode Output characteristics (a) Charge (b) Voltage Output (c) Current Output
(d) Power Output
Case Study I Dielectric-Dielectric Contact Mode

From the analysis result, as shown in Figure2 (a), at different load conditions for small R, Q increases and the
trend line shows saturation when the electrode movement stops at a particular ti me. Whereas for a higher value
of resistance there will be an absence of saturation due to continuous discharge value. The dynamic response of
current, voltage and power profile shows a similar trend line with the resistance values. Current value shows an
increase in value with less value of resistance and vice versa. The tribo material properties, the surface charge
densities, the geometrical parameters of the tribo pairs, velocity and distance of separation all have contributed
to the variation in the output values [17].

Case Study II Conductor Dielectric Contact Mode

As shown in Figure3.Conductor-Dielectric output characteristics profile shown similar trend with the dielectric-
dielectric mode where the power output and the current output shows a slightly higher value due to more charge
collected in the metal electrode than in the other mode. The comparative study of both the contact mode
numerical analysis on instantaneous power done as shown in Fig4(a) which shows a slight increase in the power
in the conductor dielectric. Surface charge density is an important factor which determines the output
characteristics.

The dynamic power with varying surface charge densities are numerically evaluated and compares as shown in
Figure4(b) which shows a tremendous increase in power with the increase in surface charge density.
The numerical simulation obtained using MATLAB is validated using the Comsol simulation where a set of

611
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure4.Power vs Resistance (a) Conductor Dielectric & Dielectric Dielectric comparison (b) varying surface charge
densities (c) Comsol simulation for V vs x (d) MATLAB numerical result & Comsol experimental result

dielectric-dielectric contact mode input parameters are used in Numerical simulation based on the Eqn.[9] to
find the Voc with respect to the varying x distance. The same input parameters are used in Comsol simulation
to get the Voc with the distance the top dielectric with the dielectric moves. The results show the theoretical
value matched with the computational value as shown in Fig4 (d).

4. CONCLUSION

The mathematical model of vertical triboelectric contact mode in conductor-dielectric and dielectric- dielectric
mode is investigated. The mathematical model based on Gauss theorem analyzed numerically in MATLAB to
evaluate the output characteristics of TEGs under a set of input data. A comparative study performed in both the
modes where the conductor dielectric mode shows a slight increase in instantaneous power and current at their
peak which further decays with the time due to the decay in the surface charge density within the dielectrics
when it moves and reaches the top electrode where the surface charge density decay can be further investigated.
The dynamic power generation with respect to the increase in surface charge density from 1µC/m2 to 100 µC/m2
shows an increase in power from 1mW to 1.6mW .The numerical analysis in MATLAB further validated in
Comsol simulation which showed the numerical simulation generated in MATLAB can be extended further to
study TEGs for optimizing their performance at any input parameters of TEG pairs.
ACKNOWLEDGEMENT

This research is supported by the Expo Live Innovation Impact Grant Programme 2020 who granted the
University of Bolton RAK Academic Centre AED 50,000 for winning the research and innovation university
summit of Expo 2020. We would also like to show our gratitude to Jikui Luo, Research Supervisor, University
Of Bolton, Bolton, BL3 5AB, United Kingdom, for sharing and mentoring the pearls of wisdom with us during
this research.

612
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] Siddique, A. R. M., Mahmud, S., Van Heyst, B..A comprehensive review on vibration based micro power generators using
electromagnetic and piezoelectric transducer mechanisms. Energy Conversion and Management (2015); 106, 728-747.
https://doi.org/10.1016/j.enconman.2015.09.071
[2] Abdelwahed, A., Amin, M., Elosairy, M., Abbasy, N. Theoretical modelling for enhancing contact-separation triboelectric
nanogenerator performance:In 2016 Annual Connecticut Conference on Industrial Electronics, Technology & Automation (CT-IETA)
(October 2016),IEEE, pp. 1-5.
[3] Zi, Y., Guo, H., Wen, Z., Yeh, M. H., Hu, C., Wang, Z. L..Harvesting low-frequency (< 5 Hz) irregular mechanical energy: a possible
killer application of triboelectric nanogenerator. ACS nano (2016) ;10(4), 4797-4805. https://doi.org/10.1021/acsnano.6b01569
[4] Dhakar, L. Triboelectric devices for power generation and self-powered sensing applications. Springer 2017.
[5] Han, M., Zhang, X.,Zhang, H. (Eds.).Flexible and Stretchable Triboelectric Nanogenerator Devices: Toward Self-powered Systems.
John Wiley & Sons, 2019.
[6] Wei, C., Jing, X. A.. comprehensive review on vibration energy harvesting: Modelling and realization. Renewable and Sustainable
Energy Reviews (2017) ; 74, 1-18.https://doi.org/10.1016/j.rser.2017.01.073
[7] Hossain, N. A., Razavi, M. J., Towfighian, S.. Analysis of mechanical deformation effect on the voltage generation of a vertical
contact mode triboelectric generator. Journal of Micromechanics and Microengineering (2020) ; DOI: 10.1088/1361-6439/ab6c74
[8] Wu, C., Wang, A. C., Ding, W., Guo, H., Wang, Z. L.. Triboelectric nanogenerator: a foundation of the energy for the new era.
Advanced Energy Materials (2019), 9(1), 1802906.https://doi.org/10.1002/aenm.201802906
[9] Niu, S., Wang, S., Lin, L., Liu, Y., Zhou, Y. S., Hu, Y.,Wang, Z. L.. Theoretical study of contact-mode triboelectric nanogenerators
as an effective power source. Energy & Environmental Science (2013),6(12), 3576-3583. https://doi.org/10.1039/C3EE42571A
[10] Dharmasena, R. D. I. G., Jayawardena, K. D. G. I., Mills, C. A., Deane, J. H. B., Anguita, J. V., Dorey, R. A., & Silva, S. R. P..
Triboelectric nanogenerators: providing a fundamental framework. Energy & Environmental Science (2017), 10(8), 1801-1811.
https://doi.org/10.1039/C7EE01139C
[11] Shafeek, S., Luo, J. Theoretical and numerical analysis of triboelectric nanogenerators for self-powered sensors :In 2016 5th
International Conference on Electronic Devices, Systems and Applications (ICEDSA) (2016, December), IEEE,pp. 1-4.
[12] Varghese, S., Shafeek, S., Kumar, R. S.,Mini, R. S. Computational investigation of material combinations in triboelectric generators :
In 2017 International Conference on Electrical and Computing Technologies and Applications (ICECTA), (2017, November),
IEEE,pp. 1-4.
[13] Niu, S.,Wang, Z. L.. Theoretical systems of triboelectric nanogenerators. Nano Energy (2015), 14, 161-
192,https://doi.org/10.1016/j.nanoen.2014.11.034
[14] Abdelwahed, A., Amin, M., Elosairy, M., Abbasy, N.. Theoretical modelling for enhancing contact-separation triboelectric
nanogenerator performance. In 2016 Annual Connecticut Conference on Industrial Electronics, Technology & Automation (CT-
IETA), (2016, October), IEEE,pp1-5.
[15] Shao, J., Jiang, T.,Wang, Z.. Theoretical foundations of triboelectric nanogenerators (TENGs). Sci. China Technol. Sci. (2020),
https://doi.org/10.1007/s11431-020-1604-9
[16] Wang, Z. L., Lin, L., Chen, J., Niu, S., Zi, Y.. Triboelectric nanogenerators. Berlin, Germany: Springer International Publishing,
2016.
[17] Wang, Z. L.. Triboelectric nanogenerators as new energy technology and self-powered sensors–Principles, problems and
perspectives. Faraday discussions (2015); 176, 447-458. https://doi.org/10.1039/C4FD00159A

613
 MICROCONTROLLER-BASED, LOW COST IMAGE-GUIDED
SOLAR TRACKING

Rughooputh Hridayesh Deepak

University of Mauritius, Reduit, Mauritius, [email protected], ORCID: 0000-0002-2758-9535

Gooroochurn Mahendra
University of Mauritius, Reduit, Mauritius, [email protected], ORCID: 0000-0003-0734-9144

Rughooputh, H.D. and Gooroochurn, M. Microcontroller-based, low cost image-guided solar

Cite this paper as:
tracking. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: In this paper, a cost-effective camera base solar tracking system was designed and
implemented. The processed image acquired by a wide-angle camera allows extraction
of accurate information about the position of the Sun both in terms of elevation and
azimuth. This system was implemented using off-the-shelf components in the aim to
reduce cost and is independent to its geographical location. Photographic films were
placed in front of the camera as a mean to eliminate unwanted artifacts, protect the
image sensor and to reduce the amount of processing to the acquired image. The results
obtained was satisfactory and this camera-based solar tracking helped simplified the
circuity of the system as well as cost. Only two DC linear actuators, a camera with a
wide-angle lens and a control unit were required which are readily available in the
market.

Keywords: Solar tracking, Image-guided, Image Processing and Analysis

© 2020 Published by ECRES

1. INTRODUCTION

With the beginning of the new millennium, our energy consumption has continued to increase mainly due to the
increase in population, transportations, growth of industries, unexplored renewable energy, and outdated
infrastructures. Exploitation of fossil fuels to meet the growing energy needs and unexplored renewable energy
sources has resulted in the depletion of non-renewable energy sources and an increase of the Earth’s temperature
(global warming) due to the emission of greenhouse gases into the atmosphere. As a result, solar energy has
attracted high interest, and given the variation of solar radiation at a given location over the course of a day and
over days of a year, solar tracking has been considered as a means to increase output from a solar PV installation.

A solar tracker is a device used to orient solar panels to face the Sun as closely as possible. Several studies have
been carried out to compare different types of tracking mechanism. Tracking can be done by moving only one
(single) axis of the solar panel (SAST) or by using a dual-axis solar tracker (DAST). Al-Rousan et al. [1]
proposed a broad classification of solar trackers as active and passive, and among the active solar tracking
systems, five categories were used: (1) sensor driver systems, (2) microprocessor driver systems, (3) open-closed
loop driver systems, (4) intelligent driver systems, and (5) a combination of two or more of these driver systems.
With the development of machine learning algorithms, intelligent driver systems are receiving increasing
attention. Al-Rousan et al. [2] used an adaptive neural fuzzy algorithm for controlling the solar tracker. The
method allows to track the position of the sun accurately, and thereby maximize energy output harnessed from

614
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the panels. It was concluded that five membership functions gave better results, which can be applied to both
single and dual-axis systems.

Lee et al. [3] used photo-resistor sensors along with a microcontroller to control two linear actuators and
compared the performance of a fixed solar system (horizontal plane) with a DAST in Malaysia. The results
obtained on an overcast day varied from 24.91% to 82.12% on a clear day. Abdallah et al. [4] showed that the
DAST produces an average of 41.34% more energy compared to a fixed panel (tilted 32 degrees for Jordan). In
this case, the tracker was independent of the weather condition.

Despite the proven benefit of solar tracking, its cost effectiveness has limited its application in installation.
Mohammad et al. [5] carried an investigation in Iran and found that the use of solar tracking leads to a
considerable reduction in the number of solar panels needed, and yet does not prove to be cost effective. The
vertical tracking system was shown to be most cost effective with a 23% improvement in output power whereas
a dual axis tracking system yielded a 32% improvement. Jamroen et al. [6] developed a low cost tracking system
based on digital logic design in a closed-loop mode using a Light Dependent Resistor (LDR) as input. They
compared the proposed tracking system against a fixed flat-plat system and determined an increased efficiency
of 44.89% on average. The price of electricity is a determinant factor in the cost assessment of the solar tracking
system and the payback period was estimated at 2 to 3 years for costs of $ 0.3/kWh to $0.2/kWh.

To deal with heterogeneous terrain and shading of solar panels, Fernandez-Ahumada et al. [7] proposed a novel
tracking technique, which implements a maximum irradiance algorithm in normal unshaded conditions. The
proposed method also allows to track solar systems for which no prior framework has been developed, e.g. non-
rectangular collectors. On the other hand, Zhu et al. [8] proposed a single-axis tracking structure for which
simulated results have shown ability to harness 96.4% of the annual solar radiation of a dual-axis system.

Computer vision approaches have also been considered to track the position of the Sun, and thereby position
solar panels. Jazi et al. [9] developed a computer tracking system based on image processing of a shadow bar.
The shadow produced in the presence of sunlight was processed on a computer to infer the Sun’s position.
Kadmiri et al. [10] on the other hand, used an omnidirectional computer vision that utilizes a catadioptric camera
and a convex mirror with a radius of 30 mm. Compared to a fixed panel with this DAST, the proposed system
yields an increase of 30% in power generated on average. Furthermore, it was also concluded that the camera-
based tracker was consistently producing more power than the LDRs based on the experiment performed. This
is mainly because the LDRs system uses the shading effect to determine the position of the Sun and therefore is
dependent on light intensity. In this case, when it is cloudy and not enough shading is produced, the LDRs
system stops working and if the Sun is out of the tracking field of the LDRs, the tracking process stops
completely.

This proposed research uses a double-axis system to develop a cost effective, low resolution, image-guided solar
tracking system, controlled by a micro-controller, which can be used on any system configuration, even for
unmodelled systems on heterogeneous surfaces. The simple algorithm allows accurate tracking of the sun under
different weather conditions, and based on previous research findings on the low sensitivity of loss in output
power with respect to accuracy of tracking the Sun [11], the proposed can be used to improve the cost
effectiveness and reliability of solar tracking. The study was conducted in Mauritius, which is a tropical island
located about 2000 km off the east coast of mainland Africa and about 855 km east of Madagascar. Mauritius
has two seasons: a warm humid summer (October to April) and a relatively cool dry winter (May to September),
where Ramgolam et al. [12] obtained an annual peak irradiance of 1152 𝑊/𝑚 2 for Mauritius, thus showing the
great potential solar energy represents for improving the energy mix.

In this paper, a cost-effective camera base solar tracking system was designed and implemented. The processed
image acquired by a wide-angle camera allows extraction of information about the position of the Sun both in

615
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

terms of elevation and azimuth. This system was implemented using off-the-shelf components in the aim to
reduce cost and is independent to its geographical location. Photographic films were placed in front of the camera
as a mean to eliminate unwanted artifacts, protect the image sensor and to reduce the amount of processing to
the acquired image.

2. SYSTEM DESIGN

In this work, a DAST for a 100 W solar panel was constructed based on the load requirement. The system was
designed and constructed to test the image acquisition, processing, analysis and control algorithm, as shown in
Figure 1. Two DC linear actuators with low speed (5 mm/s) were used to orient the panel and relays were used
to control the motion of the actuators. An off-the-shelf wide lens webcam was used to acquire images, which
was processed by a Raspberry Pi 3 Model B+ and an Arduino was used to control the actuators. The prototype
construction is shown in Figure 2.

Figure 1: Final Design of DAST system

Figure 2: Prototype of the DAST

616
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. SOFTWARE DESIGN

Photographic films were placed right in front of the lens of the camera to prevent specular reflection and other
unwanted artifacts. The images acquired by the camera and stored in the Raspberry Pi for further processing.
Based on the image analysis, appropriate signals were sent to the Arduino to control the actuators. This closed-
loop control system allowed to continuously capture images until the Sun’s centroid is in the middle of the frame
as shown in Figure 3.

Figure 3: Control System of the DAST System

The Raspberry Pi is also used to initiate start and stop of the tracking, which has been programmed to be between
8 a.m. and 5 p.m. At any point during the day, if the Sun is not visible, the solar panel will remain at its last
position until the Sun reappears and the tracking will continue due to the wide field of view of the camera.

Image Processing and Analysis

The open source computer vision library, namely OpenCV, was used for image processing. OpenCV library and
WiringPi library were installed on the Raspberry Pi before starting programming. The code was written in C
language with the WiringPi library.

The image processing sequence performed by the Raspberry Pi with camera equipped Photographic Films in
front of the lens is as follows (see Figure 4):

1. Frame capture from camera:

The raspberry will start recording the feed coming from the camera and each frame will be captured
for processing.

2. Segmentation of the frame (Grayscaling):

This step will convert the image from color to binary in order to reduce the processing power dealing
with a colored image. The image is converted from RGB (Red, Green, Blue) to HSV (Hue, Saturation,
Value).

3. Applying threshold value on the frame.

The threshold was set manually initially, and the chosen threshold was found to be reliable to
automatically segment the Sun from the background, due to the low detail image obtained with
filtering by the Photographic films.

4. Eroding and Dilating the frame:

This process was applied to remove speckles and isolate the Sun in the image. At this point, the Sun in
the image will gain a binary value of 1 and the dark part will become zero.

5. Determining the frame moment:

This will find the center of the pixel of value 1. In other words, the center of the Sun will be extracted
from the processed image.

6. Applying a centroid to the frame moment:

Once, the center is determined by the image moment, a centroid is then drawn which will be used to
take appropriate action on the actuators.

7. Applying an x-axis and y-axis onto the frame:

617
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The centroid thus determined is referenced to the arbitrary x and y axes, whose origin is the targeted
location for the Sun’s center, and hence the position for maximum irradiance for the panel.

Figure 4: Image Processing on Raspberry Pi

Control of the positioning of the solar panel

Image processing sequence above by the Raspberry Pi determines the location of the Sun. This information is
then sent to the Arduino, and the latter controls the actuators to aim to place the centroid at the origin of the
arbitrary coordinate frame.

Implementation and Results

The Neutral-Density filter (ND filter) optimum for the tracking was determined during the implementation -
enough to track the Sun even when light clouds were passing as shown in Figure 5 as revealed by the bright spot
in the image. As expected, it can be noticed that the light from the Sun is concentrated in the middle of the Sun
and the light being reflected from nearby clouds were reduced by the filter.

Figure 5: Examples of Processed Images with and without Clouds

The webcam was disassembled and the Photographic films were placed between two acrylic sheets and covered
with black tape to reduce unwanted light leaks and artifact by the added lens. An enclosure was made to make
the camera module water resistance as shown in Figure 6. The webcam was equipped with a mobile phone wide-
angle lens which provided a field of view of 120 o.

618
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 6: Camera Module

DISCUSSION & CONCLUSION

The accuracy of the system is dependent on the latency of the camera use and hardware use. As a result, a small
dead zone was implemented within the x-axis and y-axis of the image where the actuators will stop moving. The
band was adjusted on the image by offsetting the x and y-axis by a few pixels with respect to the axis. From
Figure 5, the centroid of the Sun is not exactly in the center but within the band. Without it, the actuators would
continuously have to adjust themselves due to the small lag in the whole system. Once the Sun moves out of the
band, the appropriate actuator will be actuated to re-center the Sun.

The program was successful in recovering the position of the Sun even when light clouds were passing by using
the light intensity of the clouds as shown in Figure 5. Solar tracking is notoriously known to consume energy
and thereby offset the energy savings. The energy consumption for the Raspberry Pi, the Arduino and the relays
were measured to be 4.4 W when executing the program. Since only the vision components will be used
throughout the day, the energy consumption of the solar tracking system is expected to be less than 4W.

The overall aim of the research was to devise an image-guided method, which uses low cost, low resolution off-
the-shelf camera to determine the position of the Sun which can be used to guide any solar tracking system in
uncertain terrain topography and different climate conditions. The results obtained were robust, and the use of
filtering using Photographic films was found to be a robust step to reduce image artifacts. Furthermore, this
method of tracking the Sun using a camera can also be implemented in other design, for example, to control a
large array of solar panels. The same information can be sent to several actuators with the same specifications
to orient the solar panel towards the Sun.

REFERENCES
[1] Nadia AL-Rousan, Nor Ashidi Mat Isa, Mohd Khairunaz Mat Desa, Advances in solar photovoltaic tracking systems:
A review, Renewable and Sustainable Energy Reviews, Volume 82, Part 3, 2018, Pages 2548-2569.
[2] Nadia AL-Rousan, Nor Ashidi Mat Isa, Mohd Khairunaz Mat Desa, Efficient single and dual axis solar tracking system
controllers based on adaptive neural fuzzy inference system, Journal of King Saud University - Engineering Sciences,
2020.
[3] Lee, J. F. & Nasrudin Abd.Rahim, 2013. Performance Comparison of Dual-Axis Solar Tracker vs Static Solar System
in Malaysia. Malayia, IEEE Conference on Clean Energy and Technology (CEAT).
[4] Salah Abdallah, Salem Nijmeh, Two axes sun tracking system with PLC control, Energy Conversion and Management,
Volume 45, Issues 11–12, 2004, Pages 1931-1939.
[5] Mohammad Amin Vaziri Rad, Ashkan Toopshekan, Parisa Rahdan, Alibakhsh Kasaeian, Omid Mahian, A
comprehensive study of techno-economic and environmental features of different solar tracking systems for residential
photovoltaic installations, Renewable and Sustainable Energy Reviews, Volume 129, 2020, 109923.
[6] Chaowanan Jamroen, Preecha Komkum, Sompol Kohsri, Wuttinan Himananto, Siriwat Panupintu, Sorawit Unkat, A
low-cost dual-axis solar tracking system based on digital logic design: Design and implementation, Sustainable Energy
Technologies and Assessments, Volume 37, 2020, 100618.

619
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[7] L.M. Fernández-Ahumada, J. Ramírez-Faz, R. López-Luque, M. Varo-Martínez, I.M. Moreno-García, F. Casares de la

Torre, A novel backtracking approach for two-axis solar PV tracking plants, Renewable Energy, Volume 145, 2020,
Pages 1214-1221.
[8] Yongqiang Zhu, Jiahao Liu, Xiaohua Yang, Design and performance analysis of a solar tracking system with a novel
single-axis tracking structure to maximize energy collection, Applied Energy, Volume 264, 2020, 114647.
[9] H. Arbab, B. Jazi, M. Rezagholizadeh, A computer tracking system of solar dish with two-axis degree freedoms based
on picture processing of bar shadow, Renewable Energy, Volume 34, Issue 4, 2009, Pages 1114-1118
[10] Kadmiri, Z. E., Kadmiri, O. E., Masmoudi, L. & Bargach, M. N., 2015. A Novel Solar Tracker Based on
Omnidirectional Computer Vision. Journal of Solar Energy. Volume 2015. Article ID 149852
[11] Wafa Batayneh, Ahmad Bataineh, Ibrahim Soliman, Saleh Abed Hafees, Investigation of a single-axis discrete solar
tracking system for reduced actuations and maximum energy collection, Automation in Construction, Volume 98, 2019,
Pages 102-109.
[12] Yatindra K. Ramgolam, K.M.S. Soyjaudah, Unveiling the solar resource potential for photovoltaic applications in
Mauritius,Renewable Energy, Volume 77, 2015, Pages 94-100.

620
 NUMERICAL INVESTIGATION OF NACA 2412 AIRFOIL USING
CFD TECHNIQUE

Köten Hasan
Istanbul Medeniyet University, Mechanical Engineering Department, Istanbul, 34700, Turkey, ORCID: 0000-0002-
1907-9420

Akkaya Kerim
CVS Air Co., Istanbul, 34700, Turkey, ORCID: 0000-0002-4521-6396

Koten, H. Akkaya K., CFD Validation of NACA 2412 Airfoil. 8th Eur. Conf. Ren. Energy Sys.
Cite this paper as:
24-25 August 2020, Istanbul, Turkey

Abstract: The NACA 2412 cambered airfoil experimental model of has been analyzed and validated to
determine the impact of aerodynamic performance at lower Reynolds number and constant
velocity. Validation has been carried out with steady state flow around NACA 2412 airfoil with
230 mm chord length at 30m/s, from 0° to 16° angle of attack (AOA) using commercially available
CFD software. The main aim of this study is to understand the flow behavior around the airfoil in
the flow separation region (or) stalling region solved under Navier Stroke equation. The
aerodynamic efficiency depends on the linearity of the shape and dimension of the object piercing
through air. C- type topology intended with 34.350 cells in un-structured manner, emerged with
non-slip far-field is employed. Fine grid reinforcement is given to the airfoil walls and surrounding
areas in-order to enclose boundary layer approach.
Keywords: Aerodynamic performance, airfoil, flow behavior, lower Reynolds number, validation.
© 2020 Published by ECRES

1. INTRODUCTION

Boundary layer flow separation control is one of the predominant techniques for achieving better L/D at higher
AOA for low speed and low Reynolds number driven aircraft. Flow over airfoil at transition region from laminar
to turbulent plays the vital role in determining the airfoil performance at certain AOA and velocity. The
efficiency of CFD simulation are hindered by the turbulence factor. Thus, turbulence modelling was developed
to determine the primary focus. The main classes of turbulence simulations are:
1) Reynolds Average Navier Strokes (RANS), 2) Large Eddy Simulations (LED), 3) Detached Eddy simulations
(DES).
In this study RANS turbulence model has been chosen to determine time-average fluid flow to determine
Reynolds stresses. The steady state governing equation is solved by coupling with any one of the following
RANS turbulence model 1) Spalart-Allmaras,
2) K–ε Realizable,
3) K–ω SST model, aiming to show comparison to the free field experimental measurements
for NACA 2412 airfoil.

2. METHODOLOGY

The theoretical methods of flow separation can be predicted through Karman’s integral
momentum equation as in Eqs 1-3.

(1)

621
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(2)

(3)

Where is momentum thickness, 𝛿 ∗ is the displacement thickness, and H is momentum shape.

The aerodynamic performance of individual RANS model has been investigated and validated in conjunction
with the performance outcome of Matsson [5] on NACA 2412 airfoil.

3. MODELLING AND SIMULATION SETTINGS

3.1. Computational domain

The domain is modelled based on the number of reference frame. As in the case of 2D airfoil, C-type 2 frame
domain with aft and forward far field enclosure is designed around the airfoil with dimension of 10C. Both
structural [8] and unstructured [9] meshing has been incorporated in order to eliminate the skewness on the wall
nearer to the model. The domain is then projected into 4 divisions in order to study the flow behavior over the
airfoil.
Fine and smooth quadrilateral mesh is given to the complete domain topology where the surface is face meshed
separately. In order to study the boundary layer separation the airfoil wall is meshed separately with very fine
refinement [9] by dividing the airfoil surface into 320 divisions. The walls are assumed to have non-slip
condition. The meshed view is shown in Fig.1.

Fig. 1. C- Topology fine mesh and no slip far field view

3.2. Boundary conditions

Turbulence were maintained at 1% to 5% [9] under RANS model coupled with Spalart - Allmaras model. In
order to match the validation, result to the experimental data the following boundary conditions are
implemented, where the air density is 1.225kg/m³, dynamic viscosity at 1.7894x10-5 kg/ms, gauge pressure at
0 Pa and operating pressure of 101325 Pa. Followed up with pressure velocity coupled with SIMPLE scheme
first order upwind, to generate the reasonable solution and the simulation is generated from inlet. These boundary
conditions have been implemented in all the three-turbulence model. The drastic change in flow separation for
16° AOA is shown in Fig. 2.

622
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig. 2.1. K- ω SST model flow separation behavior.

Fig. 2.2. K- ω SST model flow vectors.

4. RESULT AND DISCUSSIONS

4.1.Validation of NACA 2412 airfoil
In-order to validate the model, simulations were done on steady state incompressible flow on Reynolds
number 4.4 x 105 at various AOA ranging from 0° to 16°. Single turbulence model is used due to the limitation
of computational fluctuation in Navier Stroke equation. The Spalart - Allmaras turbulence model showed
linear performance to that of the literature experimental result, whereas the K–ε and K–ω models produced
inaccurate results as the AOA increases (Fig 3).

Fig. 3. Turbulence models error difference.

4.2.Turbulence model comparison

There is a minimal error observed between the article experimental result and validation numerical result, that
is mainly due to the high correspondence of wind tunnel outflow disturbance. The validation result of NACA
2412 airfoil for certain AOA to that of the experimental result is compared and shown in Fig. 4. The Spalart-
Allmaras model projects out the average minimal error compared to other turbulence model.

623
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fig. 4. Lift to AOA curves showing the variation between the turbulence models to the experimental model.

It is observed that the stalling point incidence angle increases the mean CL and CD, thereby shows greater
performance for numerical validation when compared to experimental results. Performance of the lift shows
good and constant hike up till 14° AOA, then falls by 22% creating a stalling point beyond that. From the
constant study, result shows increasing the AOA increases the wake and separates the boundary layer far apart
towards the aft section.

5. CONCLUSION

In this study, CFD simulation is used to predict the lift and drag by analyzing 2D external airflow. The validation
study was successfully carried out with RANS turbulence model coupled with Spalart-Allmaras model, which
represented better physical flow behavior to the experimental work than other models with least error of 6.04%.
The maximum error to that of the experimental work is recorded at 14° AOA of 19% due to its numerical stalling
point. After attaining the critical AOA (i.e) 14° , the flow over the airfoil get detached before 0.25C resulting in
a huge wake formation. The observed validated result on smooth airfoil will be used in the future studies on
modified airfoil surface, with an aim to generated better aerodynamic performance at higher AOA by neglecting
flow separation bubbles. Future work will be incorporated with the current addressed issues.

REFERENCES

[1] D. C. Eleni, T. I. Athanasios, and M. P. Dionissios, "Evaluation of the Turbulence Models for the Simulation of the Flow over a
National Advisory Committee for Aeronautics (NACA) 0012 Airfoil," Journal of Mechanical Engineering Research, vol.4, no. 3, pp.
100-111, March 2012.
[2] G. N. Vanderplaats and R. M. Hicks, "Numerical Airfoil Optimization Using a Reduced Number of Design Coordinates," NASA,
Moffett field, CaliforniaJuly 1976.
[3] B. Rajasai, R. Tej, and S. Srinath, "Aerodynamic Effects of Dimples on Aircraft Wing," in Intl. Conf. On Advances in Mechanical,
Aeronautical and Production TechniquesMAPT, 2015.
[4] H. Seetharam, E. Rodgers, and W. Wentz Jr, "Experimental Studies of Flow Separation of the NACA 2412 Airfoil at Low Speeds,"
presented at the ASEE Annual Conference & Exposition, Orleans, June 26 - 29, 1997.
[5] J. E. Matsson, J. A. Voth, C. A. McCain, and C. McGraw,"Aerodynamic Performance of the NACA 2412 Airfoil at Low Reynolds
Number," in 2016 ASEE Annual Conference & Exposition, 2016
[6] A. S. Al-Obaidi and Z. Pei Soh, "Numerical Analysis of the Shape of Dimple on the Aerodynamic Efficiency of NACA 0012 Airfoil,"
presented at the International Grand Challenges Engineering Research Conference (6th eureca), Kuala Lumpur, Malaysia, 2016.
[7] M. T. Islam, A. M. Arefin, M. Masud, and M. Mourshed, "The Effect of Reynolds Number on the Performance of a Modified NACA
2412 Airfoil," in AIP Conference Proceedings, 2018, vol. 1980, no. 1, p. 040015: AIP Publishing.
[8] L. Manni, T. Nishino, and P.-L. Delafin, "Numerical Study of Airfoil Stall Cells Using a Very Wide Computational Domain,"
Computers & Fluids, vol. 140, pp. 260-269, 30 september 2016.
[9] A. Lopes, "A 2D Software System for Expedite Analysis of CFD Problems in Complex Geometries," Computer Applications in
Engineering Education, vol. 24, no. 1, pp. 27- 38, 2016.
[10] Merryisha, Samuel, and Parvathy Rajendran. "CFD Validation of NACA 2412 Airfoil." (2019).
[11] Merryisha, S., & Rajendran, P. (2019). CFD Validation of NACA 2412 Airfoil.
[12] MERRYISHA, Samuel; RAJENDRAN, Parvathy. CFD Validation of NACA 2412 Airfoil. 2019.

624
 FIRE MODELLING AND SMOKE EXTRACTION IN THE
ENCLOSED CAR PARK ZONE USING CFD TECHNIQUE

Koten Hasan
Istanbul Medeniyet University, Mechanical Engineering Department, Istanbul, 34700, Turkey, ORCID: 0000-0002-
1907-9420

Akkaya Kerim
CVS Air Co., Istanbul, 34700, Turkey, ORCID: 0000-0002-4521-6396

Koten, H. Akkaya K., Fire Modelling and Smoke Extraction in The Enclosed Car Park Zone
Cite this paper as:
using CFD Technique. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The most important way to obstruct the smoke, during a fire and after the fire has been
extinguished, is by using the method of smoke extraction. In this study intended to reduce density
of smoke, to limit temperature according to standards, and to assist escape points. The second
considering is to limit the smoke movement in order to provide clear way for fire fighters to
intervention the fire. In this study, fire, smoke and ventilation system of the enclosed car park zone
are examined, according to the BS 7346-7 and ASHRAE Standard by using computational fluid
dynamics (CFD) technique. In addition, jet fan locations, the equipments are located in the car
park and their properties, defining mesh model, showing the smoke diffusion, visibility,
temperature gradient, velocity gradient on the 1.7 meter, which is accepted for the standard of a
person’s tall and comparing them with the standard are studied. With the research, we see that the
results can fulfill the standard.
Keywords: Fire, smoke, CFD, car park.
© 2020 Published by ECRES

1. INTRODUCTION

Car fires in enclosed car parks are an very important issue. Other than firefighting systems (like sprinkler
system), a proper ventilation system is required to extract the smoke and reduce temperature. The car park fires
affect humans’ health, finances and the environment. The fire smoke can flow rapidly through the enclosed area
and this issue means the dangers of fires in enclosed car parks. Fire may be difficult to fight because smoke may
fill up the whole car park and obstruct visibility at human level. The generally cause of death of humans is the
negative effect of smoke, not by injuries. The most important way to obstruct the smoke, during a fire and after
the fire has been extinguished, is by using the method of smoke extraction. In this study intended to reduce
density of smoke, to limit temperature according to standards, and to assist escape points. The second
considering is to limit the smoke movement in order to provide clear way for fire fighters to intervention the fire
[1-8].
In this study, fire, smoke and ventilation system of the enclosed car park zone are examined, according to the
BS 7346-7 and ASHRAE Standard by using computational fluid dynamics (CFD) technique. In addition, jet fan
locations, the equipments are located in the car park and their properties, defining mesh model, showing the
smoke diffusion, visibility, temperature gradient, velocity gradient on the 1.7 meter, which is accepted for the
standard of a persons tall and comparing them with the standard are studied. With the research, we see that the
results can fulfill the standard.

625
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2. METHODOLOGY

In this study, Computational Fluid Dynamics(CFD) is a numerical calculation method using computer based
simulation tool for the analysis of complex systems. The systems that could be calculated using CFD include
fluid flow, particles dispersion and heat/mass transfer. CFD is a very important and special tool for assisting in
the ventilation system design and optimization of systems. It could be used to model and simulate smoke
diffusion, visibility, airflow, air quality and temperature within a space for case of fire. In this study, car park
has been modeled and simulated using the a CFD software Pyrosim Fire Dynamic Simulator software [9-12].

A pressure-based solver has been defined for steady-state model the car park. The volumetric transport model
was accepted to describe the fire from cars with the inclusion of diffusion energy source options. Large Eddy
Simulation (LES) turbulence model has been used as a turbulence model.

2.1. CAR PARK LAYOUT

The car park geometry model has been designed as three-dimensional according to boundary conditions. Closed
spaces such as columns and rooms were modeled as closed volumes because they would have a negative effect
on flow. Vehicle entrances and exits are defined as atmospheric openings. Figure 1 shows a three-dimensional
car park design that is used CFD simulations.

Figure 1: 3D model of car park

2.2. BOUNDARY CONDITIONS

Once the closed parking geometry has been completely designed in CFD software, the model is ready to apply
the simulation input data. The parking environment is 20°C and atmospheric pressure. In a CFD analysis
process, boundary conditions are important parameters and they should define to model. There are two
important equipments in car parks:

1. Axial fans
2. Jet fans

626
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b) (c)

Figure 2: Car Park Ventilation Equipments: (a)Axial Fan, (b)Axial Jet Fan, (c)Radial Jet Fan

For a car park CFD analysis, some of important boundary conditions parameters are axial fans and jet fans
flow rate values, locations of fresh air and exhaust shafts, jet fans locations and fire point. Table 1 shows the
axial fans flow rate values. Also Table 2 shows the jet fans properties.

Table 1: Axial Fans Properties

Type of Axial Fan Flow Rate Values (m³/h) Quantity
Smoke Exhaust Fan 46.500 2
Fresh Air Supply Fan 32.500 2

Table 2: Jet Fans Properties

Type of Jet Fan Flow Rate Values (m³/h) Thrust Values (N) Quantity
Axial Jet Fan 5400 50 3
Radial Jet Fan 6690 60 4

An important issue for car park CFD analysis is the heat release rate (HRR). The amount of energy released
during a vehicle fire varies with time. Table 3 shows the heat release rates for different parking conditions
according to BS-7346-7.

Table 3: Heat Release Rate(HRR) Values

Indoor car park without Indoor car park with 2 car stacker with
Fire parameters
sprinkler system sprinkler system sprinklers
Dimensions Perimeter 5 m x 5 m x 20 m 2 m x 5 m x 14 m 2 m x 5 m x 14 m
Heat release rate 8 MW 4 MW 6 MW

In this study, there is a sprinkler system in the parking lot. For this reason, car fire heat release rate can be
defined as 4MW to CFD model according to BS-7346-7 standard. Figure 3 shows the CFD model after the
boundary condition definition. The design flame symbolizes a burning car for unknown reasons. This 4MW
fire will reach its peak 20 minutes (1200 seconds) after the fire starts and will be extinguished after 70 minutes
(4200 seconds). The proposed fire (heat dissipation rate) in the underground garage is selected from the
literature. As obtained from literature, car fires are considered rapid fire. In addition, a fast fire grows as
shown in the diagram Figure 3.

627
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3: HRR-Time Graphic

Figure 4: CFD Model with Boundary Conditions

2.3. MESH MODELLING

The area of the simulation model (car park) is divided into several small cells. The method used by the CFD
and FDS software is to calculate air temperature, velocity, optical density, smoke and fire emission for each
small cell. Therefore, it is important to define the size of these cells to achieve true-like results. Furthermore,
the net term encompassing all these small cells is 'Mesh'.
An increase in the number of cells results in a car park with smaller cells. Although smaller cells provide more
detail for simulation, they require more computational time. The main objective at this point is to find the
optimal balance for cell sizes. As is seen from Figure 5 mesh sizes are 0,4000m x 0,4026m x 0,5000m and we
have 1.008.432 numerical cells in this study.

628
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 5: Mesh Modelling of CFD StudieS

3. RESULTS

Smoke diffusion in the car park, temperature gradient, visibility, carbon-monoxide mass fraction and air
velocity gradient has calculated with finite elements software. We can see and compare the results. We are
aiming to obtain these answers;

1. Does the jet fans are located in the correct places?

2. Is there any dead space in the car park? (The areas which has 0 m/s air movement) (BS 7346-7)
3. Is the temperature under the specified temperature in the standard? (BS 7346-7)
4. Is the visibility enough? (BS 7346-7)
5. Is the air velocity in the car park under the acceptable value? (5 m/s is the limit)(BS 7346-7)
6. Is the Carbon-Monoxide level under the acceptable level? (ASHRAE- Principles of Smoke Management)

3.1. SMOKE DIFFUSION

(a) (b)
Figure 6: Smoke Diffusion Results: (a)0. Second, (b)1500. Second

3.2. VISIBILITY

(a) (b)
Figure 7: Visibility Results: (a)0. Second, (b)1500. Second

629
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3.3. CARBON-MONOXIDE DIFFUSION

(a) (b)
Figure 10: Carbon-Monoxide Diffusion Results: (a) 0. Second, (b)1500. Second

4. CONCLUSION

In this study, smoke discharging works successfully employed. Visibility at the escape ways is at least 10
meters which is satisfying the BS-7346-7 standard. Temperature around emergency exits is within the
boundaries of BS-7346-7 Standard. Maximum air velocity is at most 3 m/s and it is acceptable for BS-7346-7
Standard. There is no dead space in the car park. Carbon-monoxide mass fraction is under the limit
determinated in the ASHRAE Standard.

REFERANCES

[1] BS 7346-7:2013 Components for smoke and heat control systems – Part 7: Code of practice on functional
recommendations and calculation methods for smoke and heat control systems for covered car parks.
[2] NFPA 92B:2009 Standard for Smoke Management Systems in Malls, Atria, and Large Spaces
[3] 1999 ASHRAE Handbook for HVAC Applications, Chapter 12.
[4] ANSI/ASHRAE Standard 62-1989, Ventilation for Acceptable Indoor Air Quality
[5] ASHRAE. 1992. ANSI/ASHRAE Standard 55�1992, Thermal Environmental Conditions for Human Occupancy,
Atlanta: American Society of Heating, Refrigerating, and Airconditioning Engineers, Inc.
[6] Fanger, P O. 1970. Thermal Comfort. Danish Technical Press.
[7] Brundrett, G W. 1990. Criteria for Moisture Control. Butterworths.
[8] Hanzawa, H, Melikov, A K, and Fanger P O. 1987. Air flow characteristics in the occupied zone of ventilated spaces.
ASHRAE Transactions. Vol. 93 (1), pp 1020.
[9] Kimura, K and Tanabe, S. 1993. Recommended air velocity against combinations of temperature and humidity for
sedentary occupants in summer clothing, Proceedings of the 6th International Conference on Indoor Air Quality and
Climate - Indoor Air '93, Vol 6, pp 6166.
[10] Raw, G J, Aizlewood, C E, Llewellyn, J, et al. 1999. Indoor air quality and health in 10 office buildings in the UK - a
multi-disciplinary study. BRE Report 9911332.
[11] Sexton, K and Ryan, P B. 1988. Assessment of human exposure to air pollution: methods, measurements and models.
In Air Pollution, the Automobile and Public Health, A Y Watson, R R Bates and D Kennedy, eds. Washington DC:
National Academy Press.
[12] British Standards Institute. "Components for Smoke and Heat Control Systems – Part 7: Code of Practice on
Functional Recommendations and Calculation Methods for Smoke and Heat Control Systems for Covered Car Parks."
London, British Standards Institute, 2006.

630
 DEVELOPMENT OF HEMP SHIVE WALLBOARD WITH
INTEGRATED PHASE CHANGE MATERIAL

Edgars Kirilovs
RTU, Faculty of Material Science and Applied Chemistry, Institute of Design Technologies, Riga, Latvia,
[email protected], ORCID: 0000-0002-9133-9514

Gusovius Hans-Jörg
The Leibniz Institute for Agricultural Engineering and Bioeconomy, Potsdam, Germany,
[email protected], ORCID: 0000-0002-8563-1391

Inga Zotova
RTU, Faculty of Material Science and Applied Chemistry, Institute of Design Technologies, Riga, Latvia,
[email protected], ORCID: 0000-0001-8339-0339

Silvija Kukle
RTU, Faculty of Material Science and Applied Chemistry, Institute of Design Technologies, Riga, Latvia,
[email protected], ORCID: 0000-0001-7856-7179

Kirilovs, FN, Gusovius, FN. Development of hemp shive based wallboard with integrated
Cite this paper as:
phase change material. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Over the past few decades, climate change and the search for renewable energy sources have
become hot topics within the research community. About 30% of the world's energy consumption
is in the heating and cooling sector of residential buildings. Such materials can reduce the
temperature variations, leading to an improvement in human comfort and decreasing at the same
time the energy consumption of buildings, due to their capability to absorb and release energy
from/in the environment.
This paper assesses the integration of a Phase Change Materials Microcapsule concentrated
dispersion (PCMs) in hemp shive intended for building indoor wallboard. The key characteristic
of PCMs is their capacity to absorb energy and restore it. Our research focuses of Phase Change
Materials as a type of Smart Materials. The interior wallboard was developed with hemp shive
board integrated with the commercial bio-based MikroCaps PCM-S50 with a 23-28°C melting
point. In RTU Institute of design technology experimental laboratory 400x400 mm size board
samples are produced and the tests are performed to determine such parameters as thermal
conductivity, specific heat capacity, phase transmission heat.
Keywords: Phase Change Materials; Hemp shive; specific heat capacity; phase transmition heat;
© 2020 Published by ECRES

1. INTRODUCTION

High interest for energy consumption has occurred, therefore the technologies keep developing and the range of
useful solutions to improve the level of comfort increases. European Union and the whole world is now facing
climate change caused by the emissions of greenhouse gases [1]. According to the action plan developed by
United Nations "Agenda 21" regarding sustainable development, environmentally friendly technologies are
considered those that protect and reduce the adverse environmental impact throughout the material life cycle-
from the extraction of raw materials and ending with the product / material further use or recycling. The building
sector is responsible for approximately two-third of halocarbon and approximately 25-30% of black carbon
emissions [2]. Renewable raw materials, including natural fiber sources, are the future of storage resources with
a variety of positive effects on both the planet ecosystem and the living and working environment, and the energy
consumption of delivering the required functionality. Natural fibres are selected to be used in ecological

631
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

buildings where the choice of materials is based on such parameters as recyclable, renewable raw materials and
low-resource production techniques. In addition, mainly cellulosic materials are recommended for old timbered
houses as cellulosic insulations have higher moisture regain compared to inorganic materials [3,4]. Beside the
traditional way to separate fibre crop stems and use its structural main components fibres and shives the whole
crop utilization of flax and hemp became of interest since few years. The described supply chain for wet
preserved hemp raw materials is based on long term research activities at Leibniz Institute of Agricultural
Engineering and Bioeconomy (ATB) in Potsdam (Germany) [5,6]. There are done studies to analyze different
natural resources usability in board materials. [7,8,9] One of very important targets is to get good parameters for
thermal conductivity where to reach high utility there are used innovative technologies- Nano-encapsulated
phase-change material (PCMs). Phase change materials are recognized as an effective way to improve building’s
energy management. Such materials are able to store a large amount of energy due to a transition of phase that
is most frequently the solid/liquid one. This can enhance the building thermal mass and thus leads to energy
efficiency. Studies about phase-change material (PCM) as latent thermal energy storage have been made for
more than 50 years, but the low energy prices and high raw material costs did not motivate to start the product
design. Currently the situation has significantly changed since the decreasing PCM costs and increasing energy
prices, their use as latent thermal energy storage for the indoor microclimate self-regulation depending on the
environmental temperature [10,11]. The PCM can be incorporated in construction materials and elements by
direction corporation, immersion, encapsulation, shape-stabilization and form stable composite PCMs [12,13].

2. MATERIALS AND METHODS

Materials
For the production of wall board material has chosen three different components due to following criteria:
•Functional concordance;
•Impact on the environment (extraction, manufacturing, use);
•Local availability.
Hemp (Cannabis sativa L.) was grown at the experimental plot of Agriculture Science Center of Latgale (ASCL)
in 2014. Varietie Polish Bialobrzeskie (owner: Institute of Natural Fibres & Medicinal Plants, Poznan) has been
chosen for experimental part. Industrial hemp variety Bialobrzeskie is a very promising, economically
perspective cultivated plant, which permits producing high harvests of hemp stems in the Latvian soil and
climatic conditions too and it is also a regenerating plant to the environment. For raw material preparation was
used traditional preparation technology. Among other European locations they have been cultivated under
agricultural conditions as well in Vilani, Latvia (56°33′09″N 26°55′29″E). The harvest was carried out August
2014 by means of a forage harvester. The phase change material microcapsule (PCMs) MikroCaps Ltd.
(Slovenia) products have been selected and incorporated into the insulation wall board material prototypes.
According to MikroCaps PCM-S50 has a melting temperature of 23 -28°C, a latent heat of 160-180 J/g, pH 6,0
- 9,0, density 900-970 g/L, viscosity (at 25°C): 100-1.000 cPs, appearance off white a concentrated aqueous
dispersion of formaldehyde free slurry and average microcapsules particles size: 1-15 μm, type of membrane
polyurethane (formaldehyde-free), type of PCM Biobased. As adhesive for the material prototype used
commercial Kleiberit 862.0 Urea Formaldehyde Resins (UF) with hardener for cold pressing.

Cold pressing technology

Using cold pressing (matrix and punch) 400x400 mm samples are made. These linear dimensions are selected
taking into account the planned test methods and the necessary samples configurations for the tests. The
prototypes of the material are made on the basis of layers, accordingly each filler – hemp shive the binder
separately. Resin powder with hardener and water are dosed at the ratio of 2:1, according to the technical data
sheet of the producer. Dosing cups are used to determine the binder components. After mixing the components
stirring is carried out for at least two minutes to ensure smooth addition of the hardener to epoxy resin. When
the required homogenous mixture of binder is obtained the Phase Change Material microcapsule concentrated
dispersion is added in the amount of 5% of the starting material dry weight. The stirring of the components is

being done for 300 sec to obtain homogenous consistency. The binder is added gradually using a spray gun for
shives to obtain homogenous mixture and stirring is performed with an electric mixer over a period of 300 sec.
Both hemp shives and binder mixture is formed in the 400x400 mm template. This is done in two steps: during
the first step a part of the mixture is pressed to the thickness of 50 mm, according to the required thickness, and

632
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

during the second step the second part of the mixture is pressed and the material is compressed to the required
thickness of 25 mm. Matrix and punch form is immersed in a hydraulic press and pressed for 4 hours in 22±2°C
under 10 bar pressure, taking into consideration the polymerization time of the epoxy resin and hardener.

Figure 1. Preparation of the sample – schematic drawing

The efficiency of insulation materials is described by thermal conductivity λ (W/(m·K)). One approach for the
measurement of this parameter in steady-state conditions is the guarded hot plate method [14]. The NETZSCH
HFM 446 Lambda Small was used to determine the parameters. The measurement principle is simple: a
specimen is placed between two surfaces that are kept at constant temperatures, and the heat flow passing
through the plates is registered after stationary conditions are reached; a so-called ‘compensations zone’ with
the same temperatures surrounds the flow meters to avoid heat losses in other directions. Dependence of thermal
conductivity on mean temperature can be estimated by conducting the measurements with different surface
temperatures [15].

3. RESULTS AND DISCUSSION

Distribution of particle masses

The seven sieves used for sieving enabled us to determine the distribution of the particle masses according to
the dimensions of the sieve’s eye (Figure. 2). The largest proportion is fractions exceeding 3150μm - 57% of the
total amount (Figure 3). Significant proportion is also in sizes 500μm to 1000μm- 12% of total. The smallest
amount is for fractions with sizes less than 500μm, and those less than 150μm are just 3.5% from total. From
the visual observation it can be concluded that particles in a fraction above 3150μm are uneven by their structure
and size. There is a tendency for the particles to join (stick) when removing them from the sieve. The properties
of this fluttering material flow may interfere the mixing process by prolonging the stirring time until a smooth
mixture is obtained [16].

60
% of total weight

50
40
30
20
10
0

Particle size >3150µm >90µm

Figure 2. Distribution of particle masses (Bialobrzeskie shive) Figure 3. Hemp shive particle size
Thermal conductivity
Thermal conductivity of the board samples changed within the range from 0.064 to 0.074 W/m.K. In the sample
with added PCM, the thermal conductivity of the nanocapsule is 7% worse than that of a cannabis-free hemp
slab, 7% better than that of a preserved hemp-shive mixiture board. In the tests performed, it can be observed
that if the density is 8% lower, then the coefficient of heat conduction is 7% better. To put it simpler, in this case

633
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the increase in conductivity is determined by the decrease in the porosity of the insulating material. An example
of temperature dependence is shown on Figure 5, where line slope for all the materials is practically the same.
Thermal conductivity for similar unconventional natural raw materials like flax, reed, wood fibre, cement wood
wool in range 0.6…1.0 W/(m·K) [17,18], which is very close to the measured values.

Figure 4. Thermal conductivity of experimental and reference boards [17,18]

0.073
Thermal conductivity

0.071
(W/m/K)

0.069
y = 0,0002x + 0,0652
0.067

0.065
10 15 20 25 30
Sample mean temperature (oC)

Figure 5. An example of thermal conductivity dependence on mean temperature (Hemp shive with PCM)

Specific heat capacity

Another important performance indicator is high specific heat capacity. This property can be beneficial in
buffering indoor temperature – during the summer it can reduce the need for cooling. As seen in the Figure 5
the specific heat of the materials varies from 1444 – 2369 J/kg·K. According to the data analysis it can be
concluded that the thermal conductivity of the material samples is mainly influenced by the quantity of the
components of the material. The highest value of material-specific heat is for hemp shive with integrated
encapsulated phase change material and this value is 63% higher than material without phase change material.
Specific heat capacity for similar unconventional natural raw materials like flax, wood fibre, cement wood wool
in range 1280...2100 J/kg·K [17,18], which is very close to the measured values. Hemp shive sample has 12%
higher specific heat capacity than commercial material with a higher value.

634
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 5. Thermal conductivity of experimental and reference boards[17,18]

The heat capacity at 35 ° C on a NETZSCH equipment is determined by measuring the amount of heat that must
be applied to a sample of material to raise its temperature from 30 to 40 ° C, i.e. by 10 ° C. It is further used to
determine the heat of phase transition. Applying this average calorific value to the entire 10 ° C range, we obtain
a calorific value of 15.5 [J / g]. Since the heat capacity changes little and practically linearly with temperature,
such assumption is appropriate. There is also a phase transition between 20 ° C and 30 ° C, so the amount of
heat supplied in this temperature range is 62% higher than the heat capacity of the material alone. The measured
heat input between 20 ° C and 30 ° C is 2.37 [J / (g · K)]. The difference between the amounts of heat determined
by interpolation and those measured in this interval thus arises due to the change of phases. This difference is
equal to 8200 [J / kg]. Given that 5% PCMs are incorporated in the material sample, the phase change heat of
the material is equal to 163 J / kg only for PCM.
Specific heat capacity

2500

2000
(J/g*K)

1500

1000

500

0
0 10 20 30 40
Sample mean temperature (oC)

Figure 6. Specific heat capacity of experimental boards (Hemp shive +PCM)

4. CONCLUSION

Thermal conductivity is one of the most important parameters of any insulating material. The thermal
conductivity of the experimental samples is the range of 0.064- 0.074 W/(m∙K), that fits in commercially
available hemp insulation product class thats thermal conductivity is in the range of 0.066-0.074 W/(m∙K) at the
density limit rom 315 – 490 kg / m3 – i.e. falls within the thermal conductivity ranges of all generic groups of
the insulation and is comparable with that of other fibrous insulation materials.
As the material tests show, the calorific value of the boards, in which 5% encapsulate phase change material is
mixed in the mass during the manufacturing process, has increased by 62%, reaching 2369J / (g · K). This

635
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

property can be beneficial in buffering indoor temperature – during the summer it can reduce the need for
cooling.
In the experimental samples there was also a phase temperature transition at 25 ° C, which is within the PCM
melting area of 23 - 28 ° C specified by the manufacturer.

ACKNOWLEDGMENT

This work has been supported by the European Regional Development Fund within the Activity 1.1.1.2 “Post-
doctoral Research Aid” of the Specific Aid Objective 1.1.1 “To increase the research and innovative capacity of
scientific institutions of Latvia and the ability to attract external financing, investing in human resources and
infrastructure” of the Operational Programme “Growth and Employment” (No. 1.1.1.2/VIAA/1/16/152)

REFERENCES

[1] Pacheco-Torgal F., Eco-efficient construction and building materials research under the EU Framework Programme
Horizon 2020. Construction and Building Materials 2014, pp. 51-162.
[2] Wazna M. E., Gounni A., Bouari A.E., Alami M.E., Cherkaoui O. Development, characterization and thermal
performance of insulating nonwoven fabrics made from textile waste. Journal of Industrial Textiles, vol. 48, 2018.
pp. 1167-1183.
[3] Murphy DPL, Behring H, Wieland H. The use of flax and hemp materials for insulating. Proceedings of flax and other
bast plants symposium 1997. pp. 79–84.
[4] Kymalainen H. R., Sjoberg A.M. Flax and hemp fibres as raw materials for thermal insulations. Building and
Environment Vol. 43, 2008. pp. 1261–1269. DOI: 10.1016/j.buildenv.2007.03.006
[5] Lühr C., Pecenka R., Budde J., Hoffmann T., Gusovius H-J. Comparative investigations of fiberboards resulting from
selected hemp varieties. Industrial Crops & Products 2018, Vol. 118, pp. 81- 94. DOI: 10.1016/j.indcrop.2018.03.031
[6] Pecenka R., Fūrll C., Idler C., Grundmann P., Radosavljevic L. Fibre boards and composites from wet preserved hemp.
Materials and Product Technology 2009, Vol 36, pp. 208-220.
[7] Ivanovs S., Rucins A., Valainis O., Belakova D., Kirilovs E., Vidzickis R. Research of technological process of hemp
slab production, Engineering for rural development 2015, pp 202-209.
[8] Rofie S. Novel Low Density Particleboard from Hemp Shives, Doctoral Thesis. University of Wales, 2005.
[9] Lee S., Shupe T.F., Hse C.Y. Mechanical and physical properties of agro-based fiberboard. European Journal of Wood
and Wood Products 2006. pp. 74-79. ISSN 0018-3768.
[10] Zalba B., Marin J., Cabeza L.F., Mehling H. Review on thermal energy storage with phase change: materials, heat
transfer analysis and applications. Applied Thermal Engineering 2003, Vol. 23. pp. 251–283.
[11] Subramanian E. Integrating phase change materials in building materials: Experimentation, characterization and
numerical simulation, Doctoral Thesis. Clemson University, 2011.
[12] Rao Z., Wang S., Zhang Z. Energy saving latent heat storage and environmental friendly humidity-controlled materials
for indoor climate. Renewable and Sustainable Energy Reviews 2012, Vol. 16. pp. 3136– 3145. DOI:
10.1016/j.rser.2012.01.053
[13] Memon S. Phase change material sintegrated in building walls: A state of the art review. Renewable and Sustainable
Energy Reviews 2014, Vol 31. pp. 870–906. DOI: 10.1016/j.rser.2013.12.042
[14] LVS ISO 8302: 2001. Thermal insulation - Determination of steady-state thermal resistance and related properties –
Guarded Hot Plate apparatus.
[15] Gendelis S., Jakovičs A., Engelhardt M., Thermal and moisture adsorption/desorption properties for a selection of
vegetal insulation materials. 4th Central European Symposium on Building Physics, MATEC Web Conference 2019.
pp. 1-5.
[16] LVS EN 933-1:2012. Tests for geometrical properties of aggregates - Part 1: Determination of particle size distribution
- Sieving method.
[17] Kozlowski R., Wladyka-Przybylak M. Uses of natural fiber reinforced plastics. In: Wallenberger F.T and Weston N.
Natural fibres, plastics and composites. Springler, 2004. pp. 249-262.
[18] Pfundstein M., Gellert R., Spitzner M., Rudolphi A. Insulating Materials: Principles, Materials, Applications.
Regensburg: Auműller Druck, 2008.

636
 PERFORMANCE EVALUATION OF POWER CONVERTERS &
ARTIFICIAL INTELLIGENCE BASED MPPT TECHNIQUES IN
GRID-TIED PV SYSTEMS

Vidhya K Viswambaran
University of Bolton, Ras Al Khaimah, United Arab Emirates, [email protected], ORCID:
https://orcid.org/0000-0002-4341-9843

Akram Bati
University of Bolton, Greater Manchester, United Kingdom, [email protected], ORCID: https://orcid.org/0000-
0001-8548-0418

Erping Zhou
University of Bolton, Greater Manchester, United Kingdom, [email protected], ORCID: https://orcid.org/0000-
0002-0568-294X

Viswambaran, Vidhya, Bati, Akram Zhou, Erping. Performance Evaluation of Power

Cite this paper as: Converters & Artificial Intelligence based MPPT techniques in grid-tied PV systems. 8th
Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: With steadily increasing global energy consumption, renewable energy sources are gaining more
attraction. Various power converters and control strategies are employed in renewable energy
systems to improve efficiency. The generated power output from PV panels are nonlinear and
are to control. In order to increase the utilization effectively various power converters utilized
in a PV system and MPPT techniques are reviewed in this paper. This paper review various
power converter used in PV systems and performance evaluation of AI-based MPPT techniques
for grid-tied PV systems. Artificial Neural network (ANN) based soft computing tools will be
utilized for approximating the nonlinear dynamic nature of PV panels. Data set for training
feedforward Neural Networks are generated by conventional hill-climbing Perturb & Observed
based MPPT algorithm simulated using MATLAB. ANN based MPPT controller is evaluated
using the MATLAB Neural Network Tool Box. The results indicate that the ANN MPPT
controllers are efficient in tracking MPP during fluctuating environmental conditions. Finally
ANN MPPT controller is interfaced with a grid tied PV system and performance is evaluated.

Keywords: Photovoltaic (PV), Control strategies, Maximum power point techniques (MPPT), DC-DC
converter Artificial Intelligence (AI).

© 2020 Published by ECRES

I. INTRODUCTION

Sustainable energy is an essential requirement of mankind due to rising world population. Environmental impact
of conventional energy sources has boosted the demand for clean energy [1]. Economic and environmental
benefits of Renewable energy technologies are promising. The power generated by Photovoltaic (PV) systems
is clean, noise-free and eco-friendly [2]. The DC power generated PV system is being utilized for residential
and industrial purpose. The power generated by PV varies with varying climate. Power electronic converters are

637
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

crucial in integrating PV power output into the utility grid. Power converters regulate the panel output and
increase the efficiency of PV systems [3]. Maximum power point should be tracked (MPPT) continuously for
impedance matching due to varying climate. The paper will review various power converters topologies utilized
in the PV system. This paper will also analyze the performance of an AI optimized MPPT controller will be
designed to track the MPP of a 215W grid-tied PV system.

II. THEORY

A Photovoltaic system mainly involves PV panels, power converters and electronic controllers for the
conversion of solar radiations into DC electricity with high efficiency. Parallel connected PV systems are more
efficient than the series-connected in case of partial shading conditions [4]. The varying climatic conditions and
shadowing affect the PV power output. Various power converters are utilized in the PV system for maximum
power transfer between the PV panel and load [5]. Efficient power electronic technologies can grant efficiency
improvements in renewable energy evolution. To address this issue charge controllers with MPP tracking facility
is utilized. An MPPT controller works by regulating the DC-DC converter output power despite the fluctuation
in the input and the load [6]. In case of off-grid PV systems, these PV arrays charge the rechargeable batteries
for powering various DC appliances [7].

Figure 1. PV System

Selection of DC-DC converters:

Choosing an appropriate converter will optimize the effectiveness of the Power system [8]. DC-DC converters
controls can be current control or direct duty-cycle control [9].The boost converter performs adequately for PV
arrays in a parallel configuration. The boost converter has improved dynamic performance and cost effectiveness
[10]. However, at conventional boost converter suffer from large input current ripples. The performance of boost
converters can be improved by modified using voltage multipliers, cascading multiple boost converters and
inter-leaved boost converters [12].In isolated converter type DC-DC converters isolation transformers are
employed to provide isolation from the grid [13]. Tracking region of SEPIC, CUK, buck-boost and zeta
converter covers all the quadrants over the VI curve. Whereas the tracking region of the boost converter is
limited to the first quadrant and due to this tracking MPP fails under varying irradiance, temperature and load
[14]. SEPIC, Zeta and Boost converters output polarity is same as input where Buck-boost converter provides
inverted output [15].

DC-AC Converter in PV Applications:

DC-AC converters or inverters are employed in a PV system to transform the DC to AC. PV panels connected
to a centralized DC-AC converter will be cost effective because of the presence of only one inverter. However,
power losses can be high. In string inverters, a string is connected to an inverter. Losses associated with the
module mismatches and partial shading conditions are eliminated and MPP tracker can be employed on
individual strings. This is useful when different strings are placed in different orientations. In micro inverters
modules, each module is connected to a separate inverter. This configuration is more reliable as it avoids single-
point failure of the system [16].

638
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

METHODOLOGY

A grid-tied PV system will be designed and modeled using Simulink. Toolbox in MATLAB. A 48 V ,200Watts
boost converter will be designed .An ANN-based MPPT controller will be modeled using Neural network
Toolbox in MATLAB and it will be employed to track the MPP under varying climatic conditions. The regulated
DC output will be fed to a single-phase inverter for converting to AC voltage for Grid. A Phase-locked loop
(PLL) will be employed for synchronization of grid with inverter .A PID controller will ensure voltage and
current are controlled to match the grid requirements.

Inverter. The proposed system will be simulated using Simulink Toolbox of MATLAB software. The proposed
model is demonstrated in Figure 2.

Figure 2.Proposed System

PV Array parameter estimation:

1Soltech 1STH- 215-P, 215 Watt PV panel will be used as the source.PV panel specification is given in Table
1. The PV and IV characteristics plotted under varying temperature (20◦C to 30◦C) conditions are shown in
Figure 3. Table 2.gives panel voltage .panel current and duty cycle ratio estimated.

Figure 3. PV Array characteristics

639
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1. PV panel datasheet specification

Item Item Description Value
1no Pmp (W) 215

2 Voc (V) 36.3

3 Vmp (V) 29
4 Isc (A) 7.84

5 Imp (A) 7.35

Table 2.Parameter Estimation

Design of ANN-based MPPT controller

A Neural Network (NN)will be designed and trained using Neural network toolbox from MATLAB.NN predict
the duty cycle for boost converter. ANN structure designed is shown in Figure 4. The inputs for the network will
be PV voltage (V) and PV current (I) and output will be Duty Cycle (D).

Figure 3. Feed-Forward Neural Network

Figure 5.Regression for testing

640
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 6. Mean Square Error

Performance evaluation of ANN-based MPPT Controller:

The ANN is trained using two inputs PV voltage and PV current. Sample of training data is given in Table
1.Panel current and voltage will be the training inputs for ANN and the target output will be the Duty cycle D.
Levenberg-Marquardt (LM) is used as a training function. The predicted output fits the training data with a
higher value of the regression coefficient of 0.99972 can be seen in Figure 5. The mean square error (MSE) and
regression coefficient are analysed to find how close ANN can approximate the function for MPPT. Figure 6
shows the Mean Square Error (MSE) value of 1.0779*10-05, and its best performance achieved at epoch 13.
For validating the performance of ANN MPPT controller, the system was tested using a 48V, 200W DC-DC

Figure 7.ANNbased MPPT controller

boost converter. This arrangement is presented in Figure 7.

Maximum panel power recorded is 219W and the power tracked by MPPT is 193.4W in 0.804 seconds. The
efficiency of MPPT is calculated using the following equation and obtained as 89.1%.
𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝑝𝑜𝑤𝑒𝑟 𝑡𝑟𝑎𝑐𝑘𝑒𝑑
Efficiency= 𝑋 100% (1)
𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝑝𝑎𝑛𝑒𝑙 𝑝𝑜𝑤𝑒𝑟

Boost converter output is shown in Figure 8 .Figure 9 sows the power output with MPPT and without MPPT.
This indicates that the ANN-based MPPT controller can improve the PV power tracking.

641
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 8.Boost Converter Output

Figure 9.ANN MPPT performance

ANN MPPT based On grid PV system

An on grid PV system is designed and simulated in MATLAB shown in Figure 10. A 215 Watt PV panel,
1Soltech 1STH- 215-P, is coupled to a 48 V DC, boost converter. Boost converter with the specifications: C1 =
100 µF, C2 = 550 µF and L = 300µH is designed DC voltage regulation.

Figure 10. ANN MPPT based On Grid PV system

Inverter output current and voltage waveforms are shown in figure 11.It is observed that Inverter voltage is
synchronised at 220Volt, 50 Hz. Inverter current peak is recorded as 0.5A.

642
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 11. ANN MPPT based Grid-tied PV system

CONCLUSION

A 215 W PV panel IV and PV characteristics were recorded at varying irradiation and temperature levels. The
characteristics show that PV output fluctuates with varying climatic conditions. Then an ANN-based MPPT
controller was designed and simulated. The simulation results show MPPT controller designed using ANN can
efficiently track the MPP .This in turn improves the power conversion efficiency of the system. This was
experimentally validated using a PV system which employs a boost converter and a resistive load. Finally, the
PV system is integrated with grid. A PLL has been used for grid synchronization and a PID controller is used
for inverter control. The inverter output is synchronised with grid voltage and frequency by the PLL.

REFERENCES

[1] Blaabjerg, F., Iov, F., Teodorescu, R. and Chen, Z.. Power electronics in renewable energy systems. In:2006 12th
International Power Electronics and Motion Control Conference, August 2016, IEEE, Portoroz, pp. 1-17.
[2] Panda, A., Pathak, M.K. and Srivastava, S.P., A single-phase photovoltaic inverter control for the grid-connected
system. Sadhana, 41(1), pp.15-30. https://doi.org/10.1007/s12046-015-0459-2
[3] Blaabjerg, F., Teodorescu, R., Chen, Z. and Liserre, M..Power converters and control of renewable energy systems. In:
Proceedings of the 6th International Conference. Power Electron, October 2004, IEEE, pp. 1-20.
[4] Revathi, B.S. and Prabhakar, M., 2016. Non-isolated high gain DC-DC converter topologies for PV applications–A
comprehensive review. Renewable and Sustainable Energy Reviews, 66, pp.920-933.
[5] Sanjeevikumar Padmanaban, Frede Blaabjerg, Patrick Wheeler, Joseph Olorunfemi Ojo & Ahmet H. Ertas.High-
Voltage DC-DC Converter Topology for PV Energy Utilization—Investigation and Implementation, Electric Power
Components and Systems,2017, 45:3,221-232, DOI: 10.1080/15325008.2016.1248251
[6] Viswambaran, V.K., Ghani, A. and Zhou, E., Modelling and simulation of maximum power point tracking algorithms
& review of MPPT techniques for PV applications. In 2016 5th International Conference on Electronic Devices,
Systems and Applications 6-8 December. 2016: IEEE, pp. 1-4.
[7] Baharudin, N.H., Mansur, T.M.N.T., Hamid, F.A., Ali, R. and Misrun, M.I., 2017. Topologies of the DC-DC
converter in solar PV applications. Indonesian Journal of Electrical Engineering and Computer Science, 8(2),
pp.368-374. DOI: 10.11591/ijeecs.v8.i2.pp368-374
[8] Raghavendra, K.V.G., Zeb, K., Muthusamy, A., Krishna, T.N.V., Kumar, S.V.S., Kim, D.H., Kim, M.S., Cho, H.G.
and Kim, H.J., 2020. A Comprehensive Review of DC-DC Converter Topologies and Modulation Strategies with
Recent Advances in Solar Photovoltaic Systems. Electronics, 9(1), pp.31. https://doi.org/10.3390/electronics9010031
[9] Singh, V. and Tiwari, A.N., 2018, April. Study and Comparison of various types of Converters used for Solar PV: A
Review. In 2018 International Conference on Power Energy, Environment and Intelligent Control IEEE.pp. 658-664.
[10] Revathi, B.S. and Prabhakar, M., 2016. Non-isolated high gain DC-DC converter topologies for PV applications–A
comprehensive review. Renewable and Sustainable Energy Reviews, 66, pp.920-933.
[11] Amir, A., Amir, A., Che, H.S., Elkhateb, A. and Rahim, N.A., 2019. Comparative analysis of high voltage gain DC-
DC converter topologies for photovoltaic systems. Renewable energy, 136, pp.1147-1163.

643
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[12] Nathan, K.S., A Novel DC-DC Converter for Photovoltaic Applications.Doctoral dissertation, University of
Cambridge. Cambridge .2019.
[13] M. Das and V. Agarwal, "Novel High-Performance Stand-Alone Solar PV System With High-Gain High-Efficiency
DC-DC Converter Power Stages," in IEEE Transactions on Industry Applications, vol. 51, no. 6, pp. 4718-4728,
Nov.-Dec. 2015, DOI: 10.1109/TIA.2015.2454488.
[14] V.C. Kotak1, Preti Tyagi. DC To DC Converter in Maximum Power Point Tracker, International Journal of
Advanced Research in Electrical, Electronics and Instrumentation Engineering, Vol. 2, Issue 12, DECEMBER 2013
[15] Singh, S.N., 2017. Selection of non-isolated DC-DC converters for a solar photovoltaic system. Renewable and
Sustainable Energy Reviews, 76, pp.1230-1247. DOI: https://doi.org/10.1016/j.rser.2017.03.130
[16] Bayoumi, E.H., 2015. Power electronics in renewable energy smart grid: a review. International Journal of
Industrial Electronics and Drives, 2(1), pp.43-61. DOI: 10.1504/IJIED.2015.068769

644
 SUITABLE SITE SELECTION TO WIND POWER PLANTS FOR
ENVIRONMENTAL AND SOCIAL IMPACTS BY USING GIS AND
MCDM: A CASE STUDY DEVELI KAYSERI
Fatih KARİPOĞLU
Department of Energy Systems Engineering, Izmir Institute of Technology, Izmir, Turkey, 35430,
[email protected], ORCID: 0000-0002-3314-134X

Mustafa Serdar GENÇ

Wind and Aerodynamic Research Center, Department of Energy Systems Engineering, Erciyes University, 38039,
Kayseri, Turkey, [email protected], ORCID: 0000-0002-6540-620X

Kemal KOCA
Wind and Aerodynamic Research Center, Department of Energy Systems Engineering, Erciyes University, 38039,
Kayseri, Turkey, [email protected], ORCID: 0000-0003-2464-6466

Karipoğlu, F, Genç, MS, Koca, K. Suitable Site Selection to wind plants for environmental and
Cite this paper as: social impacts by using GIS and MCDM: A case study Develi, Kayseri. 8th Eur. Conf. Ren.
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Wind energy has major benefits these days such as providing the increasing energy demand and tackling
climate change problems. Detailed planning processes such as technical, social, environmental, various
agents, and political concerns are essential for the development of wind energy projects. This study aimed
to develop a tool combining Geographic Information System (GIS) and Multi-Criteria Decision Making
(MCDM) methodologies and implementation for Kayseri, Develi in Turkey as a case study. For the
analyzes, CORINE CLC 2000 and other data sources were employed for data acquisition to unlock
fragmented and hidden onshore data resources and to facilitate investment in sustainable coastal and inland
activities. For the determination of feasible site selection, different constraints and criteria were recognized
based upon a literature review and regulations that were gathered from different agencies. After detailed
analyses, near to the Havadan (7.87 MW) and Kulpak (9.22 MW) villages were found to be the most
suitable locations for the installation of a potential onshore wind farm. The output of this study could be
used by energy power planners to estimate the extent that the wind could be developed based on public
perception, administrative, energy political rules, technical, environmental, and social impacts.

Keywords: Wind Energy; Environmental Impacts; GIS; MCDM; Suitable Site Selection
© 2020 Published by ECRES

1. INTRODUCTION

Energy has become the basic input in many areas such as factories, workshops, electronic devices in homes,
street lighting, railway transportation even electric vehicles. Inherently, Energy consumption is inevitably
increasing because of the rising of the world population and the development of technological devices rapidly.
Hence, many concerns such as increasing energy demand, fossil fuel depletion, and environmental issues have
led countries to search for ways to develop Renewable Energy capacities [1]. These caused the Renewable
Energy Sources (RES) to gain more reputation [2]. It can be said that RES consists of wind, solar, geothermal,
ocean, and bioenergy. Of these resources, wind energy plays a crucial role due to its potential in energy
generation, its market value, its wide application range, and its economic attributes [3–6]. Regarding the
information of Turkey, it is located 360-420 Northern parallel with 260-450 Eastern meridians, and it is
surrounded by seas including the Mediterranean Sea at south part, the Aegean Sea at west part, the Marmara
Sea at the northwest part and the Black Sea at north part, resulting in an appearance of a peninsula. It has also a
lot of natural resources such as geothermal energy, water sources, and boron. Therefore, the Turkish Government

645
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

determined the aims for 2023. Specifically, inside of aims, Turkey will meet 30% of energy demand from RES.
In this respect, Turkey is executing an innovative energy policy for the future so that RES plays an important
role [7]. Due to geographical location, onshore and offshore wind energy potential were identified 48000 MW
and 17393,2 MW, respectively [8].

Concerning the most suitable regions for wind farm installation, it was important that the determination of
criteria must be taken into consideration how these criteria would affect the site selection regions. Geographic
Information Systems (GIS) technologies enable the identification of factors, affecting the installation of energy
power plants and their collection in a spatial database along with the effective evaluations of relevant analyses
[9]. GIS can determine the most suitable regions for energy power plants after the determination of crucial
factors for the selected area. In addition to GIS, Multi-Criteria Decision Makers (MCDM) was conducted at a
suitable region selection process for energy power plants [10]. Many researchers were interested in studying
suitable site selection examined by utilizing these two methods [11-23]. Within the scope of Turkey’s 2023
targets, it is aimed that the most suitable regions for the electricity production from wind energy are determined
and suitable wind energy conversion systems are installed. With the reach of these targets, the dependence on
external sources and fossil fuels of Turkey will be gradually decreased.

Develi County at Kayseri Province is selected for the studied area in terms of the potential and environmental
assessment for the most site selection of wind farm. The average wind speed and capacity are above 7 m/sec
and 40% in this region since Develi had a lot of high hills and mountain ranges. This wind speed and capacity
values are enough for economical wind energy investment [14]. Furthermore, the criteria were determined for
the construction of wind energy power plants at the most suitable areas and the process steps and required
analyses were conducted using databases from different sources such as Global Wind Atlas, Emodnet,
Copernicus, Wind Energy Atlas. Multiple data sources were newly utilized for accurate analysis results.
Databases from different sources, processed in GIS, and important MCDM methods, spatial analysis queries
were performed for the most appropriate site selection. Consequently, this study determined the most suitable
regions in Develi for wind energy power plants with detailed analysis. Using GIS and MCDM, all necessary
technical, social, environmental, and political details were considered at a suitable region selection process for
wind energy conversion system installation.

2. METHODOLOGY
In this study, the proposed methodology for suitable site selection of wind energy power plant in Develi County
was illustrated in Fig. 2. The diagram was handled under four steps. According to the information obtained in
the literature review, the data sources and databases required for the potential, environmental, and social analyses
were found in the first step. On the next step, according to received wind speed data from Global wind atlas,
regions with the wind speed less than 3m/sec were excluded since the wind turbine cut-in speed was generally
3 m/sec. In the third step, environmental and social analyzes for the remaining regions were performed according
to the 12 different restrictions. As seen in Fig. 1, the red-colored restrictions were very important for the wind
energy plant investment. After the environmental and social analysis, it was determined the suitable regions and
wind energy potential in the study area.

646
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. The methodology of study

2.1. Multi-criteria Decision-Making Methods (MCDM)

As mentioned before, this method is very important for suitable site selection for energy power plants. The
application of MCDM techniques supplies renewable energy management [15]. The method is tools allowed the
best choice to be made from among multiple and concurrent criteria. The international studies in the field of
solar energy evaluating the criteria effective in the installation of solar PV power plants using GIS have often
preferred MCDM to determine the weights of the parameters [16-18]. Hence, MCDM was performed in this
study. It was examined the two steps of the multi-criteria. In the first step, it was examined the wind energy
potential. Environmental and social restrictions were investigated in the next step.

2.2 Studied Area

This studied area was conducted in Develi, located southwest of the Kayseri. Develi is located at a latitude of
38.40 and a longitude of 35.50. The face measurement is 1887 m2 and its height above sea level is 1150 m. Develi
is at the foot of the high mountain (Erciyes Mountain). According to the wind energy potential atlas, Develi,
Yahyalı, and Pınarbaşı are the places with the highest wind potential in Kayseri. There are wind farms in Yahyalı
but there were not any wind farms in Develi and Pınarbaşı. Moreover, Develi have not any fault lines in its
underground. For these reasons, Develi can be chosen as a pilot region due to either a suitable geographical
situation or high wind potential. The study area is shown in Fig. 2. Also, the Develi Municipality gives
importance to renewable energy systems, and Develi has 3 solar energy power plants.

Figure 2. The Study Area

647
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2.3.The affecting factors for suitable regions and obtained data

Data are among the most basic requirements of site selection for wind energy power plants. Therefore, it was
used in many different data sources in this study. Firstly, the restrictions and demands were determined. In Table
1, the criteria of the restrictions and data sources were given. For this study, it was determined 12 different
criteria by using the GIS. These criteria consisted of wind speed, forests, military regions, civil and military
aviation, designated regions, agricultural, water sources, roads and ports, fault lines, bird’s migration paths,
energy transmission lines, which are important for wind energy installation. In addition to criteria, it was
determined the buffer zones for 9 restrictions. In Table 2, buffer zone distances and sources of these distances
were ensured for the determined restrictions. The buffer zone distances for the 9 restrictions were described in
Section 2.3.

Layer Source of Data

Wind Speed and Power Global Wind Atlas 2020
Forests CORINE CLC 2000 (311, 312, 313)
Military Regions Turkish Army Forces
Airports and Aviation CORINE CLC 2000 (124), Flight Radar 24
Urban Regions CORINE CLC 2000 (111,112)
Designated Regions Emodnet 2018, UNESCO 2020
Agricultural Regions CORINE CLC 2000 (212,213,222,223)
Water Sources CORINE CLC 2000 (511,512,132)
Roads CORINE CLC 2000 (122), OpenStreet Map 2020
General Directorate of Mineral Research and Exploration 2013, Kandilli
Fault Lines Observatory and Earthquake Research Institute (KOERI 2020)
"Decision Making based on environmental
Bird Migration Paths impacts" (Degirmenci 2018)
Energy Transmission TEİAŞ (TEIAS 2020), REPA 2020
Table I. Map layers for restrictions and data sources

Spatial constraints Buffers Sources

Agricultural Regions Outside Malczewski 2010
Military Regions 5 km Turkish Army Forces 2020
Roads 0,1 km Çolak 2020
Designated Regions 5 km Law on National Parks 1983; Law on Protection of Cultural and
Natural Properties 1983
Urban Regions 3 km Regulation on the Technical Assessment of Applications related to
Wind Power Generation 2015
Fault Lines 150 m Demirtaş 2005
Energy Transmission 0-5 km TEİAŞ 2020
Airports 3 km Legislation for Construction Criteria 2012
Bird Migration Path 3 km Zimmerling 2013
Table II. Buffer zones and its sources

2.3.1. Wind Energy in Develi

Wind speed was generally below 10 m/sec for most of the year in the study area [3]. In the winter, stronger
winds exceeding 20 m/sec and the summer winds were much weaker than the winter winds, rarely exceeds 10
m/sec. The wind energy potential atlas is shown in Figure 4. The average wind speed of the studied area at 100
m was between 7-9.5 m/sec. For the accurate wind power plant investment, the wind speed should exceed 6
m/sec for a high level of power generation and should indicate good potentiality for feasible wind farms [19].
For the wind potential analysis, data were received from Global Wind Atlas at 100 m elevation.

2.3.2. Forests
The total area of Turkey was 78 million hectares and 21.7 million hectares were designated as forest areas [20].
Kayseri has a few forest areas. In general, the number of forest areas in the districts is low. Southeast of Develi
is forested areas for natural sources and facilities areas. Also, in order to minimize forest loss, the forest areas
don’t use for different purposes.

648
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2.3.3. Military Regions

Some lands can be used by the army as military exercise and settlement areas. The military regions are very
important for every country. Therefore, these regions are closed off to use for different purposes. The army can
use these regions as military prohibited, exercises regions, and security zones. On the other hand, the wind
industry thinks that many economically feasible projects are limited due to military reasons [21]. There is one
military region in Develi borders. This region is located in Develi province. Therefore, installation of the wind
power plants is not prohibited regions.

2.3.4. Civil and military aviation

Wind farms may sometimes be dangerous for civil or military planes, helicopters. Since the wind tower and hill
heights are approximately 100 and 400 m respectively, the total height from the ground is 500 meters. This
height is dangerous when the planes or helicopter taken-in or taken-off. Therefore, wind farms must not install
near the civil or military airport regions.

2.3.5. Urban regions

The population has migrated from village to city for 10 years. Therefore, cities are being crowded day by day,
and the land area used for the installation of houses, factories, and shopping centers in the agricultural lands.
This causes irregular urbanization to occur especially in big cities. As considering economically, the land prices
are very high level in urban regions. Because of the reasons, energy plants are located outer part of urban regions
[22].

2.3.6. Designated Regions

The designated regions are very important to determine the suitable regions for energy power plants. Around
the world, the designated and the wonder of natural regions are determined by the European Council (UNESCO).
These regions are protected by this council and they are not used. Therefore, these regions are the excluded
regions in any feasible site selection study. Due to geographical location and the old volcano mountain Erciyes,
Kayseri has a lot of the natural wonders.

2.3.7. Agricultural Regions

Turkey has been a marvelous agricultural country for a long time. Because, Turkey has a lot of efficient lands.
The located in the middle of Turkey, the soil is better than other regions. Although Kayseri is expressed as a
trade center in Turkey, there are many agricultural areas in Kayseri. Also, Develi has a lot of regions related to
agriculture such as honey pumpkin. In the study, the data with regards to agricultural regions were received from
CLC 2000 and were then proceeded in GIS.

2.3.8. Water Sources and Paths

Because of global warming, the world is heating, therefore the man needs water. For the future, water sources
will be very important, maybe the wars will stem from water sources. Develi is a rich position with regards to
water sources because the snow on the Erciyes mountain is melting on sunny days and accumulates to water
sources in Develi. While the water sources and paths are very important, these regions should not be used for
any other purpose. So, these waters regions are not suitable for wind energy plants.

2.3.9. Energy Transmission Lines

The energy should be consumed where it is produced for the minimum energy losses. Turkey has an
interconnected system for energy lines from production plants to energy demand points. Considering the
distance to substations and power transmission lines in the selection of locations will provide an advantage in
reducing energy loss. Develi has Sızırlı transformer center and energy collection. Yeşilhisar transformer center
and energy transmission line are close to Develi.

649
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. RESULTS
Regarding the suitable site selection of onshore wind power plants for Develi city, it was received that much
data was related to potential, environmental and social impacts from different sources and proceed in GIS. Table
1 shows that the effective impacts are 12 for onshore wind farms installation. These data is analyzed as 12
different map layers in GIS. Finally, suitable site map layers were obtained by using all of the map layers. For
determined of the suitable site selection in Develi, it was prepared in Figure 3 and it were obtained final analysis
map layer. Firstly, according to the environmental and potential impacts, it was determined 5 suitable regions
for wind power plants. But, as indicated in Table 2, only 1 region was very suitable for wind power plants
because of the energy transmission line constraints buffers. Because other regions were much far than 5 km to
Develi energy transmission line. In Figure 4, two suitable regions for wind farms investment could be seen.
These regions were two different villages. It was examined from Google Maps to determine the suitable region,
these regions were not used as a human activity. And these regions had many mountains, hills and near the main
roads. The first determined regions were near Havadan village and the suitable region was nearly 7500 m 2. The
second region was near the Kulpak village and a suitable region was 12300 m2. Determined suitable regions for
wind energy power plants shown in Figure 5. These regions are important for wind energy power plants
installations. When only wind speed is evaluated, almost all of Develi are suitable according to Figure 5. But
when environmental and social impacts are included, only two are determined as suitable regions in Figure 18.
So, detailed and versatile analysis is very important for the determined of suitable site selection. In this
comprehensive detailed analysis study, it were analyzed technical, environmental and social impacts. After the
analysis studies, it were obtained two region with total 19800 m2 suitable region. If the restrictions set with
buffer zone distances change over time, the determined of suitable site selection and total suitable regions may
change.

4. CONCLUSIONS

In this study, the suitable site selection of wind power plants was investigated for Develi city in Turkey. For the
suitable site selection, employing 11 criteria of Develi borders were assessed in this comprehensive study. These
criteria consisted of technical, environmental, social impacts. Data was received from different sources in the
first step. The received data was proceeded in GIS, and regions with wind speed below 3 m/s were excluded
from the suitable site map. There were two regions which were wind speed below 3 m/s in Develi. The third
step consisted of 11 criteria for a suitable site selection process. This data was analyzed and finally, two regions
were determined for wind power plant installation according to wind potential, technical, environmental, and
social impacts. The first region’s wind power density was determined 1050 W/m2 and the suitable region was
7500 m2. The first region’s wind energy potential was 7.875 MW level and this wind power level was the low
rate for any wind energy plant. The second suitable region was near Kulpak village. According to mean power
density, for the second region’s wind power density, it was determined 750 W/m2, and the suitable region was
12300 m2. Therefore, the second region’s wind energy potential was 9.225 MW level and this potential value
was better than the first region. The total suitable region’s wind energy potential was determined 17.1 MW level.
The GIS and MCDM combination method provides to create a lot of map layers for the analysis of suitability
of areas according to environmental and social impacts. For the determined of suitable site selection in Develi,
it were created with 12 map layers by using GIS and MCDM. In this study, the specific conclusion of this study
is to prepare a detailed and versatile analysis study for the suitable site selection. Thanks to this study, fast,
virtual and detailed results were obtained for wind energy potential evaluation studies in Turkey. Analysis
studies were shown suitable and excluded regions and also were shown map layers with buffer zone distances.
With this study, investors, decision makers, authorities and planners can easily get information and this study
can easily adapt to changing updates.

650
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. The Final Analysis Map Layer

Figure 4. Determined Two Suitable Region in Develi

Figure 5. The Google Maps Review for Determined Regions

REFERENCES
1- Satir M., Murphy F., McDonnel K., “Feasibility study of an offshore wind farm in the Aegean Sea, Turkey”, Renewable
and Sustainable Reviews, 2018, 2552-2562.
2- Emeksiz C., Demirci B., “The determination of offshore wind energy potential of Turkey by using novelty hybrid site
selection method”, Sustainable Energy Technologies and Assessments, 2019, (36)100562.

651
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3- Adelaja A, McKeown C, Calnin B, Hailu Y. Assessing offshore wind potential. Energy Policy, 2012,42,191–200.
4- Coskun AA, Türker YÖ. Wind energy and Turkey. Environ Monit Assess, 2012, 184, 1265–73.
5- Limpo JR. Assessment of offshore wind energy in Portuguese shallow waters site selection, technical aspects and
financial evaluation. (A Master’s Thesis Instituto SuperiorTécnico). Lisbon, Portugal, Universidade Técnicade Lisboa,
2011.
6- Snyder B, Kaiser MJ. A comparison of offshore wind power development in Europe and the U.S. patterns and drivers
of development. Apply Energy 2009,86,1845–56.
7- Çolak E., Memişoğlu T., Gercek Y., “Optimal site selection for solar photovoltaic (PV) power plants using GIS and
AHP: A case study of Malatya Province, Turkey”, Renewable Energy, 2020, 565-576.
8- Altuntaşoğlu, Z. T. 2012. “Wind Energy in Turkey, Current Status, Problems” Journal of Engineer and Machine, vol.
52, issue 617, pp. 56-63.
9- M. Jahangiri, R. Ghaderi, A. Haghani, O. Nematollahi, “Finding the best locations for establishment of solar-wind power
stations in Middle-East using GIS”, a review, Renewable Sustainable Energy Rev. 66 (2016) 38e52.
10- J. Brewer, D.P. Ames, D. Solan, R. Lee, J. Carlisle, “Using GIS analytics and social preference data to evaluate utility-
scale solar power site suitability”, Renewable Energy 81 (2015) 825e836.
11- Mahdy M., Bahaj A., “Multi-criteria decision analysis for offshore wind energy potential in Egypt”, Renewable Energy,
2018, 278-289.
12- Saleous N., Issa S., Mazrouei J., “GIS-based wind farm site selection model offshore ABU DHABİ EMİRATE, UAE”,
Remote Sensing and Spatial İnformation Sciences, 2016 XXIII ISPRS Congress, 12-19 July 2016, Prague, Czech Republic.
13- Cavazzi S., Dutton A.G., “An offshore wind energy Geographical Information System (OWE-GIS) for assessment of
the UK’s offshore wind energy potential”, Renewable Energy, 2016, 212-228.
14- Akdağ S.A., Güler Ö., “Evaluation of wind energy investment interest and electricity generation cost analysis for
Turkey”, Applied Energy, 2010, 2574-2580.
15- Pohekar S., Ramachandran M., “Application of multi-criteria decision making to sustainable energy planning—a
review.”, Renewable Sustainable Energy Rev, 2004, 8(4):365–381.
16- Uyan M., “Optimal site selection for solar power plants using multi-criteria evaluation: a case study from the Ayranci
region in Karaman, Turkey”, Clean Technology Environment, 19 (9) , 2017, 2231-2244.
17- Fang, J. Li, Song W.Y., “Sustainable site selection for photovoltaic power plant: an integrated approach based on
prospect theory”, Energy Conversation Management 174, 2018, 755-768.
18- Doorga J.R.S., Rughooputh S.D.D.V., Boojhawon R., “Multi-criteria GIS-based modelling technique for identifying
potential solar farm sites: a case study in Mauritius”, Renewable Energy, 133, 2019, 1201-1219.
19- Baban, S., Parry, T., “Developing and applying a GIS assisted approach to locating wind farms in the UK”, Renewable
Energy, 2001, 59-71.
20- TUIK, Turkish Statistical Institute, 2014. Turkey in Statistics, (http://www.turkstat.gov.tr), (Accessed March 2020).
21- Lindgren F, Johansson B, Malmlöf T, Lindvall F. Siting conflicts between wind power and military aviation –
problems and potential solutions. Land Use Policy 2013;34:104–11.
22- Watson JJ, Hudson MD (2015) Regional scale wind farm and solar farm suitability assessment using GIS-assisted
multi-criteria evaluation. Landscape Urban Plan 138,20–31.

652
 RESEARCH OF ENERGY AND EXERGY MODEL OF
THERMOSIPHON CIRCULATION SOLAR COLLECTORS WITH
THERMAL PUMPS FOR HEATING AND HOT WATER SUPPLY
Yedilkhan Amirgaliyev
Institute Information and Computational Technologies CS MES RK, Almaty, Kazakhstan, [email protected],
ORCID:0000-0002-6528-0619

Murat Kunelbayev
Institute Information and Computational Technologies CS MES RK, Almaty, Kazakhstan, [email protected],
ORCID:0000-0002-5648-4476

Aliya Kalizhanova
Institute Information and Computational Technologies CS MES RK, Almaty, Kazakhstan, [email protected],
ORCID:0000-0002-5979-9756

Omirlan Auelbekov
Institute Information and Computational Technologies CS MES RK, Almaty, Kazakhstan,[email protected],
ORCID:0000-0002-2903-9056
Nazbek Katayev
Institute Information and Computational Technologies CS MES RK, Almaty, Kazakhstan, [email protected],
ORCID:0000-0003-0501-3719

Ainur Kozbakova
Institute Information and Computational Technologies CS MES RK, Almaty, Kazakhstan, [email protected],
ORCID:0000-0002-5213-4882

Yedilkhan Amirgaliyev, Murat Kunelbayev, Aliya Kalizhanova, Omirlan Auelbekov, Nazbek

Katayev, Ainur Kozbakova Theoretical and mathematical analysis of double-circuit solar
Cite this paper as
station with thermo siphon circulation . 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: The work herein presents research of thermosiphon circulation flat solar collectors with thermal pumps
for heating and hot water supply. To analyze the proposed system of a new heat pump with a flat solar
collector, a heat pump model was developed. that takes into account the energy consumption of fans and
circulation pumps. As a result of the study, it was shown, that with an increase in the flat collector area
and absorption coefficient, the cooling capacity, power consumption for the heat pump compressor and
the heat capacity of the heat pump led to an increase in the productivity of the heat pump system. The heat
balance equation has been solved, the dependence of the temperature of the coolant has been derived, and
the rate of heat transfer from the heat pump to the tank, the energy consumption for the fans and the
collector efficiency have been computed. The proposed flat solar collector system, with a heat pump as an
autonomous system for heat and hot water supply, has a number of advantages: significant increase in
working time, a simple system configuration with low capital costs.
Keywords: flat solar collector with thermosiphon circulation, heat pump, coolant, efficiency
© 2020 Published by ECRES

1. INTRODUCTION

Heating, air conditioning and lighting systems, water heating are the fourth largest energy-consuming part in the
building sector [1]. Domestic water heating systems are equipped with conventional heaters, where the

653
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

efficiency of conversion into thermal energy is relatively low [2]. To reduce the dependence of the building on
electricity consumption, solar energy can be used to heat water for domestic consumption and space heating [3].
A promising technology of water heating systems, using a heat pump can significantly improve the performance
of solar heating systems [4]. A sufficient number of studies have been devoted to solar heating systems, where
structural features, thermodynamics, numerical modeling, and economic possibilities were considered, and they
were summarized in extensive reviews by Hepbasli and Kalinci [1] and Chu and Cruickshank [5].

2. RESEARCH METHODOLOGY. SYSTEM DESCRIPTION

Figure 1 shows a schematic diagram of a dual-circuit solar system with thermosiphon circulation. The system,
in this study, is used for hot water supply and space heating during the winter period and mainly consists of a
flat solar collector, siphon metering tank, a liquid heat pump and a hot water storage tank. With the help of a
thermosiphon pipe, a constant circulation takes place, which allows the flat collector capture thermal energy,
even if there is no solar radiation. Therefore, a thermosiphon circulation flat solar collector has the advantage of
significantly increasing the operating time period compared to operation with a conventional flat plate.

The work of the proposed installation is as follows. Solar energy E with a temperature t0 is absorbed by a solar
collector 1, with a temperature t1, heating a stream of solar energy, passes through a translucent insulating glass
2. Heat, received from a solar stream heats the liquid in the coils 3, which is removed from the collector, and
cold water comes in its place from the pipeline with a cold water valve 8, and from the siphon of the dispenser
tank 7, a constant thermal siphon circulation takes place, using the circulation pipe 10. Next, the liquid enters
the heat pump 11, which consists of a condenser evaporator 12 with temperature t2, in which the heat exchanger
is made in the spiral form, absorbing the heat of the coolant, lowering its temperature below the temperature of
atmospheric air (Q2), using a throttling valve 14, thereby contributing to the additional absorption of heat from
atmospheric air. The diagram also shows the solar radiation, reflected from a translucent coating (Q0) and the
surface of the absorbing panel (Q1). In the heat pump, the energy of the coolant, with a relatively low temperature
is transferred to the coolant of the heat exchanger of the condenser 15 in the form of a spiral with a higher
temperature t2, which increases the area, as well as the heat exchange rate. To carry out such a cycle, a
compressor 13 with a temperature of t3 is used with an electric drive 17. Further, by means of a heat exchanger
of a condenser 15 with a temperature of t4, heat from the heat pump (Q5) is transferred to the tank from the heat
exchanger Q6 with a temperature t6 of heating system 18. Since the installation has two circuits, it is equipped
with automatic circulation pumps 19 and 20 for circulating liquid between the solar collector and the evaporator,
condenser and storage tank. The water temperature is brought to the required technological level and is supplied
to the consumer for the purpose of hot water supply and heating.

When there is no solar radiation in cloudy days or at night, from the siphon of the metering tank, connected to a
flat solar collector, the received heat is transferred to the coils and the walls of the flat collector pipe, then using
a circulation pipe, the liquid enters the heat pump by forced convection.

654
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. Principal diagram of dual circuit soalr installation with thermosiphon circulation

Table 1. Selected performance capabilities of flat solar collector

Parameters Value
Absorbing plate material copper
Absorber plate dimensions 2 m×1 m
Plate thickness 0.4 mm
Glazing material Hardened glass
Glazing sizes 2 m×1 m
Glazing thickness 4 mm
Insulation Foam plex (foam polyurethane)
Collector tilt 450
Absorber heat conductivity 401 W/(m K)
Insulation heat conductivity 0.04 W/(m K)
Transmittance-absorption factor 0.855
Apparent sun temperature 4350 K
Environmental temperature 303 K
Irradiation intensity 1000 W/m2

3. OUTCOMES. HEAT PUMP OF «LIQUID-LIQUID» TYPE

For solving the analysis of the thermal pump, we will formulate the aggregate of thermal pump contour processes
by means of the thermal balance equation:
𝑑𝑡5 (1)
𝑄3 + 𝑊𝐸𝐿 𝜂𝐸𝐿 𝜂𝑘 + 𝑄4 − 𝑄5 − 𝑄6 = 𝐶2 𝑑𝜏
,

where: WEL EL К – the energy entering the contour from operation of a compressor; Q4 –
utilizable heat, emitted with a compressor during the working process; Q5 and Q6 – accordingly, heat
losses through the condenser heat exchanger surface and the energy, supplied by the thermal pump to
absorber tank; С2 – heat capacity of condenser contour transfer medium; t5 – transfer medium current
temperature at the output from a contour.

The main element of the circuit under study is the heat pump. Figure 5 shows a diagram of a new
technical solution of the thermal pump in section.

The proposed device allows you utilize the heat, generated by the compressor during operation and
at the same time cool it.

655
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. Scheme of new technical solution of heat pump in section.

Hereby 1 is the heat exchanger of the evaporator; 2 - condenser heat exchanger; 3 - compressor; 4 - throttle
valve; 5 and 6 - tubes for the coolant of the evaporator and condenser; 7- hole for introducing tubes into the
capacitor housing; 8 - mesh shell-divider of air flow.

The heat exchangers of the evaporator 1 and the condenser 2 are made in the form of ring-shaped containers,
formed by the inner and outer cylindrical shells with radii R1 and R2 and are mounted coaxially above each
other, at the bottom of the evaporator, above the condenser, forming an internal cylindrical air cavity. A
compressor 3 is installed in the cavity of the evaporator heat exchanger. To ensure optimal heat transfer from
the compressor to the evaporator, a cylindrical mesh shell-divider 8 is placed in the gap between them. The
coolants of the 1st and 2 nd circuits circulate inside the ring tanks, removing heat from tubes 5 and 6. Thus, the
compressor is practically located inside the “cold pan”, the walls of which are cooled by the coolant of the tubes
5. As a result, the heat, generated by the compressor is absorbed by the TP evaporator, increasing its
performance, and the compressor is simultaneously cooled without the use of a fan.

Table 2. Performance capabilities of a new thermal pump

# Design parameters Dimensions
1 Evaporator and condenser external diameters (D2), mm 375,0

2 Internal diameter (D1), mm 325 and 307

3 Heat exchangers jackets height, mm 355 and 382
4 Heat exchangers jackets width, mm 20 and 30
5 Heat exchangers jackets volume, mm 0.52 and 0,8
6 Shell-divider diameter, mm 238
7 Surpassing the evaporator height over compressor height, mm 80,0

4.PERFORMANCE INDICATORS

COP system is specified as

𝑄𝑐
𝐶𝑂𝑃𝐻 = 𝑊 (2)
𝑐𝑜𝑚𝑝 +𝑊𝑓𝑎𝑛 +𝑊𝑝𝑢𝑚𝑝,𝑠𝑐 +𝑊𝑝𝑢𝑚𝑝,𝑤

In addition to the COP system, heat output and annual hourly workload, hourly instantaneous collector efficiency
and solar energy utilization are also used to study solar collector performance. Hourly instantaneous collector
efficiency is defined as
(𝑇𝑠𝑐,𝑖𝑛−𝑇𝑎,𝑖𝑛)
𝐸𝑓𝑓𝑖𝑐𝑒𝑛𝑐𝑦𝑐𝑜𝑙𝑙𝑒𝑐𝑡𝑜𝑟 = 𝐹𝑅 (𝜏𝛼 ) − 𝐹𝑅 (𝑈𝐿 ) 𝐺𝛽
(3)

656
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The utilization factor of solar energy is defined as the ratio of the total increase in energy from the collectors to
the product of the total global solar radiation and the area of the collector and can be written as
∑ 𝑄𝑒
𝑅𝑠𝑜𝑙𝑎𝑟 = ∑ (4)
𝐺 𝛽 𝐴𝑐

Figure 3. Dependence of thermal pump heat amount on water heating temperature

Figure 3 shows the dependence of a heat pump heat amount on the temperature of water heating. It can be seen
from the graph, that in the winter period the heat pump’s heat amount drops according to weather conditions,
and also depends on the specific heat of the substance and the temperature change in the evaporator and
condenser. With temperature change, the amount of heat decreases according to the first and second
thermodynamics laws.

Figure 4. Dependence of thermal pump capacity on time and date

As can be seen from Figure 4, the thermal power was calculated in the winter period from 01.11.2019 to
02.02.2020. It can be seen from the graph, that at a power of 1500 Wt. during the winter period (November,
December, January and February), the power drops due to the climatic conditions of the given region.

657
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 5. Dependence of exergy efficiency of dual circuit flat solar collector with thermosiphon circulation on the
temperature

Figure 5 shows, that in the summer period, exergy efficiency increases with increasing ambient temperature at
lunchtime, when the total solar activity is at its peak value.

5. CONCLUSION

The experimental results of the study show, that the energy and exergy model of solar collectors with
thermosiphon circulation with heat pumps efficiently work in heating mode with accurately measured outdoor
temperature, typical of the climatic zone of Almaty (Kazakhstan). Energy analysis shows, that with the new flat
solar collectors, the average annual values of 𝑄𝑐 and 𝐶𝑂𝑃𝐻 were 2.5 kW and 8 kW, respectively. The efficiency
of the flat solar collector was also high, the COP system in winter from November to February was improved
by 4%. Experimental data showed the feasibility of the proposed system with flat solar collectors and a new heat
pump as an autonomous system for hot water supply due to the advantages of a significant extension of the
annual working time, simple system configuration and low capital costs.

ACKNOWLEDGMENT

The work was supported by the project of the Ministry of Education and Science of the Republic of Kazakhstan
№ 055236693 contract No. 318 of 30.03.2018

REFERENCES

[1] Hepbasli A, Kalinci Y. A review of heat pump water heating systems. Renew Sustain Energy Rev 2009;13:1211–29.
[2] Kim M, Kim MS, Chung JD. Transient thermal behavior of a water heater system driven by a heat pump. Int J Refrig
2004;27:415–21.
[3] Sterling SJ, Collins MR. Feasibility analysis of an indirect heat pump assisted solar domestic hot water system. Appl
Energy 2012;93:11–7.
[4] Chua KJ, Chou SK, Yang WM. Advances in heat pump systems: a review. Appl Energy 2010;87:3611–24.
[5] Chu J, Cruickshank CA. Solar-assisted heat pump systems: a review of existing studies and their applicability to the
Canadian residential sector. ASMEJ SolEnergy Eng 2014;136:041013.

658
 DRY COOLING OPTION FOR CONCENTRATING (CSP SOLAR
POWER) PLANT FOR THE DESERT REGION, INDIA

Satya Sekhar Bhogilla

Indian Institute of Technology Jammu, Jammu, India, [email protected], ORCID: 0000-0002-0607-2764

Rama Sreekanth P.S

Vellore Institute of Technology-AP, Andhra Pradesh, India, [email protected], ORCID: 0000-0001-6444-
0345

Bhogilla, Satya Sekhar, P.S, Rama Sreekanth. Dry cooling option for concentrating (CSP
Cite this paper as: solar power) plant for the desert region, India. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
2020, Istanbul, Turkey

Abstract: The cogency of dry cooled CSP plant to assess its odds or accretion against the wet cooled plant
is investigated in this work in Jodhpur, North-western town of India. Dry cooling proves to be a
flipside option over the wet cooled system in locations where water resource is scarce. In the
current work, Andasol Parabolic trough CSP Power plant, Spain, which has 50 MWel nominal
capacity with TES hours of 7.5 at full load is considered as a reference for design. Simulation is
performed for both dry and wet cooling options using Greenius software initially at the original
plant location in order to validate results by considering actual plant values. Post this, simulations
were performed at Jodhpur with the help of the local metrological data. The parameters such as
annual mean overall efficiency, levelized cost of electricity, and water consumption with dry
cooled for the reference plant at study site Jodhpur stands at 15.6 %, 0. 1347 €/kW h, 74,793 m3/a
respectively, whereas the same for wet cooled reference power plant yields 17%, 0. 1173 €/kWh
LCOE, and 5, 71, 365 m3/a at an annual energy yield of 163.24 GWhel net.

Keywords: CSP, solar energy, dry cooling, parabolic trough

© 2020 Published by ECRES

1. INTRODUCTION

Renewable energy sources used for green energy generation directly impacts the economy of any nation. At a
global perspective, fossil fuel-based energy has raised its serious environmental concerns and thus forcing the
energy sector to take strides towards embracing renewable energy (RE) generation. Although claiming high
initial investments, RE has expanded its scope over the last two decades in both the developing and under-
developing countries due to its low running costs and unsmudged benefits to the environment [1-4]. Despite
being ambitious, many nations need to frame and discern the RE policies and self-driven targets of specific
regions/states. The reason for this is attributed to the nature of RE sources, which are often gravely dependent
on the weather patterns and distributions across the nation. So assessing these weather patterns would be decisive
in the rate of amplification of RE capacity of any country, be it Solar PV, Wind, or Concentrated solar power
(CSP).

CSP technologies use Direct Normal Incidence (DNI). Arid and semi-arid areas are most suitable for setting up
the CSP plants due to clear sky and intense sunshine almost throughout the year, and its estimated global power

659
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

generation is 2946 PWh/year. India makes a suitable destination for CSP due to its climatic conditions with DNI
of at least 1800 kWh/m2/year for most of its area [5]. It may be noted that India has an area of 14,00,000 sq.km
receiving DNI between 1865- 2007 kWh/year (moderate potential locations), and an area of 7,50,000 sq. Km
receiving DNI of more than 2,375 kWh/year (high potential areas) for CSP setup. Overall it is observed that
around 50 % of total India’s area receives DNI, which is more than the required minimum of 1800 kWh/year
for setting up a CSP, thus making it a potential destination [6]. The DNI in Northwest India receives the highest
DNI compared to the rest of India. Based upon the detailed DNI values and the appraisal of suitable land
resources, North-West India has a CSP potential of 2000 GW, with the technical and economic potential of 1700
GW and 700 GW, respectively [7]. The Jawaharlal Nehru National Solar Mission (JNNSM) setup by the Ministry
of Renewable Energy, (MNRE), Govt. of India, targeted the phase-wise development of Solar-based RE in 3
phases; in a span of 3, 4 and 5 years blocks. Initial 3-year Phase –I (2010-2013) with utility grid power 1-2 GW,
four year Phase –II (2013-2017) with 4 -10 GW grid utility, and finally, the 5-year Phase- III (2017-2022) with
20 GW.

The cooling unit of the CSP forms another essential component for assessing its techno-economic viability for
any given location. Liqreina et al. [8] performed a comparison of technical performance by considering dry and
wet cooling options for a CSP plant at a representative site in the MENA region. The study concluded that the
plant performance should be improved at an optimized plant configuration. Qoaider et al. [9] optimized the dry
cooled parabolic trough CSP plants for the MENA region. The study concluded that dry cooled CSP plants
produce power at competitive prices in desert regions, further, it also added that increased plant size decreased
LCOE due to increased operating efficiency. The real-time CSP plant installation is truly estimated by its
characteristics such as economic, operational, and energy. Still, the related decisions on investments and designs
of plants need simulation support.

The power block efficiency has proven to be better and more advantageous with wet cooling than in dry cooling
of CSP as in the case of any steam power plant. As per the elementary rules of thermodynamics of steam power
cycles, the exit steam from the turbine, in case of wet cooling, gives a lower temperature than that of dry cooling
and cools down at a faster rate. Even parabolic trough based CSP power plants work in a way identical to
conventional steam power cycle. However, discordant to the fossil-fuelled conventional power plants which
require large quantities of water and are usually built in the coastal regions, CSP are built in the desert regions
which fall in the earth’s Sun Belt. These areas receive high DNI, like the North-West part of India, Jodhpur
region, which promise the most out of a CSP plant, with very little water utility. This forces us to thoroughly
investigate and weigh out the economic competitiveness of erecting dry cooled CSP power plants in the dry and
arid regions of the North-West part of India.

Dry cooling offers the advantage of saving water up to 90 %, but nonetheless, it offers reduced overall
performance of plant under higher ambient temperatures, however, higher DNI could indemnify for such losses
by offering higher efficiency at low ambient temperatures. Therefore, it behaves that the CSP plant design,
particularly its solar field, depended on the type of cooling system being employed. A dry cooled CSP required
a bigger solar field than that of a wet cooling system for the same given power output. This forces for high initial
investment cost for dry cooled CSP than a wet cooled one with coequal generation capacity. Although it is a
throwaway indication to consider a cooling for CSP, the rationale for choosing dry or wet cooling largely
depends on securing project finance. CSP plants by themselves require high investments, and a definite trade-
off between all options has to be carefully considered for each solar field to decide on the use of dry cooling.
Although seemingly for many regions, dry cooling proves viable and affordable options, nonetheless, it has to
be contemplated only after economic evaluation.

Evaluation on the use of dry and wet cooling parabolic plants for Jodhpur, a town in the North-western state of
India, forms the main objective of the present work. The work proceeds by comparing the Dry and wet cooling
options based on the design of the Andasol plant in southern Spain as a reference plant. The techno-economic
plant performance is compared to establish that higher DNI values could indemnify for loss of efficiency
conferred by a dry cooled power plant.

660
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2. Methodology

The site chosen for the current study is Jodhpur, which is the second-largest state in Rajasthan. With hot, dry,
and semi-arid climatic conditions near about the whole year with mild showers from late June to September and
high DNI, Jodhpur (26.8 N, 73.02 E) offers a beckoning location for installation of CSP in India to achieve the
prophesied socio-economic impacts on the area. Since the mid-1980’s southern California based 354 MW, CSP
Plants have been linked to the power grid and therefore represent a CSP plant technology in its prime [9]. Off
late, nine plant operations in the power generation capacity range from 14 to 80 MW with more than 100 plant-
year experience. The mostly used solar concentrator is a parabolic trough, and hence in the present paper, the
same is considered as shown in Fig. 1. The plant output is simulated using metrological data, while parameters
like thermal & optical collector efficiency, power block efficiency, and the corresponding field losses were
calculated as described by [10] for the full year for assessing yearly plant efficiency appraisal.

Figure.1. Scheme of a CSP plant with TES [10, 13]

System Advisor Model is adopted to collect the ambient temperature and also the wind velocities in the exact
site selected for this illustrative example, which is close to Jodhpur. Greenius plant simulations were performed
using a Metadata file generated from this data. The equations used for this study are available in Ref. [10].

3. Dry cooling as a viable option

The necessity for a CSP parabolic trough plant to use s dry cooled power block arises from the water scarcity in
the selected region due to the highest solar irradiation values. The chance of conducting a more realistic
investigation is dependent on the availability of highly accurate metrological information. Dry bulb temperature
(DBT) is one such parameter in dry cooling plants that affects the plant performance. Usually, high DBTs yield
low annual energy, while low temperature during the night- tide hours leads to solidifying of HTF. As mentioned
earlier, the reciprocal effect of ambient temperatures on plant efficiency made us poll on the occurrences of such
high temperature. It is evident that such high temperatures occur when DNI values are high at noon hours. This
DNI/temperature values combination at the chosen study site offers an obvious advantage of better sizing of
plant for CSP plant installation since the plant performance reduction due to high temperatures simultaneously
reduces the release of dumped energy to surroundings.

4. Technical specifications and inputs for simulation

Andasol Plant design (wet cooled) is considered as a bench-mark reference plant for the study. Actual inputs
were considered while performing simulations for the plant yield. The thermal storage capacity, which is known
to be banked upon the turbine thermal input, is adjusted to reach 7 h. Eurothough model is considered for the

661
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

solar plant with 156 collector loops and 510,120 m2 of total net collector area [11]. All the costs taken are in
accordance with [12, 13].

5. Results and discussion

CSP plant with two examples is given in this section, using the Greenius software, which is intended for easy
and quick annual performance simulations. The model considered is 50 MW CSP plant at the same site in
Jodhpur.

Validation

In order to testify the soundness of the simulations, this section is plighted to validate Greenius simulation
outputs with de-facto plants as published by plant owners. The plant data was resourced from Herrmann et al.
[14], which used a reference and inserted it into Greenius to perform simulations. This paper considered CSP
with Parabolic trough power plants and a photovoltaic power plant, as shown in Fig. 1. The CSP plant has 50
MW gross outputs and a TES with 7.5 hours of full load capacity, allowing for electric power generation even
during the hours when sunshine is not available. This requires a large solar field to charge TES during daytime
by producing excess heat for power block demand. With tracking axis oriented to South-North, and tracking
Sun from its East-West course; a solar field is made consisting of 624 Eurothrough collectors in an arrangement
of 156 loops of 4 collectors. The TES is of 2 tank molten salt type, which is widely used in several parabolic
trough plants globally. The obtained simulation results are compared with the owner’s published values and are
given in Table 1. This indicates that the reference plant models dish out reliable data that conservatively vie with
real-time situations. The following section presents the simulations of dry and wet cooling options undergone
in the study site.

Table. 1 Validation
Parameter Simulated Published data [14] % error
Annual electricity generation (MWh/a) 146,665 157,206 -7.18
Thermal efficiency peak (%) 67.9 70 -3.09
Overall efficiency peak (%) 24.1 25 -3.7
Mean annual field efficiency (%) 44.8 46.1 -2.9
Total efficiency (%) 13.7 14.7 -7.22

Reference power plant simulation in Jodhpur

Figure. 2. Monthly power generation by CSP plant in Figure.3. Efficiency of CSP plant in Jodhpur, India
Jodhpur, India

In this section, simulations are performed by attaching dry and wet cooling systems to the reference design of
Andasol and simulated in Jodhpur by swapping between the 2 power blocks for the same metrological data. The
reference plant assessed by the net power production (NPP) in the study location is shown in Figure 2, where it
is observed that higher power production was given by wet cooling due to higher turbine efficiency. Nonetheless,

662
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

during the winter (Nov-Jan) due to low ambient temperature, slightly higher power production was given by wet
cooling. The turbine performance is also improved by using the dumped energy in the dry cooling, which
otherwise would be lost in case of wet cooling. Jodhpur geographical location offers higher DNI as compared
to southern Spain and therefore also offers higher NPP in both dry and wet cooled conditions; however higher
ambient temperatures in summer in Jodhpur resulted in the variation of NPP between the two cooling systems.
Figure 3 depicts the overall efficiency, which shows higher efficiency of wet cooling than dry cooling at all
times. The difference narrowed to a minimum when the atmosphere temperature reached low. In rainy months,
overall efficiency reduced for both cooling systems because of low DNI. As Jodhpur has a higher DNI than the
Andasol region, the design configuration was not fit and thus needs to be optimized. This could be achieved by

Table 2. Technical comparison for all cases

Parameter Unit Andasol plant [8] Jodhpur reference plant Plus Jodhpur (%)
Cooling Wet type Wet type Dry type Wet type Dry type
Annual solar field MWhth 469100 506795 506795 8 8
Annual
thermal net electricity
power (Qtha) MWhel 146665 163247 149586 11.3 2
Annual
output
power
(Wnet)block MWhel 5440 6892 8944 26.6 64.4
Annual
parasitics
water m3/a 513327 571365 74793 11.3 -85.4
Levelized electricity
consumption 17.29 11.73 13.47 -32.1 -22
the storage’s full
optimizingcosts loadh hours and the solar field for dry cooled systems as higher thermal input is
c€/kW
required for the turbine. The three simulated cases are shown in Table 2 for Spain (wet), Jodhpur (wet), and
Jodhpur (dry) with their annual values. In addition, Table 2 also compares the cooling options at each site.

Water requirements

Reducing water consumption is an expected and important outcome of the present work by considering the dry
cooling option. As indicated by the simulation, it is observed that for the proposed CSP plant with dry cooling,
the water consumption is lowered by 85 % (571365 m3/a for wet cooled to 74793 m3/a for dry cooled) as
compared with the wet cooled plant of same output capacity. This, in turn, reduced the water costs from
285,682.5 € per year in the wet cooling to 37,396.5 €/a for dry cooled CSP (water charges: 0.5 €/m3) (see Tables
3). The average monthly savings in water consumption are shown in Table 6 for wet cooled power block option
compared to a dry cooled. The data in the table were obtained as detailed in [14]; which arrived at 3.5 m3/MWh
for wet cooling and 0.5 m3/MWh for dry cooling system.

Table 3. Water savings with dry cooling plant in Jodhpur India

Month Wet cooled type (m3) Dry cooled type (m3) Savings (%)
Jan 43253 5719.5 0.867766
Feb 46371.5 6089.5 0.86868
Mar 63322 8483.5 0.866026
Apr 59871 7803.5 0.869661
May 59199 7595.5 0.871695
June 52213 6754.5 0.870636
July 38031 4963 0.869501
Aug 37296 4884.5 0.869034
Sep 49416.5 6453 0.869416
Oct 52122 6829.5 0.868971
Nov 37509.5 4914.5 0.86898
Dec 32760 4303 0.868651

6. CONCLUSIONS
Based on the technical simulation and comparison of Jodhpur plant with the reference plant at Andasol, the
following conclusions were made,
 Shifting the reference plant to Jodhpur in north-west India and by replacing it with a dry cooling system
would give 11.3 % higher energy yield resulting in a 32 % reduction of LCOE (11.73 €c/kWhe).
 Considering the technical and economic viability of dry cooled CSP units in the arid region of Jodhpur,
it is observed that the plant would affirm its realizable ability based on the extant cost of power

663
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

generation. Moreover, such plants are superior on all fronts compared to plants with wet cooling systems
in other regions of low DNI.
 Contrasting the reference plant in Spain with a dry cooled unit in Jodhpur, the expenditure and utility of
water in Jodhpur has decreased by 85 % with a simultaneous reduction in LCOE by 22 % at a 2 % gain
in energy yield.

The study concludes that Jodhpur with its high DNI is an attractive destination for CSP Plants and can consider
dry cooled units for its future projects as it offers an agreeable techno-economic viability.

Acknowledgements

Many thanks to the German Aerospace Center (DLR) for organizing 'Capacity Building Course' in the domain
of Concentrated Solar Thermal (CST) at NETRA-NTPC, India, and software available to conduct this work.

References

[1] Scolan, S., Serra. S., Sochard, S., Delmas, P., Reneaume, J.M., 2020. Dynamic optimization of the operation of a solar
thermal plant. Solar Energy 198, 643–657.
[2] Rathore, P.K.S., Rathore, S., Singh, R.P., Agnihotri, S., 2018. Solar power utility sector in india: Challenges and
opportunities, Renewable and Sustainable Energy Reviews 81, 2703-2713.
[3] Alhaj, M., Al-Ghamdi, S.G., 2019. Why is powering thermal desalination with concentrated solar power expensive?
assessing economic feasibility and market commercialization barriers. Solar Energy 189, 480–490.
[4] Ma, Z., Bao, H., Roskilly, A.P., 2018. Feasibility study of seasonal solar thermal energy storage in domestic dwellings in
the UK. Solar Energy 162, 489–499.
[5] Purohit, I., Purohit, P., 2019. Solar Thermal power Generation. Earth Systems and Environmental Sciences.
[6] Purohit, I., Purohit, P., 2017. Technical and economic potential of concentrating solar thermal power generation in India,
Renewable and Sustainable Energy Reviews 78, 648–667.
[7] Purohit, I., Purohit, P., Shekhar, S., 2013. Evaluating the potential of concentrating solar power generation in
Northwestern India. Energy Policy 62, 157–175.
[8] Liqreina, A., Qoaider, L., 2014. Dry cooling of concentrating solar power (CSP) plants. An economic competitive option
for the desert regions of the MENA region. Solar Energy 103, 417–424.
[9] Qoaider, L., Liqreina, A., 2015. Optimization of dry cooled parabolic trough (CSP) plants for the desert regions of the
Middle East and North Africa (MENA). Solar Energy 122, 976–985.
[10] Quaschning, V., Kistner, R., Ortmanns, W., 2002. Influence of direct normal irradiance variation on the optimal parabolic
trough field size: A problem solved with technical and economical simulations. ASME J. of Solar Energy Engineering
124, 160-164.
[11] Turchi, D., 2010. Parabolic Trough Reference Plant for Cost Modeling with the Solar Advisor Model (SAM). National
Renewable Energy Laboratory, Golden.
[12] Schenk, H., Eck, M., 2012. Yield analysis for parabolic trough solar thermal power plants – a basic approach. DLR-
enerMENA project.
[13] Dersch, J., Dieckmann, S., 2015. Techno-Economic Evaluation of Renewable Energy Projects using the Software
GREENIUS. Int. J. of Thermal & Environmental Engineering 10, 17-24.
[14] Herrmann, U., Geyer, M., Kistner, R., 2002. The Andasol project- Workshop on Thermal Storage for Trough Power
systems. Solar Millennium AG; FLABEG Solar International GmbH.

664
 POWER MANAGEMENT SYSTEMS FOR AN INTELLIGENT
HOME

Nabeel Shaway Shyaa Al–Atwan

Southern Technical University, Faculty of Engineering Technology, Basra, Iraq, [email protected], ORCID:
0000-0001-8274-0763

Mircea Nitulescu
University of Craiova, Department of Mechatronics and Robotics, Craiova, Romania, [email protected],
ORCID: 0000-0002-4702-3328

Nabeel Shaway Shyaa, Al–Atwan, Mircea, Nitulescu. Power management systems for an
Cite this paper as:
intelligent home. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: This paper is focus on an intelligent home system which increase the automation capability of the
home. The proposed system was applied and tested on a classical real home. The target is the
power consumption reduction and different energy parameters monitoring (current, voltage, real
power, reactive power, and apparent power). Also, in this paper, an automatic detection of the
power factor and its associated correction system is designed and implemented to eliminate or
reduce reactive energy consumption. This was done using a specialized automatic power factor
correction based on two microcontrollers which finally control a capacitor banks, which can add
an appropriate capacitive load to balance the inductive load at home and to improve the power
factor in the intelligent home. Several experiments have been analyzed, and the results are included
in the paper. The designed system is also provided with a display system for immediate reading of
the main energy values monitored in the smart home.
Keywords: energy management system, intelligent home, home automation
© 2020 Published by ECRES

1. INTRODUCTION

Increasing electricity consumption is a current reality today, both in the industrial and domestic environment.
To build an intelligent home, monitoring energy consumption is an extremely important task, along with many
others, which can bring economic benefits to the owner but also contribute to reducing overall electricity
consumption, one of the major goals of today. The increase in electricity consumption in the industrial field, but
also in the domestic one as well, is generally due to the connection of many inductive consumers of electricity.
The best known in this category (but not only) is the electric motor, under various construction variants, present
in many industrial or domestic applications. Excessive inductive consumers affect the so-called "consumer
power factor", with effects on energy costs. So, the presence of many inductive consumers in an intelligent home
(electric motors, transformers, long power lines with their own inductance, certain types of electric lamps with
gas discharges, etc., switched on and off in a random manner, depending on needs, is an important reason for
the permanent monitoring and control of the power factor attached to the energy consumption of the house in
order to maintain it within certain desired limits. A low power factor gives rise to many technical problems for
an intelligent home, such as: high apparent power consumption of the house, increased active power
consumption, the need to oversize the conductors in the house's mains, oversized general power supply
transformer (therefore, more expensive), higher consumption of materials needed to make the electrical wiring
of the intelligent home (copper, for example) and thus increase the value of the initial investment in the house,
increase network voltage losses, poor regulation of voltage and fluctuations in the house, and so on. Basically,
the improvement of the power factor can be done in two ways: by organizational measures (with which the
power factor is increased without any investment) or by producing reactive power on the spot (by using
specialized sources of reactive power. In general, solutions are available that involve high costs, or slightly
cheaper but also less efficient methods [1]. The method developed in this paper controls the reactive power of

665
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

an intelligent home. A less expensive method was desired compared to other methods [2] but which would
remain very effective as well. The general idea is quite well known: the aim is to introduce an adequate electrical
capacity through a group of capacitors so that the power factor of the house is adjusted as close as possible to
the unit. The capacitors introduced in the energy system of on intelligent home will provide a compensation type
function. They will consume a capacitive reactive power necessary to compensate for the inductive reactive
consumption of an intelligent home. The system is based on an architecture designed with two microcontrollers
that ensure the automatic adjustment of the power factor of the intelligent home whenever it falls below a specific
level preset by the user [3].

2. SOME THEORETICAL ASPECTS

The power factor (PF or cos∅) of an alternating current

(AC) electrical system is defined in electrical engineering
theory as the ratio between the actual power (P) absorbed
by the load and the apparent power (S) consumed in the
same circuit. This is illustrated in the power triangle in
Figure 1. The actual power (P) is also called the active
power of the AC electrical circuit. The power factor can
also be interpreted as the degree of utilization of the actual
Figure 1. Power triangle (leading power factor case). power compared to the total power supplied. It is a number
that theoretically can vary in the closed range [−1, 1]. As
can be seen from Figure 1, the apparent power may be greater than the actual power. This is due to the energy
stored in the charge and returned to the source or to a nonlinear charge that distorts the waveform of the current
extracted from the source. For example, if the power factor is cos ∅= 0,5, only 50% of the apparent power
absorbed is used as active power. A negative power factor only occurs when the device (which is normally the
load) generates power, which then flows back to the source. Power factor correction aims to increase a
consumer's power factor, improving the efficiency of the distribution system to which it is attached [5], [6]. The
devices used to correct the power factor can be located either at a central station, or spread along the electricity
distribution system, or incorporated into energy-consuming equipment [7], [8].
Controlling electricity consumption is a goal of the intelligent home. For this purpose, an advantageous power
factor, i.e. as high as possible, presupposes a reactive power consumed as low as possible. And if many
household consumers have an inductive behavior, the solution is to control the acceptable limits for the reactive
energy consumed. The reactive power of high values in a house leads to high energy bill costs that often penalize
this type of consumption, the need to oversize the electrical circuits in the house, including the general power
transformer of the house, and therefore increased consumption of relatively expensive materials.

3. SYSTEM FOR AUTOMATIC CORRECTION OF THE POWER FACTOR

We will assume that an intelligent home has a total inductive load whose power factor is to be corrected. The
proposed system for automatic power factor correction (APFC) in an intelligent home uses a microcontroller as
a central element. It provides the necessary adjustment of this parameter. Its task is ultimately to manage a
controlled process of proper connection of a group of capacitors that will improve the power factor by the value
of the capacitive reactance that is introduced by them. The APFC equipment is connected to the main power
supply of the intelligent home, so close to the variable load of the house. In the tested minimum form, the APFC
equipment contains a group of capacitors that can be connected or not, relays connected in series with the
capacitors and the APFC system controller. For experiments, three capacitors were used in the APFC structure.
The task of the controller is to read the power factor, voltage and current of the system. Then, according to a
known algorithm, the controller decides how many capacitors should be inserted to bring the power factor to the
desired value. Thereby the main supply is being compensated with shunt technique. The main equipment tha t
composed the APFC module are reactive power elements (capacitors), switching elements (relays), contactors
connected in series with the capacitor (for the protection of the relay in case of high current consumption)
specialized interface circuits (voltage, current and factor power) and the controller itself. The APFC unit consists
of the following components: capacitor banks, 4x20 LCD display, two microcontrollers, main supply, current
transformer, sensing current circuit, potential transformer, sensing voltage circuit, contactors, and relays. Our
system takes 220V supply and steps down the voltage to 9V through a potential transformer, then converts this

666
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

9V AC into the less 5V input to the microcontroller. A signal of current is also obtained from the main power
supply by a current transformer (CT) and processed by a current sensor circuit for microcontroller input. The
microcontroller performs power factor calculations and controls capacitors. The results are displayed on a 4x20
liquid crystal display. The block diagram of the developed APFC system is shown in the following Figure 2(a).
The main elements that appear in the block diagram have the following characteristics:
Voltage
L
Transformer
220V Current
Transformer Load
50HZ
N

Transformer Transformer Capacitor

Power
Supply Unit
Circuit Circuit Bank
5V

Conductors

First Second
Microcon Microcon- Relay
troller troller Shield
20 x 4 LCD
Display

(a) (b) (c)

(d) (e)
Figure 2. (a) The diagram of the APFC proposed system, (b) The potential transformer, general aspect, (c) The curent
transformer, general aspect, (d) Voltage sensing circuit diagram, (d) The circuit for current sensing.

The Potential Transformer

The potential transformer (PT) is shown in Figure 2(b). It transforms the mains voltage of house (220V, AC) to
9V AC. To be able to read and interpret the instantaneous value of the house supply voltage, the value found
must be compatible with the accepted range on the analog inputs of the microcontroller. It is therefore necessary
to convert from 9V AC to the range 0÷5V. A root mean square (RMS) voltage measurement is required to
determine the power factor (PF), the real power (P), and the apparent power (S). This measurement can be
completed by utilizing a PT to scale down and separate of the main high voltage. A signal conditioning circuit
is used to adapt the signal obtained from the PT output. The positive peak of a 9V AC is +12.727V and its
negative peak is -12.727V. Respecting the specifications of the microcontroller analog inputs, the signal
conditioning needs to convert the transformer output to signal ranged from 0V to 5V i.e. the positive peak < 5V
and the negative peak > 0V. The sinusoidal waveform can be scaled down using a voltage divider connected
crossways the adapter's terminals. The shifting the signal level is achieved by adding to its output of another
voltage divider connected across the microcontroller’s power supply. Figure 2(d) shows voltage sensing circuit
diagram. Resistors R1 and R2 form a voltage divider that scales down the output AC voltage of the PT. Resistors
R3 and R4 deliver the voltage bias. The capacitor C1 provides a low impedance path to ground for the AC signal.
Its value can be chosen between 10μF ÷ 100μF. Resistors R1 and R2 will be chosen to give a peak voltage output
of ~1V. For an AC to AC converter with a 9V RMS output, the resistor combination chosen is R1=10KΩ and
R2=100KΩ. So, the peak output voltage (corresponding to the peak input voltage of 12.727V) will be 1.156V.
The voltage bias providing by R3 and R4 should be half of the second microcontroller supply voltage. As such,
R3 and R4 need to be of equal resistance (R3=R4=470KΩ). If the second microcontroller is running at 5V, the
resultant sinusoidal waveform has a positive peak of 2.5V + 1.156V = 3.656V, and respectively negative peak
of 1.344V, satisfying the microcontroller input voltage requirements. Higher resistances produce a lower APFC
energy consumption (as it is intended to run on batteries for many months).

Current Transformer
In our proposed system, we used the current transformer (CT) with a ferrite material and input/output ratio about
2000:1, model YHDC, from Figure 2(c). The input current range is 0÷100A, which gives output current of
0÷50mA. The input frequency range can be 50Hz÷150KHz and the device works at a range temperature -250C
to +700C.

667
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Circuit for Current Sensing

To join a current transformer sensor to the second microcontroller, see Figure 2(e), the output signal from the
current transformer (CT) needs to be conditioned, so it meets the input requirements of the microcontroller
analog inputs. If the CT is a "current output" type, the current signal needs to change to a voltage signal with a
burden resistor. If it is a "voltage output" type, then this step can be skipped because the load resistor is attached
inside. The CT has an input current range of 0A ÷ 100A. Let us select the value of 100A as the maximum current
accepted at the power supply of the intelligent home. For this value, the maximum peak current will be √2
higher, so 141.4A. The peak current in the secondary coil (CT ratio 2000:1) will be 2000 times smaller, so
0.0707A. To respect the resolution, the voltage across the burden resistor at peak current should be one half of
the microcontroller analog reference voltage (2.5V), so the ideal burden resistance is 35.36 Ω. If we connect one
of the CT wires to ground and measure the voltage of the second wire, the voltage will vary from positive to
negative concerning ground. However, the microcontroller analog inputs require only a positive voltage. By
connecting the CT lead not to the ground but to a source of half the supply voltage instead, the CT output voltage
will now swing above and below 2.5V thus remaining positive. Resistors R1 and R2, from Figure 2(e) are a
voltage divider that delivers the 2.5V. The capacitor C1 (100µF) has a little reactance and provides a path for
the AC current to bypass the resistor.

Capacitor Bank
The capacitor bank is the collection of different values capacitors. Theoretically, parallel and/or series
combinations of various capacitors types provide a range of capacitance required to compensate the reduced
power factor of the intelligent home. The sizing of capacitors is determined based on the required reactive power
demand by the load network. In our proposed system and experiments we used three capacitors of 5μF, 10μF
and 20μF.

4. AUTOMATED POWER FACTOR CALCULATION

To know the value of power factor of a house must be calculated the delay between the voltage and the current.
This is done by the first microcontroller. Then, the value is transfer to the second microcontroller, where is
calculate the power factor. To can read the sinusoidal wave of the voltage and of the current by the first
microcontroller, it is necessary to convert these sinusoidal waves to square waves. Figure 3(a) shows the
implemented voltage circuit for transforming the voltage sine wave into a corresponding square wave signal,
while Figure 3(b) shows, similarly, the implemented current circuit for transforming the current sine wave into
a corresponding square wave signal. These solutions for establish the power factor has been many times
simulated (using Electronics Workbench Multisim and Proteus software) and then tested to prove their
functionality. PT circuit has been tested for the input and the output to show that the main supplied voltage of
the home has been stepped down to the desired voltage.

(a) (b)
Figure 3. (a) Converter of the sinusoidal voltage signal to a square wave, (b) Converter of the sinusoidal
current signal to a square wave.

In addition to the power factor value, the first microcontroller also sends a logic signal to inform the second
microcontroller about the nature of the power factor (lead or lag). The software developed for calculating the
state of power factor is summarized in the flowchart shown in Figure 4. The reactive power introduced by a
capacitor (Qc) can be calculated by the well-known relation: Qc=(2πfC)×V2, where V is the home voltage
(220V) and f are the frequency of the power system used by the home (50Hz). So, for the three capacitors C of
5μF, 10μF and 20μF, the individual reactive power values that can be compensate are respectively 76VAR,
152VAR and 304VAR.

668
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

5. SOFTWARE DEVELOPMENTS

This paragraph presents the algorithm of APFC module based to microcontrollers. As inputs, the process needs
to receive the values for power factor, current and voltage at the end of the main supply used by the house. The
algorithm is summarized in the flowchart illustrated in Figure 5, where, P = real power [W], S = apparent power
[KVA], Q = reactive power [VAR], Irms = current [A] and Vrms = voltage [V].

Find current Find voltage

raise edge raise edge

N N
Raising =1? Raising =1?

Ti = T Tv = T

Td = Tv - Ti

N
Td < 0 ?

Y
Power factor Power factor
is Lead is Lag

Sent power factor state to the

second microcontroller

Figure 4. In the first state power factor (lag or lead Figure 5. Flowchart of proposed system
microcontroller). (second microcontroller).

6. EXPERIMENTAL SET UP OF THE SYSTEM

The basic components that have been used in different experiments to complete this work are the source of
electricity supply, different variable inductive loads, the bank of capacitors, interfacing circuits, switching
circuits and two microcontrollers. Figures 6 show the complete experimental construction of the automatic
power factor correction (APFC) installed on the power supply of our intelligent home. So, Figure 6(a) present a
general image of the complete developed APFC module and the entire energy monitoring system. Figure 6(b)
presents the microcontroller circuit ATMEGA 328. Figure 6(c) presents the diagram of the current circuit, while
Figure 6(d) presents the diagram of the voltage circuit.
Table 1 provides the data recorded from an APFC home system. During the tests, the electrical loads used for
experiments in the intelligent home where: fluorescent light sources (FL), ceiling fans in the rooms, a
refrigerator, the water pump motor, and a television set (TV). As shown in Table I, we note that the value of the
power factor before the addition of the capacitive reactance in our experiments was low, i.e. less than 0.9, while
at the addition of the capacitive reactance we note that the power factor value is close to 1.
Besides automatically correcting the power factor of the home electrical load, the designed system also performs
power monitoring and displaying the results for the intelligent home. For this purpose, an integrated 4 x 20 LCD
shows permanently the electrical status of the home. It shows to the user the instant consumption for the real
power (P), the apparent power (S), the reactive power (Q), the power factor (PF), the input voltage (Vrms) and
the current flow (Irms). Figure 7(a) and Figure 7(b) present images of this display system and respectively
different values of monitoring parameters. So, the condition of the home's load before activating the APFC
module is shown in Figure 7(a), while, for the same consumers integrated into the home, Figure 7(b) presents
the condition of the home's load after activating the APFC module.

CONCLUSION
Automatic power factor correction techniques described here for an intelligent home can be also applied in
industry, commercial lines, and general power distribution system to increase stability and efficiency of the
system. Our solution was done using a based microcontrollers technique to control a capacitor banks, which can

669
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

add an appropriate capacitive load to balance the inductive load at home. Care must be taken into account to
ensure that the capacitors are not subjected to fast connection / disconnection connections, as the occurrence of
over voltages leads to a considerable decrease in the life of the capacitor bank used by the APFC device. This
APFC device helps to pull down the high current of an electrical system, reduced power consumption and so to
reduce charges on utility bills.

Table I. The experimental results of APFC

on the real home.

(a) (b)

(c)

(d)
Figure 6. (a) An image of the complete APFC module and the energy
monitoring system, (b) Microcontroller circuit.
(c) Current circuit, (d) Voltage circuit.

(a) (b)
Figure 7. (a) The condition of the home's load before activating the APFC module current signal to a square wave,
(b) The condition of the home's load after activating the APFC module (for the same consumers).

REFERENCES

[1] Ware, J. Power Factor Correction (PFC), IEE Wiring Matters, pp. 22–24, 2006.
[2] Kabir, Y., Mohsin, Y.M., Khan, M.M. Automated Power Factor Correction and Energy Monitoring System, IEEE,
Energy and Technology, 3rd International Conference, pp. 1-5, 2017.
[3] Cho, J.G., Back, J.W., Yoo, D.W., Lee, H. S. Reduced Conduction Loss Zero-Voltage-Transition Power Factor
Correction Converter with Low Cost, IEEE Transactions on Industrial Electronics., vol. 45, no. 3, pp. 395–400, 1998.

670
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[4] Jiang, Y., Lee, F.C., Hua, G., Tang, W. A Novel Single-Phase Power Factor Correction Scheme, IEEE, Eighth Annual
Applied Power Electronics Conference and Exposition pp. 287–292, 1993.
[5] EL-Sharkawi, M.A., Chen, M., Vadari, S.V., Fissel, G.W., Venkata, S.S. Development and Field Tasting of A Closed
Loop Adaptive Power Factor Controller, IEEE, Transactions on Consumer Electronics, Vol.3, No. 2, pp. 235-240, June
1988.
[6] Nalbant, M.K. Power Factor Calculations and Measurements, IEEE, Fifth Annual Applied Power Electronics
Conference and Exposition, America, New York, p
p. 543–552, 1990.
[7] Tiwari, A.K., Sharma, D., Sharma, V.K. Automatic Power Factor Correction Using Capacitive Bank, Journal of
Engineering Research and Applications, vol. 4, no. 2, pp. 393–395, 2014.
[8] Jamge, S.B., Devanpalli, P.A. Automatic Power Factor Controller Using PSoC3, International Journal of Engineering
Research & Technology (IJERT), vol. 3, no. 5, pp. 1056–1058, 2014.

671
 METAL ORGANIC FRAMEWORK MATERIAL (MOF) FOR
ATMOSPHERIC WATER HARVESTING IN ARID REGIONS

Ibrahim Al-Hashemi
University of Birmingham, Birmingham, United Kingdom, [email protected], ORCID: 0000-0003-
3864-3928

Raya Al-Dadah
University of Birmingham, Birmingham, United Kingdom, [email protected], ORCID: 0000-0001-
9410-1993

Saad Mahmoud
University of Birmingham, Birmingham, United Kingdom, [email protected],ORCID: 0000-0002-
4421-1577

Al-Hashemi, Ibrahim, Al-Dadah, Raya, Mahmoud, Saad. Metal Organic Framework

Cite this paper as: Material (MOF) for Atmospheric Water Harvesting in Arid Regions. 8th Eur. Conf.
Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: Water scarcity is a growing public health concern in arid regions. With the large amount of
water vapor in the atmosphere, atmospheric water harvesting technology which is based on
condensing the water vapor to produce potable water, offers potential in addressing the
water scarcity problem in these regions. Metal Organic Frameworks (MOF) is a new class
of microporous material with exceptional high porosity, uniform pore size, well-defined
molecular adsorption sites and large surface area (up to 5500m2/g). They have high affinity
to water vapor which can be exploited for atmospheric water harvesting. This paper
investigates the atmospheric water vapor adsorption capacities of three MOF materials in
terms of water uptake and kinetics. The performance of CPO-27 (Ni), MIL-100 (Fe),
Aluminum Fumarate MOF materials were studied and the predicted daily water production
was calculated. It was found that CPO-27 (Ni) outperformed the other MOFs where an
amount of 1.37 Lwater kgMOF-1 day-1 could be produced.
Keywords:
© 2020 Published by ECRES

INTRODUCTION

Water scarcity in arid regions has received a considerable attention in the last few years. With 96.5% of
water on earth is contained within oceans [1], seawater desalination technologies such as Reverse
Osmosis (RO) and Multi-Stage Filtration (MSF) have been used to supply freshwater in arid regions.
However, energy, operational, and maintenance costs as well as capital investment are the major
drawbacks that make these mature water purification technologies challenging to implement (2). In
many water stressed landlocked countries (Figure 1) (3), the humidity of the earth’s atmosphere can be
an accessible source to provide freshwater where an equivalent 10% of all surface water is present in
the form of vapor (4). A considerable amount of literature has been published on atmospheric water

672
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August
2020

Figure 1 Water Stress level in the world.

generators (AWGs) such as artificial rain collection (5), fog water collection (6,7) and dew water
collection have been developed and are in operation in villages in more than 25 countries around the
world (2). However, AWGs and fog harvesting systems tend to be impractical, as they are extremely
energy-intensive or require a demand for high relative humidity (typically 100%), which restrict the
geographical areas suitable for their applications (8,9).
Recently, researchers have shown an increased interest in sorbents-based atmospheric water harvesting
to provide freshwater in arid regions. Sorbent materials such as nano porous inorganic materials and
metal organic frameworks (MOFs) were investigated in adsorption heat pumps (AHPs) and atmospheric
water harvesting systems (AWH) to overcome their energy and environmental challenges (10). Porous
materials such as silica gel, activated carbon, zeolite and MOFs can harvest water from the air by
adsorption over a wide range of Relative Humidity (RH) values (11). However, conventional adsorbents
(i.e. Silica gel and Zeolite) have relatively low water vapor uptake or require high regeneration
temperatures. MOFs, which consist of organic and inorganic building blocks, have higher surface area
(up to 5500m2/g) and pore volume which result in a higher water capacity (12-14). MOFs also require
relatively lower regeneration temperatures than conventional adsorbents (15-18).

Kim et al (15) demonstrated a device based on MOF-801 that can harvest water vapor from the ambient
air with low relative humidity (i.e. ~20% RH) and driven by natural sunlight assisted by photothermal
materials. Also, Li et al (16) demonstrated a device that is based on anhydrous salt, that can harvest
water from the ambient air with low relative humidity (i.e. ~15% RH) and could release the water with
a solar flux as low as 0.7kW/m2. The developed devices performed only one Water Harvesting Cycle
(WHC) per day (adsorption at night and desorption during daytime). Thus, a small amount of harvested
water was delivered and limited by the water sorbent capacity of the material. A more recent work has
focused on the kinetics of the adsorbent material rather than its water capacity (12). This approach is
used to develop AWH systems that work continuously through multiple cycles per day, therefore,
delivered higher amounts of water. Hence, for a system to work efficiently and deliver a higher amount
of water per day, sorbents are required to have fast kinetics for water capture and release. This paper
investigates the suitability of three MOF materials namely CPO-27 (Ni), MIL-100 (Fe) and Aluminum
Fumarate for atmospheric water harvesting in terms of water uptake, kinetics and daily water production.

673
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August
2020

MATERILAS INVESTIGATED
In water harvesting applications, MOFs have been identified as promising sorbents due to their high-
water capacity, chemical flexibility and tunability which comes from the principles of reticular
chemistry. CPO-27 (Ni), also known as MOF-74 (Ni) is a good example of such material for the water
harvesting system, with a water uptake of 0.43 (gwater/gMOF at 20% RH (19). In addition, it has an
inflection point at 7% RH which makes it a viable water adsorbent in arid regions. CPO-27 (Ni) is an
isostructural MOF such that the same ligands are used with multiple divalent metal ions (Ni, Mn, Mg,
Zn and Co) (18). Aluminum Fumarate is also a well-known microporous adsorbent which is widely
investigated in adsorption heat pump applications. Aluminum fumarate is known for its hydrothermal
stability and low desorption temperature, with a water uptake of 0.35 (gwater/ gMOF) at 30% RH. While
MIL-100 (Fe), commercially known as Basolite F300, is known as environmentally benign material. It
has a high surface area of about 2800 m2 g-1 and a pore volume of 1.2 cm3g-1 [19]. The above-defined
hydrophilic materials were deemed to have a potential in the water harvesting application, as all of them
have sufficient water uptake under low relative humidity (below 40% RH) as evident by their respective
sorption isotherms (Figure 2).

Figure 2. Water sorption isotherms at 25 ℃ of the water harvester sorbent candidates [19].

METHADOLOGY AND RESULTS

To assess the viability of sorbents for water harvesting, dynamic sorption tests have been conducted for
the above described sorbent materials at constant air humidity of 20%, 30 and 40%. Samples of the
material were prepared in layers of height of 3 mm and packing porosity of 0.7 (ideal value for moisture
transfer (20)). Before each adsorption measurement, the MOF materials were dehydrated at 100 C over
night (12 hours). Then, the MOF material was allowed to cool down naturally at ambient temperature
of 25 ℃ and the water uptake was measured by accurately weighing the MOF material at time step of
30 minutes.

Figure 3 shows the rate of water adsorption of the MOF at different relative humidity values. At 20%
RH, CPO-27 (Ni) reached saturation level of 0.44 after 11 hours. As for MIL-100 (Fe), it reached
saturation of 0.19 after 15 hours while Aluminum Fumarate did not reach the saturation level during the
20 hours of testing. At 30%RH, CPO-27(Ni) reached saturation level at 9 hours. MIL-100 (Fe) reached
its saturation at 12 hours while Aluminum Fumarate reached saturation of 0.35 after 11 hours. At 40%
RH, CPO-27 (Ni) reached saturation at 7 hours, and MIL-100 (Fe) reached saturation at 13 hours while
Aluminum Fumarate reached saturation level of 0.4 after 12 hours.

674
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August
2020

Figure 3. Rate of water adsorption at 25 ℃ and (a)20% RH. (b) 30% RH. (c) 40% RH

To compare the adsorption kinetics of the tested materials, a normalized rate of adsorption R 0, Normalised
(Normalised to 1.0 for CPO-27 (Ni) at 20% RH) was established using Eq. 1 (Table 1). A faster
adsorption rate for all the tested materials was observed with the increase in the relative humidity. At
higher relative humidity (30% and 40%), CPO-27 (Ni) performed better than the other sorbents (1.35
and 1.65, respectively). Whereas, MIL-100 (Fe) and Aluminum Fumarate adsorption rate were found to
be less favorable than CPO-27 (Ni), as the adsorption rate was relatively slow (0.67 and 0.85,
respectively at 40% RH), this was attributed to their respective inflection point which lies at a relatively
high partial vapor pressure (Figure 2).
𝑋𝑖 (1)
𝑅0 = 𝑋
𝐶𝑃𝑂−27 𝑎𝑡 20 𝑅𝐻

Where Xi is the ratio of the water uptake of the material over the corresponding time to reach saturation
(Hours).

675
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August
2020

Table 1. Behavior of Dynamic Water Adsorption Kinetics at Different Relative Humidities.

MOF RH% (AT 25 C) R0, NORMALISED
20 1.00
CPO-27 (NI) 30 1.35
40 1.65
20 0.29
MIL-100 (FE) 30 0.64
40 0.67
20 0.18
ALUMINIUM
30 0.69
FUMARATE
40 0.85

As for the rate of the desorption, initially, the samples were left to reach equilibrium at 25 ℃ and 30%
RH overnight. Afterward, samples were placed in the oven and the rate of desorption was monitored at
a temperature of 80 ℃. At 80 ℃, MIL-100 (Fe) and Aluminum Fumarate revealed a very fast
dehydration of less than 25 minutes (Figure 4a). Whereas, CPO-27 (Ni) required almost 1 hour. Another
desorption rate test was conducted with a temperature of 110 ℃ (Figure 4b). As anticipated, the time
for complete desorption for the sorbents was shortened with the higher temperature. As for the CPO-27
(Ni), a complete dehydration was achieved within 35 minutes.

Figure 4. Rate of water desorption at a) 80 ℃ . b) 110 ℃.

Eq. 2 was used to predict the amount of water that can be produced daily per kg of MOF with tested
materials (Table 2). The predicted water production was calculated for each sample testing condition
(relative humidity and desorption temperature).

Daily Water Production = 𝑵 × 𝛚𝒄 (2)

676
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August
2020

Where N is the number of WHCs that can be achieved per day, and ω𝑐 is the water capacity of the MOF
(kgwater/ kg-1MOF)

At low relative humidity (20%), the water that could be produced is limited for MIL-100 (Fe) to 0.28
Lwater kg-1MOF Day-1, and negligible for Aluminum Fumarate. However, as for the CPO-27 (Ni) at 20% RH,
an amount of 0.85 Lwater kg-1MOF Day-1 could be produced daily in 2 WHCs a day. At higher relative
humidity (40%), Aluminum Fumarate could complete 2 WHCs that would produce up to 0.81 Lwater kg-
1 -1 -1 -1
MOF Day , while MIL-100 (Fe) would be able to produce up to 0.63 Lwater kg MOF Day in 2 WHCs.
Interestingly, CPO-27 (Ni) could complete 3 WHCs (RH = 40%, desorption temperature of 110 ℃) per
day which produce 1.37 Lwater kg-1MOF Day-1.

Table 2. Predicted daily water production of the MOF materials

Desorption temperature 80℃ 110℃

Adsorption RH (%) 20% 30% 40% 20% 30% 40%
(Lwater kg-1MOF Day-1)

CPO-27 0.85 1.08 1.26 0.88 1.12 1.37

Water prodcution

MIL-100 (Fe) 0.28 0.60 0.63 0.28 0.60 0.63

Aluminum Fumarate - 0.71 0.81 - 0.74 0.83

I. CONCLUSIONS
This paper investigated the use of three different MOF materials for atmospheric water harvesting
applications to tackle the issue of water scarcity in arid regions. Published work have shown that for
sorbent-based atmospheric water harvesting systems, materials should not only be optimized for high
water uptakes at low relative humidity but also the kinetics of sorbents should be considered. Indeed,
the findings reported here showed that CPO-27 (Ni), a commercially available MOF with fast adsorption
and desorption kinetics can perform multiple WHCs per day which could produce 1.37 L water kgMOF-1
day-1. This work proves MOFs as a highly promising class of materials to make sorbent-based
atmospheric water harvesting systems feasible technology.

REFERENCES

1- Van der Bruggen B, Vandecasteele C. Distillation vs. membrane filtration: overview of process evolutions in
seawater desalination. Desalination. 2002 Jun 10;143(3):207-18.
2- Pinto FS, Marques RC. Desalination projects economic feasibility: A standardization of cost determinants.
Renewable and Sustainable Energy Reviews. 2017 Oct 1; 78:904-15
3- A Quarter of Humanity Faces Looming Water Crises [Internet]. Nytimes.com. 2020 [cited 20 April 2020].
Available from: https://www.nytimes.com/interactive/2019/08/06/climate/world-water-stress.html.
4- Schneider SH. Encyclopedia of climate and weather. Oxford University Press; 2011 Jun 9.
5- Bruintjes RT. A review of cloud seeding experiments to enhance precipitation and some new prospects. Bulletin
of the American Meteorological Society. 1999 May;80(5):805-20.
6- Zhang L, Wu J, Hedhili MN, Yang X, Wang P. Inkjet printing for direct micropatterning of a superhydrophobic
surface: toward biomimetic fog harvesting surfaces. Journal of Materials Chemistry A. 2015;3(6):2844-52.
7- Wang Y, Zhang L, Wu J, Hedhili MN, Wang P. A facile strategy for the fabrication of a bioinspired hydrophilic–
superhydrophobic patterned surface for highly efficient fog-harvesting. Journal of Materials Chemistry A.
2015;3(37):18963-9.
8- Olivier J, De Rautenbach CJ. The implementation of fog water collection systems in South Africa. Atmospheric
Research. 2002 Sep 1;64(1-4):227-38.
9- Estrela MJ, Valiente JA, Corell D, Millán MM. Fog collection in the western Mediterranean basin (Valencia
region, Spain). Atmospheric Research. 2008 Mar 1;87(3-4):324-37.
10- Shine KP, Sturges WT. CO2 is not the only gas. Science. 2007 Mar 30;315(5820):1804-5.

677
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August
2020

11- Kim H, Rao SR, Kapustin EA, Zhao L, Yang S, Yaghi OM, Wang EN. Adsorption-based atmospheric water
harvesting device for arid climates. Nature communications. 2018 Mar 22;9(1):1-8.
12- Hanikel N, Prévot MS, Fathieh F, Kapustin EA, Lyu H, Wang H, Diercks NJ, Glover TG, Yaghi OM. Rapid
Cycling and Exceptional Yield in a Metal-Organic Framework Water Harvester. ACS central science. 2019
Aug 27;5(10):1699-706.
13- Canivet J, Bonnefoy J, Daniel C, Legrand A, Coasne B, Farrusseng D. Structure–property relationships of
water adsorption in metal–organic frameworks. New Journal of Chemistry. 2014;38(7):3102-11.
14- Plessius R, Kromhout R, Ramos AL, Ferbinteanu M, Mittelmeijer‐Hazeleger MC, Krishna R, Rothenberg G,
Tanase S. Highly selective water adsorption in a lanthanum metal–organic framework. Chemistry–A European
Journal. 2014 Jun 23;20(26):7922-5
15- Kim H, Yang S, Rao SR, Narayanan S, Kapustin EA, Furukawa H, Umans AS, Yaghi OM, Wang EN. Water
harvesting from air with metal-organic frameworks powered by natural sunlight. Science. 2017 Apr
28;356(6336):430-4.
16- Li R, Shi Y, Shi L, Alsaedi M, Wang P. Harvesting water from air: using anhydrous salt with sunlight.
Environmental science & technology. 2018 Apr 2;52(9):5398-406.
17- Tu Y, Wang R, Zhang Y, Wang J. Progress and expectation of atmospheric water harvesting. Joule. 2018 Aug
15;2(8):1452-75
18- Hussein E. Numerical and experimental evaluation of advanced metal-organic framework materials for
adsorption heat pumps (Doctoral dissertation, University of Birmingham).
19- Horcajada P, Surblé S, Serre C, Hong DY, Seo YK, Chang JS, Grenèche JM, Margiolaki I, Férey G. Synthesis
and catalytic properties of MIL-100 (Fe), an iron (III) carboxylate with large pores. Chemical
Communications. 2007(27):2820-2
20- Kim H, Yang S, Rao SR, Narayanan S, Kapustin EA, Furukawa H, Umans AS, Yaghi OM, Wang EN. Water
harvesting from air with metal-organic frameworks powered by natural sunlight. Science. 2017 Apr
28;356(6336):430-4.

678
 THE EFFECT OF CATHODE CAVITIES TO THE MODES OF A
CAVITY MAGNETRON

Yunus Yıldız
Gazi University, Technology Faculty, Department of Electrical and Electronics Engineering, Ankara, Turkey, E-mail:
[email protected], ORCID: 0000-0002-4638-5490

Erol Kurt
Gazi University, Technology Faculty, Department of Electrical and Electronics Engineering, Ankara, Turkey, E-mail:
[email protected], ORCID: 0000-0002-3615-6926

Yıldız, Y., Kurt, E., The Effect of Inner Cavities to the Modes of a Cavity Magnetron, 8th Eur.
Cite this paper as:
Conf. Ren. Energy Sys., 24-25 August 2020, Istanbul, Turkey

Abstract: In this study, a new hole type cavity magnetron design with 8 cavities in both anode and cathode
has been designed and electromagnetic simulation results are reported. Firstly, the design
parameters using the empirical equations are introduced. Then, we performed the simulation
studies by using the finite element method over the certain mesh structures. The simulation study
consists of two parts: Eigen- frequency and charged particle tracing analyses. Eigen-frequency
spectra have been performed for two different designs (i.e. with inner cavities and without) in order
to clarify the effect of cavity cathode structure on the resonance frequencies. Eigen-frequencies
have been explored around the operating frequency, namely 9 GHz. According to the data
obtained, the cavity cathode structure has been shown to affect the resonance frequencies of the
magnetron. Charged particle tracing analysis has been conducted to observe the electron dynamics
under a specific voltage rate. The maximum electron velocities are found as 5.36x108 m/s and
4.58x108 m/s for two structures.
Keywords: Magnetron, cavity, cathode cavity, trajectory
© 2020 Published by ECRES

1. INTRODUCTION
Magnetrons are unavoidable devices for the microwave industry. They play key role for many different
engineering fields, where a certain amplitude and frequency dependent electromagnetic waves are required. One
of the most encountered fields is the fusion energy [1,2]. In fusion plasmas, one requires an exciting microwave
source in order to stabilize the cold regions of the plasma in the contained vacuum region. Of course, the
frequency range can change from radio frequency (RF) to higher frequencies (i.e. microwave) depending on the
oscillations of the particles in the plasma. However, one needs to design the magnetron and its appropriate
waveguide to eject the electromagnetic wave on the particles. Strictly speaking, while the electrons can oscillate
with higher frequencies due to their light masses, protons have low frequency motion compared to them.

First magnetrons have been made by vacuum industry. Historically, there exist two types of microwave electron
vacuum tubes: One of them is microwave linear beam tube (O-type) and the other one is microwave cross field
tube (M-type). Magnetron is a high power microwave cross field device, usually cylindrical, which produces
high frequency oscillations with a magnetic field parallel to its axis [3]. It has been used extensively in various
industrial, scientific and medical applications, as well as in the aerospace, food and communications industries
since the Second World War. In addition, its importance is increasing day by day in applications requiring high
frequency and power. Some of the reasons why magnetrons are getting more and more important are their
efficient, compact, low cost and high output power [4].

Generally speaking, magnetrons consist of structures such as anode, cathode, an interaction region (i.e. region
between anode (A) and cathode (C)), cavities and coaxial coupling. The anode part of the magnetron is

679
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

frequently made of Copper and is generally grounded. It is also the unit, where microwave oscillations take
place. On the other hand, the cathode part is the source of electrons and it is made of metals such as Copper and
Tungsten. It has negative voltage due to the grounding of the anode block. The region between the anode and
the cathode is called the interaction zone where a plasma structure is formed. RF or microwave production takes
place by transferring the energy of electrons to RF or microwave oscillations. The required magnetic field in a
magnetron is generated by a solenoid or external magnet placed parallel to the axis of the device [3]. Thanks to
the magnetic field, electrons are directed in the desired direction inside the interaction zone. The working
frequency of the magnetron is determined by the shape and dimensions of the cavities. Also operating mode of
magnetron is determined by number of cavities and their physical dimensions. Because of the closed structure
of a magnetron resonator system, the total phase shift in one turn must be the integer multiple of 2π [3].

Eq. [1] shows the phase shift between the two adjacent cavities [5]. The power obtained from the magnetron is
transferred from the coaxial coupling outside the structure via an RF probe.
2𝜋𝑛 𝑁
∆𝜃 = 𝑁
, 𝑛 = 1,2, … , 2 , (1)

where N is the number of cavities and n is the mode number.

Factors such as output power, efficiency, quality factor and working frequency determine the performance of
magnetrons. Many studies have been achieved to develop both these factors and the parameters affecting them
from the past to the present [6-11]. On the other hand, although there are no simple design methods related to
magnetron design in the past, studies to eliminate this negative situation are presented today [12,13].

In the present study, two magnetron designs have been provided and a comparison between the results have
been reported. Indeed, one of these designs has 8 cavities in both anode and cathode, whereas other design does
not have a cathode cavity. Therefore we motivate ourselves to find the differences in eigen-frequency spectra
between two designs. In the paper, we have also reported the electric field distributions and electron motions as
well for a concrete result. The design parameters are explained in Section 2. Simulations and results are presented
in Section 3. Finally, Section 4 concludes the study.

2. DESIGN PARAMETERS

In order to determine the dimensions of a magnetron, initially, the operating frequency is determined and then
Eqs. [2-5] are considered to examine the parameters cathode radius (r c), anode inner radius (rai), anode and
cathode height (h), voltage between anode and cathode (Vapp), number of resonators (N), magnetic field applied
between the anode and cathode (Bapp), critical magnetic field (Bc) and critical voltage (Vc) [12].
𝑟𝑐
𝑟𝑎𝑖
≤ 0,5 (2)

𝑟𝑐 𝑁−4
≈ (3)
𝑟𝑎𝑖 𝑁+4

4𝜋𝑓𝑚 1
𝐵𝑐 = 𝑟 2
(4)
𝑛𝑒 (1− 𝑐 2 )
𝑟𝑎𝑖

𝑒 𝑟𝑐 2
𝑉𝑐 = 𝑟𝑎𝑖 2 (1 − ) 𝐵𝑐 2 (5)
8𝑚 𝑟𝑎𝑖 2

Bapp and Vapp are chosen lower than the critical magnetic field and voltage in order to increase the magnetron
efficiency. Since we focus on a magnetron with the operating frequency 9 GHz, the parameters have been
calculated in terms of Table 1. In this study, unlike the normal cathode structure, 8 cavities like the anode are
used in the cylindrical cathode block as the first design (called as D1). The other design without the cathode
cavities is denoted by D2. In the top of Fig. 1(a,b), the upper and three dimensional views of the magnetron are
given, respectively. The mesh view of the magnetron is shown in Fig. 1(c).

680
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)

(c)
Figure 1. (a) Top, (b) 3D and (c) mesh views of the magnetron D1.

(a) (b)

(c)
Figure 2. (a) Top, (b) 3D and (c) mesh views of the magnetron D2

From Table 1, it is obvious that 16 cavities in total has been put into cathode and anode in a circular orientation.
While the desired frequency is 9 GHz, The critical magnetic field has been calculated as 0.18 T. Note that
between cathode and anode, the plasma structure is assumed to be formed. That is the a radial distance of 3 cm.
This distance gives a very large plasma space for the formation of higher modes, too.

Table 1. Design parameters of magnetron

681
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Item no Item Description Value

1 Operating frequency (f) 9 GHz
2 Critical magnetic field (Bc) 0.18 Tesla
3 Critical voltage (Vc) 1.28 MV
4 Anode inner radius (rai) 4.5 cm
5 Anode radius (ra) 6 cm
6 Cathode radius (rc) 1.5 cm
7 Anode and cathode height (h) 3 cm
8 Anode hole radius (rha) 0.5 cm
9 Anode slot width (slotwa) 0.3 cm
10 Anode slot length (slotla) 0.326 cm
11 Cathode hole radius (r hc) 0.3 cm
12 Cathode slot width (slotwc) 0.3 cm
13 Cathode slot length (slotlc) 0.233 cm

Note that Table 1 is valid for also the other design i.e. D2 except the cathode slots. For the magnetron design,
the calculation of governing modes are critical, since the wave frequency is formed by the modes, indeed. It has
been known that the amplitude of the governing mode should be maximized in any design. Therefore the
performance of the design for especially -mode is vital.

3. SIMULATIONS AND RESULTS

3.1. Eigen-Frequency Analysis

In this section, eigen-frequencies from the 3D finite element analysis (FEA) of electromagnetic equations have
been reported. For this, FEA package have been first used to draw the 3D geometry with all components of the
magnet. The proposed mesh numbers have been increased with a thin meshing procedure in order to have better
theoretical results for modes. Both D1 and D2 structures have been analyzed separately. In the geometry,
relevant cylindrical coordinates have been adjusted with the variables of , ϕ and z. Note that they denote the
radial, angular and height component to their plane. In the material part, Copper anode and cathode have been
used. The material in the interaction space (between anode and cathode) is high vacuum with P=10-6 mbar. Note
that we have explored the solutions at the vicinity of f = 9 GHz. For the magnetron cavity, the mesh numbers of
cathode structures w. and w.o. cavity are given as 152702 and 63437 elements, respectively. Maximum and
minimum element sizes are measured as 0.42 cm and 0.018 cm as the adequate accuracy, respectively.
According to the simulations, it is found that the cavity cathode structure affects the resonance frequencies of
the magnetron. Eigen-frequency values of magnetron are presented for cathode structures (with cavity and
without cavity) in similar electric field patterns in Fig. 3(a-h).

682
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)

(c) (d)

(e) (f)

(g) (h)
Figure 3. Electric field distributions for D1 design at (a) 8.7228 GHz, (c) 9.2651 GHz, (e) 9.2729 GHz, (g) 10.232
GHz, and for D2 design at (b) 8.6944 GHz, (d) 9.1649 GHz, (f) 9.1731 GHz, (h) 10.209 GHz.

It is observed that the minimal electric field strength occurs at the cathode cavities at design D1 and elongates
at the lateral regions of the magnetron in radial direction for the first mode as in Fig. 1(a). The same behavior
occurs on the cathode surface of design D2, too. The electrons which are ejected from the cathode will be
directed in angular direction due to the magnetic field and toward the inner surface of anode at ground due to
the negative potential. Note that since the field changes minimum to maximum successively in angular
directions, electrons will have a motion in z direction following the potential hills and valleys, too. The field
solutions at larger frequencies initially rotates the pattern with 22.5 degrees in angular direction, thereby
minimum and maximum field values exchange as shown in Figs. 3(c,d). At this point, we can lead to a result

683
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

that adding cavities to cathode does not affect the pattern orientation. It affects the frequcies as seen in Table 2.
While the frequency of the mode for D1 is 8.72 GHz, the frequency is decreased a little to 8.69 GHz. Thereby,
adding the cavity to the cathode makes a reduction at the frequency within the range 3.25 ‰. A close look to
the electric field strength gives the following maximal values for the geometries with cathode cavity and without:
According to Figs. 3(a,c,g,e), the maximum values of electric fields are 1.77x10 3 V/m, 1.98x103 V/m, 1.9x103
V/m and 2.14x103 V/m for the cathode cavity included geometry, respectively. Without cathode cavity, the
maximum values yield to 1.16x103 V/m, 1.31x103 V/m, 1.26x103 V/m and 1.41x103 V/m, respectively. It is seen
that the geometry with cathode cavity has larger field value, which is important to accelerate the charges. With
a rough calculation, field increases with a factor of 1.52.
Table 2. Eigen frequencies of patterns in Fig. 3.
Number of Patterns D1 D2
1 8.7228 GHz 8.6944 GHz
2 9.2651 GHz 9.1649 GHz
3 9.2729 GHz 9.1731 GHz
4 10.232 GHz 10.209 GHz

3.2. Electron Motion Analysis

In this part, charged particle tracing study is conducted to observe electron motions. In order to compute this
study, the time dependent analysis has been adjusted to start at time t=0 and end t=2.16 ns. Electrons are
accelerated by the electric field and deviated by the magnetic field. The voltage between the anode and cathode
is chosen as -1.2 MV and the electric field was set to -40 MV/m. Also applied magnetic field is chosen 0.099
Tesla in z direction. For this analysis, high mesh number is selected. For D1 and D2, total mesh elements are
1103411 and 651821, respectively. Maximum and minimum element sizes are adjusted as 0.24 cm and 0.0024
cm, respectively. According to the simulation results, the maximum electron velocity is found for D1 as 5.36x108
m/s and for D2 as 4.58x108 m/s. Electron motions and velocity magnitudes at different times can be seen in Figs.
4(a-f) for D1 and Figs. 5(a-f) for D2 designs.

684
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)

(c) (d)

(e) (f)
Figure 4. (a) Electron motion and velocity magnitude at 0.774 ns, (b) Electron motion and velocity magnitude at 0.882
ns, (c) Electron motion and velocity magnitude at 0.9 ns, (d) Electron motion and velocity magnitude at 1.008ns, (e)
Electron motion and velocity magnitude at 1.71ns, (f) Electron motion and velocity magnitude at 2.052 ns.

685
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)

(c) (d)

(e) (f)
Figure 5. (a) Electron motion and velocity magnitude at 0.774 ns, (b) Electron motion and velocity magnitude at 0.882
ns, (c) Electron motion and velocity magnitude at 0.9 ns, (d) Electron motion and velocity magnitude at 1.008ns, (e)
Electron motion and velocity magnitude at 1.71ns, (f) Electron motion and velocity magnitude at 2.052 ns.

4. CONCLUSIONS

Magnetron resonant frequencies and electron motions and electron velocity magnitudes have been investigated
using finite element method for cavity cathode structure (D1) and without cavity cathode structure (D2).
According to the data obtained, D1 has been shown to affect the resonance frequencies and electric fields of the
magnetron. It also affects particle velocities. For the first structure with cathode cavity, the electron mobility
values increase due to large electric fields and give a more spread geometry in the gap. In the second structure,
the electrons behave more compact due to lower electric fields and they are mostly accumulated around the
cathode region with certain rotation. In the first geometry, electrons travel at the lateral regions of the magnetron
at the same time rotate around the axis.

REFERENCES

[1] Dursun, B., Kurt, E., Tekerek, M., A power circuit design for the poloidal field coils in a torus shaped plasma system,
Journal of Energy Systems, 2019, 3(3); 123-128, DOI: 10.30521/jes.609667
[2] Dursun, B. and Kurt, E. Many-body solution to the D2 gas filled inertial electrostatic confinement device. International
Journal of Hydrogen Energy 2016, 41(29), 12546-12554.
[3] Collins, G. B. Microwave Magnetrons. New York : McGraw Hill Book Company, Inc., 1948.

686
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[4] Vyas, S. K., Maurya, S. and Singh, V. P. Efficiency enhancement of cw magnetron by ferrite material filling. IEEE
Transactions on Plasma Science 2016 , 44(12) , 3262-3267. DOI : 10.1109/TPS.2016.2625339
[5] Benford, J., Swegle, J. A. and Schamiloglu, E. Boca Raton, FL: CRC Press, 2016.
[6] Yang, W., Dong, Z., Yang, Y. and Dong, Y. Numerical investigation of the relativistic magnetron using a novel
semitransparent cathode. IEEE Transactions on Plasma Science 2014 , 42(11) , 3458-3464. DOI :
10.1109/TPS.2014.2359434
[7] Wang, X., Fan, Y., Shi, D. and Shu, T. A high-efficiency relativistic magnetron with the filled dielectric. Physics of
Plasmas 2016, 23(7), 073103-1 – 073103-4. DOI: 10.1063/1.4956460
[8] Dingh, S., Zangh, R., Zhou, H. and Wang, B. Impact of magnetic field distribution on performance of 18-vane 5,8 GHz
magnetron. Journal of Electromagnetic Waves and Applications 2016, 30(7), 880-888. DOI:
10.1080/09205071.2016.1164089
[9] Choi, J. J., Lee, H. S., Jang, K. H., Sim, S. H. and Choi, H. S. Development of a 14-vane, double-strapped, 5,8-GHz
magnetron oscillator. Journal of the Korean Physical Society 2016, 69(4), 632-639. DOI: 10.3938/jkps.69.632
[10] Vyas, S. K. and Tiwari, T. Study of hole and slot type S-band pulsed power magnetron. Journal of Electromangetic
Waves and Applications 2017, 32(8), 1008-1017. DOI: 10.1080/09205071.2017.1412358
[11] Li, L., Aranganadin, K., Hsu, H. Y. and Lin, M. C. Design and development of field emission based magnetron for
industrial applications using conformal finite-difference time-domain particle-in-cell simulations. Journal of Vacuum
Science & Technology B 2020, 38(2), 023205-1 – 023205-5. DOI: 10.1116/1.5140723
[12] İsenlik, T. and Yeğin, K. Tutorial on the design of hole-slot-type cavity magnetron using CST particle studio. IEEE
Transactions on Plasma Science 2013, 41(2), 296-304. DOI: 10.1109/TPS.2012.2235466
[13] Vyas, S. K., Maurya, S., Verma, R. K. and Singh, V. P. Synthesis and simulation studies of a 10-kW 2.45-GHz cw
magnetron. IEEE Transactions on Plasma Science 2015, 43(10), 3615-3619. DOI: 10.1109/TPS.2015.2475182

687
 OVERVIEW OF PHOTO-BIOREACTORS USING
COMPUTATIONAL FLUID DYNAMICS TECHNIQUE

Yucel Necla
Department of Nanoscience and Nanoengineering, İstanbul Medeniyet University, Istanbul, 34700, Turkey, ORCID:
0000-0003-0093-2521

Koten Hasan
Istanbul Medeniyet University, Mechanical Engineering Department, Istanbul, 34700, Turkey, ORCID: 0000-0002-
1907-9420

Kılıcaslan Bulent Serkan

Department of Nanoscience and Nanoengineering, İstanbul Medeniyet University, Istanbul, 34700, Turkey,
ORCID: 0000-0002-0842-3636

Yucel N., Koten, H., Kılıcaslan B. S., Overview of Photo-Bioreactors using Computational
Cite this paper as: Fluid Dynamics Technique. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,
Turkey

Abstract: Phototrophic microalgae cultures have an important place among biomass raw materials.
Photobioreactors that offer suitable habitat for growing microalgae are divided into two as open
pools and closed systems. Computational Fluid Dynamics (CFD) technology is a powerful tool for
simulating analyzing, evaluating and optimizing of these photobioreactor structures. The quality
of the solutions obtained in simulation is acceptable and offers important advantages in evaluating
the design and performance of photobioreactors. The main aim of this research is to reveal the
studies on improving the designs and optimization of computational fluid dynamics (CFD)
photobioreactors and to focus on CFD's optimization and performance evaluation applications of
photobioreactors. In this study, two kind of photobioreactor structures were investigated and
reported.
Keywords: Microalgae, Biofuel, Photobioreactors, Computational Fluid Dynamics (CFD)
© 2020 Published by ECRES

1. INTRODUCTION

Microalgae are eukaryotic or prokaryotic, single-celled or multi-celled creatures that can grow rapidly and live
in harsh conditions. Microalgae are photosynthetic organisms grown in open and closed photobioreactors
(PBRs) that use light energy to convert carbon dioxide gas into oxygen and biomass [1]. These creatures release
high-energy fatty acids that they contain in their body and are used in biofuel production [2, 3]. Algae-based
biofuels do not contain sulfur and reduce CO 2, hydrocarbon, sulfur oxide, and nitrous oxide emissions [4].
The most important advantage of biofuel resources, which are renewable energy reserves, is that they are
environmentally friendly and therefore do not pose a threat to global warming. There is a focus on these
resources in developed and developing countries. Biofuels are produced in Asian and European countries,
primarily in the United States [5], and this fuel obtained from microalgae biomass is used in the industry.
In computer-aided simulation studies, hydrodynamic characterization and optimization of photobioreactors are
performed by using computational fluid dynamics (CFD) technique. CFD provides the analysis of basic equation
systems such as continuity equations, transport equations in a three-dimensional system [6].
The main aim of this research is to reveal the studies on improving the designs and optimization of computational
fluid dynamics (CFD) photobioreactors. In algae planting systems, a light source is created using sunlight,
artificial light, or a combination of both [7]. Photobioreactors are ideal systems used to produce large amounts

688
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

of biomass from microalgae. These structures are divided into open pools with large lighting surface areas and
closed systems where the light source can be applied in a controlled and maximum manner [8].

2. BACKGROUND and METHODOLGY

Open channels
The oldest known culture medium in microalgae cultivation is open channels (Figure 1). Open systems can be
classified as natural waters, pools, lakes, rivers, or artificial pools [7]. One of the main advantages of outdoor
pools is that they are easier to build and cost less than most indoor systems. However, there are some restrictions
on open culture pools. Some limitations such as the obstacles during the use of the light source in these channels,
low mass-transfer rate due to inadequate mixing mechanisms, diffusion of CO 2 gas with air and water losses
may decrease biomass yield [7].

Figure 1: Open ponds [9]

Open pools are generally referred to as channel pools (raceway ponds, RWP) in the literature, and mixing is
mostly done through paddles or chambers [10, 11, 12].
In the CFD modeling study conducted by Sawant et al. [10], the effect of the duct length/depth (L / W) ratio on
shear stress in the open photobioreactors was investigated.
In modeling, turbulent kinetic energy (k) was calculated from the equation below:
𝜕 𝑘2 1 𝑘2
(𝜌𝑘) + 𝛻(𝜌𝑢𝑘) − 𝛻 ((𝜇 + 𝜌𝐶𝜇 ) 𝛻𝑘) = 𝜌𝐶𝜇 (𝛻𝑢 + (𝛻𝑢 )𝑇 )2 − 𝜌𝜀 (1)
𝜕𝑡 𝜀 2 𝜀
The rate of dissipation of turbulent kinetic energy was calculated with this equation:
𝜕 𝑘2 1 𝜀2
(𝜌𝜀) + 𝛻(𝜌𝑢𝜀) − 𝛻 ((𝜇 + 𝜌𝐶𝜇 ) 𝛻𝜀) = 𝜌𝐶𝜀1 𝑘(𝛻𝑢 + (𝛻𝑢 )𝑇 )2 − 𝜌𝐶𝜀2 (2)
𝜕𝑡 𝜀 2 𝑘

Using the ANSYS CFX 14 software and the k- model the constant shaking speed for 3 different values of L /
W ratio was simulated (Figure 2).

Figure 2: Speed contours of the L / W ratio at various values [10]

689
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

According to the CFD modeling result, while a heterogeneous velocity distribution was observed in the channel
at a low-value L / W ratio, as the L / W ratio value increased, homogeneity increased in the distribution [10].
Closed systems
Closed photobioreactors are advantageous in many ways compared to open channels. For example, it is possible
to grow most algae species in closed systems, while in open pools it is possible to grow only a limited number
of algae species with a high tolerance to contamination. While a certain standardization of parameters such as
nutrients, CO2 ratio, temperature, light source, turbulence, and environmental conditions required in the
cultivation of microalgae in closed systems can be achieved, this is not possible in open systems. Besides,
biomass efficiency of closed photobioreactors is higher than in open pools [7]. However, parameters such as
difficult to clean closed photobioreactor systems, continuous controls to ensure optimum conditions, and
excessive oxygen produced by the culture medium reduces growth are the disadvantages of these reactors [13].
Taking into consideration some features of microalgae strains, flat panel, tubular (vertical or horizontal), airlift
(airlift) or their modified configurations are closed photobioreactor types used in algal biofuel production.

Flat panel photobioreactors

Flat-panel photobioreactors are reactors with a large surface area illuminated by a light source and designed for
mass production of photosynthetic microorganisms with low energy requirements (Figure 3) [14, 15].

Figure 3: Flat panel photobioreactor from Arizona Center for Algae Technology and Innovation (AzCATI) [16]

In flat-panel photobioreactors, optimization studies of culture conditions by using computational fluid dynamics
(CFD), modification of the photobioreactor, and modeling some parameters (gas profile, flow rate, etc.) in the
reactor have been carried out. In the study presented by Guler et al. [6], the culture conditions of flat panel
photobioreactors were examined with CFD simulation.

690
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 4: Gas distribution contours in flat panel photobioreactors, a. Air volume fraction; b. Gas holdup image; c. Air velocity vectors
[6]

According to the simulation images of the study (Figure 4), air bubbles were concentrated in the area between
dissolved oxygen and pH probes, and the air volume fraction ranged from 2% to 12.5% for flow conditions. In
most sections of the reactor, the local gas volume fraction was 8% excluding the highest values throughout the
sparger nozzles, and this value was reported to be compatible with current studies. When the gas volume fraction
and the air velocity contours are compared among themselves (Figures 6a and 6c), it has been observed that they
exhibit a parallel behavior with each other. In Figure 6c, it is concluded that strong local circulation occurs in
regions marked with red circles, and the diffusion rate adversely affects mixing efficiency [6].

Airlift Photobioreactors
Air-lift photobioreactors (Figure 5) consist of a tank in which the “riser” region where the gas injection is
performed and the “downcomer” region where the gas injection is not performed [17]. Due to the buoyancy of
the liquid, the bubbles rise and cause a movement in the liquid. As these bubbles formed due to the gas introduced
into the system rise in water, the gas dissolves in liquid [18]. In short, the movement of the fluid is governed by
the gas distribution delivered to the reactor [19].

Figure 5: Airlift photobioreactor profile

691
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Optimization studies of air-lift photobioreactors were done using computational fluid dynamics (CFD).
Optimization studies with this method can save both economically and time. Ebrahimifakhar et al. [18]
investigated the effect of a draft tube on reactor performance in an air-lift photobioreactor using CFD. The
circulation rate at various times (Figure 6- I- II) and the distance between the reactor wall and the draft tube
(concentric) were simulated.

Figure 6: Velocity fractions of gas at different times, the distance between the draft tube and the wall (I) 0.02 m; (II) 0.032 m; liquid
speed profiles of two reactors. The distance between the draft tube and the wall (III-a) 0.02m; (III-b) 0.032 m [17]

The gas, which was homogenously introduced into the system, formed air bubbles in the reactor and moved
upwards. After a while, the fluid movement was stable.
It was observed that the circulation speed was high when the distance between the draft tube and the reactor tank
wall was high, and the mixing speed of the gas and liquid decreased when the distance was low (Figure 6-III)
[17].

Novel Designed Photobioreactors

Systems that integrate the useful design features of different types of photobioreactors for the production of any
microalgae species in closed systems are called photobioreactors [20]. Design photobioreactors are designed to
achieve maximum efficiency for biofuel production. These photobioreactors are generally evaluated by CFD
simulations, design performance and the effect of different design patterns on hydrodynamic properties are
investigated [21].
Soman and Shastri [21] have developed a new photobioreactor design by combining airlift and flat panel
systems. The proposed new design has been optimized using CFD modeling (Figure 7).

Figure 7: CFD simulation results: (a) gas holding contours, (b) axial fluid velocity contours, (c) special case lighting
contours, (d) reactor light-receiving walls. [21]

692
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Soman and Shastri [21] concluded that this new photobioreactor, which they have proposed, can utilize the light
energy at the maximum level due to the cuboid flat surface, and that it spreads homogeneously in the heat
because of its increased surface area. In addition, the new design photobioreactor has been found to perform
better than the airlift photobioreactor.

CONCLUSION

Microalgae cultures are high-yielding biomass sources used in biofuel production. Biofuels from microalgae are
a renewable energy source that can replace petroleum and its derivative. The efficiency of biofuel production
depends on the design and performance of photobioreactors as seen in similar studies.
Photobioreactors are systems with a large surface area containing a light source. Microalgae cultures are
designed as open photobioreactors and closed photobioreactors. Improving these systems increases algae-based
biodiesel efficiency. Depending on photobioreactor design parameters it can be easily increase the biomass fuel
efficiency production capacities. Here, CFD studies play an important role for overall efficiency optimization
studies.
Design, development, and optimization of photobioreactors could be provided by computational fluid dynamics
techniques. The mixture and movement of the algae-culture, the use of different shapes to improve their designs,
parameters such as CO2 gas given to the culture, are defined in CFD software and optimization studies are carried
out. In this short review, optimization studies with CFD in photobioreactors are discussed.

REFERENCES

1. Chisti, Y. (2008). Biodiesel from microalgae beats bioethanol. Trends in biotechnology, 26(3), 126-131.
2. Sawayama, S., Inoue, S., Dote, Y., & Yokoyama, S. Y. (1995). CO2 fixation and oil production through microalga.
Energy Conversion and Management, 36(6-9), 729-731.
3. Roessler, P. G., Brown, L. M., Dunahay, T. G., Heacox, D. A., Jarvis, E. E., Schneider, J. C., ... & Zeiler, K. G. (1994).
Genetic engineering approaches for enhanced production of biodiesel fuel from microalgae.
4. Delucchi, M. (2003). A lifecycle emissions model (LEM): lifecycle emissions from transportation fuels, motor vehicles,
transportation modes, electricity use, heating and cooking fuels, and materials.
5. Wang, T. (2019). Leading biodiesel producers worldwide in 2018, by country (in billion liters). Statista, 8.
6. Guler, B. A., Deniz, I., Demirel, Z., Oncel, S. S., & Imamoglu, E. (2019). Comparison of different photobioreactor
configurations and empirical computational fluid dynamics simulation for fucoxanthin production. Algal research, 37,
195-204.
7. Ugwu, C. U., Aoyagi, H., & Uchiyama, H. (2008). Photobioreactors for mass cultivation of algae. Bioresource
technology, 99(10), 4021-4028.
8. Hase, R., Oikawa, H., Sasao, C., Morita, M., & Watanabe, Y. (2000). Photosynthetic production of microalgal biomass
in a raceway system under greenhouse conditions in Sendai city. Journal of bioscience and bioengineering, 89(2), 157-
163
9. Bitog, J. P., Lee, I. B., Lee, C. G., Kim, K. S., Hwang, H. S., Hong, S. W., ... & Mostafa, E. (2011). Application of
computational fluid dynamics for modeling and designing photobioreactors for microalgae production: a review.
Computers and Electronics in Agriculture, 76(2), 131-147.
10. Sawant, S. S., Gajbhiye, B. D., Mathpati, C. S., Pandit, R., & Lali, A. M. (2018, September). Microalgae as Sustainable
Energy and Its Cultivation. In IOP Conference Series: Materials Science and Engineering (Vol. 360, No. 1, p. 012025).
IOP Publishing.
11. Huang, J., Qu, X., Wan, M., Ying, J., Li, Y., Zhu, F., ... & Li, W. (2015). Investigation on the performance of raceway
ponds with internal structures by the means of CFD simulations and experiments. Algal research, 10, 64-71.
12. Chiaramonti, D., Prussi, M., Casini, D., Tredici, M. R., Rodolfi, L., Bassi, N., ... & Bondioli, P. (2013). Review of
energy balance in raceway ponds for microalgae cultivation: re-thinking a traditional system is possible. Applied
Energy, 102, 101-111.
13. Masojídek, J., & Torzillo, G. (2014). Mass cultivation of freshwater microalgae.
14. Sierra, E., Acién, F. G., Fernández, J. M., García, J. L., González, C., & Molina, E. (2008). Characterization of a flat
plate photobioreactor for the production of microalgae. Chemical Engineering Journal, 138(1-3), 136-147.
15. Huang, J., Kang, S., Wan, M., Li, Y., Qu, X., Feng, F., ... & Li, W. (2015). Numerical and experimental study on the
performance of flat-plate photobioreactors with different inner structures for microalgae cultivation. Journal of applied
phycology, 27(1), 49-58.
16. Wang, J., Rosov, T., Wensel, P., McGowen, J., & Curtis, W. R. (2016). A preliminary implementation of metabolic-
based pH control to reduce CO 2 usage in outdoor flat-panel photobioreactor cultivation of Nannochloropsis oceanica
microalgae. Algal Research, 18, 288–295. doi:10.1016/j.algal.2016.07.001

693
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

17. Massart, A., Mirisola, A., Lupant, D., Thomas, D., & Hantson, A. L. (2014). Experimental characterization and
numerical simulation of the hydrodynamics in an airlift photobioreactor for microalgae cultures. Algal Research, 6,
210-217.
18. Ebrahimifakhar, M., Mohsenzadeh, E., Moradi, S., Moraveji, M., & Salimi, M. (2011). CFD simulation of the
hydrodynamics in an internal air-lift reactor with two different configurations. Frontiers of Chemical Science and
Engineering, 5(4), 455-462.
19. Posten, C. (2009). Design principles of photo‐bioreactors for cultivation of microalgae. Engineering in Life Sciences,
9(3), 165-177.
20. Rebolledo-Oyarce, J., Mejía-López, J., García, G., Rodríguez-Córdova, L., & Sáez-Navarrete, C. (2019). Novel
photobioreactor design for the culture of Dunaliella tertiolecta–Impact of color in the growth of microalgae.
Bioresource technology, 289, 121645.
21. Soman, A., & Shastri, Y. (2015). Optimization of novel photobioreactor design using computational fluid dynamics.
Applied energy, 140, 246-255.

694
 DESIGN OF ARTIFICIAL NEURAL NETWORKS FOR
FUNCTION APPROXIMATION OF MAXIMUM POWER POINT
TRACKING IN PV SYSTEMS

Vidhya K Viswambaran
University of Bolton, Ras Al Khaimah, United Arab Emirates, [email protected], ORCID:
https://orcid.org/0000-0002-4341-9843

Akram Bati
University of Bolton, Greater Manchester, United Kingdom, [email protected], ORCID: https://orcid.org/0000-
0001-8548-0418

Erping Zhou
University of Bolton, Greater Manchester, United Kingdom, [email protected], ORCID: https://orcid.org/0000-
0002-0568-294X

Viswambaran, Vidhya, Bati, Akram Zhou, Erping. Design of Artificial Neural Networks for
Cite this paper as: function approximation of Maximum Power Point Tracking in PV systems. 8th Eur. Conf. Ren.
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: This paper presents a function approximation problem by Artificial Neural Networks. An
Artificial Neural Network (ANN) will be utilized for approximating the Maximum Power Point
(MPP) problem for the PV system. The neural network learns to create the map the underlying
relationship between the input and output variables. Tracking the maximum power point is
essential in a PV system to ensure maximum power is transferred between the panel and load.
Initially, the nonlinear PV characteristics of the panel are modelled and simulated using MATLAB
software. The maximum power point captured by Perturb and Observe algorithm will be given as
the target data for training the ANN. The ANN is designed and trained using Neural Designer
software. The NN accepts irradiance and temperature as variable parameters and predicts the
voltage at maximum PowerPoint. The network is tested and the accuracy of the results are
evaluated in terms of correlation coefficient of predicted values. The ANN also provided a
function approximation that maps between input and output. The results demonstrate the ability
of ANN to generalize the non-linear input-output relationship from the provided input patterns.
The network was able to accurately track the MPP for PV. Deeper networks were designed and
tested for comparing the accuracy in predicting output with shallow networks. Results
demonstrated that deeper networks showed superior performance in predicting more accurate
results.
Keywords: function approximation, Maximum Power Point Tracking, Deep learning, Photovoltaics
© 2020 Published by ECRES

1. INTRODUCTION

The current scenario of low energy conversion efficiency mandates the use of Maximum power point tracking
MPPT based charge control [1]. An MPPT controller improves the energy conversion efficiency of PV systems
by selecting the appropriate reference voltage for the DC-DC converter [2]. Artificial neural networks are
biologically inspired information processing systems [3]. These soft computing tools can learn from examples
without being programmed with task-specific rules and can make machines perform tasks like humans. Neural
network-based MPPT tracking has the ability to track global maxima [4]. PV system coupled with the neural
network will estimate the optimal reference voltage corresponding to maximum power point [4]. These networks
can be trained to perform useful functions such as function Approximation, prediction, pattern recognition, etc.

695
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[5]. Function approximation focuses on the mapping between input and output variables to generalize a function
[6]. For a function approximation problem, the neural network learns and creates a mapping between the
independent variables and the independent variables provided as training data set [7].

2.THEORY

Various ANN based MPPT techniques based on input variables and control architectures are reviewed in [8].
Back-propagation algorithms demonstrate good performance under partial shading conditions [9]. ANN-based
MPPT can overcome the drawback of conventional MPPT by avoiding local maxima as in conventional Hill
climbing MPPT techniques [10-11].Neural network-based MPPT tracking techniques are faster than
conventional MPPT techniques such as Perturb and observe algorithm based MPPT. They also provide better
accuracy than P& O [1]. Due to its ability to adapt changes ANN-based MPPTs can provide good predictions
for partial shading conditions as well [12]. Different types of ANNs such as Feedforward NNs and Recursive
neural networks (RNN), can be trained with the backpropagation to predict PV power is discussed in [8]. Deep
learning neural network architectures contain many layers and can be employed as a universal function’s
approximation tool [13]. Neural networks are employed widely in optimization (regression) and classification
problems [14].

Figure .1 Deep learning Network

Figure 2.PV System

In Regression, the network tries to approximate a function for the set of input parameters (Independent variables)
and out parameters (Target outputs or dependent variables) are provided at the input layer [15]. The regression
process focuses on adjusting the weights attempt to predict a real value output which is close to the target output
[16]. The attempt continues until the loss function (error) is minimum. In linear regression, the network tries to
map the inputs and outputs to approximate a function [17].

696
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. PV SYSTEM

A PV system as shown in Figure 2,consists of a PV panel, DC-DC converter and an MPPT controller. The
purpose of MPPT controller in the system is to maximize the power transfer from PV panel to the load. Neural
network is used to predict the reference voltage for PWM waveform generator which in turn adjusts the duty
cycle. This will ensure that DC-DC converter operating point is set to optimal value and maximum power is
transferred from the PV panel to the load.

4. DATA COLLECTION

The proposed ANN architecture is shown in Figure 3.The two input parameters to the network are Irradiance
(G) and Temperature (T) and the target output is Vm. The output parameter will be the predicted Vm. For data
collection, the PV panel behavior was modeled and simulated using MATLAB software and the Maximum
power points were extracted using Perturb and Observe algorithm. Sample data set obtained from the simulation
is shown in Table 1. 63.2% of the data is set aside for training purposes, 18.4% for validation and 18.4%% for
testing. PV panel and maximum power point tracking is modelled and simulated using MATLAB.

Figure 3. Proposed ANN for function approximation

Table 1. Sample Data Set.

Test T 0C G Vm
Point KW/m2
1 25 1 17.752
2 27 1 17.251
3 25 0.8 16.96
4 29 0.8 16.657
5 25 0.6 16.837
6 25 0.5 16.735
7 42 0.3 13.498
8 27 0 10.804
9 27 0.7 16.756
10 40 0.7 13.999

Selecting the Architecture

Because the mapping between the solar irradiations, temperature and panel output is highly nonlinear, a
multilayer network architecture is selected to learn the mapping. As the problem under investigation has two
inputs and one output, a two-layer feedforward architecture is selected with the respective number of neurons in
the input layer and the output layer. The hidden layer is selected with 3 neurons. As the problem needs function
approximation, activation function for hidden layer employs hyperbolic tangent as activation function and output
layer employs linear activation function. Network is shown in Figure 4

Figure 4. Network Architecture 2-3-1

697
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Training Strategy
The training strategy decides the learning procedure in order to obtain the best possible loss. The type of training
is determined by the way in which the adjustment of the parameters in the neural network takes place. The loss
index will be evaluated using the mean square error method. Regularization is applied for better generalization
of output. In each iteration, one hidden perceptron will be added to the network. A maximum of 10 perceptrons
will be added for evaluation.

Input Correlations
The scaling method for the input layer is the Automatic. The minimum and maximum correlation between the
input variables G and T is 0.406205. The unscaling method for the output layer is the minimum and maximum.
The input selection algorithm chosen for this problem is growing inputs. With this method, the inputs are added
progressively based on their correlations with the targets. The less correlated variables with the targets can
confuse the model with data that could have no influence on the predicted variable. This task removes the inputs
with a lower correlation than the given one. The correlation coefficient is calculated between inputs and outputs.
Correlation coefficient closer to 1 mean that a single target is closely related to a single input and a value closer
to 0 mean that there is no relationship between input and target variables.
Data Statistics

Table 2. Gives the correlations between all input and outputs. The correlation coefficient of input T with output
Vm is -0.0979442 and with G and Vm is -0.280412. This indicates there is no correlation between inputs G &
T and output Vm. Table 3 given below shows the data statistics of all inputs in the data set in terms of mean and
distribution.

The minimum and maximum is the smallest and largest value in the data set. Mean gives the center of the data
set and distribution indicates how often the data occurs.

Neural Network Parameters & Statistics

Table 4. Shows the statistics of the neural network parameters. The network is trained using Quasi Newton
method used for solving nonlinear optimization problems. A training rate of 0.001 and a minimum goal of 1e-12
was selected.

Table 5. Illustrates the statistics of the neural network parameters. This neural network parameter will be
associated with input parameters to give the output. Figures 5 and 6 illustrates the target Vm with respect to G
and T.

698
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Output layer
Table 6. Shows the statistics of the output Vm. Mean for output Vm is o and deviation is 1.

Table 6. Statistics of the output Vm

Item Minimum Maximum Mean Deviation
Vm 12.7 17.8 0 1

Final Architecture
Figure 7. Illustrates the resulted deep architecture. Connections and weights are adjusted after the training and
deployment of the model.

Figure. 7 Final Architecture

Table 7.Linear Regression Parameters

Item Value
Intercept -0.0717
Slope 1.01

5. TESTING ANALYSIS
Linear regression analysis (LRA) is used here to test the loss function of the trained neural network model. Table
7 shows the result of LRA on the neural network. Figure 8 shows the linear regression for the output Vm. The
plot indicates the best linear fit for the problem. Figure 9.shows the error histogram for the output, Vm and it’s
distribution for every test instance. The minimum frequency is 0%, which corresponds to the bins with centers
-3.6341%, -2.7256. The maximum frequency is 33%, which corresponds to the bin with center 0.90853%.

699
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Losses of the tested Model

This section gives the measures of all the errors of the model and evaluates the model for training, validation,
and testing. Table 8. Illustrates the errors of data for each use of them.

Vm Error Statistics
The table 9.shows the error statistics of Vm in terms of Mean and deviation. Absolute error is between 0.03 and
0.1 which is in an acceptable range for the problem.

6. DEEP NETWORKS

Two deep networks were designed and trained for the same problem. A 2-3-3-1 network as shown in figure 10
was modeled and trained using same training strategy to extract Vm. Figure 11 illustrates the final network
architecture after training.

Finally, a 2-4-4-1 network as shown in figure 12 is modeled and trained for the same data set and output Vm is
predicted.

Figure 13 final architecture of 2-4-4-1 Network

700
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 15.Regression for Vpred

7. DISCUSSIONS

The three networks 2-3-1, 2-3-3-1 and 2-4-4-4-1 were tested for same set of irradiation and temperature values.
Results indicated that the prediction of output Vm by 2-3-3-1 network is closer to the actual values. Figure 14
shows the function approximation by neural network by mapping input and out parameters .Correlation
coefficient is 0.9807 as shown in figure 15 whereas Correlation coefficient for 2-3-1 network is only 0.7074
.This shows that neural network with more layers can predict the output better. However increasing number of
neurons in the hidden layers did not show improvement performance. Correlation coeffiecnt for 2-4-4-4-1
network is only 0.00133. The equation gives the mathematical model approximated by neural network to predict
Vm. The neural network has mapped two input G and T and modelled a linear equation for finding the output
Vm. The table 10 shows the comparison of predicted output for three different neural architectures namely 2-3-
1, 2-3-3-1 and 2-4-4-1 with actual target values

Table 10. Losses of the tested model

scaled_G = G;
SL Network G T Vpred Vactual scaled_T = T;
No structure (KW/m2) oC Volts Volts y_1_1 = tanh (0.698608+ (scaled_G*-0.760437)+
1 231 1 25 17.6679 17.752 (scaled_T*-0.350159));
2 2331 1 25 17.7694 17.752 y_1_2 = tanh (0.787842+ (scaled_G*0.797974)+
3 24441 1 25 17.7583 17.752 (scaled_T*-0.597046));
y_1_3 = tanh (-0.108887+ (scaled_G*0.88031)+
4 231 0.8 25 17.0484 16.96
(scaled_T*-0.749878));
5 2331 0.8 25 16.9846 16.96 y_2_1 = tanh (-0.416077+ (y_1_1*-0.555786)+
6 24441 0.8 25 16.9687 16.96 (y_1_2*0.26947)+ (y_1_3*0.120117));
7 231 0.6 25 16.7989 16.837 y_2_2 = tanh (0.363281+ (y_1_1*0.248047)+ (y_1_2*-
0.15271)+ (y_1_3*-0.967041));
8 2331 0.6 25 16.9732 16.837 y_2_3 = tanh (0.122864+ (y_1_1*0.72821)+ (y_1_2*-
9 24441 0.6 25 16.8987 16.837 0.720764)+ (y_1_3*-0.46521));
10 231 0.8 26 17.4636 16.96 scaled_Vm = (-0.154602+ (y_2_1*0.917603)+
11 2331 0.8 26 17.0753 16.96 (y_2_2*0.214417)+ (y_2_3*0.576416));
Vm = (0.5*(scaled_Vm+1.0)*(17.752-12.748)+12.748);
12 24441 0.8 26 16. 949 16.96 Figure 14.Function approximation by neural network

8. CONCLUSION

This paper presented the function approximation by Artificial Neural Network .ANN approximated the non-
linear relationship of PV input and output with higher accuracy. Artificial Neural Network was designed and
trained to implement the operation of the MPPT algorithm for PV systems. Results indicate that ANN can track
the maximum power point efficiently. Then an investigation was carried out to check the feasibility of employing
deeper networks for the implementation of MPPT controllers for PV systems. Two deep network architectures
were implemented and tested for same data. Results indicate that deep networks with more layers can predict

701
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the output with a better accuracy. Hence higher-order networks will be more useful for modeling the non-linear
characteristics of PV for Maximum power point tracking.

REFERENCES

[1] Khanam JJ, Simon YF. Modeling of a photovoltaic array in MATLAB simulink and maximum power point tra cking using neural network. Electric
Electron Tech Open Acc J, 20182(3):40-46. DOI: 10.15406/eetoaj.2018.02.00019
[2] Y. H .Liu, C. Liu, L. Huang, J. W., J. H. Chen, Neural-network-based maximum power point tracking methods for photovoltaic systems operating
under fast changing environments.Solar energy,Vol 89, pp.42-53, 2013. https://doi.org/10.1016/j.solener.2012.11.017
[3] Elobaid, Lina M. et al. Artificial neural network-based photovoltaic maximum power point tracking techniques: a survey. Iet Renewable Power
Generation ,2015, 1043-1063. DOI:10.1109/IECON.2012.6389165
[4] S. Haykin & N. Network,A comprehensive foundation. Neural networks, 2.41. Prentice Hall 2004
[5] K .Suzuki, (Ed.).Artificial neural networks: Architectures and applications. BoD–Books on Demand, 2013
[6] Buduma, N., Locascio, N, Fundamentals of deep learning: Designing next-generation machine intelligence algorithms ,O'Reilly Media, Inc. 2017.
[7] S.Liang, & R.Srikant,Why deep neural networks for function approximation?. Ar,Xiv preprint arXiv:1610.04161, 2016.
[8] V. Lo Brano, G.Ciulla ,M. Di Falco, Artificial neural networks to predict the power output of a PV panel. International Journal of Photoenergy, 2014.
[9] O.Veligorskyi, R.Chakirov, Y.Vagapov,Artificial neural network-based maximum power point tracker for the photovoltaic application. 1st
International Conference on Industrial Networks and Intelligent Systems (INISCom) .pp. 133-138. IEEE. 2015
DOI:10.4108/icst.iniscom.2015.258313
[10] H. S.Agha, Z. U .Koreshi, M. B. Khan,Artificial neural network based maximum power point tracking for solar photovoltaic. In 2017 International
Conference on Information and Communication Technologies (ICICT) ,pp. 150-155. IEEE, 2017. DOI:10.1109/ICICT.2017.8320180
[11] A Anzalchi, & A. Sarwat, Artificial neural network based duty cycle estimation for maximum power point tracking in photovoltaic systems.
In SoutheastCon .pp. 1-5. IEEE. 2015. DOI: 10.1109/SECON.2015.7132988
[12] K.Ishaque & ZA .Salam, review of maximum power point tracking techniques of PV system for uniform insolation and partial shading
condition. Renewable and Sustainable Energy Reviews, 2013,Vol.19, pp.475-488., https://doi.org/10.1016/j.rser.2012.11.032
[13] Josh Patterson ,Adam Gibson ,Deep Learning-A PRACTITIONER'S APPROACH. Gravenstein Highway , United States of America, O’Reilly
MediaNorth, August 2017.
[14] A.Elgharbi, D.Mezghani, & A. Mami,A maximum power point tracking method based on artificial neural network for a pv system. International
Journal of Advances in Engineering & Technology, 2012,Vol.5 ,No (1), pp.130.
[15] Z.Zainuddin,O. Pauline,Function approximation using artificial neural networks. WSEAS Transactions on Mathematics, , 2008,Vol. 7.No 6,
pp.333-338
[16] S .Ferrari,R. F. Stengel, Smooth function approximation using neural networks. IEEE Transactions on Neural Networks, 2005.V.ol 16,No.1, pp.24-
38
[17] L. Ljung, Prediction error estimation methods. Circuits, Systems and Signal Processing, 2002,Vol.21,No.1,pp. 11-21.

702
 MACHINE LEARNING MODELS FOR PREDICTING PV
GENERATED POWER

Penka V. Georgieva
Burgas Free University, Burgas, Bulgaria, [email protected],ORCID: 0000-0001-9598-5997

Alexander Ivanov
Burgas Free University, Burgas, Bulgaria, [email protected],ORCID: 0000-0002-5870-6714

Georgieva, P. V., Ivanov, A., Machine learning models for predicting PV generated power.
Cite this paper as:
8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Renewable energy sources are attracting strong interest in the last decades as an alternative to fossil
energy sources. One important aspect of utilizing the renewable energy sources is achieving
effectiveness in power flows management and predicting the generated amount of power is a key
building block in the management process. Machine learning technics are suitable for achieving
feasible predictions because of their predictive accuracy, low computational and implementation
cost. In this paper, three machine learning technics are implemented and compared in terms of
their performance and accuracy. The presented in this paper research is the first step in the
development of a real-time software application for managing photovoltaic facilities.
Keywords: photovoltaic system, machine learning, power generation prediction
© 2020 Published by ECRES

Nomenclature
PV (system) photovoltaic (system)
ML machine learning
BFU Burgas Free University

1. INTRODUCTION

Photovoltaic (PV) systems transform solar radiation into electrical power by photovoltaic elements. The capacity
of photovoltaic systems varies from several kilowatts to hundreds of megawatts and the main types are grid-
connected, off-grid and stand-alone PV systems. [1]
The off-grid PV systems are autonomous and are designed to provide electricity needs where there is no network.
They can be located both on the roof and on the ground. These are mainly residential and commercial sites
outside the city, to which a network cannot be built, or the investment is too large and unjustified. For upgrading
the system for more consumers, it is necessary to add inverters, batteries and photovoltaic panels.
The grid-connected PV systems are being built where there is a large supplier of traditionally generated
electricity. In this case PV systems can be useful for reducing the electricity bills, serve as a backup power
supply to certain consumers (in case of failure or poor-quality power supply) or as an investment to sell the
produced electricity to the supplier. They can be mounted on the roof or installed on the ground.
The small photovoltaic installations for own consumption (micro-grid PV systems), unlike photovoltaic power
plants, do not produce electricity in large volumes. The purpose of this type of installation is to provide electricity
to meet the electricity needs of a specific building(s). They can be installed on the roofs of houses, roofs of
various buildings or roofs of enterprises, as well as on the ground. Improving energy efficiency is most often
understood as thermal insulation of a building, replacement of joinery, improvement of heat supply, electricity
supply or in general energy saving through various energy saving measures and actions.
At the same time, the production of electricity from RESs, such as the sun, is seen as something complex and
inaccessible, requiring large investments, large areas of land or roofs on which to place photovoltaic panels, a

703
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

lot of time to build their, complex documentation, and a number of other prejudices. However, every watt of
electricity produced by the sun replaces and saves the consumption of the same, but produced in a conventional
way (nuclear energy, natural gas, coal, etc.). In addition, conventional electricity is further increased by
transmission and access charges on the existing transmission and distribution network to the place where it will
be consumed. It should also be considered that every watt saved from the grid reduces the release of carbon
dioxide (CO2) and the release of other harmful emissions into the atmosphere.
In the case of self-contained photovoltaic installations installed on site, all energy produced by the sun is used
only for own consumption. This is achieved by limiting the power of the inverter so that it does not allow the
return of electricity to the grid at higher production than photovoltaics than necessary. Mass enterprises and
businesses buy the electricity they need on the free market. This makes them dependent on electricity producers
and constantly rising prices. The unpredictable situation forced the owners of the enterprises to buy the electricity
on the Bulgarian energy exchange through the so-called deals in the “day ahead” segment at unrealistically high
prices. The sharp rise in the price of electricity on the free market can be a great burden for some small businesses
because this has a direct impact on their competitiveness.
A micro grid-connected PV system consists of three modules: (1) a power source including the photovoltaic
panels and the rechargeable batteries; (2) a controller and an AC inverter; (3) consumers - the standard AC
consumers and the DC consumers.
The photovoltaic panels are configured in separate blocks in which they are distributed in parallel to achieve
operating currents and/or in series to achieve the required operating voltage. Similarly, the batteries are
configured in parallel for reaching the required current and/or in series for the required working voltage. The
batteries accumulate the excess energy generated by PVs during the daylight and to give it overnight and thus
the power supply is kept continuous. The batteries in use must be able to be deeply diluted to be fully utilized.
For this purpose, lithium batteries are particularly suited, but their cost, performance and safety features, for
now, deter producers from using them massively. Lead accumulators are cheap, but they cannot be discharged
completely because they are irretrievably destroyed (the allowable discharge is up to 20% of their capacity). The
direct current consumers can be connected to DC buses. Such are some types of lights, TVs, radios, refrigerators
etc.
The inverter converts the constant voltage into a voltage with an industrial frequency, and thus a direct power
supply to all standard consumers is provided. Few systems are limited to DC consumers only (with no inverter).
Such are, for example, the systems designed to power portable equipment - illuminated signs, temporary
luminaires and more. Autonomous photovoltaic systems without an inverter are the cheapest. The use of an
inverter enables the autonomous photovoltaic system to provide energy with indicators identical to those of the
general power supply. The quality of the electricity is acceptable, both in frequency and level. However, in the
micro power systems, the inverters are a major cost factor.
Loads are divided into two main groups –consumers receiving energy directly from the batteries and consumers
receiving energy inverted into AC voltage. Each load has some specific features to be considered. Parameters
such as start current, rated power, peak power, and power supply voltage formats are common. For example,
these parameters are not critical for LED lights, but for laptops, the operating limits are narrow and strictly
defined. For some of the AC consumers, the voltage change sometimes may be interpreted by the protection unit
as an electric shock or overvoltage in the network. It is not permissible to introduce boilers in this group as they
are very powerful consumers and require a sharp increase of the declared power to the AC converters. Modern
refrigerators have high energy class and this problem is avoided. [2]
For the correct operation of the PV panels and their proper coordination with the consumers, the systems are
provided with a controller for monitoring the proper charging/discharging of the batteries, the operation of the
photovoltaic panels and detecting emergencies. In some applications, the controller also provides optimum
battery charge. The controller is the most important element of the system and the proper choice and appropriate
software are key for the smooth operation of the system. [3]
The operational characteristics of the electricity produced by the PV system are out of the scope of this paper.
The amount of power produced by PV panels is an important parameter in the management process and its
successful prediction increases significantly the systems’ efficiency. The predicting model, proposed in this
paper, aims to support a more efficient management system of a micro grid PV system. The research is part of
a project for optimizing the energy consumption of a building with the use of independent alternative renewable
energy sources. The input data for solar radiation, wind speed and air temperature are received from BFU
METEO - a meteorology station specially designed and installed on the building of Burgas Free University for

704
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the needs of this project. BFU METEO consists of sensors for monitoring the input variables, Fronius Sensor
Box, and FroniusDatalogger Web. Data is collected in a 5-minutes interval and stored in a database in a server.
The micro grid PV system consists of:
• 20 polycrystalline PV panels, with max power approx. 235Wp each, an invertor SolarMax 13MT2
with the following technical specifications: Input values: MPP voltage range 250 … 750 V, Minimum
voltage for rated power 370 V, Maximum DC voltage 900 V, Maximum DC current 2 x 18 A, Number
of MPP-Trackers 2, Max. PV generator output per MPP tracker: 9 000 W, String connections 2 x 2,
Connection type MC 4; Output values: Rated output power at cos(ϕ)=1 is 13 000 W, Maximum
apparent output power 13 000 VA, Nominal mains voltage 3 x 400 V, Maximum AC current 3 x 20,
Mains nominal frequency / range 50 Hz / 45 Hz…55 Hz; Grid connection Three-phase (3 / N / PE);
• 10 monocrystalline PV panels, with max power approx. 150Wp each; an invertor IBC Serve Master
3300 MV with the following technical specifications: PV input 450 VDC, max 2×10 A; 100-350 VDC
MPP; Output 230 VAC, 50 Hz, Class I, 3300 W/3600 W, 15.5 Amax.
The data are collected from the BFU METEO sensors every 5 minutes and then automatically stored in the
database. The data for the quantity of the electric power produced by two types of installed PV panels
(polycrystalline and monocrystalline) is also collected and stored in the database at 5-minute intervals.

2. AI TOOLS FOR PROGNOSIS

Predicting is a key element of the process of decision making. The ability to analyze the uncontrollable aspects
of different types of events is an important factor in this process. The development of computer technologies
leads to the emergence of new methods of predicting, such as fuzzy systems. There is intensive research in the
area of forecasting solar PV power. For example, different methods used for solar PV power forecasts with a
focus on the aspects that are unique to solar forecasting are outlined in [4]. The authors explored the nuances
related to solar forecasting and its interaction with load forecasting, with the performance of solar forecasting
summarized.
In addition to the traditional forecasting methods, the modern tools Artificial Intelligence (AI) can be
successfully used in this direction. AI is a branch of computer science and the term was first used by John
McCarthy in 1956 as “the science of creating intelligent machines”. Historically, AI developed from the attempts
to formalize human knowledge by the tools of mathematical logic and originally was applied to theorems proof
and games modelling. The base of AI was philosophy, mathematics, algorithms, logic, psychology, informatics,
linguistics. Gradually the traditional AI, mainly focused on imitation of human behaviour in language form or
symbolic rules, was enriched with new ideas and thus the modern concept of AI developed.
At present, AI is a combination of traditional AI enriched with different methodologies for numerical calculation
and the subject of AI are mainly those problems for which no direct mathematical or logical algorithms exist or
can be resolved only intuitively. AI, alongside with the traditional basic sciences, includes some new ones:
neuroscience, cognitive science, ontology, operations research, economics, probability, and optimization.
Unlike in traditional computing (hard computing), the keystone in modern AI is the understanding that accuracy
and security have a high price and tolerance for imprecision and uncertainty in calculations, reasoning and
decision making should be admitted (when and where possible). The computing paradigms are Fuzzy Logic
(FL), Neural Networks (NN), approximate conclusions and non-differential optimization methods such as
Genetic Algorithms (GA) and Simulated Annealing (SA). [5]
А detailed overview of applications of Artificial intelligence (AI) technics as “alternate approaches to
conventional technics or as components of integrated systems” can be found in [6]. Artificial neural networks
have been used for prediction of meteorological data: solar radiation, wind speed, air temperature, insolation
and diffuse fraction, solar radiation; weather forecasting; sizing of photovoltaic (PV) systems; modelling and
simulation of PV systems. Fuzzy modelling has been applied for prediction solar radiation and sunshine
duration; control of PV systems. Genetic algorithms are successfully applied for sizing of different types of PV
system.
Nowadays AI technics are widely used in managing various stand-alone PV plants, off the grid PV systems and
different hybrid systems for generating electricity by using photovoltaics. [7-11]
This study is focused on the predicting the amount of electric power produced by photovoltaic panels installed
on a single building and connected with the grid. The purpose is to use this forecast to effectively manage the
energy flows of the building. The forecasting model is a neural network, trained on real data collected from the
micro energy system, installed on the building of Burgas Free University.

705
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. MACHINE LEARNING TECHNICS FOR POWER GENERATION PREDICTION

Machine learning (ML) is a building block of the modern Artificial Intelligence and is acknowledged by both
academia and business as the leading tool for data analysis, especially when managing big data. Recently,
statistical technics are extensively applied for analysis of processes in various fields including but not limited to
physics, chemistry, engineering, finance, medicine. Python programming language is often chosen for
implementation of ML models for the relative simplicity of its syntax, the number of available mathematical
and ML libraries, performance enhancements such as TensorFlow, etc.
In this section of the paper a comparison of several predictive ML technics is presented. Three ML models are
trained on the same data in a few different settings and their prediction errors are compared. Such kind of analysis
can be used as a guideline for creating real-life software that can be implemented in PV microgrid installation.
Presented models are implemented in Python using the popular scikit-learn library and are trained on the dataset,
described in the previous section of the paper.
ML aims at creating a programing and/or mathematical model that fits some real-life data. The process of
learning is inspired by how living organisms learn – by experience. The three main types of learning are:
• supervised learning - some functional relation between variables is hypothesized, and a learning
algorithm tries to automatically compute the model parameters to fit data that is usually taken from
past measurements; supervised learning can be used for regression and classification tasks;
• unsupervised learning - the structure of the data is analyzed, and this information can be further used
for principal component analysis, dimensionality reduction, noise reduction, generation of similar data;
• reinforcement learning - a utility function is optimized by providing positive and negative feedback
to an intelligent entity such as software agent or artificial neural network.
The task of predicting future values of some variable is a standard problem in supervised learning. This paper
presents results from three supervised models– a Decision Tree, a Feedforward Neural Network and a Support
Vector Machine.
Decision Tree is a non-parametric model for supervised learning that represents a tree-like structure of rules for
data classification and regression. It consists of two types of nodes – decision nodes and leaf nodes. Decision
nodes implement if-then clauses based on some threshold value thus creating splits in the data. Each split is a
subset of data points belonging to a specific group, defined by their similarity. The decision rules are extracted
from the features during the learning process. Leaf nodes represent final decisions (for example class labels).
Decision trees can work with both categorical (classification) and scalar (regression) data. Advantages of
decision trees include relative simplicity, low computation cost and easy interpretation of the results.
Disadvantages are possible overfitting, instability (small change in the training dataset can result in significantly
different tree structure), possible bias towards a dominant class in the data. [12]
Feedforward Neural Networks are mathematical models that imitate the biological neural networks. They consist
of interconnected simple calculation units called neurons. Each neuron sums its weighted inputs (represented as
numbers) and performs an activation function over their sum. The parameters of the network are called synaptic
weights and they represent connection sensitivity between neurons. There are many activation functions such as
the logistic, the hyperbolic tangent, rectifying linear unit (ReLU), softmax and others, each of them applicable
for different problems. Feedforward networks do not have recursive connections and they are usually trained
with supervised learning algorithms (e.g. gradient descent). The advantages of feedforward networks are their
ability to approximate a broad class of functions and usually they generalize well with unseen data. The
disadvantages include high computational cost and the fact that they are black-box models, which makes the
interpretation of the results hard. [13]
Support vector machines are linear parametric models that calculate the hyperplanes (decision boundaries)
between classes in a dataset. The learning algorithms that are used for this calculation try to find a line
(hyperplane) that passes at the longest average distance (margin) from the points that represent the classes.
Although support vector machines are linear models, they can make non-linear classification and regression by
using kernel functions. In that approach features are projected (mapped) to a higher-dimensional space by taking
their dot product. This „trick“ does not require to explicitly compute the coordinates of the mappings which is
an advantage in matters of performance. The kernel function serves the purpose of calculating similarity between
data points, which is why sometimes this type of learning is designated as instance-based learning. When the
trained SVM is presented with new unseen data it calculates its label by using the kernel function to estimate
the similarity of the new point to the previous ones. [14]

706
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. EXPERIMENTAL RESULTS

The database contains data for – solar radiation, wind speed, air temperature, date and time of the measurement
and the generated amount of electrical power. The inputs in the three ML models are solar radiation, wind speed,
air temperature and the target variable (output) is the generated amount of electrical power. The database consists
of 2880 records of measurement, stored in a CSV file. The design of the experiment is as follows: for each of
the models a series of trainings is performed using a different split ratio between training and testing sample
sizes. The starting point is the split 2000:880 (training: testing) samples which is roughly the classic ratio of 7:3.
After the model is trained, the mean squared error is calculated for the testing set. At the final training the whole
dataset is used for training and mean squared error is calculated for the same. The step of the increasing of the
training sample size is different for the three different models because of the computational cost of the training
– training a feedforward neural network is more computationally expensive than training a decision tree. The
feedforward neural network has 1 hidden layer with 100 neurons, performing ReLU activation function and it
is trained for 150 learning epochs. For the decision tree a step of 15 samples increase is used, while for the other
two models the step is 80.
After the final training of each model a plot is made that shows the original datapoints compared to the model
predictions. In each of figures 1, 2 and 3 the original data points and the predicted values are visualized. On each
of the three graphs, the 𝑦 axis shows the value for the generated electrical power and on the 𝑥 axis are
respectively: on the upper plot is the amount of solar radiation, on the middle plot is the wind speed, and on the
lowest - the air temperature.
There is a separate plot for each input variable (Fig. 4)

Figure 1. Decision tree predictions

Figure 2. Feedforward neural network predictions

707
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Support vector machine predictions

Figure 4. Plotting the average mean squared error (y axis) for each model as a function of training sample size (x axis).
Top to bottom: Decision tree, Feedforward neural network, Support vector machine

5. CONCLUSIONS

The results obtained from the training of the ML models show that the least error is achieved when using all the
data (including the zeros during the night), the least training error is achieved from applying the decision tree
model. A useful development of the presented forecasting model is to create an autonomous software for
managing photovoltaic systems. Such a program will be widely used given the pace of development of this
technology around the world. Optimizing PV systems with respect to their cost is another priority in the
development of the proposed models.
ACKNOWLEDGMENT

This research is partially funded by Bulgarian National Science Fund, agreement No. KP-06-COST-8, August
6, 2019, "Prediction of Characteristics and Optimization of a PV System by Artificial Intelligence Tools" for
providing national co-financing for participation of Bulgarian collectives in approved actions under the
European program for cooperation in the field of research and technology COST.

708
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] P. V. Georgieva, R. Dolchinkov, “Fuzzy Models for Managing a Micro Grid PV System,” 42nd International
Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO), pp. 1066-
1071, 2019
[2] E. Skoplaki, J. A. Palyvos, “Operating temperature of photovoltaic modules: A survey of pertinent correlations.”
Renewable Energy, Vol. 34 № 1, pp. 23-29, 2009
[3] R. Dolchinkov, P. Georgieva, “Effectiveness of solar tracking systems“, Yearbook of BFU, Vol. XXVII, 2012 (in
Bulgarian)
[4] A. Tuohy, and coauthors, “Solar forecasting: Methods, challenges, and performance,” IEEE Power and Energy
Magazine vol. 13 (6), pp. 50–59, 2015
[5] P. V. Georgieva, “Fuzzy rule-based systems for decision-making,” Engineering Sciences, Bulgarian Academy of
Sciences, vol. 53 (1), pp. 5–16, 2016
[6] A. Mellit and S. Kalogiroub, “Artificial intelligence techniques for photovoltaic applications: A review,” Progress in
Energy and Combustion Science vol. 34, pp. 574–632, 2008
[7] E. Kardakos, M. Alexiadis, S. Vagropoulos, C. Simoglou, P. Biskas and A. Bakirtzis, "Application of time series and
artificial neural network models in short-term forecasting of PV power generation," 48th International Universities'
Power Engineering Conference (UPEC), pp. 1-6, 2013
[8] R. Welch, S. Ruffing and G. Venayagamoorthy, “Comparison of feedforward and feedback neural network architectures
for short-term wind speed prediction,” Proceedings of international joint conference on neural networks, 2009
[9] Y. Huang, J. Lu, C. Liu, X. Xu, W. Wang, and X. Zhou, “Comparative study of power forecasting methods for PV
stations,” International Conference on Power System Technology (POWERCON), Hangzhou, China, pp. 1 – 6, 2010
[10] T. Wang and Su-Yuan Tsai, “Solar Panel Supplier Selection for the Photovoltaic System Design by Using Fuzzy Multi-
Criteria Decision-Making Approaches,” Energies vol. 11(8), 2018
[11] A. Lee, H.-Y.Kang, C.-Y. Lin and K.-C. Shen, “An integrated decision-making model for the location of a PV solar
plant,” Sustainability, vol. 7(10), pp. 13522–13541, 2015
[12] L. Rokach, and O. Maimon, (eds), “Decision Trees.,” In: Data Mining and Knowledge Discovery Handbook. Springer,
Boston, MA, 2005
[13] A. Smola, B. Scholkopf, “A tutorial on support vector regression”, Statistics and Computing 14: 199–222, 2004
[14] A. Ivanov, “Fractional activation functions in feedforward artificial neural networks”, 20th International Symposium
on Electrical Apparatus and Technologies (SIELA), 2018

709
 NEW ANALYTICAL METHOD DEDICATED TO CALCULATING
THE ENERGY GAP VALUE OF NIO/FTO THIN-FILM
MATERIAL OF PHOTOVOLTAIC SYSTEMS

Slimani Hamza
Department of Physics, Fac. Exact Sciences, University of El Oued, El Oued, Algeria, [email protected]
ORCID: 0000-0003-2088-3907

Bessous Noureddine
Department of Electrical Engineering, Fac. Technology, University of El Oued, El Oued, Algeria, [email protected]
ORCID: 0000-0002-7290-6683

Hamza S, Noureddine B. New analytical method dedicated to calculating the energy gap value
Cite this paper as: of NiO/FTO thin-film material of photovoltaic systems. 8th Eur. Conf. Ren. Energy Sys.24-25
August 2020, Istanbul, Turkey

Abstract: This paper presents the thin films were fabricated using zinc acetate (Ni) precursors by the spray
pyrolysis technique deposited onto fluorine tin oxide (FTO). This study is dedicated to
photovoltaic system applications. The placement of the zinc oxide (NiO) solution was made on
FTO under 380 ˚C using a spray pyrolysis technique. The NiO films were grown in different
concentration solution ranges varying from 0.1 M to 0.5 M. This work used many analysis
techniques. The present paper has proposed a new analytical model that ensures a unified formula
which defines the energy gap (Eg) value. This new formula used logical steps that allow us to
guarantee the variation between the different Eg values. In addition, this study took a real example
of NiO/FTO to validate our model. This study made the characterization of the thin layers prepared
under the influence of the concentration of the zinc solution. The thin films prepared show average
optical transparency between 60 % and 90 % in the visible range and their band gap values are
ranged between 3.38 eV and 3.74 eV. These thin films preparation are dedicated to photovoltaic
system applications.
Keywords: Photovoltaic system, ultra violet visible spectroscopy, energy gap, analytical model,
transmittance curve
© 2020 Published by ECRES

1. INTRODUCTION

The development of modern science is based on thin-film devices [1-3]. Currently, thin films have a large portion
of the needs of the more technologically developed integrated circuit industry [4-5]. As new generations of
integrated circuits require advanced materials in order to have smaller and faster devices. So, thin layers play an
important role in the technology domain with new processing techniques [6,7].

So, thin film technology can play an important role in achieving several goals. As a two-dimensional system,
thin layers are of great importance to many real-world problems. The advantage lies in the cost of thin-film
materials which is very low when compared with other materials.

In addition, knowledge of the functions and / or applications of thin films under new properties can be used for
the development of new technologies for many future applications [8-9]. It is known that the thin film technology
is based on three foundations: preparation, characterization and applications successively. Among the
applications of thin films are: microelectronics, communication, optical electronics and catalysis, coatings of all
kinds and strategies of production and conservation of energy, etc. On the other hand, thin-film technology is
reducing the size of two-dimensional semiconductor devices.

710
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Recently, transparent conductive metal oxides (TCO) such as fluorine-doped tin oxide (FTO) and tin-doped
indium oxide (ITO) are very interesting materials. The reason is because of their many technological applications
such as optoelectronic devices and the production of electrical energy based on solar cells and [10,11]. For this,
the optical, electrical, structural and morphological properties of the TCO films are implicit in the determination
of performance. Generally, the family of TCOs are used in solar cells such as a window layer.

It is known that NiO films are the most commonly used window layer materials. NiO has a wide band gap
energy, high energy radiation stability, and photochemical properties [12,13]. It is also important to note that
solar cell electrodes, gas sensors, light-emitting diodes, acoustic devices, etc. can build them based on the NiO
film to discover new photoelectric materials. It has properties that allow its use in many applications [14,15]; as
it has very interesting electromechanical properties. Among the semiconductor oxide materials that is the most
multifunctional used in different fields for the manufacture of optoelectronic devices operating in the blue and
ultraviolet (UV) regions, there is NiO. In other ways, its high electrical conductivity and high transmission in
the visible region allows it to be one of the most likely materials for TCOs. NiO can be doped with various
metals [16]. The good electronic transport properties of NiO, allows it to use as an electron transport material in
hybrid solar devices.

Numerous studies have shown that high quality NiO thin films can be manufactured in several ways according
to different techniques and in different growth conditions. We will study in this work to define the structural and
optical characteristics of NiO / FTO thin films. The NiO films were prepared using the spray pyrolysis technique.
The spray pyrolysis technique is simple and inexpensive and can easily be applied to large scale deposits
compared with other techniques such as vacuum evaporation and sputtering. It is important here to note that the
FTO films were manufactured by a spray pyrolysis technique.

It has also been observed that other surface treatments, such as argon ion milling and polymer coating, reduce
the deep emissions and produce increased band edge emission in NiO nanostructures [17]. The reduced emission
of defects should make it possible to improve the performances of hybrid nanostructures based on NiO.

In this paper, the solution of nickel was carefully prepared for different concentrations (0.1M, 0.2M, 0.3M, 0.4M
and 0.5M) and under a fixed temperature which is plays a major role to determine the sample properties.

This paper is interested to extract a new formula of the energy gap (Eg) value from a transmittance curve. The
novelty of this work is to present an analytical model that leads to a new unified formula of Eg value. This
formula will lead us to ensure the good reading of the Eg value for several samples.
For this reason, an example of a test which is based on our films in order to show the accuracy of the formula
by comparing the obtained results with the previous studies (literature).
Finally, this work is dedicated to the photovoltaic systems applications.

2. EXPERIMENTAL TESTS

2.1 Preparations of thin film samples

Before starting to demonstrate the energy gap formula; we firstly indicate to the experimental preparation thin
conductive and transparent layers of NiO. Our preparation made different solution concentrations by spray
pyrolysis processing on glass substrates with fluorine-doped tin oxide (FTO). These films are obtained from
solution of nickel nitrate dissolved in methanol. Under a fixed temperature (T =480°C) and with different
concentration of solution which varies from 0.1 M to 0.5 M.
Figure 2 shows the UV-Visible transmittance spectrum of NiO/FTO layers at different concentrations. It is clear
that the energy gap will be different from one layer to another.

2.2 Measurements
In this case, we used a spectrophotometer (UV-Vis). Whereby we were able to plot curves representing the
variation in the transmittance depending on the wavelength in the field of UV-visible 300-900 nm (see Fig. 2).
By exploiting these curves; it is possible to estimate several quantities such as the thickness of the film, optical
characteristics; the optical absorption threshold, the absorption coefficient, the width of the forbidden band and
the refractive index.
Our proposed approach here is to determine the energy gap value using the transmittance spectra. The next

711
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

section will try to put the calculation in detail.

II. 3. ANALYTICAL MODEL OF ENERGY GAP CALCULATION

Everything in nature is subject to the laws that govern it. We therefore thought of a law to determine the energy
gap (Eg) value based on the transmittance curves. There are in the UV-Vis field some points are almost aligned
that would build a straight line. The intersection between the line and the curve above, noting it point B, and
below is the point A.
Theoretically the line formed is parallel to the transmittance axis and perpendicular to the wavelength axis. But
in reality the straight line is inclined, that is to say λB ≠ λA.
Point B is projected on the horizontal line which passes by point A; we notice this new point by C. As point B
and C have the same wavelength λB (see Fig. 1).
We found some researchers in this field that take the E g in point B; but other researchers take it in point A. This
led us to reflect on a realistic formula for defining Eg and comparing the results obtained between the practice
and the theory values.

Figure 1. Illustrative scheme of the transmission spectra

From the figure 1, we obtain a right triangle ACB, right angled at C; we determine the center of gravity G of the
triangle (see Fig. 2).

According to Fig. 1, we assume that the triangle points A, B, C have equal masses m. By applying the law of the
gravity center which is given by:

mOA  mOB  mOC

OG  (1)
mmm
or,
OA  OB  OC
OG  (2)
3
we can write also,
OA  A  i  TA  j (3)
OB  B  i  TB  j (4)
OC  B  i  TA  j (5)
so,
 A  2B   2TA  TB 
OG  i  j (6)
3 3

712
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

We are interested in this study on the part of the wave axis. The wavelength λG of the gravity center is written
in this case as follows:

 A  2B 
G  (7)
3

It is known that:
1240
Eg  (8)

According to equation (7) which defined the wavelength λG, we can extract the wavelength λ which
introduces in the calculation of Eg value by:
 A  2B 
  G 
3 (9)
so,
1240 1240 3720
Eg    (10)
  A  2B   A  2B 
3
The final formula for determining the Eg value is:
3720
Eg  (11)
  A  2 B 

This method allows us to determine the energy gap Eg value based on the transmittance curve simply, efficiently
and closer to the results known from the literature.

We have demonstrated by mechanical laws from the beginning that we consider the selected points as mass
centers as we have determined the center of mass of the proposed system in order to have a unified formula of
the Eg.

3.1 Applying the new formula on a real example

In Fig. 2, we presented the transmittance spectrum of the FTO / NiO films under different concentrations (to the
left for 0.2 M).

Figure 2. Transmittance spectra of the FTO and the NiO thin films grown on FTO substrates at various sol
concentrations.

The band gap or energy gap (Eg) values of our NiO/FTO layers are between 3.38 eV and 3.74 eV. According to
Fig. 2, we can see a variation of the Eg value with the concentration of the solution. From a concentration of 0.2
M presented in Fig. 2 (to the left); the Eg = 3.68 eV value is very logical with the literature.
3720 3720
Eg  
 319.3  2  (345.4)  1010.1
so,

713
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Eg  3.68 eV
Note: In this occasion, we can define another formula that is based on the calculation of the average value of
each energy gap value, as follow:
1240
E gA  (12)
A
and,
1240
EgB  (13)
B
E gA  EgB
E g  Mean  Eg  (14)
2
or,
 A  B 
Eg  Mean  Eg  620  (15)
 A  B 
In this case, we can found the E g value as follow:
 319.3  345.4
Eg  Mean  Eg  620 
 319.3  345.4
so,
Eg Mean  Eg  3.736 eV
It is clear that both values are not equal (3.682 eV with 3.736 eV). For this reason, our analytical model shows
a very good approximation when comparing it with literature.
3.2 Discussion of Results
Pankove [18] proposed a technique based on traced optical absorbance data to determine the energy gap. The
work of Tauc et al. was applied to the study of the optical and electronic properties of amorphous germanium.
The Tauc diagram is a graph commonly used to determine the optical band gap of a material, usually a
semiconductor. The optical band gap can determine using the Tauc’s relation [18, 19] which is given by:
 h  p  A h  Eg  (16)

Where, α is the absorption coefficient, h is the Planck's constant, A is a constant, ν is the transition frequency,
Eg is the band gap corresponding to a particular transition occurring in the film and p can take the values
according to:
- Direct allowed transitions: p=2,
- Direct forbidden transitions: p =2/3,
- Indirect allowed transitions: p= 1/2,
- Indirect forbidden transitions: p= 1/3.
Generally, the allowed transitions dominate is based on the absorption processes.
so,
p=2 or p=1/2, in case when the transitions are direct and indirect, respectively.
This new approach has led us to find a well-defined formula of Eg. The logical steps during the demonstration
confirm the accuracy of the results. In addition, the example proposed in this study (section 4.1) shows the
effectiveness of this formula when we compare it with other works.
In this paper, we have tried to put a new approach that helps to avoid the overlap in the E g calculation.
The transmittance of samples and their evolution in terms of wavelength are represented below.
Initially we compared single-layer (FTO) and multi-layer samples (NiO/FTO) with different
concentrations to achieve our goal.

UV-visible and FTIR spectra have been measured in the chemical and process engineering department of El-
Oued University.
The wavelengths belong to 200nm and 900nm.
According to Fig. 2, the optical band-gap is seen to decrease from 3.74 eV to 3.38 eV when the sol concentration
is increased or decrease from 0.1 to 0.5 M. The optical band-gap energies of the NiO thin films with sol
concentrations is 3.38 eV in 0.5 M. But the transmittance value for 0.5 M is very low comparing with 0.1 M
which has Eg=3.74 eV. We can verify that similar results were found by other works. The changes in the optical
band-gap have been suggested to be due to changes in the thickness and the strain in the films. We compared

714
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the energy gap values for sol concentrations between 0.1 and 0.5 M based on theclassical method and the new
formula as showed the table as follows:

Table 1. Energy gap Value of Different NiO Concentrations

NiO Concentration 0.1M 0.2M 0.3M 0.4M 0.5M FTO
Eg for Classical Method (eV) 3.93 3.92 3.86 3.79 3.75 3.67
Eg for New Method (eV) 3.74 3.68 3.57 3.52 3.38 3.62

It is clear that the results obtained are in correspondence with the results of the conventional calculation method.
This confirms the accuracy of the proposed model.
It is also important to note that the verification was made for other materials such as ZnO, CuO, etc.

5. CONCLUSION
In conclusion, we have proposed a new analytical model based on the transmittance curve to calculate the Eg
value. Our study was based on a real example under different concentrations in order to calculate the Eg value
precisely. This example is composed of a thin film which contains NiO/FTO. In addition, we carefully checked
the values obtained with other results. We found the value of Eg = 3.68 eV which was found in correspondence
with the literature. Equation (11) gives a new formula which has been found according to well-defined steps.
It may be important to facilitate the calculation of the Eg value in the subjects that use the transmittance curves
in the materials which are dedicated to the photovoltaic systems or others uses.

6. ACKNOWLEDGMENTS
This paper was partially supported by VTRS laboratory in university of El-Oued, Algeria. We thank all
colleagues in the VTRS laboratory for any help to us.

7. REFERENCES
[1] Jagannathan T. Introductory Chapter. Thin Film Processes-Artifacts on Surface Phenomena and Technological Facets.
The Prominence of Thin Film Science in Technological Scale, 2017.
[2] Orkut S. Technological Background and Properties of Thin Film Semiconductors. In : Surface Science. IntechOpen,
2020.
[3] Rao, M. C, Shekhawat, M. S. A brief survey on basic properties of thin films for device application. International Journal
of Modern Physics 2013; p. 576-582.
[4] Douglas B. F. Chip design in China and India: Multinationals, industry structure and development outcomes in the
integrated circuit industry. Technological Forecasting and Social Change 2014, vol. 81, p. 1-10.
[5] Dan B. Integrated Circuit Design and Software in Taiwan. Global Taiwan: Building Competitive Strengths in a New
International Economy 2005, p. 194.
[6] Se-Hwa W. The dynamic cooperation between government and enterprise: The development of Taiwan's integrated
circuit industry. Taiwan's Enterprises in Global Perspective 1992.
[7] Bernard G, Rabiner L. Parallel processing techniques for estimating pitch periods of speech in the time domain. The
Journal of the Acoustical Society of America 1969, vol. 46, p. 442-448.
[8] Paola V, Michele M. Using text processing techniques to automatically enrich a domain ontology. In : Proceedings of
the international conference on Formal Ontology in Information Systems-Volume, October 17-19, 2001, Ogunquit,
Maine, USA, p. 270-284.
[9] Bellas I, Tassou S. A. Present and future applications of ice slurries. International Journal of Refrigeration, 2005, vol.
28, no 1, p. 115-121.
[10] Jenny P, Ian B. New product development from past research to future applications. Industrial Marketing Management,
1998, vol. 27, no 3, p. 197-212.
[11] Kyusang L, Jaesang L, Mazor B. A. Transforming the cost of solar-to-electrical energy conversion: Integrating thin-
film GaAs solar cells with non-tracking mini-concentrators. Light: Science & Applications 2015, vol. 4, no 5, p. e288-
e288.
[12] Hongxun Y, Miaoliang H, Jihuai W.U. The polymer gel electrolyte based on poly (methyl methacrylate) and its
application in quasi-solid-state dye-sensitized solar cells. Materials chemistry and physics 2008, vol. 110, no 1, p. 38-
42.
[13] Helmut F, Friedrich S, Jürgen R. Cationic 2‑Azaallenylidene Complexes, Novel Compounds with Unusual
Photochemical Properties. Angewandte Chemie International Edition in English 1985, vol. 24, no 2, p. 121-123.
[14] Dorota C. D, Marta Z. B, Magdalena S. The chitosan–Porphyrazine hybrid materials and their photochemical
properties. Journal of Photochemistry and Photobiology B: Biology 2018, vol. 181, p. 1-13.

715
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[15] Marguerite R. Chitin and chitosan: properties and applications. Progress in polymer science 2006, vol. 31, no 7, p.
603-632.
[16] Huber D. L. Synthesis, properties, and applications of iron nanoparticles. Small 2005, vol. 1, no 5, p. 482-501.
[17] Jae-ha M, Tae Ho S, Xiubing H. Enhancement of redox stability and electrical conductivity by doping various metals
on ceria, Ce1− xMxO2− δ (M= Ni, Cu, Co, Mn, Ti, Zr). International journal of hydrogen energy 2015, vol. 40, no 35,
p. 12003-12008.
[18] Karthik K, Dhanuskodi, S, Gobinath C. Nanostructured CdO-NiO composite for multifunctional applications. Journal
of Physics and Chemistry of Solids 2018, vol. 112, p. 106-118.
[19] Tauc J, Grigorovici R, Vancu A. Optical Properties and Electronic Structure of Amorphous Germanium. Physica status
solidi 1966, vol. 15, pp. 627-637.
[20] Pankove J. I. Optical Processes in Semiconductors 1971.

716
 SIMULATIVE EVALUATION OF TRIBOELECTRIC POLYMER
PAIRS IN VERTICAL CONTACT MODE

Shimna Shafeek
University of Bolton, United Arab Emirates, s.shafeek @bolton.ac.uk, ORCID: https://orcid.org/0000-0001-7396-6808

Sibgatulla Sharieef
University of Bolton, United Arab Emirates, [email protected], ORCID: https://orcid.org/0000-0002-9816-9124

Shafeek, Shimna, Sharieef, Sibgatulla. Simulative Evaluation of triboelectric polymer pairs in

Cite this paper as:
Vertical Contact mode. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract:
Triboelectric nanogenerators are shown a recent development in the energy field in various
applications powering sensors to biomedical applications. The research development of
tribogenerators is trending in the renewable energy area as it can harness waste mechanical energy
due to the friction. Studies have shown various mathematical modeling done on the triboelectric
principle based on Gauss electric field principle. Triboelectricity generation due to contact
electrification depends on various factors which include the surface charge density, materials, the
geometrical features of the tribo pairs, the mode of operation in terms of velocity etc. The
significance of nanomaterials in the generation of triboelectricity is a research area where polymers
have shown good results. Here in this study a detailed computational and numerical simulation is
done on selected pairs of triboelectric material combination chosen from the triboelectric series.
Computational simulation is performed using Comsol multiphysics to evaluate the output
performance in terms of Voc and Qsc. Numerical simulation is performed using MATLAB to
evaluate the output performance Current, Power, Voltage with respect to time for selected input
parameters. The method adopted will be a useful tool for determining the output characteristics of
any triboelectric pairs.
Keywords: numerical, computational, simulation, triboelectric generator, energy harnessing
© 2020 Published by ECRES

1. INTRODUCTION

In the era of harnessing energy from various sources which lead to the production of waste, energy has been
trending in the research field. Due to the depletion of fossil fuels and an increase in the demand of it, the
replacement of fossil fuels by the renewable energy source attracted the world at a high level [1]. The world is
moving towards artificial intelligence and IoTs where the demand for a huge number of the energy source as the
battery is required [2]. The limited lifetime, maintainability, cost, continuous monitoring of the performance of
the devices are some of the key drawbacks of the batteries. The idea of powering devices from its mechanism
leads to the emerging sustainable solution of systems to work as self-powered [3]. This sustainable energy
solution made the industry to think about the sensors to work as a self-powered using an active autonomous
power source.
Various energy harnessing mechanical methods are adopted during the years which includes electromagnetic
generator, electret generators and piezoelectric generators for exploiting the ambient environment by harvesting
mechanical energy and providing self-powered devices. The requirement of heavy permanent magnets in
electromagnetic generators, the necessity of pre-charging and low electrical power in electret generators and the
low output and toxicity in piezoelectric generators are considered as the drawback and thereby commenced on
researching on the new energy source triboelectric generators as a cheap and efficient energy source for low
powered systems[4-5]. From the time of the invention of Triboelectricity and Triboelectric Generators (TEG)by
Wang (2012), the studies are going on powering the low power devices like sensors, its power management
circuits and optimal output performance behaviour.

717
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Triboelectric generators work on the principle of contact electrification and electrostatic charge generation
where two materials having varying electron affinities come in contact will generate the charges thereby
developing an electric potential between them. If the system is assumed as a closed-circuit electron flow takes
place thereby electricity generation happens. The contact between the materials are made by small movements,
vibration, environmental changes thereby charges are produced by friction [6]. The material pairs used in the
TENG are termed as tribo pairs where the materials with a maximum difference in electron affinity are chosen.
With the development of four modes of operation of triboelectric generators to generate electric potential,
vertical contact mode where tribo materials are vertically moved apart; sliding mode where the tribo materials
are slide apart, single electrode mode and free-standing mode the TENGs are used in various applications as
energy recovery and regeneration system. Considering the mechanical wear of the materials in contact, vertical
contact mode is of high significance in the research area for its improvement[7].
Studies have shown the efficiency of the TENG depends on the material aspects in terms of the morphology,
geometry and properties. Triboelectric materials are arranged in a triboelectric series based on the electron
affinities. Dielectric polymer tribomaterials are showing good usage in the industry for TENG [8]. There are
various analytical methods generated theoretically to find the output performance of TENG. In this study, the
dielectrics as tribopairs will be analyzed theoretically and numerical, the computational simulation will be
performed on different recently used tribopairs in applications thereby finding the optimal pair for a set of input
parameters.

2. POLYMERS AS TRIBOELECTRIC PAIR

Triboelectric generators work when a pair of material rub against each other thereby there will be a tendency of
one material to donate electrons while the other to accept electrons. When these materials are having a chance
to mechanically separate each other, it will create an air gap which results in dipole moment on the charges
generated. To equalize the electric potential generated a small current will flow when an electric load is
connected between the contacting surfaces. If continuously the materials are in contact and separated by a
distance, it will provide small alternating current[9]. For the material selection as tribopairs, a high difference in
electron affinities are considered as highly good as a choice for good performance. Metals like Al, Cu, Au etc.
are good electron donors to act as tribo positive pair, but due to the oxidation and highly reactive tendency of
these materials to form an oxide layer made to select polymers as tribo positive layer[10]. Among the polymers,
studies shown PA6/Polyamide6 as a good tribo positive layer due to its excellent donor characteristics whereas
polymers like PTFE, FEP, PDMS, KAPTON, PVDF are tribo negative layers due to negatively charged
characteristics. Polymers superpower of charge transfer and capturing capacity when the tribomaterials come
into contact thereby generating friction enabled it to use for a wide range of applications in the triboelectric
field[11]. Though a huge number of tribo combinations can be made with an enormous number of materials, all
combinations will not be able to provide good results. The various properties which determine the output
characteristics of TEG pairs include the surface charge density, dielectric constant of materials, work function,
frictional coefficient, the geometry of material combinations etc. Studies have shown that the surface charge
density depends on the chemical potential difference between the tribo pairs and is given by the equation[12],
𝑡
[(𝑊−𝐸0 )𝑒](1+ )
ℇ𝑧
𝜎=𝑡 (1)
⁄ℇℇ +(1√𝑁𝑠 (𝐸)𝑒 2 )(1+𝑡⁄ℇ𝑧 )
0

where W is the work function of the metal, E0 is the charge-neutrality level of the surface states, e, t, ε, ε0 and
z are the charge of an electron, distance of space, relative permittivity of dielectric, vacuum permittivity of free
space and thickness of the dielectric film and Ns (E) is the density of surface states. Based on this since the work
function between tribo material contact increases surface charge density increases thereby increase in output
performance. In addition to work function, the dielectric constant is an important parameter based on which the
TEG performance vary. The mathematical expression for effective surface charge density which can be
transferred (σ ‘) in terms of the initial material surface charge density (σ 0) is given by [13]
−𝜎0𝑑𝑔𝑎𝑝
𝜎′ = 𝑑 (2)
𝑑𝑔𝑎𝑝 + 1⁄ℇ
1
where dgap is the gap distance,d1, ℇ1 are the thickness and dielectric constant of tribo material pair. The equation
shows the effective surface charge density increases with the increase in dielectric constant.

718
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. THEORETICAL OUTPUT CHARACTERISTICS OF TEG IN VERTICAL CONTACT MODE

The output characteristics Voltage, Current and Power are derived based on Gauss electric field equation where
the internal electric field of each part of TENG is calculated first. Since the size of the dielectric is much larger
than the air gap and the thickness of the electrode, it can be assumed that the electric field is like an infinite
electric plate. The equations are derived from considering three dielectric areas, of which two are dielectric pair
of εr1 and εr2 as the dielectric constants, d1 and d2 their thickness.

Figure.1. TEG in vertical contact mode

𝑄
𝐸𝑑𝑖−𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐1 = 𝐴𝜀 (3)
0 𝜀𝑟1

𝑄
𝐸𝑑𝑖−𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐2 = 𝐴𝜀 (4)
0 𝜀𝑟2

1 𝑄
𝐸𝑎𝑖𝑟−𝑔𝑎𝑝 = 𝜀 (𝐴 − 𝜎) (5)
0

Where Q is the charge generated by material in coulomb per meter square, A is the contact surface area of the
dielectric material, ε0 is the relative permittivity of vacuum [14-15].

For a separation distance of x(t), of surface charge density σ, the electrical potential generated across the
electrode given by

𝑄 𝑑 𝑑 𝜎
𝑉(𝑡) = − 𝐴𝜀 (𝜀 1 + 𝜀 2 + 𝑥(𝑡)) + 𝜀 𝑥(𝑡) (6)
0 𝑟1 𝑟2 0

Assuming for open circuit (Q=0) and short circuit (V=0);

𝑉𝑜𝑐 = 𝜎𝜀0 /𝑥(𝑡) (7)

𝑄𝑠𝑐 = 𝐴𝜎𝑥(𝑡)/(𝑑0 + 𝑥(𝑡)) (8)

Equating in Ohms law, 𝑉 = 𝑅𝑑𝑄/𝑑𝑡

Where assuming the motion to be linear with velocity v, maximum distance as x max [16-18]

Assuming time taken to reach xmax as tmax , where tmax = xmax/v,

Using the below displacement velocity concept condition, x (t) = vt (t ≤ tmax) and x(t) = xmax (t ≥ tmax ) charge
generation with respect to time is given by
2
𝜎𝑣0 2 𝑒 𝐴𝑡+𝐵𝑡 𝐴 𝐴 1
𝑄(𝑡) = 𝑅𝜖0
𝑒 −(𝐴𝑡+𝐵𝑡 ) × [ 2𝐵
− 𝑀. 𝑒𝑟𝑓𝑖 (√𝐵𝑡 + 2√𝐵 ) + 𝑀. 𝑒𝑟𝑓𝑖 (2√𝐵 ) − 2𝐵 ] (9)
Error function

719
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2
2𝑒 𝑥
𝑒𝑟𝑓𝑖(𝑥) = ∫ 𝑑𝑥 (10)
√𝜋

Constant A
𝑑0
𝐴= 𝑅𝐴𝜀0
(11)

Constant B
𝑣0
𝐵= 2𝑅𝐴𝜀0
(12)

Constant M
𝐴2
𝐴√𝜋
𝑀= 𝑒 4𝐵 (13)
4𝐵√𝐵

From Q, instantaneous current, Voltage and Power can be found by the following equations.

Instantaneous Current
𝑑𝑄(𝑡)
𝐼(𝑡) = (14)
𝑑𝑡

Instantaneous Voltage

𝑉(𝑡) = 𝑅𝐼(𝑡) (15)

Instantaneous Power

𝑃(𝑡) = 𝑉(𝑡)𝐼(𝑡) (16)

4. SIMULATIVE EVALUATION

From the above discussions, the performance of triboelectric generators is influenced greatly by the material
properties and thereby the type of material used for the implementation. Usually, materials used for the
tribogenerators are nanomaterial which can give good surface contact and hence the necessity of the right choice
of materials required because of its cost. Though a vast range of materials from polymer itself there which can
give output, the right combination can be found from numeric and computational analysis and evaluation. For
the validation of the simulation tools, Comsol and MATLAB with the theoretical concept of TEG are done
initially for a pair of tribo materials in the first phase. With the same methodology, five pair of TEG is chosen
with PA6 as the positive tribo pair, the other polymers PVDF, PDMS, PTFE, FEP, and Kapton as tribo negative
pair to find the optimal pair showing the good output characteristics.

Phase Study I Validation of Simulation tools

For the phase I study, a set of input data as shown in Table I. are used to perform the numerical and computational
analysis of open circuit voltage short circuit discharge based on the theoretical data from Eqn. [7] and [8].
Table I. The Input data for phase I
Serial no Material Dielectric Thickness
Constant m
1 Tribomaterial1PVDF 𝜀𝑟1 = 7.4 d1 =0.1 × 10−4
2 Tribomaterial 2PA6 𝜀𝑟2 = 4 d2 =0.3 × 10−4
3 Contactarea A,m 2
4 × 10−4
4 Maximum 4 × 10−3
distance(x)m

The other assumption taken for the model is the surface charge density of 34.2µC/m2 for the study.

720
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)

(c) (d)
Figure2.Numerical and Computational study (a) Qsc vs distance (b)Voc vs distance (c)Voc distance 246.5µm spacer gap
(d) Voc distance 396.5µm spacer gap

The result showed a good agreement between the numerical and Comsol results. The Voc and Qsc value for
both the analysis results are shown in Figure 2(a) and (b). The theoretical value matches with the experimental
values which show the Comsol and Matlab simulation in agreement with the theoretical concept of TEG.

Phase Study II Simulation Study of Polymer Tribopairs

Numerical study on tribopairs of five sets of polymers is done in Matlab. The input data chosen for the study is
based on the material properties and polymer characteristics fit for the tribo device especially in the case of
mechanical properties wear and tear of the materials. The assumptions taken for the study includes the surface
charge density as 20 μC/m2 and all the combination is subjected to a resistive load of 100Mohm. The contact
area A is taken as 50x10-4m2, dielectric thickness 1 and 2 taken as 0.1x10-3m. All the combinations are subjected
to vertical contact mode for a maximum spacer distance of 5mm, the velocity of 0.1m/sec for a time period of
20ms. PA6 is taken as tribo positive pair for all other materials in Table II.
Table II. The Input data for phase II
Item no Material Relative Permittivity
1 PA6 4
2 PVDF 7.4
3 PDMS 2.75
4 PTFE 2
5 FEP 2.1
6 KAPTON 3.34

721
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)

(c) (d)

(f)
(e)
Figure3.Numerical & Computational simulation for polymer tribo pairs (a)Qsc vs x (b) Comsol Qsc vs x (c) Q vs t
(d)P vs t (e) V vs t (f) I vs t

Simulation is done based on the theoretical Eqs. [9-16].From figure3. (a)-(f) PA6 PTFE combinations shown
good output performance characteristics in terms of output current, voltage and power.PA6 FEP combination
also showed near performance to that of PTFE pair. Due to the presence of strong C-F bond and hence its high
electronegativity results in the strong pull of electrons during the material contact. The PA6 PTFE simulation
showed a maximum power of 0.75mW/m2 with the assumed surface charge density of 20µC/m2 when a load of
100Mohm connected in the circuit. More power can be expected by improving the material morphology and
properties thereby improving the surface charge density. Since high surface charge density produces high output
power, by patterning the tribo pairs at the contact surface can improve the contact area, ionized air injection can
also enhance the tribo pair output power performance.

722
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

5. CONCLUSION

Triboelectricity being a promising solution for harnessing waste mechanical energy, the study to enhance the
output performance of the device is an emerging area in research. The flexibility of choosing a variety of
materials from organic to inorganic and to polymers, composite as tribo material pairs made the technology
cheap and reliable. Polymer pairs as tribo materials are giving good performance characteristics compared to all
other material combinations. Efficient tribo material combination by simulative evaluation shown
fluoropolymers PTFE, FEP with PA6 good output performance. Further studies can be done by evaluating the
output power with varying load resistance for all the tribo polymer combinations. The validated simulation
generated in the current study can be extended further to evaluate the output performance of any tribo material
pair before the experiment and hardware implementation.
ACKNOWLEDGEMENT

This research is supported by the Expo Live Innovation Impact Grant Programme 2020 who granted the
University of Bolton RAK Academic Centre AED 50,000 for winning the research and innovation university
summit of Expo 2020. We would also like to show our gratitude to Jikui Luo, Research Supervisor, University
Of Bolton, Bolton, BL3 5AB, United Kingdom, for sharing and mentoring the pearls of wisdom with us during
this research.

REFERENCES

[1] Bhamre,S,Mali,S,Mane,C. Optimization of electric vehicle based on triboelectric nanogenerator. In :E3S Web of
Conferences,2020, EDP Sciences, Vol. 170, p. 01027.
[2] Qiu, C, Wu, F, Lee, C, Yuce, M. R. Self-powered control interface based on Gray code with hybrid triboelectric and
photovoltaics energy harvesting for IoT smart home and access control applications. Nano Energy 2020, 70:
104456<https://doi.org/10.1016/j.nanoen.2020.104456>
[3] Fatma, B, Gupta, S, Chatterjee, C, Bhunia, R, Verma, V,Garg, A. Triboelectric Generator made of Mechanically Robust
PVDF Film as Self-powered Autonomous Sensor for Wireless Transmission Based Remote Security System. Journal
of Materials Chemistry A 2020, < https://doi.org/10.1039/D0TA04716C>
[4] Askari, H, Hashemi, E, Khajepour, A, Khamesee, M. B, Wang, Z. L. Tire Condition Monitoring and Intelligent Tires
Using Nanogenerators Based on Piezoelectric, Electromagnetic, and Triboelectric Effects. Advanced Materials
Technologies, 1800105 <https://doi.org/10.1002/admt.201800105>
[5] Wu, C, Wang, A. C, Ding, W, Guo, H,Wang, Z. L. Triboelectric nanogenerator: a foundation of the energy for the new
era. Advanced Energy Materials 2019, 9(1), 1802906 < https://doi.org/10.1002/aenm.201802906>
[6] Wang, Z. L, Lin, L, Chen, J, Niu, S, Zi, Y. Triboelectric nanogenerators. Basel, Switzerland: Springer International
Publishing, 2016.
[7] Yoo, D, Go, E. Y, Choi, D, Lee, J. W, Song, I, Sim, J. Y,Kim, D. S. Increased interfacial area between dielectric layer
and electrode of triboelectric nanogenerator toward robustness and boosted energy output. Nanomaterials 2019, 9(1),
71 < https://doi.org/10.3390/nano9010071>
[8] Baik, J. M,Lee, J. P. Strategies for ultrahigh outputs generation in triboelectric energy harvesting technologies: from
fundamentals to devices. Science and Technology of Advanced Materials 2019, 20(1), 927-
936https://doi.org/10.1080/14686996.2019.1655663
[9] Kim, D. W, Lee, J. H, Kim, J. K, Jeong, U. Material aspects of triboelectric energy generation and sensors. NPG Asia
Materials 2020, 12(1). doi:10.1038/s41427-019-0176-0
[10] Choi, D, Park, Y. T, Kim, S. H, Park, J. H, Woo, C. S, Lee, K. S, Kook, M. J. Nanogenerators in Korea. MDPI-
Multidisciplinary Digital Publishing Institute, 2019.
[11] Chen, A, Zhang, C, Zhu, G,Wang, Z. L. Polymer Materials for High‐Performance Triboelectric Nanogenerators.
Advanced Science, 2000186< https://doi.org/10.1002/advs.202000186>
[12] Lee, L. H.Dual mechanism for metal-polymer contact electrification. Journal of electrostatics 2005, 32(1), 1-29.<
https://doi.org/10.1016/0304-3886(94)90026-4>
[13] Lee, J. W, Ye, B. U,Baik, J. M. Research Update: Recent progress in the development of effective dielectrics for high-
output triboelectric nanogenerator. APL Materials 2017, 5(7), 073802.<
https://doi.org/10.1063/[email protected]>
[14] Barkas, D. A, Psomopoulos, C. S, Papageorgas, P, Kalkanis, K., Piromalis, D, Mouratidis, A.Sustainable Energy
Harvesting through Triboelectric Nano–Generators: A Review of current status and applications. Energy Procedia
2019, 157, 999-1010< https://doi.org/10.1016/j.egypro.2018.11.267>
[15] Niu, S.,Wang, Z. L.. Theoretical systems of triboelectric nanogenerators. Nano Energy (2015), 14, 161-
192,https://doi.org/10.1016/j.nanoen.2014.11.034

723
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[16] Abdelwahed, A., Amin, M., Elosairy, M., Abbasy, N. Theoretical modelling for enhancing contact-separation
triboelectric nanogenerator performance:In 2016 Annual Connecticut Conference on Industrial Electronics,
Technology & Automation (CT-IETA) (October 2016),IEEE, pp. 1-5.
[17] Shafeek, S., Luo, J. Theoretical and numerical analysis of triboelectric nanogenerators for self-powered sensors :In
2016 5th International Conference on Electronic Devices, Systems and Applications (ICEDSA) (2016, December),
IEEE,pp. 1-4.
[18] Varghese, S., Shafeek, S., Kumar, R. S.,Mini, R. S. Computational investigation of material combinations in
triboelectric generators : In 2017 International Conference on Electrical and Computing Technologies and Applications
(ICECTA), (2017, November), IEEE,pp. 1-4.

724
 THEORETICAL STUDY ON PRESSURE EFFECTS ON PHYSICAL
PROPERTIES OF SEMICONDUCTOR COMPOUND LICDP

Yasemin Öztekin Çiftci

Gazi University, Department of Physics, Ankara, TURKEY. [email protected], ORCID: https://orcid.org/0000-0003-
1796-0270

Çiftci,YO, Theoretical study on Pressure effects on Physical Properties of Semiconductor

Cite this paper as:
Compound LiCdP, 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Recently, due to their important features as large thermal conductivity, large thermopower, high
melting point and interesting magnetic fautures etc., Half-Heusler (HH) compounds which are
small band gaps are an interesting materials. Pressure dependence of the structural, electronic,
elastic, vibrational and optical properties of LiCdP in MgAgAs structure have been studied using
first principles calculations depend upon density functional theory (DFT). The extracted structural
parameters at 0GPa are agreed well with the other results. Calculated band structure graph indicates
a direct energy band gap of 0.23 eV at Γ Brillouin zone fort this compound. Additionally, elastic
constants, phonon frequencies and optical quantities were calculated. Elastic constants have been
derived by utilizing stress- strain technique. According to obtained elastic results, LiCdP are
mechanically stable up to 20 GPa. The calculated phonon graphs of LiCdP does not exhibit
negative frequencies any where indicating it’s dynamically stabilities. Investigations of the optical
properties have been obtained for the energy range of 0–30 eV. Our calculated structural results
are in good areement with the literature.
Keywords: DFT, LiCdP, semiconductors, electronic, elastic, vibrational, pressure
© 2020 Published by ECRES

1. INTRODUCTION

Half-Heusler (HH) compounds have been attracted increasing attention in promising new semiconducting
technology. The investigations on HH semiconductors are novel materials for the improvement of these types
of technological materials. Since Half-Heusler materials has large temperature stability, they can be used as
efficient thermoelectric materials [1]. Alloying of the Half Heusler compounds may reduce thermal conductivity
because of mass fluctuation[2]. Generally, HH compounds form in a MgAgAs phase corresponding space group
F 4 3m (space number: 216))[3]. In last years, the HH compounds which have ABC form where A and B show
a transition element and C includes the main group element earned reputation due to their potential applications,
such as thermoelectrics, green energy-related fields and spintronics [4].

The growing and utilization of optoelectronic instruments depend vitally on the existence of favorable
semiconductor materials The conformity in physical properties between the HH compounds and the pair
semiconductors provides to discover novel half-Heusler semi-conductors as seen by Kacimi et al[5]. Gruhn [6]
have studied various HH compounds to select suitable materials for optoelectronic technologies. Mehnane et al.
[7] have performed the theoretical study on the structural, electronic and optical features for 36 HH composites
indicating new applicant of the optoelectronic implementations. For I–II–V and I–III–IV types 96 HH
compounds, structural electronic, and optical properties have been examined using first-principles calculations.
Recently, the authors have studied electrical, optical, and chemical properties[7].

725
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In this work, the pressure effects on a structure, electronic, elastic, vibration and optical quantities of LiCdP
compound which has HH semiconductors family have been researched and checked our calculated results with
the other results.

2. METHOD AND CALCULATIONS

To examine the properties of structural, electronic and elastic, vibrational and optical of LiCdP, the first-
principle studies have been used by CASTEP[8,9]. For the exchange-correlation terms, it has been chosen
Perdew-Burke-Ernzerhof (PBE) parametrization due to generalized gradient approximation (GGA) [10]. The
electron-ion interactions were taken by the Ultrasoft Vanderbilt pseudopotential [11,12]. For each elements, the
electronic valence configurations were chosen as Ru: 4s24p64d75s1, Os: 5s25p65d66s2, and Hf:5d26s2 .The kinetic
energy cutoff were used as 600 eV. In the irreducible Brillouin zone, a set (12x12x12) Monkhorst-Pack [13] k-
points grid are used. To obtain a well converged total energy, these parameters are carefully tested.

3. RESULTS

Structural and electronic properties

Unitcell of LiCdP which forms in the MgAgAs-type structure with space group 216 is given in Figure 1. Here,
Li, Cd and P atoms locate on atomic positons (0.5; 0.5; 0.5), (0.25; 0.25; 0.25) and the (0; 0; 0). To obtain
structural parameters we need to find the equilibrium geometry. For this aim, the unit cell size and the atomic
locations are relaxed. To determine the Bulk modulus (B), E (V) curves obtained different with lattice constant,
have been fitted the Murnaghan equation of state. Our obtained structural parameters of title compound are given
in Table I

Figure 1. Unitcell of LiCdP

Table 1. Calculated structural parameters and other calculated results for LiCdP
a0 (Å), V0 (Ǻ3) B(GPa)
Present 6.15 233.2 52.95
Theory a 6.13 54.96
a: [5]
For LiCdP, the divided by the zero pressure structural parameters, lattice parameters ( a/a0 )and cell volume
(V/V0 ), under pressure up to 60 GPa are given in Fig. 2. When pressure rises from 0 to 60 GPa, the ratio a/a0
and V/Vo decrease with pressure approximately by 13.4% and 35.1%. Unfortunately, to compare our results there
are no experimental data. It can be obtained the following quadratic function relationships for LiCdP by adjusting
lattice parameter and volume curves with the pressures.

a/ao= 0.995 - 0.0038xP + (2.964x10-5)xP2 (1)

V/Vo= 0.986 - 0.0109xP + (9.082x10-5)xP2 (2)

726
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2: Pressure dependence of lattice parameter and cell volume for LiCdP.

The electronic properties of half-Heusler LiCdP semiconductor were investigated by the electronic band
structures with total and partial density of states and were depicted in Figures 3 and 4 at optimized lattice
constant. It can be seen from Fig. 3 that in the electronic band structure and total density of states graphs, both
the maximum of valance band and the minimum of the conduction band is located at Γ-point indicating direct
band gap material with a band gap of 0.23 eV at zero pressure (The Fermi energy is set to zero) which is smaller
than that of Kacimi et al. [5] (0.55 eV). One knows that the calculated the band gap energy are smaller than the
experimental data because of low prediction of conduction band energies within DFT.

Fig.2: Electronic band structure and partial DOS LiCdP at 0GPa

Figure 4. Partial density of states of half-Heusler LiCdP compound

Partial and total density of states for LiCdP at 0GPa are obtained and given in Fig. 4. In Fig.4, the valence bands
seperate in two regions. The lowest energy band region of valence band is occurred by Cd –p states. The second

727
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

energy bands near the Fermi level come from the hybridization of P- p states and Cd–s states. Above the Fermi
level, in the conduction band, the great addition originates from Li s-states and P p-states with Cd –s states. Due
to band gap seen in Fig.4 The half-Heusler LiCdP compound is a semiconducting property.

As a function of the Eg is obtained from band structure graphs for different pressure and demonstrated in Fig. 5,
and the relation between band gap values and pressure can be formulated following :

Eg=0.363-0.0439*P +5.024*10-4*P2

Noticeably, the value of Eg increase with increasing pressures and this compound exhibite almost exponantional
relation between the Eg and the pressure ranges from 0 to 60 GPa.

Figure 5.:Obtained band gaps energies of LiCdP under pressure up to 60 GPa.

Mechanical properties

The elastic constants show reaction of a material under extirinsic strength. The stabilities and hardness of a solid
can be determined by obtained the elastic features like shear, bulk, and Young’s moduli and Poisson’s ratio. To
calculate elastic constants, the “stress-strain” method [14] was used. The pressure influence on the elastic
variations of LiCdP between 0 and 60 GPa was obtained. At the transition point, thf variation of the elastic
constants may effectively change the macroscopic properties of solids [15]. The obtained elastic constants and
mechanical properties obtained from elastic constants with variations of pressure are given in Table 2. It is
considered that the elastic constants should provide Born’s stability criteria[16]. For mechanical stability of
cubic crystals, well known Born’s stability terms are: C11>0, C11-C12>0, C44>0, C11+2C12>0 and C12<B<C11.
MgAgAs phase of LiCdP compound are mechanically stable up to 20 GPa according to Born’s stability criteria.

Table 2: Elastic properties of LiCdP under pressure up to 50 GPa.

Press C11 C12 C44 B G E V
ure0 55.91 51.47 19.61 52.95 8.699 6.564 0.479
10
(GPa 96.32
(GPa) 93.88
(GPa) 30.852
(GPa) 94.699
(Gpa) 10.933
(GPa) 3.631
(Gpa) 0.493
20 132.104 133.08 27.917 132.754 7.6512 -1.4658 0.501
)
30 166.935 172.01
0 28.760 170.320 4.463 -7.652 0.507
40 200.115 209.80
2 21.792
1 206.577 -3.523 -14.655 0.511
50 230.351 9
245.07 22.276 240.1 -13.03 -22.31 0.515
4

From Table 2 that estimations of C11 , C12 and C44 elastic constants tend to increase with pressure from 0 to 60
GPa.

Shear modulus and bulk modulus show measure of resistance to reversible deformation and while resistance to
volume change, respectively. Bulk (B) and shear moduli(G) which use the Voigt-Reuss-Hill formula are

728
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

calculated utilizing three elastic constants obtained for cubic structure[17,18]. Also, Poisson's ratio (v) and the
Young's modulus (E) can also be obtained using elastic constants [19 ,20, 21]. Calculated elastic parameters for
each P values are given in Table 2.

CONCLUSİONS

Here, under pressure semiconductos Half-Heusler LiCdP compound have studied to illuminate the structural,
electronic, elastic, vibrational and optical properties in MgAgAs phase using DFT. For our obtained the lattice
parameters at zero pressure it has found good agreement with theoretical values. Obtained electronic structure
graphs profile indicate the semiconducting property direct band gap which has 0.23 eV. The total and partial
density of states were calculated to see the additions of the atomic orbitals near the gap. The calculated elastic
constants exhibite that title compound are stable mechanically up to 20GPa in MgAgAs structure. The calculated
elastic constants increase with pressure. These studies show a usefull knowledge of LiCdP compound and it may
be used in optoelectronic and other technological applications.

REFERENCES
[1] H.C. Kandpal, C. Felser, R. Seshadri, J. Phys. D: Appl. Phys. 39 (2006) 776.
[2] F. Casper, T. Graf, S. Chadov, B. Balke, C. Felser, Semicond. Sci. Technol. 27 (2012)063001.
[3] F. Casper, T. Graf, S. Chadov, B. Balke , and C.Felser, Half-Heusler compounds: novel materials for energy and
spintronic applications. Semicond. Sci. Technol. 27, (2012). 063001.
[4] N. Shutoh, S.Sakurada, Thermoelectric properties of the TiX(Zr0.5Hf0.5)1 − XNiSn half-Heusler compounds. J. Alloys
Compd., 389, (2005) 204-208.
[5] S. Kacimi, H.Mehnane, A. Zaoui, I–II–V and I–III–IV half-Heusler compounds for optoelectronic applications:
Comparative ab initio study. J. Alloys Compd., 587, (2014) 451-458.
[6] T. Gruhn, Comparative ab initio study of half-Heusler compounds for optoelectronic applications. Phys. Rev. B, 82,
(2010) 125210.
[7] H. Mehnane, B. Bekkouche, S. Kacimi, , et al. First-principles study of new half Heusler for optoelectronic
applications. Superlattices and Microstructures.,51, (2012), 772-784
[8] W. Zhu, H. Xiao, J. Comput. Chem. 29 (2008) 176.
[9] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.
[10] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh,C. Fiolhais,
Phys. Rev. B 46 (1992) 6671.
[11] D. Vanderbilt, Phys. Rev. B 41 (1990) 7892.
[12] L. Bellaiche, D. Vanderbilt, Phys. Rev. B 61 (2000) 7877.
[13] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188
[14] L. Page and P. Saxe, Symmetry-general least-squares extraction of elastic data for strained. Phys. Rev. B, 65,
(2002)104104.
[15] B. B. Karki, L. Stixrude, and J.Crain, Ab initio elasticity of three high-pressure polymorphs of silica Geophys. Res.
Lett. 24, (1997) 3269-3272.
[16] M. Born, and K. Huang, Dynamical Theory of Crystal Lattices, Clarendon Press, (1956).
[17] R. Hill, Proc. Phys. Soc. London A 65 (1952) 349.
[18] M.A. Ali, A.K.M.A. Islam, N. Jahan, S. Karimunnesa, Int. J. Mod. Phys. B 30, (2016), 1650228.
[19] A. Bouhemadou, Braz. J. Phys. 40 (2010) 52.
[20] A.A. Maradudin, E.W. Montroll, G.H. Weiss, I.P. Ipatova, Theory of Lattice Dynamics in the Harmonic
Approximation, Academic Press, New York, 1971.
[21] V.V. Bannikov, I.R. Shein, A.L. Ivanovskii, Phys. Status Solidi (RRL) 3 (2007) 89.

729
 COMPOSITION EFFECTS ON ELASTIC AND ELECTRONIC
PROPERTIES OF RU DOPED OSHF (RUXOS1-XHF) ALLOYS: AB-
INITIO STUDY

Yasemin Öztekin Çiftci

Gazi University, Physics Department, Ankara, TURKEY. [email protected], ORCID: https://orcid.org/0000-0003-
1796-0270

Çiftci,YO, Composition effects on Elastic and Electronic Properties of Ru Doped OsHf

Cite this paper as: (RuxOs1-xHf) Alloys: Ab-initioStudy, . 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: The Virtual Crystal Approximation (VCA) have been used by first principles
study based on density functional theory (DFT) calculation to analyze structural,
electronic and elastic properties of rutenium doped OsHf alloys (RuxOs1-xB) for
different Ru composition from 0.0 to 1.0 by 0.1. VCA method is established by
using mixture atoms that one atomic site is taken by two or more different atom.
For each x composition, it has been calculated the electronic band structures and
partial density of states (PDOS) as a electronic properties. Using calculated elastic
constants, the mechanical quantities which are shear modulus, Bulk modulus,
Young’s modulus, Poisson’s ratio, G/B ratio and hardnes have been obtained for
all x compositions. For obtained elastic constants show that RuxOs1-xHf alloys are
stable mechanically and decrease with increasing Ru composition.
Keywords: DFT, OsHf, Ru doped, electronic, elastic
© 2020 Published by ECRES

1. INTRODUCTION

Due to their superconducting features [1,2], oxidation resistance, hydrogen storage capacity [3,4], high melting
point, strength, and medical utilizations [5], intermetallic compounds are important for researchers. Among
intermetallic compounds, the refractory materials have been studied since their high-temperature potential
applications have a lot academic interests because of outstanding mechanical properties and high-temperature
stabilities[ 6-8].

Recently, the studies of the phase diagrams and structural stabilities for transition metals and and platinum group
elements have been performed [9,10]. According to obtained results for the OsX compounds, these compounds
are not super- hard but they are ductile solids [11]. Using direct synthesis calorimetry at high-temperature the
formation enthalpies of transition metals have been studied and compared with the predicted data which found
by ab initio calculations by Miedema et al [12]. Experimentally, using CALPHAD method thermodynamic
parameters, thermochemical and phase diagram parameters for Os-Hf alloys have been investigated [13].

Arıkan et al[14] have investigated structural, elastic, electronic and vibrational features of Os-based OsX(X=Hf,
Ti, Y and Zr) compounds using the density functional theory. They have found that OsX compound are stable
dynamically and mechanically and also these compounds are weak metallic and ductile materials. Wu et al[15]
have investigated the site and addition effects on the structural, elastic, and electronic properties of RuHf and
RuHf–Metal compounds.

730
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Although, there are some studies about structural, elastic, electronic features of OsHf, there are no any study on
elastic and electronic features of the RuxOs1-xHf. Here, first principle calculations of structural, electronic and
elastic properties of RuxOs1-xHf alloys from x=0 to 1 as a function of x composition have been performed.

2. METHOD AND CALCULATIONS

To investigate properties of structural, elastic and electronic of Ru doped OsHf (RuxOs1-xHf compounds (x =0,1),
the first-principle studies have been used CASTEP[16,17]. For the exchange-correlation terms, it has been
chosen Perdew-Burke-Ernzerhof (PBE) parametrization due to generalized gradient approximation (GGA) [18].
The electron-ion interactions were taken by the Ultrasoft Vanderbilt pseudopotential [19,20]. For each elements,
the electronic valence configurations were chosen as Ru: 4s24p64d75s1, Os: 5s25p65d66s2, and Hf:5d26s2 using
the virtual crystal approximation (VCA) [21]. The kinetic energy cutoff were used as 600 eV. In the irreducible
Brillouin zone, a set (12x12x12) Monkhorst-Pack [22] k-points grid are used. To obtain a well converged total
energy, these parameters are carefully tested. The RuxOs1-xHf alloy which crystallizes in the B2 structure is
modeled using the VCA method [21] where the molar ratios (x) of Ru element is scalled up 0.0 to 1.0 by 0.1
interval.

3. RESULTS
Structural and electronic properties

Here, for doping ratio from 0 to 1of Ru, the lattice parameters of RuxOs1-xHf alloys are obtained by structural
optimization, using the VCA method in the B2structure. The unit cell of OsHf with is given in Figure 1. The
locations of atoms in OsHf are (0, 0, 0) and (0.5 0.5 0.5) for Os and Hf, respectively. The structural parameters
of RuxOs1-xHf alloys which obtained by minimizing the total energy are given in Table 1. Obtained results for
OsHf and RuHf are in good agreement with other results[14,15]. Since ionic size of Ru (1.78 Å) is smaller than
that of Os(1.85 Å), lattice parameters of RuxOs1-xHf for x compositions decrease with increasing Ru
compositions.

Figure 1. Unitcell of OsHf

The formation energies of the of RuxOs1-xHf for each x value have been calculated and given in Table 1. The
obtained negative formation energies for all of the x compositions show that RuxOs1-xHf alloys are
structurally stable and synthesizable. The most stable formation energy is obtained for x=0.9.

731
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1: Structural parameters such as lattice constant, volume, bond length (Ru, Os-Hf) and formation energies for
RuxOs1-xHf
x a(A) V(A3) d (A) ∆Hf
0 3.291 35.641 2.85 -0948
OsHf
0.1 3.239(exp)
3.290 35.613 2.849 -0.636
[23] 3.257(Theory)
0.2
[14] 3.289 35.582 2.848 -0.403
0.3 3.288 35.553 2.847 -0.242
0.4 3.287 35.513 2.847 -0.150
0.5 3.286 35.483 2.846 -0.123
0.6 3.284 35.44 2.845 -0.159
0.7 3.283 35.40 2.844 -0.257
0.8 3.282 35.36 2.843 -0.419
0.9 3.280 35.295 2.841 -0.647
1 3.278 35.247 2.839 -0. 947
RuHf 3.225
[15]
The obtained band structure and DOS, which gives information about chemical bonding and electronic
properties for RuxOs1-xHf for x=0.9 are displayed in Fig.2. This graph shows that these alloys are metallic due
to absence of band gap.

Fig.2: Electronic band structure and partial DOS for Ru0.9Os0.1Hf

732
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Mechanical properties

For industrial applications, it is very important the investigations of mechanical feautures such as
elastic moduli, bulk modulus, Young’s modulus and ductile/brittle behavior. Here, to calculate the
elastic constants strain-stress method are used. Mechanical properties of RuxOs1-xHf (x=0, 1) were
obtained and tabulated in Table 2. Three elastic constants ((C11, C12 and C44) given in Table 2 should be
satisfied the mechanical stability conditions are C11+2C12>0, C44>0, C11－C12> 0 for cubic structure.Our results
show that the RuxOs1-xHf (0< x< 1) alloys in the B2 structure are stable mechanically.

Table 2: Elastic properties

x C11 C12 (GPa) C44 B G E ν G/B Hv

(GPa) (GPa) (Gpa) (GPa) (Gpa) (GPa)

0.517 18.48
394.98 149.732 118.726 231.484 120.272 312.673 0.274
0 393.69 139 110.05 224.43 116.81
Theory[14]
0.1 395.285 149.523 118.790 231.444 120.410 313.211 0.2745 0.520 18.47
0.2 394.321 148.781 118.341 230.628 120.093 312.805 0.2739 0.521 18.51
0.3 392.094 147.248 117.403 228.863 119.386 311.693 0.273 0.522 18.53
0.4 388.083 144.120 116.261 225.441 118.516 310.028 0.271 0.525 18.61
0.5 385.239 142.584 114.714 223.469 117.315 308.204 0.270 0.526 18.62
0.6 381.581 140.751 112.516 221.028 115.612 305.725 0.269 0.523 18.55
0.7 376.990 137.516 109.306 217.341 113.365 303.479 0.267 0.521 18.60
0.8 373.539 132.37 104.98 212.765 110.964 304.263 0.261 0.520 19.22
0.9 374.806 127.464 100.127 209.911 108.96 310.111 0.253 0.519 20.37
1(RuHf) 0.513 21.28
372.550 121.39 93.449 205.11 105.205 312.885 0.2458
Theory[15] 355.84 127.21 84.37

It can be seen from Table 2 that the values of C11 , C12 and C44 elastic constants tend to decrease with x
composition from 0 to 1.0 as x composition of Ru incerase.

Shear modulus and bulk modulus show measure of resistance to reversible deformation and while resistance to
volume change, respectively. Bulk (B) and shear moduli(G) which use the Voigt-Reuss-Hill formula are
calculated utilizing three elastic constants obtained for cubic structure[24,25]. Also, Poisson's ratio (v) and the
Young's modulus (Y) can also be obtained using elastic constants [26,27, 28]. Calculated elastic parameters for
each x values are given in Table 2. For OsHf, obtained bulk modulus express the average bond strength of atoms
for a given solid is 231.48 GPa showing reasonably powerful bonding strength of the atoms for RuxOs1-xHf. The
values of the bulk modulus decrease with increasing (Ru) x composition. But, shear modulus G depends on
change of shape in a solid. Obtained G values incease with x values while Young modulus E values decrease
with x values.

Covalent or ionic bonding knowledge of the materials can be evaluated by the values of poisson ratio (v) which
indicates strucure of bonding forces. Covalent and ionic solids, the value of v is 0.10 and 0.25, respectively [29]
. Our obtained values of v for the solid solutions are within line of ionic materials. For central-force, poisson

733
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

ratio are between 0.25 and 0.50 [29]. Because, the obtained Poisson's ratios don’t enter this range indicating
non-central interatomic forces for RuHf.

CONCLUSIONS

In this study, first principles thoretical study about the structural, elastic, and electronic features of RuxOs1-xHf
has been performed. The obtained lattice constants for OsHf and RuHf are good agreement with their theoretical
values and for x values, they decrease with increasing Ru contents since ionic radius of Ru is lower than Os.The
influence of Ru rations on the structural, electronic and elastic properties of the OsHf alloys have been studied
utilizing density-functional theory calculations. Among considered doping ratios, Ru0.9Os0.1Hf alloy indicate to
be the most stable structure. Investigated doped alloys are metallic nature and has a ionic bonds. Our obtained
elastic constants indicate that all RuxOs1-xHf alloys are stable mechanically. As a result, these detaied
investigations of the structural, electronic and mechanical properties of RuxOs1-xHf may be used for
technological applications in future.

REFERENCES

[1] Hein, RA, Cox JE, Blaugher RD, Waterstrat RM, Solid State Commun. 1969, 7, 381.
[2] O. Rapp, J. Invarsson, T. Claeson, Phys. Lett. 50A (1974) 159.
[3] R. Jerlerud-Perez, B. Sundman, CALPHAD 27 (2003) 253.
[4] S.W. Cho, C.S. Han, C.N. Park, E. Akiba, J. Alloy. Compd. 289 (1999) 244.
[5] L.D. Zardiackas (Ed.), Titanium, Niobium, Zirconium and Tantalum for Medical and Surgical Applications, ASTM
International, 2006.
[6] N. David, T. Benlaharche, J. Fiorani, M. Vilasi, Intermetallics 15 (2007) 1632.
[7] H. Saage, M. Krüger, D. Sturm, M. Heilmaier, J. Schneibel, E. George, L. Heatherly, C.
Somsen, G. Eggeler, Y. Yang, Acta Mater. 57 (2009) 3895.
[8] J. Lemberg, R. Ritchie, Adv. Mater. 24 (2012) 3445.
[9] K. Mahdouk, J. Gachon, J. Alloy. Compd. 278 (1998) 185.
[10] W. Xing, X.-Q. Chen, D. Li, Y. Li, C.L. Fu, S.V. Meschel, X. Ding, Intermetallics 28 (2012) 16.
[11] Q.-J. Liu, Ning-Chao Zhang, Fu-Sheng Liu, Zheng-Tang Liu, J. Alloy. Compd. 589 (2014) 278.
[12] S.V. Meschel, X.Q. Chen, O.J. Kleppa, Philip Nash, CALPHAD 33 (2009) 55.
[13] C. Guo, Z. Du, J. Alloy. Compd. 407 (2006) 188.
[14] N. Arıkan,n, O.Örnekb, Z.Charific, H.Baazizc, Ş. Uğur, G.Uğur, J. of Phys. And Chem.of Solids 96-97(2016)121–127
[15] J.Wu, S. Liu, Y. Zhan, M. Yu, Materials and Design 108 (2016) 230–239
[16] M.D. Segall, P.J.D. Lindan, M.J. Probert, C.J. Pickard, P.J. Hasnip, S.J. Clark, M.C. Payne, J. Phys. Condens. Mater
14 (2002) 2717.
[17] W. Zhu, H. Xiao, J. Comput. Chem. 29 (2008) 176.
[18] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett.
77 (1996) 3865.
[19] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh,C. Fiolhais, Phys. Rev. B 46 (1992)
6671.
[20] D. Vanderbilt, Phys. Rev. B 41 (1990) 7892.
[21] L. Bellaiche, D. Vanderbilt, Phys. Rev. B 61 (2000) 7877.
[22] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188
[23] A.E. Dwight, Trans. Metall. Soc. AIME 215 (1959) 283
[24] R. Hill, Proc. Phys. Soc. London A 65 (1952) 349.
[25] M.A. Ali, A.K.M.A. Islam, N. Jahan, S. Karimunnesa, Int. J. Mod. Phys. B 30
(2016), 1650228.
[26] A. Bouhemadou, Braz. J. Phys. 40 (2010) 52.
[66] M.A. Ali, A.K.M.A. Islam, M.S. Ali, J. Sci. Res. 4 (2012) 1.
[27] A.A. Maradudin, E.W. Montroll, G.H. Weiss, I.P. Ipatova, Theory of Lattice Dynamics in the Harmonic
Approximation, Academic Press, New York, 1971.
[28] V.V. Bannikov, I.R. Shein, A.L. Ivanovskii, Phys. Status Solidi (RRL) 3 (2007) 89.
[29] H. Fu, D. Li, F. Peng, T. Gao, X. Cheng, Comput. Mater. Sci. 44 (2008) 774.

734
 FINE STRUCTURED RED-BAND TAIL PHOTOLUMINESCENCE
(PL) SPECTRA OF NANOCRYSTALLINE CDS: IN THIN FILMS

Shadia J. Ikhmayies
Jabal El-Hussain, Amman 11121, Jordan, [email protected], ORCID: 0000-0002-2684-3300

Ikhmayies, J. Fine Structured Red-Band Tail Photoluminescence (PL) Spectra of

Cite this paper as: Nanocrystalline CdS:In Thin Films. 8th Eur. Conf. Ren. Energy Sys.24-25 August 2020,
Istanbul, Turkey

Abstract: Cadmium sulfide (CdS) is an important material for the manufacture of solar cells, and
optoelectronic devices. Investigating the photoluminescence (PL) of this material is important for
the development of these industries. In this work, nanocrystalline indium doped cadmium sulfide
thin films (CdS:In) are prepared using the spray pyrolysis method. The films are characterized
using X-ray diffraction (XRD), scanning electron microscopy (SEM), X-ray energy dispersive
spectroscopy (EDS), and photoluminescence (PL). The PL spectrum is recorded in the wavelength
range 640 – 1280 nm which covers the red band tail and part of the near infrared region. The PL
spectra showed fine structure that is related to the nanocrystalline nature of the films. A
deconvolution peak fit is performed to detect the peaks in the region of interest, where the peak
parameters are determined. The peaks are found to be collected into bands that include doublets.
A deep discussion is performed to explain the origins of the bands, the doublets and the fine
structure.
Keywords: Cadmium sulfide, Solar cells, Nanocrystalline thin films, Fine structured photoluminescence,
Spray pyrolysis.
© 2020 Published by ECRES

1. INTRODUCTION

Cadmium sulfide (CdS) is a wide band gap semiconducting material that has a cubic (zincblende) and hexagonal
(wurtzite) structures. The wurtzite structure (α-CdS) is the stable structure which has the space group p63mc [1]
and direct band gap energy equals to 2.57 eV at room temperature [2]. The cubic structure (β-CdS) is metastable,
and has a slightly lower band gap energy value of 2.40 eV at room temperature [2], and space group F-43m [1].
This material has great interest for multiple applications in integrated optics, optoelectronics, and photovoltaics,
which include but not restricted to light emitting diodes, light detectors, lasers, and solar cells. In thin film form
CdS has several applications such as the use as window layer in heterojunction solar cells.
Cadmium sulfide has intrinsic defects that are dominant native donors. These are; sulfur vacancy (VS), cadmium
interstitial (Cdi), cadmium antisite CdS, which is less favorable, and sulfur interstitial (Si). It also has dominant
native acceptor, which is the cadmium vacancy (VCd). The VS is a deep double donor, with the transition from
the +2 to 0 charge states falling 1.72 eV above the valence band minimum (VBM) of CdS. Cadmium interstitial
(Cdi) forms a localized state within the band gap, acting as a deep donor in CdS. The transition level of Cdi is
only 0.15 eV below the conduction band minimum (CBM) of CdS, making it effectively behaving as a shallow
donor. The sulfur interstitial (Si) acts as a deep donor, and it exhibits a deep transition of about 0.7 eV above the
VBM. The deep acceptor cadmium vacancy (VCd) can trap up to 2 electrons in the dangling bonds of the
adjacent S atoms. The transition for VCd is 1.10 eV above the VBM of CdS [3]. Cadmium sulfide is n-type due
to its intrinsic sulfide vacancies (VS). For the use as a window material in heterojunction solar cells CdS must
have n-type conductivity, and this can be attained by several ways such as the variation of the reagent precursors'
concentration, or doping with cationic atoms such as indium.
CdS Nanostructures have properties that are dramatically different from bulk CdS, which are important for
photochemical, and photocatalytic applications. Properties of nano structured CdS can be tuned by changing the
size, and shape of the nanocystallites [4], where the decrease in size results in quantum confinement and increase

735
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

of surface to volume ratio (surface effects). Quantum confinement effects as well as surface effects control
optical properties of semiconductor nanocrystals [5].
There are different ways to synthesize CdS nano structured thin films such as; rf sputtering [6], chemical bath
deposition (CBD) [7-10], thermal evaporation [11], pulsed laser deposition (PLD) [12,13], thermolytical growth
procedure [14], microwave-hydrothermal method [15], and spray pyrolysis [16-24]. Spray pyrolysis is preferred
in this work because it is a low cost method which enables intentional doping, and the production of highly
transparent, large area thin films.
The PL spectrum of CdS covers four fundamental bands which are labeled as Green, Yellow, Red, and Raman,
which are studied by several authors. This work focuses on the fine structure found in the tail of the red band
and the beginning of near infrared band for nanocrystalline CdS:In thin films produced by the spray pyrolysis
technique on glass substrates at a substrate temperature of 490˚C.

2. EXPERIMENTAL PROCEDURE

Nanocrystalline CdS:In thin films are produced as in previous works [16-24] on glass substrates using the spray
pyrolysis method using a precursor solution prepared by dissolving 2.24×10-2 moles of thiourea (NH2)2CS
(>97% S) and extra pure CdCl2.H2O (MERCK Art. 2011) in 350 ml of distilled water. For indium doping InCl3
(MERCK Art.12471) is used as a source of indium, where the ratio of indium ions [In+3] to that of cadmium
ions [Cd+2] in the solution is 1.0×10-4. The solution was intermittently sprayed on hot glass substrates –that
were ultrasonically cleaned with methanol- of temperature T = 490 °C for time periods that depend on the
required film thickness. The films’ thickness was estimated as described elsewhere [22].
X-ray diffraction (XRD) is used to explore the structure using a Philips PW1840 Compact X-ray diffractometer
system with Cu Kα (λ = 1.5405 Å) at a diffraction angle 2θ from 0˚ to 60˚. While scanning electron microscopy
(SEM) is used to explore the morphology using FEI scanning electron microscope (Inspect F 50). Compositional
analysis is made by X-rays energy dispersive spectroscopy (EDS) using the same SEM system. The PL spectra
are recorded at T= 23 K using a system that consists of an Air Product He cryostat DISPLEX DE-202 capable
of cooling down to 10 K. Argon ion laser of wavelength 488 nm is used as the excitation source. The laser power
is 10 mW and the diameter of the laser beam on the sample is about 2 mm. A multi-channel optical spectrometer
(an Avantes Fiberoptic Spectrometer AVS-S2000) that hosts two gratings is used to collect the PL signal, where
the grating used is that of wavelength range 640 – 1280 nm.

2. RESULTS AND DISCUSSION

Figure 1 shows the X-ray diffractograms of two films of different thickness (140 and 500 nm). The figure reveals
the nanocrystalline nature of the films and shows that they have a mixed phase (cubic and hexagonal) structure.
All prominent peaks are indexed with Millar indices as shown in the figure, where C refers to the cubic phase
and H refers to the hexagonal phase. So, the peaks C(1 1 1), C(2 2 0), and C(3 1 1) belong to the cubic phase,
and the peaks H(0 0 2), H(1 0 1), H(1 0 2), H(1 0 3), H(1 1 2), and H(2 0 2) belong to the hexagonal phase.
There is a preferential orientation related to reflections from the H(101) plane followed by that from the C(3 1
1)/ H(1 1 2) planes, where the positions of the lines C(3 1 1) and H(1 1 2) approximately coincide. The weak
peaks in the X-ray difractograms are produced by the small nanocrystallites.

736
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. X-ray diffractogtams of the CdS:In thin films. a) t = 500 nm. b) t = 140 nm.

Figure. 2 displays the SEM images for two CdS:In thin films. The nanocrystalline nature of the films is obvious,
and films' surfaces are uniform. Figure 2a, shows the SEM image of the film of thickness t = 500 nm, where the
average grain size is about 200 nm, and Figure 2b shows that of the film of thickness t = 140 nm where the
average grain size is about 100 nm. It is well known that the size of the crystallites decreases with decreasing
film thickness, so it is expected that the average size of the nanoparticles is smaller in the case of the thinner
film. It can be noticed that Figure.2a shows a smaller density of small grains of average size around 100 nm.

Figure 2. SEM images of the CdS:In thin films. a) t = 500 nm [24]. B) t = 140 nm

Figure 3 shows the EDS spectra of the same two films under study. The films contain In in addition to Cd, and
S. From the EDS report it is found that the ratio of Cd: S in the thicker film is 51.05: 46.98 at.%, or 1.09, and in
the thinner film it is 46.97: 49.75 at.%, or 0.94. This means that the thicker film is Cd - rich, but the thinner film
is S-rich. The ratio of In: Cd in the thicker film is 1.97: 51.05, or 3.86%, but in the thinner one it is 3.28: 46.97
or 6.98%. So the larger indium content in the thinner film substitutes the cadmium deficiency, where In
substitutes Cd in the crystal lattice.

737
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a)

(b)
Figure 3. EDS spectra of CdS:In thin films. a) t = 500 nm [24]. B) t = 140 nm.

It is well known that the properties of CdS films including photoluminescence are affected by the deposition
methods and preparation conditions, which may introduce different types of impurities and defects. Since the
precursor solution from which the CdS:In thin films are produced is made from thiourea (NH2)2CS and
CdCl2.H2O, the defect structure of the films will be influenced by the chemical structure of the complex
[Cd(tu)2Cl2], where (tu = (NH2)2CS [25]. That is there will be ClS (Chlorine substitutional defects) in the
produced CdS films, where chlorine replaces sulfur, or fills sulfur vacancy in the crystal lattice. These defect
species, besides intrinsic defects will act as recombination centers. In a similar manner, the doping with indium
(In+3) will produce a luminescence activator, and indium will substitute cadmium or fills cadmium vacancy,
where a shallow level will appear close to the bottom of the conduction band.
Usually the PL spectrum of cadmium sulfide shows a sharp band in the green-yellow region and a broad band
centered at in the yellow region around the wavelength λ = 600 nm, in addition to a much weaker broad infrared
band around 1100 nm (1.1 eV), where the broad red band is generally attributed to recombination of trapped
charge carriers at surface defects, and the low-energy band is highly attributed to the native defects such as sulfur
or cadmium vacancies [26]. Both these bands were studied in our previous works [18-20,22]. The current study
focuses on the fine structure in the tail of the broad red band and near infrared region. This fine structured is
produced by the nanocrystallites of indium doped cadmium sulfide (CdS:In) thin films of SEM images shown
in Figure.2.
Figure.4 shows the PL spectra for the two films under study recorded in the energy range 0.969-1.938 eV or
wavelength range 640 – 1280 nm. The range of interest in this study is 1.2-1.9 eV, which is close to or larger

738
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

than half the band gap energy of CdS at room temperature. The PL spectrum of the thinner film is weaker than
that of the thicker one. This is because of the increase of the surface to volume ratio with decreasing crystallite
size, which leads to increasing the surface band. That is, the excited electron will be captured by the surface or
defect state before the occurrence of the direct radiative recombination, which results in the reduction of the PL
intensity.
It is well known for films prepared by chemical routes such as the spray pyrolysis that the samples show size
distribution, and this is obvious in the SEM images shown in Figure.2. Hence, the PL spectrum is the summation
of the spectra of individual crystallites of all different sizes. In the case of the thicker film a larger size
distribution is present as seen in Figure.2a, so the PL spectrum of the thicker film shows sharp features with
lower intensity originating from individual crystallites. The thinner film shows the same features but with higher
intensity, which is evidence of the higher density of the smaller nanocrystallites and the stronger surface effects,
because surface-to volume ratio is larger. The positions of the sharp peaks of both PL curves coincide, and the
peaks appear as each one consists of two very close but merged peaks, where this is more apparent in the PL
spectrum of the thinner film.

t = 140 nm
t = 500 nm

4000

3000
C [counts]

2000

1000

0.8 1.0 1.2 1.4 1.6 1.8 2.0

E (eV)

Figure 4. The red band tail and the near infrared PL spectra of the two CdS:In thin films under study.

A deconvolution peak fit to Gaussian peaks was performed for the spectrum of the thinner film and presented
in Figure.5. Figure.5a shows the produced peaks, with their positions displayed in the figure. Figure.5b shows
the second derivative of the fit, where the positions of the peaks are determined by positions of the minima in
the second derivative curve. The peak parameters obtained from the fit which are the amplitude, position, and
full width at half maximum of are inserted in table 1 and table 2. Also, the energy separation between successive
peaks (ΔE) is shown in the tables. The fit shows 10 peaks in the tail of the red band and 34 peaks in the range
of interest in the near infrared band, with 8 doublets. In each doublet the peaks are separated by bout 4.6-10
meV. The observed doublets are due to the presence of the two phases (zinc blende and wurtzite) in CdS
nanocystallites. Therefore, if two types of lattices coexist side by side in one particle, the energies of localized
states in different domains assume slightly different values. This results in doublet structures for the transitions
between these states.
From Figure.4 and Figure.5 it is observed that these peaks are distributed in 7 bands, two of them in the red band
tail and listed in Table 1. The strongest one extends in the range 1.82-1.9 of the red band tail and consists of 5
peaks and 1 doublet. This band is related to surface defects [20] such as isolated S vacancies (VS) located near
to the surface of the films. The next strong band is that in the range 1.72-1.82 in the near-infrared (IR), and it
consists of 6 peaks with 2 doublets. This fine-structured band corresponds to IR0 in ref. [25], for CdS films,
which doesn't contain fine structure. Kelle et al. [27] attributed this IR0 band to the complexes of intrinsic defects
with substitutional chlorine [VCd(ClS)2], but others [25,28,29] attributed it to oxygen-related centers, which are
identified as [VCdOS]2– defect complexes, which result from oxygen incorporation in vacant sulfur sites.
The other 5 bands in the near infrared are listed in Table 2. Looking at the first three bands in the near infrared;
the first one in the range 1.59-1.72, which consists of 13 peaks with 3 doublets according to Fig.5 and table 2,

739
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the second one in the range 1.52-1.57 and contains 7 peaks with 1 doublet, and the third band in the range 1.40-
1.52 eV, which contains 5 peaks with 1 doublet. These three bands are located in the same positions of the three
bands R1, R2, and R3 in ref. [25] in the same order. So the difference between these bands and those of ref. [25]
is the fine structure. Hence they will be attributed to the same defects: [VSVCd] for the first one, and [VCdClS]-
for the second and third bands, because the films contain Cl.
The other two bands in the remaining range of interest of the near infrared region are very weak bands. The first
one is in the ranges 1.34-1.38 with 4 peaks and no doublets, and the second in the range 1.25-1.3 with 4 peaks
and no doublets too. In a previous work [20] these emissions were partially related to Cd complex (ICd +-VCd-)
and deep acceptor states such as VCd2-. But if compared with those in ref. [25], the last one of these two bands
corresponds to IR1 (wavelength range: 1030–1060) in ref. [25] which was related to [ClS]+ donor [29,30].

C:\peakfit4\PF4\CLIPBRD.WK1
Pk=Gauss Amp 44 Peaks Bg=Linear
r^2=0.997912 SE=43.1675 F=4190.56
4000 4000
3500 3500
3000 3000
PL(Counts)

PL(Counts)
2500 2500
2000 2000
1500 1500
1000 1000
500 500

3500 3500
1.657 1.8839
3000 3000
PL(Counts)

PL(Counts)
1.6483
2500 1.8075 2500
1.5525 1.6263
1.784
2000 1.5334 1.6105 2000
1.4574 1.5263 1.774
1500 1.6856 1500
1.2823 1.3451 1.4772 1.5934 1.7613 1.8549
1.3824 1.6757
1000 1.2767
1.471
1.544 1.6029 1.8321 1.9008 1000
1.2695 1.362 1.6667 1.7427
500 1.5031 1.5648 1.8951 500
1.751
1.2648 1.3569 1.4327 1.6306 1.6969
0 0
1.1 1.3 1.5 1.7 1.9
E(eV)
(a)

C:\peakfit4\PF4\CLIPBRD.WK1
Second Derivative

Second Derivative

2e+08 2e+08

5e+07 5e+07

-1e+08 -1e+08

3500 3500
PL(Counts)

PL(Counts)

2500 2500

1500 1500

500 500
1.1 1.3 1.5 1.7 1.9
E(eV)
(b)
Figure 5. PL spectrum of the film of t = 140 cm shown in Fig.4 with the deconvolution peak fit (a), and second derivative
(b).

From Table 1 and Table 2 it is noticed that the energy difference between successive peaks in most cases is in
the range ΔE <25 meV. This means-in addition to the energy difference related to different phases- that these
peaks are not from transitions at the same site but rather from different sites in different crystalline environments.

740
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. CONCLUSIONS

The PL spectra of nanocrystalline CdS:In films are recorded at T = 23 K in the wavelength range 640 - 1280
nm, which includes the tail of the red band and part of the near infrared region. The spectrum showed fine
structured bands in the region of interest, which is attributed to the nanostructure of the films. A deconvolution
peak fit is performed assuming Gaussian peaks. The fit revealed the presence of doublets separated by 4.6-10
eV, which are attributed to the mixed state of the films (cubic and hexagonal). The produced peaks are collected
in bands and the origins of these bands are explained and discussed.

REFERENCES

[1] Tan JJ, Li Y, Ji, GF. High-pressure phase transitions and thermodynamic behaviors of cadmium sulphide. Acta
Physica Polonica A 2011;120 (3): 501-506. DOI: 10.12693/APhysPolA.120.501
[2] Samantilleke, AP, Cerqueira, MF, Heavens, S, Warren, P, Dharmadasa IM, Muftah, GEA, Silva, CJR, Marí, B.
Characterisation of chemical bath deposited CdS thin films on different substrates using electrolyte contacts. Thin Solid
Films 2011; 519 (21): 7583-7586. Doi:10.1016/j.tsf.2010.12.218
[3] Varley, JB, Lordi, V. Electrical properties of point defects in CdS and ZnS. Applied Physics Letters 2013; 103:
102103. https://doi.org/10.1063/1.4819492
[4] Ikhmayies, SJ. Tuning the properties of nanocrystalline CdS thin films, JOM. 2014; 67(1): 46-60.
https://doi.org/10.1007/s11837-013-0819-5
[5] Okamoto, S, Kanemitsu, Y, Hosokawa, H, Murakoshi, K, Yanagida, S. Photoluminescence from surface-caped CdS
nanocrystals by selective excitation. Solid State Communications 1998; 105 (1): 7-11. https://doi.org/10.1016/S0038-
1098(97)10058-8
[6] Tsai, CT, Chuu, DS, Chen, GL, Yang, SL. Studies of grain size effects in rf sputtered CdS thin films. J. Appl. Phys.
1996; 79: 9105-9109. https://doi.org/10.1063/1.362645
[7] Ghosh, B, Kumar, K, Singh, BK, Banerjee, P, Das, S. Growth of CdS thin films on indium coated glass substrates via
chemical bath deposition and subsequent air annealing. Applied Surface Science 2014; 320: 309-314.
https://doi.org/10.1016/j.apsusc.2014.09.034
[8] Sasikala, G, Thilakan, P, Subramanian, C. Modification in the chemical bath deposition apparatus, growth and
characterization of CdS semiconducting thin films for photovoltaic applications. Solar Energy Materials & Solar Cells
2000; 62: 275-293. https://doi.org/10.1016/S0927-0248(99)00170-1
[9] Nair, PK, Nair, MTS, Campos, J. Photocurrent response in chemically deposited CdS thin films. Solar Energy Materials
1987; 15: 441-452. https://doi.org/10.1016/0165-1633(87)90093-1
[10] Fernando, D, Khan, M, Vasquez, Y. Control of the crystalline phase and morphology of CdS deposited on
microstructured surfaces by chemical bath deposition. Materials Science in Semiconductor Processing 2015; 30: 174-
180. https://doi.org/10.1016/j.mssp.2014.10.002
[11] Sahay, PP, Nath, RK, Tewari, S. Optical properties of thermally evaporated CdS thin films. Cryst. Res. Technol.
2007; 42 (3): 275-280. DOI: 10.1002/crat.200610812
[12] Perna, G, Capozzi, V, Ambrico, M, Augelli, V, Ligonzo, T, Minafra, A, Schiavulli, L, Pallara, M. Structural and
optical characterization of undoped and indium doped CdS films grown by pulsed laser deposition. Thin Solid Films
2004; 453-454: 187-194. DOI: 10.1016/j.tsf.2003.11.105
[13] Perna, G, Capozzi, V, Pagliara, S, Ambrico, M. Absorption and photoluminescence analysis of CdS ablated films
grown on quartz substrate. Applied Surface Science 2000; 154–155: 238-243. https://doi.org/10.1016/S0169-
4332(99)00411-0
[14] Antolini, F, Burresi, E, Stroea, L, Morandi, V, Ortolani, L, Accorsi, G, Blosi, M. Time and temperature dependence
of CdS nanoparticles grown in a polystyrene matrix. Journal of Nanomaterials. 2012; 2012: 815696.
https://doi.org/10.1155/2012/815696
[15] Soltani, N, Gharibshahi, E, Saion, E. Band gap of cubic and hexagonal CdS quantum dots – experimental and
theoretical studies. Chalcogenide Letters 2012; 9(7): 321-328. https://www.semanticscholar.org/paper/BAND-GAP-
OF-CUBIC-AND-HEXAGONAL-CDS-QUANTUM-DOTS-Soltani-
Gharibshahi/5eacd559f189d83cde99c4da45600ecb6ccaf8b1
[16] Ikhmayies, SJ, Ahmad-Bitar, RN. Effect of film thickness on the electrical and structural properties of CdS:In thin
films. American Journal of Applied Sciences 2008; 5 (9): 1141-1143. DOI: 10.3844/ajassp.2008.1141.1143
[17] Ikhmayies, SJ, Ahmad-Bitar, RN. The influence of the substrate temperature on the photovoltaic properties of spray-
deposited CdS:In thin films. Applied Surface Science 2010; 256: 3541–3545.
https://doi.org/10.1016/j.apsusc.2009.12.104
[18] Ikhmayies, SJ, Ahmad-Bitar, RN. Direct observation of Infrared Photoluminescence of Spray-Deposited CdS:In
Thin Films. J. Luminouscence 2008; 128 (4): 615-619. https://doi.org/10.1016/j.jlumin.2007.10.011
[19] Ikhmayies, SJ, Ahmad-Bitar, RN. Dependence of the photoluminescence of CdS:In thin films on the excitation
power of the laser. Journal of Luminescence 2014; 149: 240-244. https://doi.org/10.1016/j.jlumin.2014.01.046

741
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[20] Ikhmayies, SJ, Ahmad-Bitar, RN. Effects of annealing in nitrogen atmosphere and HCl-etching on the
photoluminescence spectra of spray-deposited CdS:In thin films. Applied Surface Science 2009; 255: 8470-8474.
https://doi.org/10.1016/j.apsusc.2009.05.165
[21] Ikhmayies, SJ, Ahmad-Bitar, RN. Effects of processing on the electrical and structural properties of spray-deposited
CdS:In Thin Films. Physica B: Condensed Matter 2009; 404 (16): 2419-2424. DOI: 10.1016/j.physb.2009.04.052
[22] Ikhmayies, SJ, Ahmad-Bitar, RN. Temperature dependence of the photoluminescence spectra of CdS: In thin, films
prepared by the spray pyrolysis technique, Journal of Luminescence 2013; 142: 40-47.
https://doi.org/10.1016/j.jlumin.2013.03.033
[23] Ikhmayies, SJ, Juwhari, HK, Ahmad-Bitar, RN. Nanocrystalline CdS:In Thin Films Prepared by the Spray-Pyrolysis
Technique. J. Luminouscence 2013; 141: 27-32. DOI: 10.1016/j.jlumin.2013.02.045
[24] Ikhmayies, SJ. A study of the absorption Edge of CdS:In Thin Films. International Journal of Materials and
Chemistry 2018; 8(1): 10-14. doi:10.5923/j.ijmc.20180801.02
[25] Naumov, AV, Bolgova, TG, Semenov, VN, Maiorova, TL, Klyuev, VG. Luminescence and photoconductivity of
alkali-metal-doped cadmium sulfide films. Inorganic Materials 2006; 42 (5): 23–529.
https://doi.org/10.1134/S0020168506050025
[26] Chahboun, A, Rolo, AG, Filonovich, SA, Gomes, MJM. Factors influencing the passivation of CdS quantum dots
embedded in silica glass. Solar Energy Materials and Solar Cells 2006; 90 (10): 1413-1419.
https://doi.org/10.1016/j.solmat.2005.10.006
[27] Kelle, KI, Kirs, YY, Tulva, LT. On the origin of the orange-red luminescence from cadmium sulfide, III Vsesoyuznoe
soveshchanie po problemam fiziki soedinenii AIIBVI (III All-Union Conf. on the Physics of II–VI Compounds),
Vilnius 1972; 2: 85-89.
[28] Naumov, AV, Semenov, VN, Goncharov, EG. Properties of CdS films prepared from thiourea complexes of cadmium
salts. Neorg. Mater. 2001; 37 (6): 647-652 [Inorg. Mater. (Engl. Transl.) 2001; 37(6): 539-543].
https://doi.org/10.1023/A:1017594228500
[29] Semenov, VN, Klyuev, VG, Kushnir, MA, et al. Luminescence spectra of films produced by spray pyrolysis of
thiourea cadmium complexes, Zh. Prikl. Spektrosk 1993; 59 (1/2): 114–119.
[30] Martynov, VN, Volkova, ES, Totsina, TS, Luminescence of vapor-grown CdS laser crystals. Mater. Elektron. Tekh.
1998; 1: 70-74.

Appendix

Table 1. Fit parameters of the red tail of the PL spectrum of the film of thickness 140 nm.
Peak No. in Amplitude Center (eV) ΔE (meV) Band No. FWHM
the Fit (Counts) (meV)
34 523.05 1.7427 51.5
8.3
35 329.09 1.7510 10.0
Doublet 10.3
36 492.36 1.7613 11.2
12.7
Second Band

37 466.38 1.7740 10.3

Doublet 10.0
38 457.20 1.7840 13.6
23.5
39 1425.49 1.8075 40.3
24.6
40 792.49 1.8321 24.0
22.8
41 1268.13 1.8549 30.3
29.0
42 3076.75 1.8839 57.7
First Band

11.2
43 457.28 1.8951 15.4
Doublet 5.7
44 491.43 1.9008 7.8

742
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 2. Fit parameters of the near infrared PL spectrum of the film of thickness 140
Peak No. Amplitude Center (eV) ΔE (meV) Band No. FWHM (meV)
in the Fit (Counts)
1 47.97 1.2648 64.2
4.7
2 39.65 1.2695 7.7
7.2
3 26. 60 1.2767 3.6

Fifth Band
5.6
4 99.71 1.2823 5.4
5.1
5 61.10 1.2874 4.2
57.7
6 69.35 1.3451 21.0
11.8
7 106.44 1.3569 3.6

Fourth Band
5.1
8 192.20 1.3620 7.3
20.5
9 68.90 1.3824 13.5

50.3

10 82.92 1.4327 14.9

24.7
11 159.50 1.4574 17.4

Third Band
13.6
12 723.90 1.4710 6.6
Doublet 6.2
13 847.84 1.4772 7.7
25.9
14 347.97 1.5031 29.0
23.2
15 1343.65 1.5263 7.8
Doublet 7.1
16 1390.81 1.5334 7.5
10.6
17 534.04 1.5440 6.7
8.5
18 865.05 1.5525 Second Band 13.4
12.3
19 337.14 1.5648 7.6
10
20 375.86 1.5748 27.2
18.6
21 244.07 1.5934 17.3
9.5
22 754.19 1.6029 8.3
Doublet 7.6
23 749.91 1.6105 7.5
9.3
24 363.91 1.6198 5.0
6.5
25 1005.78 1.6263 15.8
4.3
26 173.81 1.6306 4.0
17.7
27 1587.44 1.6483 10.4
Doublet 8.7
28 1470.15 1.6570 9.6
9.7
29 115.21 1.6667 3.6
9.0
30 669.85 1.6757 11.7
Doublet 9.9
31 580.58 1.6856 9.6
First Band

11.3
32 214.86 1.6969 11.0
11.1
33 197.82 1.7080 16.3

743
 ARITHMETIC STUDY ON ENERGY SAVING FOR SOME
COMMON DATA CENTRE TOPOLOGIES

Pedro Juan Roig

Miguel Hernández University, Elche, SPAIN, [email protected], ORCID: 0000-0002-8391-8946

Salvador Alcaraz
Miguel Hernández University, Elche, SPAIN, [email protected], ORCID: 0000-0003-3701-5583

Katja Gilly
Miguel Hernández University, Elche, SPAIN, [email protected], ORCID: 0000-0002-8985-0639

Roig, Pedro Juan. Arithmetic Study on Energy Saving for some common Data Centre
Cite this paper as:
Topologies. 8th Eur. Conf. Ren. Energy Sys. 24-25August 2020, Istanbul, Turkey

Abstract: Data Centre facilities are rapidly increasing as parallel computing schemes are ever growing.
Those kinds of facilities are designed according to different topologies, but many of them are
underutilized, leading to energy misuse in many of the switches providing network interconnection
for the end hosts. In this paper, some common Data Centre topologies are studied, such as Fat
Tree, Leaf and Spine, N-Hypercube and Folded N-Hypercube, in order to find out their
performances in terms of energy saving purposes for the same number of hosts hanging on each
design, hence obtaining a coherent comparison among them all. It is to be said that each topology
has its own benefits and drawbacks, meaning that each one may be the best option depending on
the initial conditions of an implementation. However, in the conditions established for steady
latency and jitter, it seems that Fat Tree topologies get better results as the size of the Data Centre
grows up, whereas Leaf and Spine does it otherwise.
Keywords: Data Centre, Fat Tree, Folded N-Hypercube, Leaf and Spine, N-Hypercube
© 2020 Published by ECRES

1. INTRODUCTION

Multi-access Edge Computing (MEC) is going to be one of the key players in the Information Technology and
Communication (ITC) field in the near future. It extends the Cloud Computing paradigm, bringing the
computing assets to the edge of the network, thus resulting in better performances than Cloud Computing in a
radio environment regarding latency, bandwidth and power consumption [1].
The overall infrastructure of MEC may be similar to that of Cloud, although the number of resources involved
may be significantly scaled down as the number of wireless users getting in will be restricted to those getting
access to the coverage area of the MEC, having the possibility of using the Cloud services located in the
worldwide network as a backup solution.
This fact allows for the use of energy saving strategies when the number of users connected to the MEC
infrastructure is small enough and the amount of resources involved in their interactions are located in certain
number of hosts, thus leaving the rest of hosts in an idle state. This condition may happen quite often as MEC
services may be offered as added value services by network operators, which users may have to pay, whereas
general wireless services may be available for other users not willing to pay an extra fee for those premium
services.
The target of this paper is to revise, in an arithmetic way, some of the most well-knows cloud topologies in order
to study their characteristics so as to see which one better fulfills the requirements of MEC traffic with regards
to power consumption. It is to be noted that some switches taking part in the MEC architecture may well be put
in an idle state as long as the hosts being used are not making use of such switches, thus resulting in energy
saving.

744
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Likewise, redundant paths may exist between any two hosts, and in such a case, an orchestrator might choose
just one path or some of them, whilst leaving the rest aside, hence activating the switches involved and leaving
out the others, depending on the load balancing policies put in place or the energy consumption scheme being
in use in the whole system, resulting in less energy usage. Needless to say that if all redundant paths are active,
all switched will do so accordingly.
The organisation of the paper is going to be as follows: first, in Section 2, interconnection topologies are
presented, then, in Section 3, case scenarios are defined, in Section 4, calculations for average distance and
energy saving are performed, and finally, in Section 5, some conclusions are drawn.

2. INTERCONNECTION TOPOLOGIES

Focusing on the topological view of parallel computing infrastructures, it has to be noted that hosts have to be
interconnected by a series of switches according to a certain topology, which may impact performance subject
to the number of switches and hosts involved in each topology. Therefore, this is not the case that there is one
topology better than others, but each one has its own pros and cons, depending on the implementation proposed.
Regardless of the use of those topologies, this is, no matter if they are used in an internetwork scope, such as
Cloud Computing, or those used in a more restricted scope, such as Fog Computing, Mist Computing or MEC,
let us consider an abstraction composed by switches being interconnected to each other according to different
patterns, where those switches will be used to interconnect the different hosts where the parallel computing is
taking place.
Therefore, focusing on interconnection networks, they are usually hierarchical structures divided into two
different categories, such as trees and graphs. Some of the most relevant instances of the former are Fat Tree
and Leaf & Spine, whereas some of the most interesting instances of the latter are N-hypercube or Folded N-
hypercube.
Regarding switching topologies applied to Data Centre architectures, tree designs are usually preferred due to
the advantages given by their modular design, whereas the graph designs are usually more devoted to
multiprocessor implementations. Furthermore, from the point of view of maintenance, a tree-like structure
makes things easier. Nonetheless, graph designs are also used in multiple deployments due to its compactness.
Additionally, there are more complex topological structures regarding network interconnections for Data Centre
involving tree and graph designs, as well as other types, which may achieve even better performances.
Nevertheless, in this paper we are going to focus our study in the ones specified above, as being the simplest
ones and the most sought after. Besides, all designs to be studied are going to be considered with an
oversubscription rate 1:1, meaning that all possible links are available in the designs, as that is the optimal way
to evaluate their optimal performances.

Fat Tree

It is a tree-like structure with three layers of switches, where a parameter K drives the whole design [2]. The
lower layer is called Edge and it is the one where hosts are hanging on, the middle one is called Aggregation
and the upper one is called Core. Besides, parameter K marks how switches in the first and second row connect
to each other forming Pods, resulting in full mesh connectivity for the K switches evenly distributed in the
different layers within a single Pod, and full mesh connectivity among Pods, which account for partial mesh
connectivity between switches in the second and third row. Moreover, each switch has K ports, where the first
and second row have half of them looking upwards and the other half looking downwards, leading to the fact
that each Edge switch is holding K/2 hosts.
In this architecture, any two hosts may be up to three hops away from each other, in a way that hosts connected
to the same switch are one hop away, hosts connected to the same Pod are two hops away and hosts connected
to different Pods are three hops away. If there is one hop away, the number of switches involved in a transaction
will be one Edge switch. On the other hand, if there are two hops away, the number of switches involved will
be three, those being the Edge switches connecting both end hosts and also a single Aggregation switch
connecting both Edge switches, although load balancing policies may allow more than one Aggregation
switches. Eventually, if there are three hops away, the number of switches involved will be five, meaning two
Edge switches, two Aggregation switches and a Core Switch linking them, although more Core switches might
be involved depending on the load balancing conditions imposed. A Fat Tree implementation for K=4 is
presented in Figure 1.

745
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. Fat Tree architecture for K=4 and oversubscription rate 1:1

Leaf and Spine

It is also a tree-like structure with two layers of switches, where there are some degrees of freedom in the design
of the topology [3], allowing for a less restrictive layout compared to Fat Tree. The lower layer is called Leaf,
whereas the upper layer is called Spine. The only condition is that there must be a full mesh connectivity among
all switches situated in the lower layer, the ones where host are hanging on, and all the switches situated in the
upper layer.
The number of ports in each switch is not determined, as well as the number of hosts being held by each Leaf
switch, and even the number of switches forming both layers, therefore, all those features are up to the designer.
An instance of a Leaf and Spine implementation is shown in Figure 2, where the number of ports in all switches
is 8, with Leaf switches having 4 ports looking upwards and other 4 looking at the hosts, permitting up to 4 hosts
per switch.
This architecture allows for less distance among hosts, as any two hosts connected to the same Leaf switch are
one hop away, whilst if this is not the case, they will be just two hops away. In the former case, the number of
switches taking part in a single transaction is just one Leaf switch, whereas in the latter case, it will be three,
such as both Leaf switches connecting the hosts involved and a single Spine switch interconnecting both,
although may be more Spine switches depending on the load balancing policies put in place in the system
implementation.

Spine

Leaf

Figure 2. Leaf and Spine architecture with oversubscription rate 1:1

N-Hypercube

It is a graph-like topology, copying the structure of a hypercube of dimension N (also known as Qn). The point
here is that a switch is situated in every vertex of the N-hypercube chosen, also called nodes, and all of them are
joined together according to the edges of the N-hypercube being used. Therefore, if a group of hosts are hanging
on each of the switches, then any two given hosts being connected to any two switches will get together by virtue
of following the shortest path within the N-hypercube between them [4]. Figure 3 depicts the simplest designs
for N-hypercube.
It is to be noted that in this graph-like design, all switches have some hosts connected, as opposed to what happen
with the tree-like designs, where only the lower layer have hosts attached. This fact makes that distances get
reduced, although maintenance tasks and complexity get higher [5]. Furthermore, the nomenclature of the nodes
of an N-hypercube helps us understand the possible paths between any two given switches, as the number of
binary digits reflect the number of dimensions, meaning N, and any two given neighbouring nodes differ in just
one binary digit, stating that there is just one movement in that dimension to reach one another, given by
swapping just one bit.
The dimension N is chosen to be the lowest available to admit the number of hosts involved in the design, and
that amount is not tied to N in any way. It is to be noted that this topology is composed by 2N nodes, being the

746
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

switches, and N∙2N-1 edges, being the available links among those switches. Also, the maximum distance
between any two given switches is N, being that the case of two nodes opposite each other.
Hence, any two given hosts may be hanging on the same switch, meaning they are just one hop away, or
otherwise, all the way to N hops away. Furthermore, the nomenclature proposed above allows to easily see the
distance between any two nodes, accounting for the number of zeros found as the result of applying a logical
AND operator, those being the dimensions to get from source node to destination. Additionally, the number of
redundant paths between any two given hosts depends on the number of hops away, therefore, load balancing
policies may be put in place.

Figure 3. N-Hypercube architecture for N=0, N=1, N=2, N=3 and N=4 with oversubscription rate 1:1

Folded N-Hypercube

It is also a graph-like structure, improving the N-hypercube design with the establishment of a new link for each
node towards its furthest node, being the opposite node. This way, the maximum distance between any two
nodes gets reduced from N to ceil(N/2), getting shorter distances between further nodes, although getting a more
complex interconnection network with more links to manage, as well as rising up costs [6]. It is to be noted the
number of nodes within the topology is the same as above, this is, 2N nodes, whereas the number of edges is
(N+1)∙2N-1 edges, being those the available links among those switches. Figure 4 exhibits the easiest designs
for Folded N-hypercube.
Following the same nomenclature as the N-hypercube, the extra links go from a node referred with a binary
number to its 1’s complement. On the other hand, the way to calculate the possible paths from one source node
to a destination node may be the same as described above. Additionally, as Folded N-hypercubes (also known
as FQn) add extra redundant paths to the design, they allow for the application of more load balancing policies.
In summary, this extra feature allows for better performances and it is worthwhile in many case scenarios,
although each one has to be assessed individually, as getting more interconnections may imply a worse
manageability, which penalizes the implementation of full mesh topologies or other deployments with a great
deal of links.

Figure 4. Folded N-Hypercube architecture for N=2, N=3 and N=4 with oversubscription rate 1:1

3. DEFINING THE SCENARIOS

In order to compare all four topologies proposed, some prerequisites have to be met in advanced. The most
restrictive design is Fat Tree, as the number of switches (nodes) and links (edges) are determined by parameter
K. It establishes the number of ports per switch to K, and even the number of hosts hanging on each switch,
resulting in just K/2. Therefore, this amount is going to be established for all scenarios in order to better compare
their performance and their energy saving schemes [7]. Focusing on Fat Tree, the number total of nodes, edges
and hosts for each switch, Pod and the overall design of Fat Tree may be found in Table 1. The scenarios to be
studied correspond to K=4, K=8 and a generic K for Fat Tree, and in turn, the rest of topologies will be set up
in order to have the same amount of hosts per switch, arranging the rest of nodes and edges accordingly [8].

747
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1. Relevant number of items for the scenarios proposed in a Fat Tree topology
Topology Item Description K=4 K=8 generic value
Fat Tree Number of Ports per switch 4 8 K
Number of hosts per switch 2 4 K/2
Number of hosts per Pod 4 16 K2/4
Number of hosts overall 16 128 K3/4
Number of Edge switches (with hosts) 8 32 K2/2
Number of Aggregate switches 8 32 K2/2
Number of Core switches 4 16 K2/2
Number of switches overall 20 80 5K2/4
Number of links overall 48 384 3K3/4

With respect to Leaf and Spine, there is not such a parameter K, but for comparison purposes, let us suppose
that the number of hosts per Leaf switch will be established to be K/2, whereas the rest of parameters are set as
stated in Table 2, where the number of Leaf switches is sufficient to interconnect all hosts and the number of
Spine switches is enough to bring together all Leaf switches [9]. Therefore, on the one hand, for K=4, there will
be 8 Leaf 4-port switches, with 2 hosts connected to each of them, as well as 2 Spine 8-port switches, providing
a full mesh topology with all Leaf switches. On the other hand, for K=8, there will be 32 Leaf 8-port switches,
with 4 hosts connected to each one, along with 8 Spine 32-port switches, giving full mesh topology. In order to
work with generic expressions, let us consider parameter L=log2(K), which will be used in Table 2 for Leaf &
Spine.
Table 2. Relevant number of items for the scenarios proposed in a Leaf and Spine topology
Topology Item Description K=4 K=8 generic value
Leaf and Spine Equivalent L value respect to K L=2 L=3 L
Number of Ports per Leaf switch 4 8 2L
Number of Ports per Spine switch 8 32 22L-1
Number of hosts per switch 2 4 2L-1
Number of hosts overall 16 128 23L-2
Number of Leaf switches (with hosts) 8 32 22L-1
Number of Spine switches 2 8 22L-3
Number of switches overall 10 40 5∙22L-3
Number of links overall 32 192 23L-1

Referring to N-hypercube, parameter N will be key in order to manage the topology and the number of switches
and links may be found in Table 3, where K=4 corresponds to N=3 and K=8 corresponds to N=5 due to the total
number of hosts in the topology, but this may be up to the designer. It is to be remarked that if two hosts are
𝑁
connected to the same node, there is just one available path through that switch, meaning ( ). On the other
0
𝑁
hand, if they are one switch away, there may be ( ) possible switches, and if they are two switches away, the
1
𝑁
possible switches may be ( ), and so on. In summary, the sequence of numbers denoting the possible switches
2
𝑁 𝑁 𝑁 𝑁 𝑁
to be reached from an initial one is given by the sequence an=( ),( ) , ( ),..,( ),( ), which are the
0 1 2 𝑁−1 𝑁
same as the i-th columns (i=0,1,..,N) of the Pascal Triangle for row N, therefore, it might be expressed as
𝑁
⋁𝑁𝑖=0 ( ). Figure 5.a shows the Pascal Triangle, where each cell is obtained by summing up its both upper cells,
𝑖
𝑁 𝑁!
or otherwise, by calculating the combination of the N-th row and the i-th column, such as ( ) = 𝑁! ⋅ (𝑁−𝑖)!
𝑖
Table 3. Relevant number of items for the scenarios proposed in a N-Hypercube topology
Topology Item Description K=4 K=8 generic value
N-Hypercube Equivalent N value respect to K N=3 N=5 N
Number of links to other switches 3 5 N
Number of Ports per switch 5 9 2N-1
Number of hosts per switch 2 4 N-1
Number of hosts overall 16 128 (N-1)2N
Number of switches overall 8 32 2N
Number of links overall 12 80 N∙2N-1

748
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Regarding the Folded N-hypercube, it is an extension of the N-hypercube with the addition of extra links
between any two opposite nodes, hence, there are more available paths to reach a certain node. As a consequence,
N values selected are the same as above, although the number of switches and links are presented in Table 4.
Regarding the amount of nodes to be reached from any given one with any number of hops away, it is greater
than above due to the extra links. In order to undertake the calculations, it is to be considered the definition of
factorial, this is, 𝑛! = 𝑛 ⋅ (𝑛 − 1)! , which yields to (𝑛 − 1)! = 𝑛!/𝑛 . From this expression, it may be deduced
that 0! = 1, and also that the factorial of a negative integer number is undefined, by means of (-1)!=0/0, which
might be skipped if Gamma function is not considered. The values required are just those shown in Figure 5.b,
which may populate the Folded Pascal Triangle for an hypercube of dimension N and the number of hops away
are given by variable j, being the result of summing up the second and the last value of a certain row, then the
𝑁
𝑐𝑒𝑖𝑙( ) 𝑁
third and the before last, and so on. Therefore, the expression desired is the following: ⋁𝑖=0 2 [( ) +
𝑖
𝑁 2 2 2
( )] ∶ [𝑓𝑙𝑜𝑜𝑟(𝑖 ⁄(𝑁 − 𝑖 + 1)) + 1], yielding the right results:𝑁 = 2 → ⋁1𝑖=0 [( ) + ( )] ∙ 1 , [( ) +
𝑁−𝑖+1 0 3 1
2 2 3 3 3 3 3 3 1
( )] ∙ 1 → 1, 3. 𝑁 = 3 → ⋁𝑖=0 [( ) + ( )] ∙ 1 , [( ) + ( )] ∙ 1 , [( ) + ( )] ∙ → 1, 4, 3.
2 0 4 1 3 2 2 2

Table 4. Relevant number of items for the scenarios proposed in a Folded N-Hypercube topology
Topology Item Description K=4 K=8 generic value
Folded N-Hypercube Equivalent N value respect to K N=3 N=5 N
Number of links to other switches 4 6 N+1
Number of Ports per switch 6 10 2N
Number of hosts per switch 2 4 N-1
Number of hosts overall 16 128 (N-1)∙2N
Number of switches overall 8 32 2N
Number of links overall 16 96 (N+1)∙2N-1

(a) (b)
Figure 5. (a) Pascal triangle for the first N values, (b) Folded Pascal triangle for the first N values.

4. CALCULATIONS FOR THE SCENARIOS PROPOSED

All four topologies are going to be compared for three case scenarios, as proposed above, such that for 16 Hosts
(2 per switch), 128 Hosts (4 per switch) and a generic value. The value presented for 0 switches means the one
source host, which obviously does not takes any switch and the value for 1 switch means the hosts hanging on
the same switch as the source host (this is, 1 hop away). Then, it is important to distinguish between tree-like
and graph-like topologies, as in the former, two or three hops away imply the use of 3 and 5 switches,
respectively, due to the tree hierarchy, whereas in the latter, the number of hops away are just the number of
switches away, due to the linear form of a graph [10].

It is also to be noted that in graph-like designs, the distance between any two switches holding a source and a
destination host is the distance between those hosts minus one, as the source switch is not being counted in the
way to the destination switch, as it is the starting point [11]. Furthermore, in those designs, the number of hosts
is the proper cell from the Pascal Triangle, multiplied by the number of hosts per switch. At this point, let us
compare the average distance among hosts (ht) through switches (sw):

Fat Tree (K=4)  Avg(distance among hosts) = (1 ht ∙ 0 sw + 1ht ∙ 1sw + 2ht ∙ 3 sw + 12 ht ∙ 5 sw) / 16ht = 3.81
Fat Tree (K=8)  Avg(distance among hosts) = (1 ht ∙ 0 sw + 3 ht ∙ 1sw + 12 ht ∙ 3 sw + 112ht ∙ 5 sw) / 128ht
= 4.68

749
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fat Tree (K)  Avg(distance among hosts) = ((K/2-1) ht ∙ 1sw + (K2/4-K/2) ht ∙ 3 sw + (K3/4-K2/4) ht ∙ 5 sw)
/ K3/4

Leaf Spine (L=2)  Avg(distance among hosts) = (1 ht ∙ 0 sw + 1 ht ∙ 1sw + 14ht ∙ 3 sw) / 16 ht = 2.69
Leaf Spine (L=3)  Avg(distance among hosts) = (1 ht ∙ 0 sw + 3 ht ∙ 1sw + 124ht ∙ 3 sw) / 128 ht = 2.92
Leaf Spine (L)  Avg(distance among hosts) = ((2L-1-1) ht ∙ 1sw + (23L-2-2L-1) ht ∙ 3 sw) / 23L-2

N-Hypercube (N=3)  Avg(distance among hosts) = (1 h ∙ 0 s + 1 h ∙ 1s + 6h ∙ 2 s + 6h ∙ 3 s + 2 h ∙ 4 s) / 16 h

= 2.43
N-Hypercube (N=5)  Avg(distance among hosts) = (1h∙0s + 3h∙1s + 20h∙2s + 40h∙3s + 40h∙4s + 20h∙5s +
4h∙6s) / 128 h = 3.49
𝑁
N-Hypercube (N)  Avg(distance among hosts) = (∑𝑁 𝑖=0 ( ) ⋅ (𝑖 + 1)) / (N-1)2N
𝑖
Folded N-Hypercube (N=3)  Avg(distance among hosts) = (1 h ∙ 0 s + 1 h ∙ 1s + 8h ∙ 2 s + 6 h ∙ 3 s) / 16 h =
2.18
Folded N-Hypercube (N=5)  Avg(distance among hosts) = (1h∙0s + 3h∙1s + 24h∙2s + 60h∙3s + 40h∙4s) / 128
h = 3.05
𝑁
𝑐𝑒𝑖𝑙( ) 𝑁 𝑁
Folded N-Hypercube (N)Avg(dist.hosts)= (∑𝑖=0 2 [( ) + ( )] : [𝑓𝑙𝑜𝑜𝑟(𝑖 ⁄(𝑁 − 𝑖 + 1)) + 1] ⋅
𝑖 𝑁−𝑖+1
(𝑖 + 1)) / (N-1) ∙2N
Next, let us compare the average number of switches being used in a given transaction. It is to be considered
that switches not being in use in a certain interval of time allow for the application of energy saving schemes, in
a way that unused switches may be put in idle state through control signals generated by an orchestrator, thus
saving resources. In order to undertake this calculations, the results obtained before are being used.

Table 5. Average distance between switches and Average switches in Energy Saving mode
Topology Total switches Total links Avg dist swit Avg Idle rate
K=4 K=8 K=4 K=8 K=4 K=8 K=4 K=8
Fat Tree 20 80 48 384 3.81 4.68 16/20 75/80
Leaf and Spine 10 40 32 192 2.69 2.92 7/10 37/40
N-Hypercube 8 32 12 80 2.43 3.49 5/8 28/32
Folded N-Hypercube 8 32 16 96 2.18 3.05 6/8 29/32

To round it up, it is to be said that some factors, such as the need of a simple internetworking so as to be able to
manage in an easy way or the need of steady values for latency and jitter may make the difference when choosing
a topology. In the first case, graph-like architectures may do a better job for small to middle networks, losing
out as it grows up, whilst in the second case, tree-like architectures may make the deal, as the number of hops
away gets steady while the network grows. Additionally, in tree-like instances, Leaf and Spine may be the winner
in small networks as the number of hops away is never greater than two, whereas Fat Tree, goes up to three.
Besides, it is to be said that the former may get extended with an extra top layer called SuperSpine, making a
three layer infrastructure without the tight restrictions endorsed by the parameter K, although it is to be said that
Fat Tree is much more popular when dealing with three layer deployments.

5. CONCLUSIONS

In this paper, different Data Centre topologies have been presented, such as Fat Tree, Leaf and Spine, N-
Hypercube and Folded N-Hypercube. On the one hand, the first and the second ones are considered as tree-like
topologies, where there are different rows of switches forming a hierarchy, such that just the lower layer holds
the different hosts hanging on them. On the other hand, the third and the fourth ones are branded as graph-like
topologies, where all switches have hosts connected to them.
In order to compare all topologies, the number of hosts has been fixed for two scenarios, leaving a third one with
generic values. The measurements obtained are average distance among switches and energy saving schemes,
all of them on an arithmetic manner. The calculations performed show that there is no such a better topology, as
results vary depending on the number of items involved, and furthermore, other key performance indicators may
be seen, such as the complexity of the network because of the amount of interconnections. Additionally, in
networks with the need of steady rates of latency and jitter, such as VoIP or video streaming, tree-like

750
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

architectures make the difference, taking into account that Leaf and Spine is the best, but it does not scale good,
whereas Fat Tree does achieves a good trade off in all circumstances.

REFERENCES

[1] Mao Y, You C, Zhang J, Huang K, Ketaief K B . A Survey on Mobile Edge Computing: The Communication
Perspective. IEEE Communications Surveys & Tutorials 2017, Vol. 19 , Issue 4, pp. 2322-2358. DOI:
10.1109/COMST.2017.2745201
[2] Al-Fares M, Loukissas A, Vahdat A . A scalable, commodity data center network architecture. In Proceedings in
SIGCOMM’2008, pp. 63-74. DOI: https://doi.org/10.1145/1402958.1402967
[3] Okafor K C . Leveraging Fog Computing for scalable IoT datacenter using Spine-Leaf network topology. In Proceedings
in JECE’2017, Vol. 1, pp. 1-11. DOI: https://doi.org/10.1155/2017/2363240
[4] Harary F, Hayes J P, Wu J-H . A survey of the theory of hypercube graphs. Comput.Math.Applic. 1988, Vol. 15, No.
4, pp. 277-289. DOI: https://doi.org/10.1016/0898-1221(88)90213-1
[5] Karthik K, Jena S, Gopal T V . Hypercube and its variants using NS2. Intern. Journal of Latest Trends in Engineering
and Technology, Special Issue, ICRACSC-2016, pp. 187-193.
https://www.ijltet.org/journal_details.php?id=916&j_id=3804
[6] Zhu, Q, Xu, J-M, Hou , X, Xu , M . On reliability of the folded hypercubes. Journal of Information Sciences 2007; Vol.
177, Issue 8: 1782-1788. DOI: https://doi.org/10.1016/j.ins.2006.11.003
[7] Roig P J, Alcaraz S, Gilly K, Juiz C . Modelling VM Migration in a Fog Computing Environment. Journal Elektronika
ir Elektrotechnika, Vol. 25(5), pp. 75-81. DOI: https://doi.org/10.5755/j01.eie.25.5.24360
[8] Roig P J, Alcaraz S, Gilly K, Juiz C . Modelling a Leaf and Spine Topology for VM Migration in a Fog Computing
Environment. In Proceedings ELECTRONICS 2020, accepted by not yet published
[9]Roig P J, Alcaraz S, Gilly K, Filiposka S, Aknin N . Formal Algebraic Specification of an IoT/Fog Data Centre for Fat
Tree or Leaf and Spine Architectures. In Proceedings ICECCE 2020, accepted by not yet published
[10] Neeman A . Algebraic and Analytic Geometry, Canberra, AUSTRALIA: Cambridge University Press, 2007.
[11] Caldwell J . Structural Induction Principles for Functional Programmers. In: Proceedings TFPIE 2013. 2º Workshop
on Trends in Functional Programming In Education; 2013: Electronic Proceedings in Theoretical Computer Science
136, pp. 16-26.

751
 PROTON EXCHANGE MEMBRANE FUEL CELL
DEGRADATION: AN EXPERIMENTAL ANALYSIS

Sujit Sopan Barhate

Department of Technology, Savitribai Phule Pune University, Pune, India,[email protected], ORCID: 0000-
0002-9895-6902

Rohini Mudhalwadkar
College of Engineering Pune, Instrumentation and Control Department, Pune, India, [email protected]
ORCID: 0000-0003-4565-2646

Barhate, SS, Mudhalwadkar, R. Article title. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
Cite this paper as: 2020, Istanbul, Turkey , Proton exchange membrane fuel cell degradation: An experimental
analysis

Abstract: Proton exchange membrane fuel cell is a clean energy generator as it emits water as a by-product.
The fuel cell has various applications in stationary power generation and transportation. However,
there is need to improve durability for transportation applications. Fuel cell durability is limited as
its performance degrades over a period. Fuel cell performance degrades due to age and fault
conditions. In this study, we have experimentally created faults to study the degradation in its
performance. We created three faults which are common in the fuel cell system. The faults are:
water flooding, reactant gas starvation and high operating temperature. Fuel cell performance
degrades more than 30% during the fault conditions. Degradation due to aging is experimented by
using a membrane which was operated for more than 2000 hours.
Keywords: PEM fuel cell; Fuel cell degradation; Faults in fuel cell
© 2020 Published by ECRES

1. INTRODUCTION

Pollution is a big concern for the world. Many countries have taken targets to reduce the pollution. The pollution
is mainly caused by burning fossil fuels. Moreover, fossil fuel levels are depleting. Hence, using renewable
energy sources is one of the solutions to arrest the issues. Proton exchange membrane (PEM) fuel cell is one of
the candidates for clean power generation. There are various other types of fuel cells like, Solid Oxide, Direct
Methanol, Molten Carbonic, Alkaline, and Phosphorous Acid [2, 4]. Among these, PEM fuel cell is better in terms
of power density, efficiency, transient response, scalability, robustness, simple construction, lower operating
temperature, and faster start-up time [9]. PEM fuel cell is an electrochemical device which generates electrical
power by reacting Hydrogen with Oxygen. By-products of the reaction are water and heat. Since, it emits water
as by-product, it can be termed as clean power generator. If the Hydrogen used in the fuel cell is generated using
renewable energy sources like solar or wind then the power generation is truly non-polluting.
PEM fuel cell has applications in stationary as well as mobile power generations. Fuel cells are being used in
automobile, locomotive and marine. A target of 5000 hours of durability is given by the US Department of
Energy for transportation applications. However, around 4000 hours of durability is achieved so far [1]. Many
researchers have studied the durability of the PEM fuel cell. They have calculated remaining useful life of the
fuel cell based on degradations. Artificial intelligence tools were used to predict the durability [1, 11]. Few used
Auto-Regressive Moving Average (ARMA) models to predict fuel cell performance and to calculate remaining
useful life [5, 7, 12]. ARMA model was also used for prognosis of fuel cell degradation [11]. Fuzzy logic [17]
and adaptive neuro fuzzy interference [18] used to degradation and prognosis respectively.
Fuel cell performance and life degrades due to faults in the fuel cell system and mechanical degradation of the
cell components. Mechanical damages in fuel cell are caused when the fault condition remains for long time.

752
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Fuel cell performs optimum in the standard conditions of operating temperature and humidity [15]. Reactant
gases are humidified before entering into the fuel cell. The gases in tern humidify the membrane. The proton
exchange membrane has a special characteristic of conducting hydrogen protons. The proton conductivity
increases when the membrane is wet. On the converse, proton conductivity drops when the membrane is dried.
The membrane dries up when either humidification is less or the cell is operating at higher temperature. Higher
operating temperature evaporates water inside the cell. Prolong such condition damages the membrane. Pin holes
form on the membrane. Hydrogen molecule directly travel to cathode side and the anode side reaction does not
take place. On the other hand, if temperature is under control but humidification is increased or water removal
from fuel cell is not effective then, water flooding situation arises. Water flooding inside the fuel cell reduces
reaction rate hence, degrades fuel cell performance. Many researchers have worked to detect faults in the fuel
cell system. Electrochemical impedance spectroscopy (EIS) is commonly used technique to detect water
flooding and degradation [8, 16]. EIS is also used for state of health in electric vehicles [13]. Other fault detection
techniques like tomography [6], fuzzy logic [14], artificial neural networks [10], Principle component analysis
[3] have been used.
This paper presents experimental analysis when the fault conditions are present in the fuel cell. It also presents
analysis in performance degradation between new and used fuel cells of same size. Rest of the paper is organized
as follow: section 2 describes fuel cell reactions and polarization curve, section 3 describes the experimental
result analysis, and section 4 concludes the paper.

2. PEM FUEL CELL

PEM fuel cell uses Hydrogen and Oxygen as reactant gases. Hydrogen is supplied at anode and Oxygen at
cathode of the cell. Both the reactant gases flow through a graphite made flow-fields. The reactant gases are
spread on the membrane surface through a gas diffusion layer (GDL). GDL is made up of carbon. Platinum
particles are deposited on the membrane. Platinum is used as a catalyst in the fuel cell. At anode, Hydrogen split
into electrons and protons in a reduction reaction [1]. The electrons flow through the external circuit and protons
travel through the membrane. Electrons and the protons combine together to become Hydrogen element again
at cathode side. Refer reaction [2] for the Hydrogen formation at cathode. The Hydrogen reacts with Oxygen at
cathode to form water. Heat is generated out of the reaction as it is an exothermic reaction. Refer reaction [3] as
a complete fuel cell reaction.

2𝐻2 = 4𝐻 + + 4𝑒 − (1)

4𝐻 + + 4𝑒 − = 2𝐻2 (2)

𝑂2 + 2𝐻2 = 2𝐻2 𝑂 + 𝑒𝑛𝑒𝑟𝑔𝑦 (3)

The fuel cell voltage drops as the current is drawn. Figure 1 (a) shows a polarization curve of the fuel cell. The
terminal voltage drops due to the losses in the fuel cell. There are three types of losses, activation loss, ohmic
loss and concentration loss. The losses are shown in figure 1(b).

753
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)
Figure 1. PEM Fuel Cell (a) polarization curve, (b) losses in the fuel cell

3. EXPERIMENTAL ANALYSIS

The purpose of the experimentation is to monitor and validate the PEM fuel cell performance degradation due
to aging and degradation at fault conditions. Various experiments were conducted to observe the performance
degradation. A PEM fuel cell tester, FCT-50S was used for the experimentation. The tester has a graphical user
interface called FC-Lab. Tests like voltage scan, EIS, current pulse, current interruption can be configured and
executed using the FC-Lab. The test setup has three compressed gas cylinders, Hydrogen, Oxygen and Nitrogen.
Hydrogen used in the experiments was 99.99% pure. Oxygen was medical grade. Nitrogen is used to purge the
cell before and after the experiments. A voltage scan test was executed on the fuel cells. We used two fuel cells
in the experimentation. One had new membrane assembly and another was aged membrane. The aged membrane
had run more than 2000 hours. The experiments are:

Voltage scan test on PEM fuel cell having new membrane assembly

A fuel cell was assembled with the unused membrane. It simulated zero hour operation that means the
performance shall be optimum at standard operating parameters. The assembled fuel cell was connected to the
FCT-50S. The cell was purged by compressed Nitrogen to ensure clean flow-fields. The voltage scan test was
executed. Voltage and current values were logged for further analysis. Similarly, another fuel cell was assembled
with the aged membrane. The degraded cell performance was expected at the standard operating conditions. The
cell was connected to the FCT-50S, the cell purged with Nitrogen to ensure clean field-flow and executed the
voltage scan test. Voltage and current values were logged to further analysis. Aged membrane can be seen with
minor mechanical damages.
Figure 2 shows the comparison between the cell performance. Solid black lines are the polarization and power
density curves for the new membrane assembly fuel cell. Whereas, dotted black lines are the polarization and

754
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

power density curves for the aged membrane assembly fuel cell. Polarization curves have primary Y-axis and
power density curves have secondary Y-axis. Current density is on the X-axis. Current and power densities of
the aged cell was degraded by almost 90% with respect to the new cell. EIS test was performed on both cells
and their result is shown in figure 3. Degradation of the aged cell can be seen from the Bode plot.

Figure 2. PEM fuel cell performance comparison when used new and aged membrane assemblies

(a) (b)
Figure 3. Bode plots of EIS tests on new (a) and old (b) PEM fuel cells

Voltage scan test on PEM fuel cell when operated in water flooding condition

Water flooding is one of the common faults in the PEM fuel cells. The fault condition arises when fuel cell goes
under frequent current cycles, humidification of reactant gases is high and water is blocked in the side the flow-
fields. Water flooding reduces the reaction rate hence drops power output.
Experimentally, the flooding situation was created by frequent current cycling tests and by blocking the water
outlet of the fuel cell. Then, a voltage scan test was executed. Voltage and current values were logged for the
further analysis. Figure 4 shows the difference in fuel cell performance when operated under water flooding
situation with respect to the fuel cell working in standard operating conditions. PEM fuel cell performance
dropped by around 30% when operated in the water flooding condition.

755
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 4. PEM fuel cell performance comparison when operated in standard condition vs water flooding condition

Voltage scan test on PEM fuel cell when operated in reactant gas starvation condition

Reactant gas starvation fault condition can arise when there is issue in gas delivery system. When the fuel cell
does not receive required quantity of reactant gas then, the cell starves. Starving cell degrades performance
because the reaction rate is slow. Prolonged starvation condition can damage the membrane.
The condition was simulated by partially opening the Hydrogen gas valve. The cell was purged with Nitrogen
to ensure there is no other fault condition present. Then, a voltage scan test was executed. Voltage and current
values were logged for the further analysis. Figure 5 shows the difference in fuel cell performance when operated
under starving condition with respect to the fuel cell working in standard operating conditions. PEM fuel cell
performance dropped by around 55% when operated in the water flooding condition.

Figure 5. PEM fuel cell performance comparison when operated in standard condition vs starvation condition

Voltage scan test on PEM fuel cell when operated in high temperature condition

The PEM fuel cell under test has standard operating temperature 500C. However, for the experimentation
purpose, it was operated at 1000C. Usually, fuel cell temperature increases if the thermal control is poor for the
fuel cell system.
The cell was purged with Nitrogen to ensure there is no other fault condition present. Then, a voltage scan test
was executed. Voltage and current values were logged for the further analysis. Figure 6 shows the difference in
fuel cell performance when operated under high temperature condition with respect to the fuel cell working in
standard operating conditions. PEM fuel cell almost did not work when operated at 100 0C. High temperature
cases multiple effects in the fuel cell. It evaporates water inside the fuel cell. It leads to dry membrane. Hence,
proton conductivity is restricted. It results in to almost no reaction at cathode side.

756
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 6. PEM fuel cell performance comparison when operated in standard condition vs high temperature condition

4. CONCLUSION

The paper has presented practical situations in the lifetime of the PEM fuel cell operation. The PEM fuel cell
degrades over time. The cell undergoes mechanical damage when operated for long time. Experimentally, it was
demonstrated that the performance was degraded around 90% in this case. Fuel cell performance also degrades
due to the fault conditions. Instantaneous fault conditions were experimentally created to study the degradation.
Most common faults were chosen in the study. Poor water and thermal management lead to water flooding and
high temperature faults. Reactant gas starvation fault is caused if gas delivery system has an issue. PEM fuel
cell performance degradation observed more than 30% during the fault conditions. Prolonged fault condition
may damage the cell further and reduce its remaining useful life.
Further the study can be extended to detect faults and isolate them based on the pattern they have shown in their
performance during the fault condition.

ACKNOWLEDGMENT

This work was supported by Department of Automobile Engineering, PSG College of Technology, Coimbatore,
India for providing resources like PEM fuel cells and experimentation setup. I sincerely thank Dr. P. Karthikeyan
for his guidance. I am thankful to his research group in helping me in the experimentation.

REFERENCES

[1] Vichard L., Harel F., Ravey A., Venet P., Hissel D. Degradation prediction of PEM fuel cell based on artificial
intelligence. International Journal of Hydrogen Energy, Vol. 45, Issue 29, 2020, pp. 14953-14963.
https://doi.org/10.1016/j.ijhydene.2020.03.209
[2] Gaikwad, S. D. and P. C. Ghosh. Sizing of fuel cell electric vehicle: A pinch analysis-based approach. International
Journal of Hydrogen Energy, Vol 45, Issue 15, 2020; 8985-8993. doi: 10.1016/j.ijhydene.2020.01.116
[3] Lin RH., Pei ZX., Ye ZZ., Guo CC., and Wu BD. Hydrogen fuel cell diagnostics using random forest and enhanced
feature selection. International Journal of Hydrogen Energy, Vol. 45, Issue 17, 2020, pp. 10523-10535.
https://doi.org/10.1016/j.ijhydene.2019.10.127
[4] Khan, S. S., H. Shareef, and A. H. Mutlag. Dynamic temperature model for proton exchange membrane fuel cell using
online variations in load current and ambient temperature. International Journal of Green Energy Vol 16 No. 5; 2019
361-370. doi: 10.1080/15435075.2018.1564141
[5] Wang Z. Lifetime Prediction Modeling of Automotive Proton Exchange Membrane Fuel Cells. SAE Technical Paper,
2019-01-0385, 2019, doi:10.4271/2019-01-0385.
[6] Ifrek L., Rosini S., Cauffet G., Chadebek O., Rouveyre L., and Bultel Y. Fault detection for polymer electrolyte
membrane fuel cell stack by external magnetic field. Electrochimica Acta 2019, 313, pp. 141-150.
https://doi.org/10.1016/j.electacta.2019.04.193

757
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[7] Detti A. H., Steiner N. Y., Bouillaut L., Same A. B., and Jemei S. Fuel Cell Performance Prediction using an
AutoRegressive Moving-Average ARMA Model. IEEE Vehicle Power and Propulsion Conference, 2019.
DOI: 10.1109/VPPC46532.2019.8952535
[8] Ritzberger D., Striednig M., Simon C., Hametner C., and Jakubek S. Online estimation of the electrochemical
impedance of polymer electrolyte membrane fuel cells using broad-band current excitation. Journal of Power Sources,
405, 2018, 150-161. https://doi.org/10.1016/j.jpowsour.2018.08.082
[9] Lue X, Qu Y, Wang Y, Qin C, Liu G. A comprehensive review on hybrid power system for PEMFC-HEV: issues and
strategies. Energy Convers Manag 2018; 171:1273-91. https://doi.org/10.1016/j.enconman.2018.06.065
[10] Abbaspour A., Yen K. K., Forouzannezhad P., and Sargolzaei A. Active Adaptive Fault-Tolerant Control Design for
PEM Fuel Cells. IEEE Energy Conversion Congress and Exposition, 2018, DOI: 10.1109/ECCE.2018.8557620
[11] Zhou D., Al-Durra A., Zhang K., Ravey A., and Gao F. Online remaining useful life prediction of proton exchange
membrane fuel cells using a novel robust methodology. Journal of Power Sources 399, 2018, 314-328.
https://doi.org/10.1016/j.jpowsour.2018.06.098
[12] Kim T. Y., Kim B. S., Park T. C., and Yeo Y. K. Development of Predictive Model based Control Scheme for a
Molten Carbonate Fuel Cell (MCFC) Process. International Journal of Control, Automation and Systems 16(2), 2018
791-803. http://dx.doi.org/10.1007/s12555-016-0234-0
[13] Pivac I., Bezmalinovic D., and Barbir F. Catalyst degradation diagnostics of proton exchange membrane fuel cells
using electrochemical impedance spectroscopy. International Journal of Hydrogen Energy, Vol. 43, 2018, pp. 13512-
13520. https://doi.org/10.1016/j.ijhydene.2018.05.095
[14] Pivac I., Simic B., and Barbir F. Experimental diagnostics and modeling of inductive phenomena at low frequencies
in impedance spectra of proton exchange membrane fuel cells. Journal of Power Sources, 365, 2017, 240-248.
http://dx.doi.org/10.1016/j.jpowsour.2017.08.087.
[15] Barhate, S. S., R. Mudhalwadkar, and A.K. Prakash. A survey on factors affecting performance and durability of
PEM Fuel Cells in Automotive applications. International Journal of Control Theory and Applications, 10(9), 2017,
pp. 659-669.
[16] Jeppesen C., Araya S. S., Sahlin S. L., Thomas S., Andreasen S. J., and Kaer S. K. Fault detection and isolation of
high temperature proton exchange membrane fuel cell stack under the influence of degradation. Journal of Power
Sources, 359, 2017, 37-47. http://dx.doi.org/10.1016/j.ijhydene.2014.04.163
[17] Davies B., Jackson L., and Dunnett S. Expert diagnosis of polymer exchange fuel cells. International Journal of
Hydrogen Energy, Vol. 42, 2017, pp. 11724-11734. https://doi.org/10.1016/j.ijhydene.2017.02.121
[18] Mao L., Jackson L., and Jackson T. Investigation of polymer electrolyte membrane fuel cell internal behaviour
during long term operation and its use in prognostics. Journal of Power Sources, 362, 2017, 39-49.
https://doi.org/10.1016/j.jpowsour.2017.07.018

758
 KINATION- STUDENTS’ EXPLORATION OF A FUTURE
RENEWABLE ENERGY SOURCE

Penka V. Georgieva
Burgas Free University, Burgas, Bulgaria, [email protected], ORCID: 0000-0001-9598-5997

Milen Baltov
Burgas Free University, Burgas, Bulgaria, [email protected], ORCID: 0000-0002-1195-7479

Maria Daskalova
Burgas Free University, Burgas, Bulgaria, [email protected], 3rd year student, Software Engineering,
bachelor

Nikolai Tsonev
Burgas Free University, Burgas, Bulgaria, [email protected], 3rd year student, Software Engineering, bachelor

Kalina Hristova
Burgas Free University, Burgas, Bulgaria, [email protected], 3rd year student, Software Engineering, bachelor

Georgi Karavasilev
Burgas Free University, Burgas, Bulgaria, [email protected], 3rd year student, Software Engineering,
bachelor

Georgieva, P. V., Baltov, M., Daskalova, M., Tsonev, N., Ahmed, G., Hristova, K.,
Cite this paper as: Karavasilev, G. KiNation- students’ exploration of a future renewable energy source. 8th
Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In this paper, several aspects of the management and technological process of creating
technological startup KiNation are presented. KiNation is focused at developing KiPath - a
technologically innovative pavement for collecting kinetic energy from vibrations that occur after
the movement of pedestrians, vehicles, industrial machines, etc. Once the energy is collected it is
transformed into electricity, which later can be used for different purposes. KiPath is equipped
with a small solar panel as an alternative energy source in case there is no movement on the
pavement. Some of the challenges that the team has faced are also described.
Keywords: Kinetic energy, renewable energy source, piezoelectric device, technological startup
© 2020 Published by ECRES

1. INTRODUCTION

In nowadays the energy sector, which is the basis for the overall economic development, takes a key role in
achieving sustainability and stability on a global scale. In correspondence with contemporary requirements and
needs, energy sector must provide fuels and energy with minimal public costs and in compliance with the
requirements for environmental protection.
One innovative method for generating energy is using piezoelectric devices, based on the concept that energy
generated from vibrations from people, vehicles, waves etc. can be collected and transformed into electricity.
[1], [2]
Kinetic energy is energy of motion and everybody possesses it. When the body is at rest, his kinetic energy
equals to zero, but when it moves, the kinetic energy always has a positive value and the greater the acceleration
of the moving body is, the greater numerical value of the kinetic energy gets. And as soon as it is the energy of
motion, therefore it could be found in every object that moves. In our daily round we are surrounded by it, people

759
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

who go for a walk or people who ride bicycle around the city. Another aspect of the use of kinetic energy is in
nuclear physics, where elementary fragments are used as projectiles for breaking up atom nuclei and other
elementary particles in order to extract even smaller particles. [3]
Another example in which kinetic energy can be found is in wind. With the help of wind turbines, it could the
transformed into electricity. Also, newer approaches exist for transforming and using such energy: for example,
kinetic plates can collect kinetic energy from human foot and to convert it into electricity. [4], [5]
In 2009, Lawrence-Cook produced Pavegen – floorings that create energy out of the electrical network via the
weight of the people who walk on it. The floorings consist of plates made of 100% recycled rubber. The sides
are connected to induction generators made of 80% recycled materials. Moving vertically through the human
weight, a rotational motion is generated, which generates electricity, independently of the surrounding electrical
network. [6], [7], [8]

2. KiNation -THE TECHNOLOGICAL STARTUP

KiNation is a technological startup aiming at providing an alternative energy source by developing specific
pavement, which collects the kinetic energy from the movement of pedestrians, vibration from vehicles,
industrial machines, waves, sound etc. This pavement can convert the kinetic energy into electricity. The kinetic
pavement provides an opportunity for generating energy in a completely environmentally friendly way, without
any risk for the people and the environment. One of the main problems when using the services of the large
electricity suppliers is the high price for consumption and installation. The implemention of KiPath would
reduce electricity costs. Moreover, the possibility of KiPath to be installed both indoor and outdoor makes it
applicable literally everywhere.
A startup is most often a temporary organization, created with the idea that it will grow and develop into a large
company. Usually the creators of technological startups consider that their product will be important for society
and will be able to change the world in their field. Here is where the main difference in the business model of
starting a small business and the founding a startup comes from. In the first case, the entrepreneur aims at finding
his place in the local market and his goals are his company to bring income for him/her and their family, whereas
the startup founder aims at creating something, which will affect the whole business –a product, which will
change the market, will draw in a lot of foreign clients, will have a massive scale. [9]
KiPath is created on a similar principle, which collects kinetic energy from pedestrians, vibrations from vehicles,
industrial machines, waves, sound and others, converting it into electricity. KiPath is also equipped with small
solar panels for maximum energy production in moments of reduced traffic. (Fig. 1)

Figure 1. The logo of KiPath

The team, standing behind the creation of every successful project, is the most essential ingredient. A successful
team means more productivity, organization and motivation. To achieve this, there are various important aspects
that a team leader must be aware of. Commitment to company goals and ideas is the first of these basis elements.
The team leader is the core of the team`s understanding. Through him/her each of the team members become
aware of their role, of their responsibility and therefore know how they should contribute to the final goal.
Communication is the element that ensures the smooth work of the team. Proper communication is manifested
in the possibility for each team member to express their opinion, to present their ideas and to ask for help.
Conflict management is another component which is crucial for good team management because there is no
team in which there are no conflicts. They are more than normal, but the key is to manage them in a productive
way. Here the role of the leader is to show the team members how they can deal with the unnecessary stress in
the work process and to set good tone of communication. He/she must be involved in all the activities, especially
those that require urgent action.

760
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. The team of KiNation

3. KiPath – THE PRODUCT

The goal of the KiNation team is to create a product that has features good enough to be presented to customers
and thus to get as much feedback from them as possible with the least resources. Conceptual scheme of KiPath
is shown in Fig. 3.

Figure 3. The conceptual scheme of KiPath – harvesting kinetic energy

At the beginning the market is defined as one that does not satisfy many types of customers and uses, such as
the one that is only niche and it is important to use the right segment of the market in order to get start-up capital
with which we can expand our startup.
One way to promote KiPath is to develop an application. Through it we can keep an eye on the data of active
users and their rating of how satisfied they are of that we develop. Using agile methodology that helps our start-
up company to work in the right way of work and adapt as quickly as possible to user`s feedback also contributes
to right management. For our future progress we aim to apply for various funding campaigns in order to scale
our company. We have already participated in a pre-accelerator program that has given us the opportunity to
work with different mentors and teams which helps to improve our knowledge in the field of entrepreneurship
and motivation.
Arduino Uno development platform with ATmega 328 microcontroller is used for implementation and it is
chosen because of its low cost and ease of construction and prototype testing. To achieve the piezoelectric effect,
piezoelectrics are connected in parallel and attached to cork. Cork is a good dielectric, possesses hydrophobicity,
has a low cost and these features make it suitable as a base for the prototype. The prototype mainboard MB102
gives an opportunity to create testing project without the need for welding. It is important component of the
whole architecture that connects and synchronizes all the technical components of the prototype. In the
implementation of the project two solar panels 1W6V with high conversion value and efficiency are used. Solar
panels are suitable for all projects implemented with battery power. The battery which supplies ithe prototype
is from MS series of MHB with a capacity of 1.3 (Ah) and voltage of 6V. (Fig. 4)

761
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 4. KiPath- the prototype

Piezoelectric harvesters generate energy by bending mechanical elements, i.e. beams or membranes. The
resultant mechanical vibrations oscillate at resonance frequencies which can range from tens to hundreds of
Hertz. There are several factors that cause vibrations on a rigid body in a dynamic system, such as system mass
unbalance, and tear and wear of the materials. Each system has a unique behavior that can be described through
the damping constant and natural frequency. The study of the dynamic characteristics of a vibrating body
associated with energy harvesting is usually done with a single degree of freedom lumped spring-mass system.
In the absence of commercially available kinetic energy, the harvesting portable devices that could be used to
collect energy from pedestrians can be described by mathematical estimations of kinetic energy as a standard
mass-spring damping system and energy collected from pedestrians’ unrestricted motions. Once the gravity is
filtered out from the raw acceleration values, the filtered acceleration is converted to proof mass displacement
using the Laplace domain transfer function:

𝐴(𝑠)
𝑧(𝑡) = 𝐿−1 {𝑍(𝑠)} =
𝑏 𝑘
𝑠2 + 𝑠 +
𝑚 𝑚

where 𝑚 is the proof mass, 𝑘 is the spring constant, 𝑏 is the damping factor, 𝐴(𝑠) and 𝑍(𝑠) denote, respectively,
the Laplace transforms of the overall magnitude of the acceleration 𝑎 (𝑡) = √𝑎𝑥2 (𝑡) + 𝑎𝑦2 (𝑡) + 𝑎𝑧2 (𝑡)
and the proof mass displacement 𝑧(𝑡). The resulted proof mass displacement 𝑧(𝑡) is limited by the limit of the
proof mass displacement 𝑍𝐿 . Finally, the generated harvested power is determined by:
𝑝(𝑡) = 𝑏𝑧̇ 2 (𝑡).

4. CHALLENGES

The challenges which every startup face, aiming to develop a technological product, unknown so far on the
market and its implementation niche, which already exists in the global market are not unimportant at all. The
development of our product, above all, consist of an initial prototype that potential buyers can get based in. It
consists of two planes of cork, to which piezoelectricts are attached, connected in parallel, forming a loop by
themselves through cables.
Simultaneously, all technical components are connected to a mainboard that is programmed afterwards. In order
to increase energy production even in the hours with minimal traffic and vibration that KiPath collects, we have
included as an alternative energy source- solar panels. Before implementing the idea and the launch of the
product on the market a research about the area with the most intensive traffic and vibration, is made.
It is also important to determine potential competitors. The time spent on development, testing and improvement,
so that our product to be innovative and unique, and not just another one on the market, is also so viral. It is

762
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

good practice researches and consultations in different fields and sources to be made so that our product does
not effect on copyright and intellectual property, which would lead to new obstacles.
We have taken into consideration with the new trends in technology and we coordinate our technical skills, along
with the marketing ones to avoid one not surpassing the other one. On the other hand, we would not inform the
community about all the features of the product before its official launch in order to keep the interest of our
customers. Releasing all the technical features during the first presentation would lead to huge interest in terms
of the event by the presents, which would help the sale of our technological product afterwards.

5. CONCLUSIONS AND DEVELOPMENT

The dynamics of modern life contribute to the need of innovations and development in all aspects. Industry 4.0
is key factor in creating modern technology startups, business innovations and new niche markets. A product
itself is not enough to earn hard-earned audience of potential buyers. The key is in the uniqueness and the
features of KiPath as well as the ability to include so many improvements and extensions, ensuring the retention
of current customers, earning new ones and expansion of the business on a national and global scale afterwards.
In the process of development it became clear that this kind of pavements can be used in different ways as power
source of street lighting , for making and maintaining illuminated sidewalks as well as for powering led lightings
used to illuminate information boards for public transport , advertising signs, billboards and others.
In the future we plan to put different sensors that detect traffic intensity, temperature changes as well as seismic
activity, which would help to inform the community on time and detect earthquakes.

ACKNOWLEDGMENTS
This research is funded by National Scientific Research Fund of Bulgaria under Contract No. KP-06-COST-8/
06.08.2019 for providing national co-financing for the participation of Bulgarian teams in approved actions
under the European program for cooperation in the field of research and technology COST under the project
"Characteristics prediction and optimization of a photovoltaic system with artificial intelligence methods”.
The research behind the content in this paper is conducted under the IRISI („Indexing the Regional Innovative
Levels in the Sectors of the Economy - scenario for the identified in the ISSS four priority thematic areas for
smart specialization and their positioning towards the circular economy”) project financed by the Bulgarian
Scientific Research Fund.

REFERENCES

[1]www.researchgate.net/publication/324008016_Harvesting_Kinetic_Energy_of_Footsteps_on_Specially_Designed_Til
es?fbc lid=IwAR1JQk0mxGduX11ZubaNfhC8tnRZOq5ogigRYeAZdami6dz_x-Su7BBZhLo
[2] OECD Green Growth Studies: Energy. Aviable on: www.oecd.org/greengrowth/greening-energy/49157219.pdf
[3] R. S. Dahiya, M. Valle, Robotic Tactile Sensing, © Springer Science+Business Media Dordrecht 2013
[4] C. Hairfield, Harnessing Kinetic Energy from Humans. Renewable Energy and the Sustainable Campus Professor
David Dethier, May 14th, 2013
[5] B. Pozo, J. Garate, J. Araujo and S. Ferreiro, Energy Harvesting Technologies and EquivalentElectronic Structural
Models – Review. Electronics 2019, 8, 486
[6] www.world-nuclear.org/information-library/energy-and-the-environment/renewable-energy-and-electricity.aspx
[7] https://impakter.com/pavegen-paving-way-sustainability/
[8] https://smartcityhub.com/energy/pavegen-energy-through-the-power-of-footsteps/
[9]https://masschallenge.org/article/how-to-start-a-tech-
company?fbclid=IwAR2zU2AE2ttAcAb_5MTM4ZprjPTSyGxkXGsCP AAkxwHnZjGnhLVClQz4V6M

763
 3D FLOW SIMULATION OF A HORIZONTAL AXIS WIND
TURBINE WITH A FLAT PLATE ON PRESSURE SIDE
Mustafa Serdar Genç
Erciyes Üniversitesi, Mühendislik Fakültesi, Kayseri, Turkey, [email protected], MSG Teknoloji Ltd. Şti.,
Kayseri, Turkey, ORCID: 0000-0002-6540-620X

Khalida Sekhoune Özden

Erciyes Üniversitesi, Mühendislik Fakültesi, Kayseri, Turkey, [email protected], ORCID: 0000-0003-
0525-4088

Tuna Murat BODUR

MSG Teknoloji Ltd. Şti., Kayseri, Turkey, [email protected], ORCID: 0000-0003-2307-4603

Genç, M.S, Özden K.S., Bodur T.M. 3D flow simulation of a Horizontal Axis Wind Turbine
Cite this paper as: with a Flat Plate on Pressure Side. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: This study presents a 3-D computational study of a new innovative horizontal axis wind turbine
design using the ground effect principle. To benefit from the cushion effect, a flat plate was
installed at a specified distance under the pressure surface of each wind turbine blade. Standard
HAWT case and two other cases with h/c= 0.1 and h/c=0.25 configurations were calculated using
SC/flow Cradle CFD commercial code. Six-Degree of Freedom rigid body motion method is
employed to calculate the rotation motion of the rotor and SST Shear-Stress Transport k-ω
turbulence model is used to simulate the fluid region. The impact of the plate on the rotation of
wind turbine and the general flow behavior around and behind the wind turbine was discussed.
Additional results of the predicted pressure and velocity magnitude were present in form of
contours. The results indicated that the choice of minimum h/c=0.1 was the best choice for the
replacement of the flat plate, further the installation of the flat plate could enhance the performance
of HAWT.
Keywords: Horizontal axis wind turbine, blade with flat plate, Cradle
© 2020 Published by ECRES

1. INTRODUCTION

The growth of interest in renewable energies increases day by day especially wind power. This source of
sustainable energy is attracting attention and became popular regarding its advantages comparing to the other
existing sources. Also named as clean energy and eco-friendly, this energy can be adaptive for domestic
installations by placing a wind turbine. This last can be divided into two categories VAWT and HAWT and
generally, in the research on these devices, optimizations are made to maximize the electrical generation.
Concentrating on the different elements of the wind turbine, blade optimization takes the most important part,
Bing Zhu et al.[1] investigated in the performance characteristics of a horizontal axis wind turbine by installing
a fusion winglet at the tip of the rotor blade and reported that it can increase the WT power performance. Seoa
and Hong [2] studied numerically the performance improvement of airfoils for wind blade using groove different
shapes and positions or grooves on blades. Beyhaghi and Amano [3] studied leading edge slots use on NACA
4412 wind turbine airfoil at different angles of attack placing slots on leading edge of blades. Matthias Schramm
et al. [4] conducted a simulation of an airfoil using leading edge slat. Oltmann et al. [5] investigated on load
reduction of wind turbines using trailing edge flaps. Zhang et al. [6], Manolesos and Voutsinas [7] investigated
on (VGs) on surface of wind turbine blades. Wing in ground effect can benefits from cushion or Ram effect, as
a result of air static pressure augmentations on the lower wing surface Qu et al. [8-13]. Ahmet et al. [14] and
Hsiun and Chen [15] conducted a study about the influence of ground effect on airfoil performance numerically.

764
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

In order to benefit from this phenomenon, in this paper 3-D computational analysis was applied to study the
performance change of a new designed horizontal axis wind turbine. The ground effect on wind turbine blade is
simulated by placing a flat plate under the pressure surface of each blade (3 blades) [16,17] the distance between
this two elements is changed and the results are compared with those obtained from simulation of the standard
wind turbine. This idea is protected with a patent application [16].

2. Methodology of the simulation

This section includes three different parts, the model preparation of the two HAWT configurations, calculation
setup part where discontinuous (dynamic) meshing details with boundary conditions are introduced. Finally, the
post-processing part that compromise the visualization of the various results.
The commercial code MSC Cradle-SC/Flow was employed in this 3-D numerical simulation. It solves
Equations of (1) and (2) which represent conservation of mass and conservation of momentum Equations. SST
(Shear-Stress Transport) k-ω turbulent model was used in this investigation and the convenient transport
equation is given by Equation (3). Incompressible transient simulation of 4000 cycles have been executed. As
shown in Figure 1 and Figure 2, computational domain was divided into two related parts, main domain which
presented stationary and the subdomain which rotational. The rotational movement of the rotor was calculated
using six-degree-of-freedom for rigid-body-motion. Figure 1 presents the two configurations (a) the standard
HAWT of 2.8 m of diameter and 5.2m of high, to get the second configuration with the flat plate were placed at
a certain distance from the pressure surface of the blades. This distance was unidimensional and set as h/c where
h was the distance and c was the average chord of the blade. In this study two different cases were simulated for
two h/c values 0.1 and 0.25 where the average chord was 0.150 mm.

(1)

(2)

(3)

(a) (b)
Figure 1. Geometric details of the used wind turbines (a) HAWT without flat plate, (b) HAWT with flat plate

Computational domain generation

Figure 2 shows the dimensions of computational domain and boundary conditions of the present CFD model,
the horizontal axis wind turbine is placed in coincidence with the fluid domains bottom wall. Center of wind
turbine is 3C (three chord lengths) away from the inlet of domain and 15C away from the outlet. Where D is the
diameter of the cylinder (fluid subdomain) which includes the hub and blades parts. All elements of the wind
turbine three blades, three flat plates, hub, nacelle and tower are set as wall. Environmental pressure at the outlet

765
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

boundary, velocity at the inlet boundary (10m/s) and finally the rest of the fluid domain is set as symmetry
condition.

Figure 2. CFD computational domain and boundary conditions

Mesh generation

Figure 3 presents the grid created around the wind turbine blades ant the fluid region stationary mesh domain in
green and the dynamic mesh domain in the purple color. Polyhedral mesh is used, and the number of total
elements is 1,550,000. Total of 12 layers of inflation were generated near the wall surfaces including the wind
turbine elements and the bottom wall of the enclosure domain. These calculations were carried out in transient
mode with total 4000 cycles and time step of 0.001s.

Figure 3. Static mesh domain (Green) and dynamic mesh domain (Purple)

3. Results and discussion

Results obtained from the simulation of the standard horizontal axis wind turbine are reported in Figure 5 and
Figure 6 in the other side simulation results of the wind turbine with the flat plate are indicated in Figure 7 and
Figure 8 for h/c= 0.1 and Figure 9 and Figure 10 for h/c=0.25. Total pressure and velocity contours around the
wind turbines are given at two different time frames 1 sec and 4 sec. It can be observed that velocity values of
the standard wind turbine are near to values of velocity magnitude of the case when h/c=0.25. Maximum velocity
values are respectively 21.01 m/s and 22.4, in the case of h/c=0.1 maximum value is smaller (14.61m/s). Further,
at the wake region of HAWT with flat plates more turbulent flow is generated at the case of h/c=0.1. On the
other hand, it is also observed that the maximum rotation velocity is 24 rpm when the plate is set at h/c=0.1 and
for HAWT with the plate at h/c=0.25 it is 14.4 rpm, whereas the standard HAWT has 22.4 rpm rotation velocity.

766
8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)
Figure 5. Total pressure of standard HAWT (a)t=1s, (b)t= 4s

(a) (b)
Figure 6. Velocity magnitude of standard HAWT (a) t=1s, (b) t=4s

(a) (b)
Figure 8. Total pressure of HAWT with flat plate h/c 0,1 (a) t=1s , (b) t=4s

(a) (b)
Figure 7. Velocity magnitude of HAWT with flat plate h/c 0.1 (a) t=1s, (b)t= 4s

767
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)
Figure 9. Velocity magnitude of HAWT with flat plate h/c 0.25 (a) t=1s, (b)t= 4s

(a) (b)
Figure 10. Total pressure of HAWT with flat plate h/c 0,1 (a) t=1s, (b)t= 4s

Figure11. presents streamlines around the three configurations Standard HAWT, HAWT with flat plate with
h/c=0.1, and HAWT with flat plate with h/c=0.25 at t=4s. Relatively high level of turbulence is observed in the
wake region when flat plates are employed.

(a) (b) (c)

Figure 11. Streamlines (a) Standard HAWT, (b) HAWT with flat plate h/c=0.1, (c) HAWT with flat plate h/c=0.25 at
t=4s

4. Conclusion

In this study, 3-D computational analysis was presented the performance of a new designed horizontal wind
turbine. Due to the placed plate under the pressure surface of the blades, the turbine take advantage from the
created cushion effect. It can be seen that velocity values of the standard wind turbine are near to values of
velocity magnitude of the case when h/c=0.25. At the wake region of HAWT with flat plates, more turbulent
flow is generated at the case of h/c=0.1. On the other hand, the maximum rotation velocity is 24 rpm when the
plate is set at h/c=0.1 and for HAWT with the plate at h/c=0.25 it is 14.4 rpm, whereas the standard HAWT has
22.4 rpm rotation velocity.

ACKNOWLEDGMENT
The authors are grateful to the Small and Medium Enterprises Development Organization of Turkey
(KOSGEB) for Research and Development Project Funding.

768
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1]. Zhu B, Sun X, Wang Y, Huang D. Performance characteristics of a horizontal axis turbine with fusion winglet. Energy
2017; 120:431-40.
[2]. Seo SH, Hong CH. Performance improvement of airfoils for wind blade with the groove. Int J Green Energy
2016;13(1):34-9.
[3]. Beyhaghi S, Amano RS. A parametric study on leading-edge slots used on wind turbine airfoils at various angles of
attack. J Wind Eng Ind Aerod 2018;175: 43-52.
[4]. Schramm, M., Stoevesandt, B., & Peinke, J. (2016, September). Simulation and optimization of an airfoil with leading
edge slat. In Journal of Physics: Conference Series (Vol. 753, No. 2, p. 022052).
[5]. Oltmann NC, Sobotta D, Hoffmann A. Load reduction of wind turbines using trailing edge flaps. Energy Procedia
2017; 136:176-81.
[6]. Zhang, L., Li, X., Yang, K., Xue, D., 2016. Effects of vortex generators on aerodynamic performance of thick wind
turbine airfoils. J. Wind Eng. Ind. Aerod. 156, 84–92.
[7]. Manolesos, M., Voutsinas, S.G., 2015. Experimental investigation of the flow past passive vortex generators on an
airfoil experiencing three-dimensional separation. J. Wind Eng. Ind. Aerod. 142, 130–148.
[8]. Qu Q, Jia X, Wang W, et al (2014) Numerical study of the aerodynamics of a NACA 4412 wing in dynamic ground
effect. Aerospace Science and Technology 38:56–63. DOI: 10.1016/j.ast.2014.07.016
[9]. Qu, Q., Wang, W., Liu, P., & Agarwal, R. K. (2015). Airfoil aerodynamics in ground effect for wide range of angles
of attack. AIAA Journal, 53(4), 1048-1061.
[10]. Qu, Q., Ju, B., Huang, L., Liu, P., Agarwal, R. K. (2016). Flow Physics of a Multi-Element Airfoil in Ground Effect.
In 54th AIAA Aerospace Sciences Meeting (p. 0856).
[11]. Qu, Q., Huang, L., Liu, P., Wang, W., & Agarwal, R. K. (2016). Numerical Study of Aerodynamics and Flow Physics
of the 30P30N Three-Element Airfoil in Dynamic Ground Effect. In 34th AIAA Applied Aerodynamics Conference
(p. 3722).
[12]. Qu, Q., Wang, W., Liu, P., & Agarwal, R. K. (2015). Aerodynamics and Flow Mechanics of a Two-Element Airfoil
in Ground Effect. In 53rd AIAA Aerospace Sciences Meeting (p. 0550).
[13]. Qu, Q., Zuo, P., Wang, W., Liu, P., & Agarwal, R. K. (2015). Numerical investigation of the aerodynamics of an
airfoil in mutational ground effect. AIAA Journal, 53(10), 3144-3154.
[14]. Ahmed MR, Takasaki T, Kohama Y (2007) Aerodynamics of a NACA4412 Wing in Ground Effect. AIAA Journal
45:37–47. DOI: 10.2514/1.23872
[15]. Hsiun C-M, Chen C-K (1996) Aerodynamic characteristics of a two-dimensional wing with ground effect. Journal of
Aircraft 33:386–392. DOI: 10.2514/3.46949
[16]. Genç M. S., Özden Sekhoune K, Karasu İ, Özden M,, “An Airfoil Located Flat Plate On Pressure Side”, TURKISH
Patent, 2017/23311, 2017.
[17]. Özden K.S., Karasu İ., Genç M.S., Experimental investigation of the ground effect on a wing without/with trailing
edge flap, Fluid Dyn. Res., 52, 1-20, (2020)

769
 FLOW-INDUCED MECHANICAL ANALYSIS OF A
HORIZONTAL AXIS WIND TURBINE WITH A FLAT PLATE ON
PRESSURE SIDE

Mustafa Serdar Genç

Erciyes Üniversitesi, Mühendislik Fakültesi, Kayseri, Turkey, [email protected], MSG Teknoloji Ltd. Şti.,
Kayseri, Turkey, ORCID: 0000-0002-6540-620X

Tuna Murat BODUR

MSG Teknoloji Ltd. Şti., Kayseri, Turkey, [email protected], ORCID: 0000-0003-2307-4603

Khalida Sekhoune Özden

Erciyes Üniversitesi, Mühendislik Fakültesi, Kayseri, Turkey, [email protected], ORCID: 0000-0003-
0525-4088

Genç, M.S., Bodur, T.M., Özden K. S., Flow-Induced Mechanical Analysis of A Wind Turbine
Cite this paper as: with a Flat Plate on Pressure Side 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: In this study, a structural analysis of a new innovative horizontal axis wind turbine (HAWT) design
using the ground effect principle. A flat plate was placed at a certain distance (h/c= 0.1 and
h/c=0.25) on the pressure side of blade, and it was modeled using MSC APEX modeler, MSC
NASTRAN solver and Cradle (SC/Flow). Six-Degree of Freedom rigid body motion method is
used to simulate the rotation of the blades. When the results were examined, it was shown that the
highest displacement was on the turbine blades and plates. These values were the maximum of
1.13x10-1mm. It was concluded that the PLA material selected for the blades and plates was usable
according to the results of this analysis.
Keywords: Fluid Structure Interaction, Wind Turbine Blade, Ground effect, MSC Software, Structural
Analysis
© 2020 Published by ECRES

1. INTRODUCTION

The utilization of wind energy has progressed rapidly in the last two decades as its energy production cost
decrease. Atmospheric boundary layer, wind gust, thermal convection stratification, unsteady and low Reynolds
(Re) number aerodynamics problems, etc. affect the cost and energy production of wind turbines. These effects
raise important design problems. To prevent negative effects on wind turbines and to increase aerodynamic
performance, there are many flow control techniques that have two branches such as active and passive [1,2].
Passive control techniques are the methods that should be used continuously in the flow, such as vortex
generators (VGs) [3-7] or roughness [8] while actives are the methods that should be used in the flow when it is
needed, such as high lift devices [9], suction and/or blowing [10-13], acoustic disturbance[14- 17], and plasma
actuator[18,19]. Both approaches have associated added “costs” such as extra drag force or increased swirl and
complexity, thus new approaches are desired. A new concept passive flow control method is partially flexibility
[20,21]. The other new idea on a wind turbine with the ground effect is protected with a patent application [22].
The ground effect on the wind turbine blade formed by placing a flat plate under the pressure surface of each
blade [22,23]. In this study, a structural analysis was conducted to investigate the strength of a newly designed
horizontal axis wind turbine with a flat plate. The ground effect on the blades of HAWT is modeled by placing
a flat plate on the pressure surface of blades based on the studies [22,23].

770
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2. METHODOLOGY

In this study, aeroelastic simulation of the small-scale horizontal axis wind turbine was done. Structural analysis
with Fluid-Structure Interaction (FSI – One-way Interaction) was conducted in MSC APEX modeler and MSC
NASTRAN solver, using pressure distributions obtained from flow analysis in Cradle (SC/Flow). For small
scale HAWT turbine structural analysis, PLA (Polylactic Acid) material was selected for the blades and stainless
steel-304 material was used for the tower. Deformation and stresses on the wind turbine blade and the flat plate
at a wind speed of 10 m/s were obtained, and critical zones where were important depending on strength were
identified. The state in which pressure distributions from time-dependent flow motion become stable was studied
as one-way like in Figure 1.

(a) (b)
Figure 1. (a) Concept of fluid structure interaction [24]
(b) Data exchange between FEM and CFD [25]

SST (Shear-Stress Transport) k-ω turbulent model was used throughout the flow analysis which was more
successful at solving low Reynolds flow separations than other RANS models. In the structural analysis, the
MSC APEX and MSC Nastran Solver in MSC Software provided the opportunity to reduce the required solution
time and not to fill the disk space unnecessarily during the solution. As a result of CFD analysis, the pressure
distributions over the HAWT were obtained, then the boundary conditions in MSC Apex, the material definition,
and the mesh correction process preparations were conducted. Finally, the analysis results from the MSC Natran
solver were taken (Figure 2).

Figure 2. MSC Software Product Tree

As shown in Figure 3, MSC Software is very capable of transmitting the flow analysis results among the software
in its product tree. In the study, the pressure distributions formed on the flat plate of the new concept wind
turbine, whose flow analysis was completed with Cradle-SC/Flow, and the structural characteristics of MSC
APEX were integrated into the preparation program. Based on the results of the flow analysis, it is the 4th second
that the turbine blade begins to rotate and stabilizes. In this case, structural analysis was performed according to
the pressure distribution. During this examination, displacements on parts, Von Misses stresses were examined

771
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

and the resistance of the turbine, whose design and material selection was made to the limit conditions were
examined.

(a) (b)

(d) (c)
Figure 3. (a) MSC Software Fluid – Sturcture Contact Converter (b) Results Contact on MSC Software (c) Fluid Area
And Structure Area (d) Pressure and Structure Interation

(c)

(a) (b)
Figure 4 (a) Pressure on WT – front view (b) Pressure on WT – back view (c) Mesh on WT blade with flat plate

Figure 4 also shows the pressure domains formed on the turbine when the flow analysis results are transferred
to the structural part. Maximum pressures was observed at the points that the flow contacted the surface, and the
pressure distribution at the remaining points caused by the rotational motion was obtained. In Table 1, material
properties used in parts of HAWT with flat plate in MSC Apex were given.

772
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1. Material of Turbine

Material Names PLA - Bioplymer 304 – Stainless Steel
Density – kg/mm3 1.29e-6 8e-6
Elasticity Modulus (MPa) 2910 1.93E+5
Poisson’s Ratio 0.3 0.29
Turbine Application Area Turbine Blades and Plates Turbine Hub, Nacel and Tower

RESULTS AND DISCUSSIONS

In the study, a structural analysis was carried out on the newly designed wind turbine with the flat plate under
flow of 10m/s. The results of the pressure distribution on the turbine in the fourth second, which was assumed
to be stable from the moment it started to spin. This study was done as one-way. The studies of flow analysis
were carried out using MSC Software's SC/Flow (Cradle), Nastran solvent and MSC Apex programs for
structural analysis were performed. MSC Software provided advantages throughout these analyses in terms of
fast mesh capacity, lack of dense space in the results, fast imaging of fluid and structural interactions, and ease
of use.
The maximum stresses due to pressure distributions formed at the connection and root areas of the turbine tower
as seen in Figure 5, and it was obtained that this value was 8.46x10-2 MPa. In Figure 6, the maximum
displacement due to the pressure distribution (in the fourth second) was observed on the turbine blades and the
flat plates. These values were 1.13x10-1 mm on the plate and 5.43x10-2mm on the turbine blade. When the results
were examined, it was shown that the highest displacement was on the turbine blades and plates. These values
were the maximum of 1.13x10-1mm. It was concluded that the PLA material selected for the blades and plates
was usable according to the results of this analysis.

(a) (b)
Fig 5. (a) Von Misses Stress on Wind Turbine – Front Shown (b)Von Misses Stress on Wind Turbine – Side Shown.

773
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(a) (b)

(c) (d)
Fig.6. Maximum displacements on the blade and the flat plate at different views.

4. CONCLUSION

In this study, a structural analysis of a new innovative horizontal axis wind turbine (HAWT) design using the
ground effect principle was conducted. A flat plate was placed at a certain distance (h/c= 0.1 and h/c=0.25)
under the pressure side of blade using MSC APEX modeler, MSC NASTRAN solver and Cradle (SC/Flow).
Six-Degree of Freedom rigid body motion method is used to simulate the rotation region of the turbine. When
the results were examined, it was shown that the highest displacement was on the turbine blades and plates. The
maximum stresses due to pressure distributions formed at the connection and root areas of the turbine tower.
Further, it was concluded that the PLA material selected for the blades and plates was usable according to the
results of this analysis.

ACKNOWLEDGMENT

The authors are grateful to the Small and Medium Enterprises Development Organization of Turkey
(KOSGEB) for Research and Development Project Funding.

REFERENCES

[1] Genç, M. S., Karasu, I., Açıkel, H. H., Akpolat, M. T. Low Reynolds number flows and transition. In: Low Reynolds
number aerodynamics and transition. InTech, 2012.

774
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[2] Koca K., Genç, M. S., Açıkel H.H., Çağdaş M., Bodur T.M., Identification of Flow Phenomena over NACA 4412 Wind
Turbine Airfoil at Low Reynolds Numbers and Role of Laminar Separation Bubble on Flow Evolution, Energy, vol
144, pp.750-764, 2018.
[3] Xue, S., Johnson, B., Chao, D., Sareen, A., Westergaard, C. Advanced aerodynamic modeling of vortex generators for
wind turbine applications. European Wind Energy Conference (EWEC), Warsaw Poland. 2010.
[4] Velte, C. M., Hansen, M. O. Investigation of flow behind vortex generators by stereo particle image velocimetry on a
thick airfoil near stall. Wind Energy, 2013, 16(5), 775-785.
[5] Betterton, J. G., Hackett, K. C., Ashill, P. R., Wilson, M. J., Woodcock, I. J., Tilman, C. P., Langan, K. J. Laser doppler
anemometry investigation on sub boundary layer vortex generators for flow control. Tenth International Symposium
on Applications of Laser Techniques to Fluid Mechanics. 2000.
[6] Yang, K., Zhang, L., Xu, J. Simulation of aerodynamic performance affected by vortex generators on blunt trailing-
edge airfoils. Science in China Series E: Technological Sciences, 2010, 53(1), 1-7.
[7] Wang, H., Zhang, B., Qiu, Q., Xu, X. Flow control on the NREL S809 wind turbine airfoil using vortex generators.
Energy, 2017, 118, 1210-1221.
[8] Genç, M. S., Koca K., Açıkel H.H., Investigation of Pre-Stall Flow Control on Wind Turbine Blade Airfoil Using
Roughness Element, Energy, vol. 176, pp. 320-334, 2019.
[9] Genç, M. S., Kaynak, Ü., Lock, G. D. Flow over an aerofoil without and with a leading-edge slat at a transitional
Reynolds number. Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering,
2009, 223(3), 217-231.
[10] Genç, M. S., Kaynak, Ü., Yapıcı, H. Performance of transition model for predicting low Re aerofoil flows without/with
single and simultaneous blowing and suction. European Journal of Mechanics-B/Fluids, 2011, 30(2), 218-235.
[11] You, D., Moin, P. Active control of flow separation over an airfoil using synthetic jets. Journal of Fluids and Structures,
2008, 24(8), 1349-1357.
[12] Huang, L., Huang, P. G., LeBeau, R. P., Hauser, T. Numerical study of blowing and suction control mechanism on
NACA0012 airfoil. J. Aircraft, 2004, 41(5), 1005-1013.
[13] Hassan, A. A., JanakiRam, R. D. Effects of Zero‐Mass “Synthetic” Jets on the Aerodynamics of the NACA‐0012
Airfoil. Journal of the American helicopter Society, 1998, 43(4), 303-311.
[14] Açıkel, H. H., Genç, M.S. Flow control with perpendicular acoustic forcing on NACA 2415 aerofoil at low Reynolds
numbers. Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering, 2016,
230(13), 2447-2462.
[15] Genç, M. S., Açıkel, H. H., Akpolat, M. T., Özkan, G., Karasu, İ. Acoustic Control of Flow over NACA 2415 Airfoil
at Low Reynolds Numbers. Journal of Aerospace Engineering, 2016, 29(6), 04016045.
[16] Zaman, K. B., McKinzie, D. J. Control of laminar separation over airfoils by acoustic excitation. AIAA journal, 1991,
29(7), 1075-1083.
[17] Collins, F. G., Zelenevitz, J. Influence of sound upon separated flow over wings. AIAA journal, 1975, 13(3), 408-410.
[18] Post, M. L., Corke, T. C. Separation control on high angle of attack airfoil using plasma actuators. AIAA journal,
2004, 42(11), 2177-2184.

[19] Huang, J., Corke, T. C., Thomas, F. O. Plasma actuators for separation control of low-pressure turbine blades. AIAA
journal, 2006, 44(1), 51-57.
[20] Genç, M. S., Açıkel H.H., Koca K., Effect of partial flexibility over both upper and lower surfaces to flow over wind
turbine airfoil, Energy Conversion and Management, 219, 113042, 2020.
[21] Açıkel, H. H., Genç, M.S., Control of Laminar Separation Bubble over Wind Turbine Airfoil Using Partial Flexibility
on Suction Surface, Energy, vol.165, pp.176-190, 2018.
[22] Genç M. S., Özden Sekhoune K, Karasu İ, Özden M,, “An Airfoil Located Flat Plate On Pressure Side”, TURKISH
Patent, 2017/23311, 2017.
[23] Özden K.S., Karasu İ., Genç M.S., Experimental investigation of the ground effect on a wing without/with trailing
edge flap, Fluid Dyn. Res., 52, 1-20, (2020).
[24] Kyoungsoo Lee, Ziaul Huque, Raghava Kommalapati, and Sang-Eul Han, The Evaluation of Aerodynamic Interaction
of Wind Blade Using Fluid Structure Interaction Method, July 2015, Vol.3 No.4, Journal of Clean Energy
Technologies, DOI: 10.7763/JOCET.2015.V3.207.
[25] Vincenzo F., Linari M., Dr F. Mohdzawawi, and Dr J. Morlier, NONLINEAR AEROELASTIC STEADY
SIMULATION APPLIED TO HIGHLY FLEXIBLE BLADES FOR MAV, International Forum on Aeroelasticity and
Structural Dynamics, IFASD 2017-56, 25-28 June 2017, Como – Italy

775
 IMAGE-BASED WATER LEVEL ESTIMATION FOR
REDUNDANCY INFORMATION USING CONVOLUTIONAL
NEURAL NETWORK

Fleury, Gabriela Rocha de Oliveira

Scientific Computer Lab - Pontifical Catholic University of Goiás, Goiânia, Brazil, [email protected],
ORCID: 0000-0002-7274-1364

Nascimento, Douglas Vieira do

Scientific Computer Lab - Pontifical Catholic University of Goiás, Goiânia, Brazil, [email protected],
ORCID: 0000-0002-9817-8864

Galvão Filho, Arlindo Rodrigues

Scientific Computer Lab - Pontifical Catholic University of Goiás, Goiânia, Brazil, [email protected]
ORCID:0000-0003-2151-8039

Carvalho, Rafael Viana

Scientific Computer Lab - Pontifical Catholic University of Goiás, Goiânia, Brazil, [email protected]
ORCID: 0000-0001-8656-5472

Ribeiro, Filipe de Souza Lima

Operation department - Jirau Hidroeletric Power Plant, Goiânia, Brazil, [email protected]
ORCID: 0000-0001-5416-5036

Coelho, Clarimar José

Scientific Computer Lab - Pontifical Catholic University of Goiás, Goiânia, Brazil, [email protected]
ORCID: 0000-0002-5163-2986

Fleury, GRO, Nascimento, DV, Galvão Filho, AR, Carvalho, RV, Ribeiro, FSL, Coelho, CL.
Cite this paper as: Image-based water level estimation for redundancy information using convolutional neural
network. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Monitoring and management of water level has become an essential task in hydroelectric powers.
Activities as water resources planning, supply basin management and flood forecasting are
mediated and defined through its monitoring. The measuring station is located on the Madeira
River in Rondônia, Brazil. The measurements are made by sensors installed on the river facilities
that carry out the estimations precisely. Weather conditions influence the results obtained by these
sensors; therefore, it is necessary to have redundant approaches besides the sensors which maintain
the high accuracy of the measured values. Conventional cameras supply this necessity measuring
it through human eyes. However, its method is not reliable and has low accuracy. It is proposed an
approach, in redundancy to sensors, to use image processing to measure the water level, with high
accuracy, and to predict the water level measured using a convolutional neural network for
regression. Results present low errors according to its prediction.
Keywords: Hydroelectric Power Plant, Energy, Water level, Redundancy information, CNN
© 2020 Published by ECRES

776
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION

Monitoring water level has become an essential task for the regulatory control of rivers in order to manage
disaster risk assessment, flood warnings, water resources planning, public and industrial supply. In hydropower
energy production, it is essential to monitoring the rainfall, inflows and water level in order to maximize the
energy revenue, while taking into account dam safety risks [1]. The continuous monitoring of inflows and water
level is an essential tool for hydropower dam operators by providing real-time data for decision making in power
generation and planning. This information, which is highly valuable, requires a maximum accuracy and must be
efficiently available to the hydropower control systems and operators.
The inflow and water level monitoring techniques have become more advanced over the years in response to
increased demands for more accurate, timely and reliable information [2-4]. Different methods are used in
redundancy in order to safely guarantee the availability of this information. Staff gauge, installed at hydrometric
stations over the river, is the most used and the simpler tool for water level detection and observation. Therefore,
a water level monitoring system can be composed of sensors and video cameras for the staff gauge, which
measures the level of the water at the control room. However, in days of storms or climate changes, the sensors
used may not be accurate presenting flawless in the acquired data [5]. Moreover, video cameras pointed at the
staff gauge, used as a secondary method of measurement, are monitored by human eyes. The problem with this
method is that human eyes are not reliable and subject to errors, which compromise the security of the system.
Therefore, defining an accurate and reliable method to monitor the water level is a challenge for hydropower
system control.
Predictive models based on a dataset of previous measurements provide an ideal solution to monitoring the water
level and streamflow in hydrology. Methods using Artificial Neural networks [6,7], Support Vector Machines
[8,9] and hybrid models [10,11] have been discussed to predict water level and streamflow. In common, they
use measurements from sensors of meteorological and hydrometric stations. However, such methods applied to
safe and accurate monitor water level at hydropower control remain an open field of research. To that end, this
study proposes to develop an automatic detection approach that can be used in redundancy of sensor techniques,
in order to assure a security system of monitoring water level. It proposes image analysis techniques and neural
networks in order to automatically measure and predict the water level considering images from conventional
cameras of the staff gauge.
The dataset used as a case study was provided by the Jirau Hydroelectric Plant, installed on the Madeira River,
in the state of Rondônia, Brazil. It consists of real time videos of the staff gauges at different time lapsing (day
and night) and weather conditions (sunny, cloudy, and raining). Digital image processing is applied to remove
noises for better visualization, and to extract the region of interest of the staff gauge. Therefore, it is possible to
determine the water level measurement, creating a relationship between the water line surface and the beginning
of the staff gauge. Convolutional neural networks (CNN) with a regression layer is applied to train and predict
the water level. The Jirau plant is managed by the Consortium Energia Sustentável do Brasil (ESBR).

2. CASE STUDY
Dataset
The dataset consists of thirty-five videos in real time of the staff gauges between 05/31/2020 and 06/04/2020
during different moments, days, nights, and weather conditions. These videos are separated into frames, totaling
3.364 images, and the water level is measured for each image obtained. Initially, is removed from all images the
symbol of date and hours preparing it for the extraction of the region of interest. Detected the water level, these
data are prepared to be trained by a CNN Regression, thus, each image is classified by its level and divided in
two sets: 2.706 images for training set and 658 for testing set.

Image Processing

Image processing consists of enhancing the image in order to extract ROI. In most of the images, the camera
angle is not suitable, and the staff gauge looks not straight which makes it complicated to detect the water level
in the final analysis. Moreover, the image quality also influences the extraction of the ROI, as depicted in Figure
1(a). To improve image quality, a vertical shearing filter is applied to make the image straight [12]. Next, the
non-uniform illumination is corrected by applying imaging segmentation. The morphological opening filter is
applied to noise reduction and gamma filter to enhance the contrast. The resulting image is depicted in Figure

777
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1(b). To extract the ROI, the enhanced image is binarized and the borders of the ROI are detected and cropped
from the image. The result is the region of interest of the image and it is depicted in Figure 1(c).

(a) (b) (c)

Figure 1. (a) Original Image, (b) Image Enhancement, (c) Extraction of the region of interest
.
Water level detection

The staff gauge contains count marks which correspond to the level of the water reached. To measure the water
level, it is necessary to define a window size of these count marks, in pixel coordinates, and obtain the number
of counts present in the staff gauge. The water level is obtained through the image obtained by the extraction of
the region of interest. The staff gauge is made up of counters that mark the value at which the measurements
are. Thus, it is necessary to define the window size of these counters in pixel coordinates and obtain the number
of counters present in the staff gauge. In addition, the size difference between the beginning of the ruler and the
river's surface line is also defined. Considering a fixed mark, in meters, the number of counts and the defined
surface line, it is possible to obtain a relationship between them according the Eq. [1]:

𝑑
𝑙 = 𝑟 − (𝑐 + 𝑠 ) ∗ 0.1 (1)

where l is the water level of the river, r is the fixed mark, c is the number of counts on the staff gauge, d is the
difference between the water line surface and the beginning of the staff gauge and s is the size of each counter
in pixel coordinates. To ensure the results are in centimeters, l is multiplied by 0.1

Convolutional Neural Networks model proposed

The proposed CNN model is composed of 19 layers in which the first layer is responsible to normalize the image
dimension in order to guarantee the output of 224 rows and columns and the three channels (RGB). The next
five layers contain a set of filters in order to perform convolutional operations with its subsequent layer. For the
set of filters, each of them produces an activation map which is stacked along the depth dimension, producing
the output volume. As a parameter it is used for all layers, a kernel with size 3x3. To avoid fast reduction of
matrix dimension it is set padding 0 after each convolutional layer. For each layer non-linearity layers involve
the use of a nonlinear activation function which takes a single number and performs a fixed mathematical
operation on it, and it is a complement of the convolutional. Rectified linear units, ReLu, is the most popular
and to be non-saturating is the fastest to apply the convergence of gradient descent. It computes the activation
function f(x), where x is the output from the convolutional layer and the input of reLU Eq. (5) which means that
its activation is threshold at zero [11].
𝑓(𝑥) = 𝑚𝑎𝑥(0, 𝑥) (2)
The pooling layer is responsible to reduce the spatial size, the number of parameters and the computation in the
network. It was used an average pooling type to reduce the activation maps. However, it was necessary to apply
a dropout layer equal to 20% in order to decrease furthermore the data. The fully connected layer performs
classification of the features and extracts all the convolutional layers [20]. To perform the predictions, a
regression layer is added at the end of the CNN architecture. The CNN architecture proposed is summarized in
Figure 2. It represents the six convolutional layers with its respective filters between brackets, the reLU and
average pooling layer (after each convolution layer), one dropout layer, the fully connected layer and the
regression layer to make the predictions.

778
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 2. Convolutional Neural Network architecture

The proposed predicted model is evaluated by three different criteria: the root mean square error (RMSE),
depicted by Eq. [8], the mean absolute error (MAE) in Eq. [9] and the determination coefficient (R²) in Eq. (10),
1
𝑅𝑀𝑆𝐸 = √ ∑𝑁
𝑖=1(ŷ𝑖 − 𝑦𝑖 )
2
(2)
𝑁

1
𝑀𝐴𝐸 = ∑𝑁
𝑖=1 |ŷ𝑖 − 𝑦𝑖 | (3)
𝑁

∑𝑁
𝑖=1(𝑦𝑖 − ŷ)²
𝑅² = 1 − ∑𝑁
(4)
𝑖=1(𝑦𝑖 −Ӯ )²

where Nis the amount of data, ŷi predicted value, yithe measured value and yi the average of the measured value.
These errors are used to measure the accuracy of the prediction model for training and testing. The RMSE and
MAE measured the magnitude of the errors. A lower value for RMSE is better than a higher one. The MAE
calculates the average over the data set between the prediction and the water level measurements. R² represents
the relationship between the variance of prediction and the total variance of data
3. RESULTS AND DISCUSSIONS

The proposed model is evaluated by training and testing images for the CNN Regressions predictions. Table 1
shows the results of RMSE, MAE and R². As can be seen, the accuracy for training images predictions was 93%,
while for testing images was 91%.
Table I. Evaluation of the prediction
Dataset RMSE MAE R²
Training 0.2668 0.1991 0.9004
Test 0.2829 0.2228 0.8868

Figure 3 shows predictions where Figure 3(a) correspond to the training dataset and Figure 3(b) the testing
dataset. The proposed model is able to produce results close to real measurements on training and testing dataset.

Figure 3. (a) Training Prediction, (b) Testing Prediction

779
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. CONCLUSIONS

This work proposes water level detection using images in a strategy based on a CNN model. This model was
trained and tested using an image dataset extracted from videos of the measuring rule provided by Jirau HPP.
The predictive capacity of the proposed model was tested in terms of RMSE, MAE and R², resulting in low
errors for training and test sets. Therefore, it can be concluded that CNN based strategy is a promising
approach for water level detection of Madeira River, and can collaborate with security regarding data integrity,
through the redundancy of information through another acquisition paradigm. This safety is essential in the
efficiency studies of Jirau Hydroelectric Power Plant and energy production as well as hydropower control and
management. For future works, it is necessary to study other configurations of the neural network and other
deep learning strategies to improve the generalization of the model.

ACKNOWLEDGMENT

Authors thank Energia Sustentável do Brasil for their support in conducting this study “Projeto regulamentado
pela ANEEL e desenvolvido no âmbito do Programa de P&D da Energia Sustentável do Brasil S.A. (PD-06631-
0007/2018)”

REFERENCES

[1] Katole S., Bhute Y. A. Real Time Water Quality Monitoring System based on IoT Plataform. A Review International
Journal on Recent and Innovation Trends in Computing and Communication 2017, 5:302-305
[2] Stech J. L., Lima, I. B. T., Novo, E. M. L. M., Silva, C. M., Assireu, A. T., Lorenzzetti, J. A., Carvalho J. C., Barbosa
C.C., Rosa R. R. Telemetric monitoring system for meteorological and limnological data acquisition. Verhandlungen
Internationalen Verein Limnologie, 2006; 29:1747-1750, DOI: 10.1080/03680770.2006.11902987.
[3] Han, K., Zhang D., Bo J., Zhan Z. Hydrological Monitoring System Design and Implementation Based on IOT.
Physics Procedia 2012; 33:449-454. DOI: 10.1016/j.phpro.2012.05.088
[4] Tsai T., Yen P. Improvement in stage measuring technique of the ultrasonic sensor gauge. Measurement 2012; 45:1735-
1741, DOI: 10.1016/j.measurement.2012.04.012.
[5] Zhang, Z., Zhou, Y., Liu, H., Gao, H. In situ water level measurement using NIR-Imaging video camera. Flow
measurement and instrumentation. 2019, 67, pp. 96-106. Doi: https://doi.org/10.1016/j.flowmeasinst.2019.04.004
[6] Daniell, T.: Neural networks. Applications in hydrology and water resources engi- neering. In: National Conference Publication -
Institute of Engineers. Australia, 1991.
[7] Kenabatho, P., Parida, B., Moalafhi, D., Segosebe, T.: Analysis of rainfall and large-scale predictors using a stochastic
model and artificial neural network for hydrological applications in Southern Africa. Hydrol. 2015; 11:1943–
1955. DOI: 10.1080/02626667.2015.1040021
[8] Deka P. C., Raghavendra S. Support vector machine applications in the field of hydrology: a review. Appl. Soft Comput 2014;
19:372–386.DOI: doi.org/10.1016/j.asoc.2014.02.002
[9] Khan, M.S., Coulibaly, P.: Application of support vector machine in lake water level prediction. J. Hydrol. Eng 2006; 11:199–
205.DOI: 10.1061/(ASCE)1084-0699(2006)11:3(199)
[10] Kim, T.W., Vald ́es, J.B.: Nonlinear model for drought forecasting based on a con- junction of wavelet transforms and neural
networks. J. Hydrol. Eng. 8(6), 319–328, 2003.
[11] Brownlee, J. Predict the future with MLPs, CNN’s and LSTM’s in Python. In: Machine Learning Mastery. Deep
learning for time series forecasting. 2019, pp. 5-6
[12] Schneider, A., Hommel, G., Blettner, M., Linear Regression Analyzes. In: Deutsches Ärzteblatt International. 2010;
107:776-782

780
 A PRELIMINARY IN-SITU MEASUREMENT AND SIMULATION
OF THE PERFORMANCE OF GROUND-COUPLED HEAT
EXCHANGER SYSTEM IN THE TROPICAL CONTEXT OF
MAURITIUS

Gooroochurn Mahendra
University of Mauritius, Reduit, Mauritius, [email protected], ORCID: 0000-0003-0734-9144

Seegobin Budhiraj
University of Mauritius, Reduit, Mauritius, [email protected], ORCID: 0000-0003-0171-4702

Jankee Lowkhaushalsing
University of Mauritius, Reduit, Mauritius, [email protected], ORCID: 0000-0001-6354-8418

Gooroochurn, M., Seegobin, B. and Jankee, L. A preliminary in-situ measurement and

Cite this paper as: simulation of the performance of ground-coupled heat exchanger system in the tropical
context of Mauritius. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: The horizontal ground-coupled heat exchanger system with air medium has been found
to be a useful green building design measure in various countries. The system helps for
cooling and heating depending on the prevailing climatic conditions and the temperature
of the ground at the level of the buried pipe network. This paper presents a study
conducted for a small scale prototype made of PVC PN10 pipes with the aim to study its
behavior for both the heating and cooling effect in the tropical context of Mauritius. An
experimental setup was deployed to collect performance data, which were used to validate
a CFD simulation model. The CFD simulation model was subsequently extended to
predict achievable performance over longer pipe lengths. The results show that the
horizontal ground-coupled technology with air medium buried at shallow depths of 1 to
1.5m can be effectively applied for the type of soil and climatic conditions prevailing in
Mauritius.
Keywords: Ground-coupled heat exchanger, Energy Efficiency, Sustainability in Built Environment
© 2020 Published by ECRES

1. INTRODUCTION

The increasing demand for energy and the depleting fossil fuels have fueled explorations in new frontiers of
renewable energy technology. Geothermal heating and cooling are new advancement in HVAC industry, where
the earth’s thermal capacity and temperature constancy are utilized, with reported electricity saving up to 51%
in HVAC energy use and reduced CO2 emissions [1]. It has been observed that the temperature of ground at a
depth of about 2 m remains almost constant throughout the year and is approximately equal to the annual average
temperature of a particular locality. In summer season, this constant temperature of the earth’s undisturbed
temperature remains lower than the ambient temperature. Thus, to meet the cooling requirement of a building a
ground-coupled heat exchanger (GCHE) system can be effectively used if its outlet temperature is low enough
[2].

781
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2. LITERATURE REVIEW

According to Guohui Gan [3], various factors such as the type, texture, moisture and temperature of a soil and
the depth of the buried pipes affect the performance of a ground heat exchanger. For example, at a depth of 1.5
m, it has been noted that the mean heat transfer is 17% higher in a sandy soil and it is 14.5% lower in clay loam
soil as compared to loamy sand soil. Furthermore, at a moisture content of 0.2m3/m3, the average heat transfer
is found to be 8%, 14% and 22% in sandy soils, loamy sand and clay loam respectively.

According to an experiment carried out by Yupeng Wu [4] for a duration of two months, it has been found that
the coefficient of performance, COP of a ground source heat exchanger decreases as the running time increases.
The COP on average was found to be 2.5. To increase the efficiency of the system, it was recommended to use
a longer heat exchanger and a large space for bigger installations. Another factor found affecting the performance
of GCHE is the size of the heat exchanger; the larger the diameter of the coil, the higher the rate of heat
extraction.

A study conducted in Chennai (India) by G.N Tiwari [5] for the design of an earth to air heat exchanger, found
that the difference in temperature of air recorded between the inlet and outlet was found to be more for winter
than summer where the limiting outlet temperature for winter was 28.6⁰C for an inlet temperature of 23⁰C.
During summer, a temperature of 29.3⁰C was obtained for an average ambient temperature of 34⁰C. The length
of the pipe was found to influence the performance on the system with an optimum length of 21 m concluded.
Another study conducted in India by N. Bordoloi [6] with the aim to measure the variation of Coefficient of
Performance (COP) for different ducting materials. The experimental values of COP obtained for PVC was
found to be between 1.8 and 2.9 for winter and summer respectively.

Till date, many studies on calculation models have been carried out for the GHCE and these studies conducted
were based on calculation models. The models are divided into three categories, which are: One-dimensional
model (1-D), Two-dimensional model (2-D), and Three-dimensional model (3-D). In Comparison to 1-D and 2-
D models, 3-D models produce more accurate results despite being more time-consuming. In 3-D models, the
performance analysis of GCHE systems using heat and energy balance equations requires the use of
Computational Fluid Dynamic (CFD). CFD is a powerful method to study mass and heat transfer, and the
corresponding partial differential equations can be solved numerically in a discretized form [7].

To investigate the influence on the pipe material, an experimental set-up of a GCHE system was constructed by
Bansal et al. [8] in Ajmer, India, which consisted of two horizontal circular pipes of 23.42 m buried lengths and
0.15 m inner diameters each. One of the pipes was made up of polyvinyl chloride (PVC) while the other one
was made up of mild steel. The tubes were buried at a depth of 2.7m in flat land and dry soil. A thermal model
of the GCHE was created using FLUENT 6.3 for CFD simulations [8]. It was observed that there was a very
small difference in the air outlet temperature between steel and PVC pipe when all other input conditions were
the same. PVC has a smaller thermal conductivity than steel. Thus, this indicates that the performance of the
GCHE system is not heavily dependent on the pipe material. Other factors like cost and durability become the
deciding factor for the pipe [8].

Similarly, a GCHE system was tested for dry and summer conditions, by researchers in Bhopal (India). The
experimental set up consisted of two PVC pipes of 0.1016 m internal diameter. Both pipes were connected in
series and the total length was 19.228 m. The pipes were buried at a depth of 2m in black cotton soil. A quasi-
state 3D model, of the GCHE was developed in CFO platform CFX 12.0. For airflow velocities of 5m/s, 3.5m/s,
2m/s, researchers deduced that a drop in air temperature was significant in the initial length of the pipe, rather
than in the remaining length of the pipe. The temperature was recorded at 8 locations throughout the pipes for
the different velocities. It was found that the minimum and maximum air temperature drops were 11.3˚C and
12.9˚C, observed at air flow velocities of 5m/s and 2m/s respectively. Researchers also concluded that the

782
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

simulated results are comparable with the experimental results [9]. The COP is a metric commonly used to
analyze the thermal performance of a GCHE. A study conducted by Monkong [10] in Thailand with a GCHE of
38.5 m length iron pipe, having a diameter of 80mm, buried at 1m depth. The COP was found to be 3.56, 2.04
and 0.77 during summer, winter and monsoon days respectively. It was also concluded that cooling CoP was
higher than heating COP [10].

Researchers at the University of Sannio, Roma [11] analyzed the performance of the GCHE for Italian climates.
This experiment yielded the following conclusions:

1) The influence of tube material is negligible on the influence of the system performance.

2) Tube lengths of values below 10m are unsatisfactory, while tube length of about 50m is a good compromise.
Pipe length greater than 50m results in a negligible reduction in cooling energy.

3) A depth of 3m is considered to be a good compromise between performance and excavating costs.

The performance of the geothermal heat pumps is partly affected by the soil conditions where the pump is
being installed. Before installing a GHP, a geothermal contractor will usually assess the land. A rocky land
reduces the possibility of installing a geothermal system. The rocks make the drilling operation difficult for the
installations of looping pipes. Gooroochurn et al. [12] showed that the length of the pipe has a more sensitive
effect on the performance of the system compared to the other parameters considered. The separation of the
pipes is also an important parameter found to affect system performance. It was noted that a separation of 2 m
between the 2 pipes is satisfactory to prevent any thermal interaction between the buried pipes. Moreover, the
cooling effect and COP of the system were directly affected by the flowrate of the air and the optimum flow rate
and COP for the system were found to be 60 L/s and 3.89 respectively.

3. METHODOLOGY

A prototype was constructed to allow air to flow in, exchange heat with the surrounding soil via the pipe wall
and produce a cooling or heating effect of the traversing air. The experimental data obtained through flow and
temperature, have been used to analyze the behavior of air when cooled or heated when passed through the
GCHE system. As a preliminary study to validate the CFD simulation model, the experimental setup had only
one pipe (6.6m long), positioned at 1 – 1.5 m below the ground level (see Figure 1).

The prototype comprising of a fan coupled with a piping system of diameter 110 mm (PN 10), buried at 1 – 1.5
m in the earth, has been used to allow air to flow in and out. The pipe rated as PN 10 chosen has the ability to
support the pressure and load that is exerted by the soil. It is perceived as being most suitable for underground
works. As advised by numerous journal papers, incorporated in the design is a water sink which will accumulate
water in the possible case of condensation in the pipe. The vertical pipes being different in length (1 m and 1.5
m) provides a slope of around 5⁰ to the bottom pipe. This allows accumulated water to flow to the water sink
where it will be collected. The entry and exit level of air is positioned at 1 m above the ground to ensure that the
temperature of the soil does not affect the air entering/exiting the system.

The extraction fan was rated at 145 Watt with nominal diameter of 200 mm can provide a constant flow of air
(open air) of 10 m/s, suitable for indoor and outdoor purposes. The fan can provide velocity exceeding 0.62 m/s,
therefore calculations have shown that turbulent flow regimes can be created. The combined velocity and
temperature sensor has 2 probes in built (See Figure 2) with one probe used to record the temperature (lower)
and the other one (upper) to record the velocity of the air. The range of value for the temperature is 0 – 100 ⁰C
and that for velocity is 0 – 10 m/s. A trench of 1 m wide, 8 m long and 1.5 to 2 m deep was dug by an excavator
for the experimental setup as shown Figure 4.

783
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The air velocity/temperature sensors were placed at the entry and exit level of the pipe with a length of 0.55 m
from the end as shown in Figure 5 in order to allow the air to properly develop inside the pipe. Two temperature
sensors (TMC 20-HD) were placed in the pipe to record how the air conditions, and two additional temperature
probes were placed in the soil, in order to record how the temperature of the soil is varying, as illustrated in
Figure 5.

Figure 1: Layout of experimental set-up

Figure 2: Extraction fan and velocity and temperature sensor

Figure 3 Air temperature sensor and data logger

To develop a simulation model that would allow to study the impact of the factors on system performance for
cases not replicable in practice, the experimental data recorded on different time and days were used to refine

784
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

and validate the CFD model, which was created in the SolidWorks software package. To enable the study of the
soil interaction with the pipe as air flows within the latter, the dimension of the soil environment was set at 3.5
x 3.5 x 12 m, although in practice the size and depth of the soil will be considered as unlimited with varying
temperature at different level with respect to the depth. The soil material employed in the flow simulation
database was gravel soil with parameters:
Density = 2050 kg/m3
Specific Heat = 18 J/(kgK)
Thermal conductivity =0.52 W/(mK)

Figure 4: Trenching works

Figure 5: Placement of the sensors

The temperature of the soil and incoming air recorded experimentally have been applied as boundary conditions
in the CFD model to generate the temperature profile within the pipe at various points, including at the
extremities. Depending on the temperature of the soil and the ambient air (T_in), the process of cooling or
heating have been successfully implemented in the CFD model.

785
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. EXPERIMENTAL RESULTS

This section provides a graphical analysis of the data obtained for the temperature measurements at various
points in the installation. The intervals at which temperatures were found to be maximum and minimum during
the day for the cooling effect as well as those with minimum temperatures that were recorded during the heating
effect of the pipe, will be of particular interest. Out of the huge volume of data recorded, the following graph
shows a general trend observed.

Figure 6: Temperature and flow measurements

The graph shows the results obtained in the afternoon from 1 p.m. to 6:20 p.m. The different variations in
temperature during that day can be observed by the series of peaks and valleys. The peaks show the maximum
temperature recorded for the temperature of air entering the system (green solid line). The maximum range of
temperature recorded during that day (peak values) was between 29 ⁰C and 32.24 ⁰C. The minimum range of
values observed for the valleys was between 28.38 ⁰C and 29 ⁰C. The range of air velocity (black solid line)
varied from 6 m/s to 9 m/s and this fluctuation in the reading was due to the turbulent flow that occurred at the
entrance. Temperature recorded by the temperature probe found in the bottom inclined pipe (solid blue line) was
seen to vary with respect to the air temperature entering the system. In fact, it followed the same trend as the
latter but at a lower temperature with a difference of 0.1 ⁰C – 1.5⁰C recorded. This indicated that cooling of air
was already taking place at that point. The temperature of the soil (blue dotted line) was 25.04⁰C and it gradually
increased to 25.67 ⁰C by the end of the experiment on that day.

5. SIMULATION RESULTS

The simulation results obtained for the CFD model corresponding to the same physical set up implemented are
compared to the experimental results in Table 1. The close agreement of the experimental and simulated values
shows that the CFD model can be taken as a satisfactory thermal model for the horizontal ground-coupled heat
exchanger system with air as the flowing medium. The next step was to extend the CFD model to predict the
expected cooling and heating effect for longer pipe lengths, which have been found to be above 50 m in most
applications. Figures 7 to 9 shows the temperature profile observed across the pipe for lengths of 20m, 40m and
60m respectively. The results show a clear gain in cooling and heating effect as the pipe length is increased,
with confirmation of 40 – 50 m as a good choice for the pipe length to obtain a temperature drop into the human
comfort zone of 27 C when external temperature close to 34 C are experienced, which are typical in peak
summer conditions.

6. DISCUSSION & CONCLUSION

For the experimental data generated, a COP of 1.94 and 0.81 was obtained for the cooling and heating effect
respectively, which can be improved further by using longer length pipes as shown by the CFD simulations for
longer pipe lengths of 20 m, 40 m and 60 m. The research findings generated from this preliminary project has

786
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

shown clear promise to allow this technology as part of green building design to reduce dependency on active
cooling systems, which have been found to yield significant increase in electrical peak power demand in summer
times. The problem of condensation of air moisture as a phenomenon of cooling within the heat exchanger is a
challenge which needs to be looked into, given that humidity levels can rise to above 90% in Mauritius and
similar tropical climates. A separate arrangement was made to collect condensate in the experimental setup and
it was indeed observed that condensation takes place within the pipe.
Table 1: Simulated and Experimental Data

I.

Figure 7: Maximum cooling and heating effect with 20 m pipe length

Figure 8: Maximum cooling and heating at 40 m pipe length

787
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 9: Maximum cooling and heating at 60 m pipe length

In terms of temperature drop, the results show good promise to use the technology for the type of soil
typically found in Mauritius at shallow levels of 1 to 1.5 m, for the peak summer conditions where the
achievable drop can be used to improve thermal comfort in unconditioned buildings and reduce air-
conditioning use otherwise. The research project will be pursued by implementing a real-life system over
50 m in pipe length to validate the CFD simulation findings and definitely showcase the applicability and
benefits of this technology as a means to decouple the effects of climate change on thermal comfort and
energy use in the built environment.

REFERENCES

[1] Dhepe, N., Raahul, K., A Review of the Advancements in Geothermal Heating and Cooling System, Journal of Alternate
Energy Sources and Technologies ISSN: 2230-7982 (Online), ISSN: 2321-5186 (Print) Volume 8, Issue 1
[2] Anon (2018) Hybrid ground coupled heat exchanger systems for space heating/cooling applications: A review.
[3] Gan, G., 2018. Dynamic thermal performance of horizontal ground source heat pumps – The impact of coupled heat
and moisture transfer. Energy, 152, pp.877-887.
[4] Wu, Y., Gan, G., Verhoef, A., Vidale, P. and Gonzalez, R., 2010. Experimental measurement and numerical simulation
of horizontal-coupled slinky ground source heat exchangers. Applied Thermal Engineering, 30(16), pp.2574-2583.
[5] Tiwari, G., 2014. Design of an Earth Air Heat Exchanger (EAHE) for Climatic Condition of Chennai, India. Open
Environmental Sciences.
[6] Bordoloi, N., 2017. A COMPUTATIONAL STUDY ON THE PERFORMANCE OF EARTH AIR HEAT
EXCHANGER (EAHE) USING DIFFERENT DUCT GEOMETRIES AN. DEPARTMENT OF MECHANICAL
ENGINEERING LOVELY PROFESSTIONAL UNIVERSITY, PUNJAB.
[7] Bisoniya, T., Kumar, A. and Baredar, P. (2018). Study on Calculation Models of Earth-Air Heat Exchanger Systems.
[8] Bansal, V., Misra, R., Agrawal, G. and Mathur, J. (2010). Performance analysis of earth–pipe–air heat exchanger for
summer cooling. Energy and Buildings, 42(5), pp.645-648.
[9] Heating potential evaluation of earth–air heat exchanger system for winter season. (2014). Journal of Building Physics,
39(3), pp.242-260.
[10] Mongkon, S., Thepa, S., Namprakai, P. and Pratinthong, N. (2018). Cooling performance and condensation evaluation
of horizontal earth tube system for the tropical greenhouse.
[11] Ascione, F., Bellia, L. and Minichiello, F. (2011). Earth-to-air heat exchangers for Italian climates. Renewable Energy,
36(8), pp.2177-2188.
[12] M. Gooroochurn, M. Z. Hossenbaccus, "Horizontal Ground-Coupled Heat Exchanger as a Passive Design Technique
for Tropical Climates", International Journal of Advances in Mechanical and Civil Engineering (IJAMCE), Volume-
5,Issue-4, Aug 2018, pp56-62.

788
 THERMODYNAMIC MODELLING TO OPTIMIZE THE PERFORMANCE
OF SMALL SCALE ORGANIC RANKINE CYCLE (ORC) SYSTEMS

Ibrahim Albaik
University of Birmingham, Birmingham, United Kingdom, [email protected], ORCID: 0000-0003-0280-
7214

Gavin Tozer
University of Birmingham, Birmingham, United Kingdom, [email protected], ORCID: 0000-0003-0248-
1015

Raya Al-Dadah
University of Birmingham, Birmingham, United Kingdom, [email protected], ORCID: 0000-0001-9410-1993

Saad Mahmoud
University of Birmingham, Birmingham, United Kingdom, [email protected], ORCID: 0000-0002-4421-
1577

Albaik. Ibrahim, Tozer. Gavin, Al-Dadah. Raya, and Mahmoud. Saad. Thermodynamic
Cite this paper as: Modelling to Optimize the Performance of Small Scale Organic Rankine Cycle (ORC)
Systems. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey.

Abstract: This paper concerns the development of a thermodynamic model to optimize the performance of
small-scale organic Rankine cycle (ORC) systems in terms of selecting the best combination of
refrigerants and evaporator/condenser heat exchangers to achieve highest power output and cycle
efficiency. The work involves investigating the use of four refrigerants namely R22, R410A,
R404A and R407C and two types of heat exchangers namely shell and tube and brazed plate types
with heat source at temperature less than 100C. Results showed that R22 produced the highest
thermal efficiency of 8.1% while R410A showed the lowest efficiency of 7.4%. As for the power
output, R410A showed the highest power output of 20kW followed by R22 and R407C with output
of 19kW while R404A produced the lowest power output of around 16kW. Also, results showed
that plate heat exchangers have higher heat transfer coefficients leading to smaller heat transfer
surface area compared to shell and tube heat exchangers. Therefore, they offer a more compact
solution to reduce the overall system size and reduce the amount of working fluids used.

Keywords: Small Scale Organic Rankine Cycle, Electronic Platform, Shell and Tube Heat Exchangers, Plate
Heat Exchangers, Thermal Efficiency, Power Output
© 2020 Published by ECRES

INTRODUCTION
The global increasing demand for power generation causing serious concerns regarding emission levels caused
by fossil fuels used to generate this power [1]. Therefore, there is a need to develop efficient technologies to
generate electricity that use renewable energy sources and the available low grade waste heat sources [2]. With
the limitations of current energy conversion processes, around half of the world energy used is discarded as
waste heat, with major part of this heat at temperatures lower than 2500C [3]. From the industrial processes of
the EU countries around 800 TWh/year of waste heat is discharged to the environment [4]. Organic Rankine
Cycle systems with their simple structure and relatively low driving temperature offer the potential to exploit
the low grade waste heat to generate electricity without the need to use fossil fuel and produce emissions [5, 6].
Large scale ORC systems are relatively mature technology, where in the last ten years many power plants have
been installed worldwide in the MW range [2,7]. However, the downsizing of this technology to develop small
scale power plants (~ 10 kW) still facing some technical challenges which hinders it widespread use and

789
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

commercialization [8].The selection of the right working fluid and the optimization of major components like
evaporator, condenser and the expander are required to improve the overall performance of small scale ORC
systems. The working fluid has major impacts on the design and sizing of the system major components,
operational temperature and pressure, cost and cycle performance [9, 10]. Generally, the expander design has
major effects on the performance of the small scale ORC systems. Different kinds of expanders were investigated
such as scroll expander [11], single and twin screw expander [12, 13], radial turbines [14], and recently, a
Wankel expander was reported [15]. Although different expanders have shown advantages and limitations at
various operating conditions and power outputs, no expander has shown a major lead on others, and commercial
availability with completive costing still a major challenge. Also the use of low cost and efficient heat exchanger
solutions were investigated. Helical coil heat exchanger for supercritical ORC systems that improved heat
transfer coefficient and cycle efficiency was reported by [16]. Commercially available shell and tube heat
exchanger to recover exhaust heat from internal combustion engine was used by [17].
In this work, a thermodynamic model to optimize the performance of small-scale organic Rankine cycle (ORC)
systems (less than 100kW power output) in terms of selecting the best combination of refrigerants and
evaporator/condenser heat exchangers to achieve highest power output and cycle efficiency is developed. The
work involves investigating the use of four refrigerants namely R22, R410A, R404A and R407C and two types
of heat exchangers namely shell and tube and brazed plate types with heat source at temperature less than 100C.

ORC THERMODYNAMIC MODEL

A typical non-recuperative small-scale ORC system is adopted to investigate the interaction of the cycle
components with ORC system performance. Figure 1 illustrates the ORC system with the T-S diagram showing
the thermodynamic cycle and working fluid phases. In process (1-2), the working fluid is typically pressurized
from saturated liquid to compressed liquid phase, and then it is heated in the evaporator reaching the superheated
vapor phase through the process (2-3-4-5). Where from 2 to 3 the working fluid is heated through the preheating
process, then from 3 to 4 the boiling process occurs and finally the vapor is superheated (point 5). In process (5-
6), the expansion process is carried on through an expander to reduce the pressure from the evaporator pressure
to that of the condenser. In the condenser, the superheated vapor is first cooled to saturated vapor and then and
condensed (process (6-1)). In this paper, four working fluids were investigated; R22, R404A, R407C and
R410A.
Temperat
ure

Entropy
Figure 1. Schematic and T-S Diagrams

The modelling of the ORC system is established using the first law of thermodynamics applied to open systems
and ignoring kinetic and potential energy changes between inlet and outlet of each component. The main inputs
to the model are fluid flowrate 𝑚̇ and evaporator operating pressure 𝑃2 , the pressure ratio 𝑃𝑟𝑎𝑡𝑖𝑜 = 𝑃2 /𝑃1 and
the evaporator outlet temperature 𝑇5 (which depends on the heat source temperature) For the pump, the power
consumption is evaluated assuming that the fluid enters the pump as saturated liquid (𝑋1 = 0) and assuming
isentropic efficiency of 80%.
(ℎ2𝑠 − ℎ1 )
𝜂𝑝𝑢𝑚𝑝 =
(ℎ2 − ℎ1 ) (𝟏)
Where ℎ2𝑠 is the enthalpy of the fluid at the exit of the pump evaluated at 𝑃2 and entropy value equal to that at
inlet to the pump 𝑠2𝑠 = 𝑠1 . ℎ2 is then evaluated using equation 1 and the assumed isentropic efficiency.

790
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑊𝑝𝑢𝑚𝑝 = 𝑚̇ (ℎ2 − h) (𝟐)

The heat input in the evaporator can be obtained as follows:

𝑄𝑖𝑛𝑝𝑢𝑡 = 𝑚̇ (ℎ5 − ℎ2 ) (𝟑)

Where ℎ5 is determined for superheated vapor at 𝑃2 and 𝑇5 .

The power produced by the expander is determined assuming an isentropic efficiency of 70% using the
following equations:

𝑊𝑒𝑥𝑝𝑎𝑛𝑑𝑒𝑟 = 𝑚̇ (𝐻5 − 𝐻6 ) (𝟒)

(𝐻5 − 𝐻6𝑠 )
𝜂𝑒𝑥𝑝𝑎𝑛𝑑𝑒𝑟 =
(𝐻5 − 𝐻6 ) (𝟓)
Where ℎ6𝑠 is the enthalpy of the fluid at the exit of the expander evaluated at 𝑃1 and entropy value equal to that
at inlet to the expander 𝑠6𝑠 = 𝑠5 . ℎ6 is then evaluated using equation 5 and the assumed isentropic efficiency.

The heat rejected at the condenser, the net power and the thermal efficiency are evaluated using equations

𝑄𝑟𝑒𝑗𝑒𝑐𝑡𝑒𝑑 = 𝑚̇ (𝐻6 − 𝐻1 ) (𝟔)

𝑊𝑛𝑒𝑡 = 𝑊𝑒𝑥𝑝𝑎𝑛𝑑𝑒𝑟 − 𝑊𝑝𝑢𝑚𝑝 (𝟕)
𝑊𝑛𝑒𝑡
𝜇𝑡ℎ𝑒𝑟𝑚𝑎𝑙 = (𝟖)
𝑄𝑖𝑛𝑝𝑢𝑡
The system now is modified by taking into consideration the pressure losses in the evaporator and condensers:

𝑃5 = 𝑃2 − ∆ 𝑃evap,loss (𝟗)
𝑃1 = 𝑃6 − ∆ 𝑃𝑐𝑜𝑛𝑑,loss (𝟏𝟎)
Where ∆ 𝑃evap,loss is the pressure losses in the evaporator and ∆ 𝑃𝑐𝑜𝑛𝑑,loss is the pressure losses in the
condenser.

A recalculation for all the working fluid properties are carried out in order to take into consideration the effect
of the pressure losses in the cycle performance.

EVAPORATOR AND CONDENSER HEAT EXCHANGERS

The most commonly used heat exchangers in the ORC application are shell and tube and plate heat exchangers.
Plate heat exchanger has higher heat transfer coefficient and more compact design compared to the shell and
tube, while shell and tube is more desirable in terms of maintenance and assembly. Therefore, both designs are
investigated to study the effect of heat exchanger type on the overall cycle performance. The total heat transfer
area A is calculated as the following equations:
𝑄𝑒𝑣𝑎𝑝/𝑐𝑜𝑛𝑑
𝐴= (𝟏𝟏)
𝑈 𝐿𝑀𝑇𝐷
Where U is the overall heat transfer coefficient and LMTD is the logarithmic temperature difference that can be
calculated as follows:
(𝑇ℎ,𝑖𝑛 −𝑇𝑐,𝑜𝑢𝑡 )−(𝑇ℎ,𝑜𝑢𝑡 −𝑇𝑐,𝑖𝑛)
𝐿𝑀𝑇𝐷 = (𝑇ℎ,𝑖𝑛 −𝑇𝑐,𝑜𝑢𝑡 )
ln( )
(𝑇ℎ,𝑜𝑢𝑡 −𝑇𝑐,𝑖𝑛 ) (𝟏𝟐)
The total surface area is restricted with the available commercial heat exchangers, where the shell and tube
selection is based on EJ Bowman brand while the plate heat exchanger is selected from HRS brand. Thus, the
pressure losses are also extracted based on the tube/plate dimensions and arrangements. The heating/cooling
water temperatures are given as follows:

𝑄𝑖𝑛𝑝𝑢𝑡 = 𝑚̇𝑤 𝐶𝑝𝑤 (𝑇ℎ,𝑖𝑛 − 𝑇ℎ,𝑜𝑢𝑡 )

(𝟏𝟑)
𝑄𝑟𝑒𝑗𝑒𝑐𝑡𝑒𝑑 = 𝑚̇𝑤 𝐶𝑝𝑤 (𝑇𝑐,𝑜𝑢𝑡 − 𝑇𝑐,𝑖𝑛 )
(𝟏𝟒)

791
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The cycle performance is firstly built ignoring the pressure drop in the heat exchangers and pipes between the
components. Then the pressure ratio in the expander/pump is specified as an input. The temperature of the
evaporator outlet is also specified corresponding to the evaporator pressure in order to satisfy the proper working
fluid phase. A displacement pump is selected as an input to identify the flow rate in order to provide the amount
of heat that is available in the heat source. The flow rate-pressure curve is used to identify the pump efficiency.
The rest of the thermal properties and powers will be calculated using the thermodynamic model. Meanwhile,
the heat exchangers and the expander are selected to calculate the actual pressure drops and efficiency
respectively. This selection will update the cycle calculations performance. The ORC model is developed in
MATLAB with fluid properties extracted from COOLPROP. The App-Designer platform in MATLAB is used
as a user interface. Figure 2 illustrates the flow chart of the ORC system.

Given the Input parameters:𝑃2, 𝑃𝑟𝑎𝑡𝑖𝑜 , 𝑇5 and working fluid type

Satisfied

Testing the given parameters to satisfy with the working fluid phase in the evaporator outlet
Not

Fluid phase satisfaction test by the user through a given library

Satisfied

Pump selection by the user corresponding to the pressure and pressure ratio inputs (mass flow rate and efficiency as outputs)

𝑄𝑖𝑛𝑝𝑢𝑡 Calculation based on mass flow rate of the pump

Over or lower
than Range

The availability test of the 𝑄𝑖𝑛𝑝𝑢𝑡 amount from the Modified 𝑃3 𝑎𝑛𝑑 𝑃1
source
Range
Withi
n

losses ∆ 𝑃evap,loss
Obtained pressure

and ∆ 𝑃𝑐𝑜𝑛𝑑,loss
Obtained the cycle parameters: 𝑇1 , 𝑇2 , 𝑇3 , 𝑇4 , 𝑇5, 𝐻1 , 𝐻2 , 𝐻3 , 𝐻4 , 𝐻5

Expander Selection by the user (efficiency is the

output)

Power and thermal efficiency calculations: 𝑊𝑒𝑥𝑝𝑎𝑛𝑑𝑒𝑟 , 𝑄𝑟𝑒𝑗𝑒𝑐𝑡𝑒𝑑 , 𝑊𝑛𝑒𝑡 𝑎𝑛𝑑 𝜇𝑡ℎ𝑒𝑟𝑚𝑎𝑙

Evaporator and condenser heat exchanger selection by the user

Modified Power and thermal efficiency

Figure 1. ORC Model Flow Chart

RESULTS

Figure 3 shows the TS diagram for ORC cycle using the four investigated refrigerants with mass flow rate of
1kg/s, evaporator outlet temperature of 70C, pressure ratio of 3 and condenser outlet temperature of 10C. It
can be seen that the evaporator and condenser operating pressures are highest for R410A and lowest for R22.

792
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. T-S Diagram of various working fluid at evaporator outlet temperature of 70C, pressure ratio of 3, condenser
outlet temperature of 10C and flow rate of 1kg/s

Figures 4 (a,b) compare the ORC thermal performance and power output for four combinations of heat
exchangers used in the evaporator and condenser of the ORC system. Case 1 is for shell and tube heat exchangers
used in both the evaporator and condenser, case 2 where shell and tube heat exchanger is used for the evaporator
and plate heat exchanger used for the condenser, case 3 where plate heat exchangers are used for evaporator and
condenser while case 4 where the plate heat exchanger is used for the evaporator and shell and tube heat
exchanger is used for the condenser. It can be seen that R22 shows the highest thermal efficiency of 8.1%
followed by R407C with efficiency of 7.5% while R410A showed the lowest efficiency of 7.4%. As for the
power output, R410A showed the highest power output of 20kW followed by R22 and R407C with output of
19kW while R404A produced the lowest power output of around 16kW.

Figures 4 (c,d) show the overall heat transfer surface area of the evaporator and condenser for the various heat
exchanger type cases and the average overall heat transfer coefficient. It can be seen that the heat transfer surface
area for using shell and tube heat exchangers in the evaporator and condenser is the highest while the case of
using plate heat exchangers for the evaporator and condenser shows the lowest heat transfer surface area. This
is due to the higher overall heat transfer coefficients experienced with the plate heat exchangers as shown in
figure 4 (d). Therefore, plate heat exchangers would offer a more compact solution to reduce the overall system
size and reduce the amount of working fluids used.

793
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

R22 R410A R407C R404A 25

8.1530%

8.1510%
8.1130%

8.1100%
R22 R410A R407C R404A
20

POWER OUTPUT (KW)

THERMAL EFFECIENCY

15

7.5850%

7.5830%
7.5640%

7.5620%

7.5620%
7.5600%
7.5510%

7.5490%
7.4520%

7.4510%
7.4470%

7.4460%
10

0
S&T/S&T S&T/PHE PHE/PHE PHE/S&T
S&T/S&T S&T/PHE PHE/PHE PHE/S&T
EVAPORATOR/CONDENSER TYPE EVAPORATOR/CONDENSER TYPE
45
R22 R410A 1800 R22 R410A
40
TOTAL SURFACE AREA (M2)

R407C R404A 1600

35 R407C R404A
1400
30 1200

U (w/(M2.K)
25 1000
20 800
15 600
10 400
5 200
0 0
S&T/S&T S&T/PHE PHE/PHE PHE/S&T S&T/S&T S&T/PHE PHE/PHE PHE/S&T
EVAPORATOR/CONDENSER TYPE EVAPORATOR/CONDENSER TYPE

Figure 4. Effect of heat exchanger types on (a)the thermal efficiency of the ORC cycle, (b) Power output, (c)
Total surface area and (d) average overall heat transfer coefficient
II. CONCLUSIONS
Small scale ORC systems (less than 100kW power output) is an effective technology to utilize low grade waste
heat sources and generate clean electricity with no fossil fuel consumption. The selection of the working fluid
and the optimization of the system components like evaporator and condenser will have a major impact on the
small-scale performance in terms of efficiency and power output. In this work, a thermodynamic model to
optimize the performance of small-scale organic Rankine cycle (ORC) systems in terms of selecting the best
combination of refrigerants and evaporator/condenser heat exchangers to achieve highest power output and cycle
efficiency is developed. The work involves investigating the use of four refrigerants namely R22, R410A,
R404A and R407C and two types of heat exchangers namely shell and tube and brazed plate types with heat
source at temperature less than 100C. Results showed that R22 produced the highest thermal efficiency of 8.1%
followed by R407C with efficiency of 7.5% while R410A showed the lowest efficiency of 7.4%. As for the
power output, R410A showed the highest power output of 20kW followed by R22 and R407C with output of
19kW while R404A produced the lowest power output of around 16kW.

Also, results showed that the heat transfer surface area for using shell and tube heat exchangers in the evaporator
and condenser is the highest while the case of using plate heat exchangers for the evaporator and condenser
shows the lowest heat transfer surface area. This is due to the higher overall heat transfer coefficients
experienced with the plate heat exchangers for the same flow rate. Therefore, plate heat exchangers offer a more
compact solution to reduce the overall system size and reduce the amount of working fluids used.

ACKNOWLEDGMENT

The authors would like to thank the British Academy (grant no. UWB190157) and the University of Birmingham
Institute of Global Innovation (IGI-IAS project 3106) for sponsoring this research.
REFERENCES

[1] L. Li, Y. Ge and S. Tassou, "Experimental study on a small-scale R245fa organic Rankine cycle system for low-grade thermal
energy recovery," in The 8th International Conference on Applied Energy – ICAE2016, Uxbridge, Middlesex, 2017.
[2] P. Collings, A. Mckeown, E. Wang and Z. Yu, "Experimental Investigation of a Small-Scale ORC Power Plant Using a Positive
Displacement Expander with and without a Regenerator," in Proceedings of the Heat Powered Cycles Conference,
Bayreuth,Germany, 2019.
[3] E. Energy, "The Potential for Recovering and Using Surplus Heat from Industry; Department of Energy and Climate Change,"

794
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

London, UK, 2014.

[4] L. Tocci, T. Pal, I. Pesmazoglou and B. Franchetti, "Small Scale Organic Rankine Cycle (ORC):A Techno-Economic Review,"
2017.
[5] S. Lecompte, H. H. M. v. d. Broek, B. Vanslambrouck and M. D. Paepe, "Review of organic Rankine cycle (ORC) architectures
for waste heat recovery," Renewable and Sustainable Energy Reviews, vol. 47, p. 448–461, 2015.
[6] H. G and N. J, "Heat recovery opportunities in UK industry," Applied Energy, vol. 116, pp. 387-397, 2014.
[7] T. Turunen-Saaresti, A. Uusitalo and J. Honkatukia, "Design and testing of high temperature micro-ORC test stand using
Siloxane as working fluid," School of Energy Systems, Lappeenranta University of Technology, Finland , 2017.
[8] B. M, B. L, D. P. A, O. V, O. S and P. M, "Experimental Performance of a Micro-ORC Energy System for Low Grade Heat
Recovery," in IV International Seminar on ORC Power Systems, ORC2017, Milano, Italy, 2017.
[9] H. Chen, D. Y. Goswami, M. Rahman and E. Stefanakos, "Energetic and exergetic analysis of CO2- and R32-based transcritical
Rankine cycles for low-grade heat conversion," Applied Energy , vol. 88, no. 8, pp. 2802-2808, 2011.
[10] T. Turunen-Saaresti, A. Uusitalo and J. Honkatukia, "Design and testing of high temperature micro-ORC test stand using
Siloxane as working fluid," Finland, 2017.
[11] P. Garg, G. Karthik, P. Kumar and P. Kumar, "Development of a generic tool to design scroll expanders for ORC applications,"
Applied Thermal Engineering, vol. 109, pp. Part B:878-88, 2016.
[12] I. Papes, J. Degroote and J. Vierendeels, " New insights in twin screw expander performance for small scale ORC systems from
3D CFD analysis," Applied Thermal Engineering, vol. 91, pp. 535-546, 2015.
[13] D. Ziviani, A. Suman, S. L. S, M. D. Paepe, M. v. d. Broek and P. Spina, "Comparison of a 25 Single-screw and a Scroll
Expander under Part-load Conditions for Low-grade Heat Recovery ORC Systems," in The 6th International Conference on
Applied Energy – ICAE2014, 2014.
[14] S. Kang, "Design and experimental study of ORC (organic Rankine cycle) and radial turbine using R245fa working fluid,"
Energy, vol. 41, no. 1, pp. 514-524, 2014.
[15] G. Tozer, R. Al-Dadah and S. Mahmoud, "Gavin paper," in 5th International Seminar on ORC Power Systems, Athens, Greece,
2019.
[16] M. Lazova, H. Huisseune, A. Kaya, S. Lecompte, G. K. and M. D. Paepe, "Performance Evaluation of a Helical Coil Heat
Exchanger Working under Supercritical Conditions in a Solar Organic Rankine Cycle Installation," in Energies, 2016.
[17] S. Bari and S. Hossain, "Waste heat recovery from a diesel engine using shell and tube heat exchanges," Applied Thermal
Engineering, vol. 61, p. 355–363, 2013.

795
 COMPARATIVE STUDY OF DISCHARGE EMISSION
AMPLIFICATION IN AR-DRIVEN GAS DISCHARGE SYSTEM
WITH MICROPOROUS ZEOLITE CATHODE

B.G. Salamov
1
Department of Physics, Faculty of Science, Gazi University, 06500 Teknikokullar, Ankara, Turkey,
[email protected], ORCID: 0000-0002-9819-8160

H. Hilal Kurt
1
Department of Physics, Faculty of Science, Gazi University, 06500 Teknikokullar, Ankara, Turkey,
[email protected], ORCID: 0000-0002-1277-5204

Salamov, BG and Kurt H. Hilal. Comparative Study of Discharge Emission Amplification in

Cite this paper as: Ar-Driven Gas Discharge System With Microporous Zeolite Cathode. 8th Eur. Conf. Ren.
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In this work the zeolite-gas discharge system (ZGDS) with microporous dielectric cathode is
studied in dependence on the physical properties of cold non-equilibrium plasma and discharge
mode. The discharge emission intensity (DEI) in the UV, visible and near IR ranges (330-850 nm)
generated by a planar Ar-driven ZGDS and the possibility of enhancing DEI in a wide range of
gas pressure pg up to one atmosphere (10-760 Torr) for given properties of a microporous zeolite
cathode (MZC) has been studied. Interesting results of the discharge emission amplification for
different gas-discharge gap dg (45-250 μm) in ambient air and Ar-driven ZGDS are presented for
the first time. Electrons entering the discharge gap with microporous and multichannels structure
of the MZC are multiplied in the electric field by the avalanche mechanism, so that current and
DEI increases in the discharge gap of the ZGDS. We studied the relative DEI to evaluate the
enhanced performance of the ZGDS as a source of stable discharge emission in a wide range of pg
up to one atmosphere.
Keywords: Gas discharge system, microporous zeolite electrode, air and argon medium, discharge emission,
atmospheric pressure, enhanced performance.
© 2020 Published by ECRES

1. INTRODUCTION

The considered gas-discharge system with a microporous zeolite cathode (ZGDS) [1-3] is designed to generate
non-equilibrium plasma in free space by creating cold (close to room temperature) and stable plasma states in
different gases that can chemically activate gaseous media and surfaces of thermally unstable materials at
atmospheric pressure. One of the effective methods of plasma-chemical activation of gases at atmospheric
pressure (AP) (for example, for cleaning from harmful impurities) is the transmission of gas flows through the
gas-discharge chamber. At the same time, in the plasma, gaseous compounds are excited and dissociated until
the formation of various radicals and the formation of harmless, easily recyclable substances. When using
microporous zeolite cathodes (MZC) in gas-discharge reactors [4] as a catalyst for increase the discharge
emission intensity, it is necessary to solve the following problems: 1) preparing of atmospheric plasma activation
on the zeolite surface, for example, removing organic templates used in the synthesis of zeolites; 2) synthesis of
the zeolite-based catalyst itself; 3) regeneration of already used zeolite-based catalysts. In solving these
problems, it is necessary not only to ensure the purity of the surface of the samples, but also to preserve the
crystalline structure of the zeolite. However, the high-temperature treatment used in all three cases (773 K and
higher) often causes undesirable effects (e.g. transition of one crystalline modification of the zeolite to another,
local heating of the catalyst) [5]. In turn, it is known that the treatment of solids with active particles of low-
temperature plasma generated in various types of electric discharges leads to various changes in their properties.

796
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The effect of low-temperature plasma on the properties of aluminosilicates has been studied very little. Attempts
have been made to use gas discharge plasma for the synthesis of calcium aluminates. Only a few works appeared
in which the possibility of using a radio-frequency discharge to remove organic templates from zeolites, as well
as to obtain a zeolite-based catalyst, were studied. Nothing is known about the modifying and regenerating effect
of plasma on aluminosilicate catalysts, and it is also not clear what effect oxygen plasma causes when processing
catalysts [6].
Glow discharge plasma was used to change the surface composition of silica gels, cements, and catalysts, which
is especially important since their structure did not change [7]. The uniformity and selectivity of the sorbents
used in gas chromatography increased after treatment with oxygen and argon plasma [8]. This plasma increased
the activity, stability, and selectivity of zeolite-based catalysts [9]. The mode of existence of a glow discharge:
low pressure (0.2–3 Torr), low temperatures (340–473 K), short exposure times (seconds, up to tens of minutes),
increased concentrations of active particles, non-isothermal plasma [10] . In [11], the formation of a large
number of various active particles as a result of oxygen excitation was substantiated. The presence of ozone and
O4+, O4-, O6+, O6- molecules was also detected. The oxygen atoms O(3p) and metastable O2(1Δg) molecules had
the highest activity. Argon plasma contains fewer active particles than oxygen. Argon atoms are mainly in an
excited electronic state. The temperature of an argon plasma is lower than that of oxygen; its effect on the surface
of a solid is mainly physical in nature [11]. In [12], it was found that argon plasma worsens the surface
characteristics of sorbents, while oxygen plasma increases their surface and adsorption capacity for benzene.
Under the action of a glow discharge in O2 and Ar on the hydrogenation catalyst of carbon monoxide of the
composition α-Fe2O3 /Zeolite (ZSM), the formation of the X-ray amorphous phase occurred during the
destruction of crystalline iron oxide. The low-temperature oxidation of copper by oxygen plasma was
accompanied by an increase in the surface by a factor of 1.5–2. A Cu1O1,7 film containing superstoichiometric
oxygen was obtained [13].

Recently microporous zeolite (Ca,K2,Na2,Mg)4Al8Si40O96.24H2O has greatly applied as a perspective new

dielectric materials used in technological and optoelectronic devices [14,15]. The unique surface, adsorption and
transport properties make it as perspective material in the plasma light sources and energy storage devices. In
recent work [16] the authors suggested a device for increasing the adsorption efficiency of microporous zeolite
by treatment it in an electric discharge. At the same time, prolonged application of the electric field considerably
changes the transport and adsorption properties of MZC due to the appearance on the surface, cavities and
channels of electric charges of different signs [17]. At the same time, we suggested that the modification and
regeneration of the MZC under the action of low-temperature argon plasma at AP causes the appearance of new
surface structures, which increase discharge emission intensity (DEI), selectivity, and stable operation range
[15]. Thus, the content of these research works indicates that the interaction of argon Glow discharge plasma
with the surface of metallic and microporous dielectric materials usually increases their activity, selectivity and
stable operation range. The main reason for these changes is the creation of new active centers of catalysis and
adsorption with a new structure and composition [18-20]. The number of studies in this area is relatively small,
and therefore the continuation of work in this direction, supplemented by the determination of the electro-optical
characteristics of plasma devices based on microporous zeolite, is quite advisable.

2. EXPERIMENTAL

Fig. 1. Scheme of the ZGD cell: (1) metallic Cu contact; (2) MZC; (3) diameter of the gas discharge gap dg; (4) thickness
of the insulating mica; (5) semi-transparent conductive SnO2 contact; (6) flat glass disc.

797
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The experimental setup of the ZGDS is shown in Fig.1. The thickness of the dg is varied in the range from 45 to
250 m. A glass disc (30 mm), covered with a conducting layer of SnO2 (5) was used as the anode. The local
electric field achieved in the vicinity of the electrodes is about 105 V/cm, strong enough for discharge generation
at AP in Ar and ambient air. When a high U0 applied to electrodes, the discharge is ignited in the dg. This
discharge mode corresponds to the Townsend or Glow discharge regime. Detailed information regarding
experimental system and MZC is given in our earlier study [21].

Fig. 2. SEM image of the clinoptilolite sample.

Due to the volume of free intracrystalline space, clinoptilolite (Ca,K2,Na2,Mg)4Al8Si40O96.24 H2O [7,15,16] is
referred to as medium-porous zeolites, its total porosity is on average 30%, and the specific surface reaches
about 105 cm2/g. The density of the mineral varies between 2.11-2.2 g/cm3. Comprehensive information
concerning location, arrangement and chemical composition of the natural zeolite (1011 Ω.cm) is presented
in Ref. [21,22]. Studies of the crystalline structures of clinoptilolite [23] showed that, due to the relatively high
Si/Al > 4 (~ 82% of tetrahedrons occupied by Si), the clinoptilolite crystalline framework is heat-resistant (in
air up to 700 °C) and is also resistant to aggressive substances and ionizing radiation.
Electron microscopic studies revealed that zeolite samples mainly have a complex micro-surface relief formed
by microcrystals and aggregates of various mineral phases. Mono-mineral aggregates of zeolites are represented
in most cases by weakly crystallized mass or microcrystals. Micro-crystal aggregates are concentrated in
microarrays and in micro-cracks fairly uniformly distributed over the surface of the sample. Clinoptilolite
crystals have a lamellar shape.

3. RESULTS AND DISCUSSION

At atmospheric pressure (AP), the process of plasma-chemical activation of gases is accompanied by the effect
of negative plasma components on the positive electrode when passing gas flows through the gas-discharge
chamber. In the case of a gas discharge in an air and argon medium, the negative components of the plasma are
electrons and oxygen ions (oxygen is present in Ar as an impurity). The mechanism of plasma-chemical
activation of gases under the action of electrons is sputtering due to evaporation. However, the electron energy
in the discharge (50-100 eV) is insufficient for sputtering. We were convinced of this by specially putting
experience on the surface treatment of the anode in an electron microscope with an electron beam of such energy.
We were convinced of this by specially conducted experiments on the surface treatment of the anode in an
electron microscope with an electron beam of such energy. Possible mechanisms of interaction of oxygen ions
with the anode surface are sputtering and the formation of oxide on the surface. Considering the high oxidative
capacity of oxygen ions and the results of our experiments, we believe that the main mechanism of plasma-
chemical activation of gases is accompanied by the interaction of the electrode surface with oxygen ions
bombarding it and the formation of oxide on the electrode surface. The electron flow of the discharge is a catalyst
for the oxidation process, creating cathode diffusivity, it induces an oxidation process on the electrode surface.
The transfer of electric charge in zeolites in constant an electric field can be simulated as electric transport in
dispersed systems. It can be carried out due to ionic, electronic and molion conductivities. The contribution of

798
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

these components is different and depends on the properties of the components included in the system. Zeolite
can conditionally be considered as a system consisting of a solid phase - an alumina-silicon-oxygen framework
pierced by capillaries of intracrystalline channels oriented in a certain way relative to the lines of the external
electric field strength and extra-frame subsystem (exchange cations and water). It is obvious that the conductivity
of the framework is negligible compared to the conductivity of the water-cationic subsystem. This is due to the
high migration ability of the forming off-frame cations and water molecules.

7
7 d = 45 m, D = 12 mm,
d = 45 m, D = 12 mm,
6 p = 760 Torr, MZC, Ar
6 p = 360 Torr, MZC, Ar
 Ar
2 mA
5
 Ar 5  Ar

Current (mA)
Ar
Current (mA)

  Air
250 V 4,6 mA
 Air
4 4  Air
 Air

3 3

2 2 400 V

1 1

200 400 600 800 1000 1200 1400 200 400 600 800 1000 1200 1400 1600
Voltage (V) Voltage (V)
Fig. 3. The hysteresis behavior of the ZGDS with MZC in an air and argon medium, (a) dependence of the current on the
applied voltage at 360 Torr and (b) I – V characteristic of the ZGDS at AP. The system parameters are dg = 45 μm and D
= 12 mm.

During of the experiments, it is established that the obtained I–V characteristics completely satisfy Ohm's law.
Fig. 3 shows the I–V characteristic of the ZGDS with MZC at a constant temperature of 300 K in the voltage
range 200–1600 V. The average residual resistance of the MZC sample is (2.1 ± 0.39) .1010 Ohm [1,2]. The
different character of the I–V characteristic can be associated with an increase in the mobility of charge carriers.
The hysteresis phenomenon was clearly observed, which reflects the process of formation of the active surface
of a zeolite under the influence of a reaction medium. It should be noted that the hysteresis phenomenon with a
sharper and more distinct peak is observed at high pressure values. The shape of the hysteresis is determined by
the type and shape of the MZC pores. Characteristically, the presence of hysteresis indicates the simultaneous
presence of both meso– and micropores [24, 25]. At the same time, we suggested that observed the hysteresis
phenomenon in the I – V characteristic of the MZC under the action of low-temperature argon plasma at AP
causes the appearance of new surface structures, which increase the transport property, selectivity, and stable
operation range (see Fig.4) [4].
Considering the Fig.3 and Fig.4 one can note the following: (i) Breakdown voltage Uign values in Ar-filled ZGDS
are lower compared with the Uign values in an air conditions, (ii) although lower voltages are applied to the Ar-
filled ZGDS, stable operation range [2] for Ar is substantially exceeds the corresponding value for ambient air;
(iii) the current in Ar is increased more than that for air at given values of the pressure pg. The porous structure
of the DZC also provides the ability to amplify the electric field at the interface between the dielectric and the
gas, where a dielectric permittivity gradient is observed, thereby possibly reducing the Uign [21]. We belived that
electron emission occurred from not only the surface but also the cross-sectional surface of microporous DZC,
therefore discharge emission is more increase when using large DZC with internal emission amplification.
Electrons entering the discharge gap with microporous and multichannels structure of the MZC are multiplied
in the electric field by the avalanche mechanism, so that current and DEI increases in the discharge gap of the
ZGDS.

799
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1200
1500 (b) D = 22 mm, d = 45 m, Ar, MZC
(a) D = 22 mm, d = 45 m, Air, MZC
1300 1000 Ud
Ud

U = 730 V
1100 800

U (Volt)
U= 666 V
U (Volt)

900 600 Stable Operation Range

Stable Operation Range
700 400

500 200
Uign Uign
Uign, min= 214 V
300 0 100 200 300 400 500 600 700 800
0 100 200 300 400 500 600 700 800
Pressure (Torr)
Pressure (Torr)

Fig. 4. Stable operation range of ZGDS with MZC: (a) in an ambient air medium, (b) in argon medium. The system
parameters are dg = 45 μm and the diameter of MZC area was D = 22 mm.

The stable operation range of ZGDS in argon medium can be expanded by increasing the conductivity of the
MZC and the conductivity can be adjusted by the feeding voltage and pressure pg. Zeolite water plays an
important role for the stability of MZC framework. It is clear that for the penetration depth, the ion flux density
and irradiation time are of primary importance, and the dose is indirect and secondary. The concentration of the
adsorbed gas is determined by the penetration depth and dose, as well as the gas permeability of the microporous
zeolite. A distinctive feature of a normal glow discharge in argon is the presence of a significant amount (up to
5–7%) of hydrogen ions H+ and H2+ in the total ion current [20], which can be considered as a consequence of
the effect of gas separation in a glow discharge, when a component with a lower ionization potential is
concentrated near the cathode. It should be noted that hydrogen is especially important for generating excess
stabilized vacancies, because even at comparable energies [21] hydrogen penetrates much deeper than argon.
This means that in the process of radiation-stimulated diffusion of an inert gas, apparently, the main role is
played by hydrogen, which not only stabilizes the radiation vacancies, but also increases their mobility, which
facilitates the inert gas penetration into the microporous zeolite through hydrogen-stabilized radiation vacancies.

4. CONCLUSION

Thus we suggested that a suitable choice of the gas medium, dg, pg and electric field is important for further
improvement of the microstructures and performance of the MZCs which considerably depend on the channel
morphology and nature of the nanopores. Electrons entering the discharge gap with microporous and
multichannels structure of the MZC are multiplied in the electric field by the avalanche mechanism, so that
current and DEI increases in the discharge gap of the ZGDS. This device can be used in plasma devices,
optoelectronics, field-emission displays, energy storage devices, catalysis and biomedical applications. At the
same time, we suggested that the modification and regeneration of the MZC under the action of low-temperature
argon plasma at AP causes the appearance of new surface structures, which increase discharge emission
intensity, selectivity, and stable operation range.
REFERENCES

[1] Koseoglu, K, Özer, M, Salamov, BG. Field-enhanced transport processes in the gas discharge system with porous
zeolite, IEEE Transactions on Plasma Science 2015, 43, N10, 3576-3581
[2] Koseoglu, K., Salamov, BG. New Approach for Charge Transport Mechanisms in the Atmospheric Pressure Cold
Plasma Device with Porous Zeolite, Plasma Process. Polym. 2016, 13, N3, 355-365.
[3] Astrov, YuA, Portsel, LM, Lodygin, AN, Shuman VB. Planar microdischarge device for high-speed infrared
thermography: Application of selenium-doped silicon detectors, J. Appl. Phys. 2008, 103, 114512-6
[4] Lebedeva, NN, Orbukh, VI., Salamov, BG. Investigation of the effect of discharge plasma stabilization by a
semiconductor, J. Phys. III 1996, 6, 797-805
[5] Pylinina, AI, Mikhalenko, II, Ivanov-Shits, AK, Yagodovskaya, TV, Lunin, VV. The Influence of plasma chemical
treatments on the activity of the Li3Fe2(PO4)3 catalyst in butanol-2 transformations, J. Phys.Chem. 2006, 80, №6,
882-885
[6] Lui, CJ, Yu, K, Zhang, YP. Characterization of plasma treated Pd/HZSM-5 catalyst for methane combustion, Appl.
Catalysis B: Environmental. 2004, 47, 95-101

800
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[7] Yagodovskaya, TV, and Lunin, VV, Zh. Fiz. Khim. 1997, 71, N5, 775 [Russ. J. Phys. Chem. 1997, 71 (5), 681]
[8] Lunin, VV, Dadasheva, EA, Yagodovskaya, TV, and Knipovich, OM, Abstracts of Papers, 2nd All-Union Meeting
“Application of Plasma in the Technology of Catalysts” (Kiev, 1991), p. 18.
[9] Lapidus, AL, Solomonik, IG, Krylova, AYu, Krashenninikov, EG, Abstracts of Papers, IV Russian Conference with
the Participation of CIS Countries “Scientific Fundamentals of the Preparation and Technology of Catalysts” (Ufa,
2000), p. 35.
[10] Aluna, R, Bratchikova, IG, Mikhalenko, II, and Yagodovskaya, TV, Abstracts of Papers, IV Russian Conference with
the Participation of CIS Countries “Scientific Fundamentals of the Preparation and Technology of Catalysts” (Ufa,
2000), p. 93.
[11] Pylinina, AI, Dobrova, EP, Mikhalenko, II, and Yagodovskaya, TV, Zh. Fiz. Khim. 2005, 79, N4, 650 [Russ. J. Phys.
Chem. 2005, 79, N4, 552]
[12] Kim, HH, Tsubota, S, Daté, M, et. al. Catalyst regeneration and activity enhancement of Au/TiO2 by atmospheric
pressure nonthermal plasma, Appl. Catalysis A: General. 2007. V.329. P. 93-100.
[13] Gavrilova, TB, Nikitin, YuS, Vlasenko EV, et. al. Gas chromatographic investigation of silica supports modified in a
high-frequency low-temperature plasma, J. Chromatogr. 1991. V.552. № 1-2. P. 179-185.
[14] Astrov, YuA, Lodygin, AN, Portsel, LM, Beregulin, EV. Comparative study of noise in low-current Townsend
discharge in nitrogen and argon Phys. Rev. E 2017, 95, 043206-7
[15] Aluna, R, Bratchikova, IG, Mikhalenko, II, and Yagodovskaya, TV. Abstracts of Papers, IV Russian Conference with
the Participation of CIS Countries “Scientific Fundamentals of the Preparation and Technology of Catalysts” (Ufa,
2000), p. 93.
[16] Gashimov, АМ, Zakiyeva, IG. Dielectric parameters of the composite based on natural zeolite treated by electrical
discharge. Tech. Phys. 2006, 51, 185.
[17] Bunyatova, U, Orbukh, VI, Eyvazova, GM, Agamaliev, Z, Lebedeva, NN, Salamov, BG. Peculiarities of unusual
electret state in porous zeolite microstructure, Superlattices and Microstructures 2017, 111, 1203-1210
[18] Mavroyannis, C. The interaction of neutral molecules with dielectric surfaces, Mol. Phys. 1963, 6, 593-600
[19] De Boer, JH, Kasperema, JH, Vandongen, RH, Brokhoff, JCP. Krypton adsorption on the surface of copper powder,
J. Interface Sci. 1972, 38, № 1, 97
[20] Haul, RAW, Swart, ERA. Argon adsorption an der Oberfläche von Eisenpulver, Z. Electrochem. 1957, B.61, №3, C.
5380-5387.
[21] Salamov, BG, Kurt, HHilal. Ar-driven gas discharge system on the basis of dielectric zeolite material, JOM The
Minerals, Metals & Materials Society, 2020 DOI: 10.1007/s11837-019-03955-1.
[22] Cai, Z, Coordrich, TL, Ziemer, KS, Warrywocla, J, and Sacco, A. J. Appl. Catal. B: Environ. 2011, 102, 323.
[23] Gottardi, G, Galli, E, Natural Zeolites, Springer, Berlin 1985.
[24] Michot, LJ, Villieras, F. Assessment of surface energetic heterogeneity of synthetic Nasaponites. The role of layer
charge, Clay Miner. 2002, 37, 39-57.
[25] Neaman, A, Pelletier, M, Villieras, F. The effects of exchanged cation, compression, heating and hydration on textural
properties of bulk bentonite and its corresponding purified montmorillonite, Applied Clay Science. 2003, 22, 153-168.

801
 THE PLASMA SYSTEM WITH INP PHOTODETECTOR
H. Hilal Kurt1
1
Department of Physics, Faculty of Science, Gazi University, 06500 Teknikokullar, Ankara, Turkey,
[email protected], ORCID: 0000-0002-1277-5204

F. Karip1
1
Department of Physics, Faculty of Science, Gazi University, 06500 Teknikokullar, Ankara, Turkey,
[email protected]

Kurt H. Hilal, Karip Fuat, The plasma system with InP Photodetector 8th Eur. Conf. Ren.
Cite this paper as:
Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: The research aims to investigate the gas discharge system with a Fe-doped InP photodetector.
Infrared Image converters are of great importance in optical research. Current-voltage curves are
obtained for different plasma parameters such as pressure, gas discharge gap under infrared
stimulation. Breakdown curves are obtained for He and air media. InP is an III-V group direct
bandgap semiconductor material. InP has high electron mobility and it causes higher plasma
current and discharges light emission when compared to other semiconductor photodetectors.
Keywords: Microplasma, InP, breakdown, He, air
© 2020 Published by ECRES

INTRODUCTION

Plasma studies focused on both academic and industrial applications, especially based on high-speed imaging
by using microplasma systems. High display speed will reveal different materials to display. The structure of
the semiconductor gas discharge cell was measured using a high-speed streak camera [1](Abroyan, Eremeev
and Petrov N N (1967)). The rough image of our semiconductor gas discharge system was seen in our system.
Different researchers have discussed various plasma structures such as optical-electronic and elastic of these
semiconductors. Glow discharge between the electrodes can be used widely in different technological plasma
processes [2]. Breakdown voltage depends on the pressure, interelectrode distance and gas type. Even though
all parameters are kept constant, (p.d) value cannot be held constant for characterize breakdown voltage and the
Paschen curves, VB depends on the product of pd exponentially [3]. On the other hand, the breakdown expresses
the transition of a non-sustaining discharge into a self-sustaining discharge when the appropriate conditions are
fulfilled in the plasma in which ionization process takes place depending on the interelectrode distance d and
gas pressure p which was discussed by Panchen’s Law. Many processes are responsible for ionization
phenomena such as pressure, temperature, plasma species, electrode configuration, and nature of electrode
surfaces and the availability of the initial conducting particle in the plasma . A new infrared image converter in
dc argon plasma has been proposed by Kasymov and Paritskii. In this system, a light-sensitive semiconductor
cathode, which has a high dielectric property, has a current density in gas discharge zones. Since the freeload
density is low in this dc Argon system, the current amount is low. The Townsend regime also occurs at very low
currents, such as 10-6 A.

Microdischarge unit is the main part of an IR image converter and it was invented in the 1970s [4]. In such
converter systems, a micro gap filled by gas, a high-resistivity semiconductor cathode (sensitive to IR region),
and a glass plate coated with a transparent film is mainly used. It has crucial importance to choose the convenient
semiconductor materials for specified by good sensitivity and high speed of operation [5]. The operation

802
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

principle of such a device is based on controlling the gas discharge via a high-resistivity semiconductor electrode
[6]. The spatial distribution of the discharge light emission and current and in the discharge gap is proportional
to the semiconductor conductivity, which is controlled through incident IR radiation projected onto the
photodetector; it is converted into the visible glow of the gas [5]. Gas discharges are mainly classified as a
Townsend, glow and arc discharges [7].

In this study, InP is used as a cathode of the plasma system. The system characteristics depend on both InP
material and plasma. InP cathode gives higher plasma current. The plasma gap is filled with He and Air gases.
He is chosen due to stable operation in a gas discharge system.

RESULT AND DISCUSSION

The qualities of the photodetectors are measured with their uniform spectral response to the optical signal, and
their high speed and low noise in the conversion process. The results show that the system characteristics
strongly dependent on both the semiconducting electrode and the plasma parameters. At the same time, IR
sensitivity is strongly influenced by pressure variation and cathode material. Those materials are called as photon
detectors since they convert IR optical signal into the current in the system and thus act as a kind of transducer.
Figure1 shows current versus voltage graphs under different cathode diameter. When the cathode diameter D
increases, lower breakdown values are reached as shown in Figure 1. It is related to the increased active
discharge area of the plasma system.

Figure 1. CVCs for different cathode diameter D for p = 480 Torr.

2000
1800
1600
1400 He
Voltage (V)

1200 Air
1000
800
D = 12 mm, d = 440 m, InP, Dark
600
400
200
0 200 400 600 800

Pressure (Torr)
Figure 2. Breakdown curves for Air and He discharges as a function of pressure

Figure 2 shows Paschen curves for Air and He gases in plasma. Those shows very different behaviours. He
results in lower breakdown values. However, Air causes a higher breakdown voltage. Using a resistor distributed
semiconductor InP significantly changes the current distribution and breakdown values. The value of the
transmitted current and the type of discharge is determined by the homogeneity of the resistance distribution of

803
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the semiconductor cathode. This is the desired situation in some practical applications related to the spatial
distribution of current density in light-sensitive semiconductor materials; because this increases the resolution,
facilitates the technical realization of reliable operation and expands the region of the material used. However,
technological plasma systems often use the discharge cell where the cathode diameter is much larger than the
distance between the cathode and the anode.

Figure 3 shows CVCs for different interelectrode distances d. As seen in the figure, gas makes the transition
from a non-conducting state to conducting state at lower voltage value for small interelectrode distances (50
m and 100 m). These two curves show different behaviour compared to other interelectrode distances (240-
525 m).

-4
3x10
p = 360 Torr, D = 12 mm, He

-4
2x10
CURRENT (A)

d = 50 m
d = 100 m
d = 240 m
-4
1x10 d = 340m
d = 440 m
d = 525 m

0
200 300 400 500 600 700 800 900
VOLTAGE (V)

Figure 3. CVCs as a function of interelectrode distance d.

The plasma current was investigated in the system with InP photocathode for the dc electric field.
One of the control parameters is the conductivity of the semiconductor electrode. Due to the semiconductor's
sensitivity to light, its conductivity can be controlled by illumination. Additionally, semiconductor electrode gas
discharge systems serve as fast converters that convert IR images into images. Therefore, it is important to
examine their operational stability for these and other technical applications.

3. CONCLUSION

The Infrared radion is absorbed by InP detector within the bandgap due to interaction of the electrons with lattice
atoms and impurities. The observed CVC values resulted from the collision of free electrons which emitted from
InP cathode into plasma interval. The optical absorption mechanism is responsible for the transport mechanism
in the system. In this study, IR stimulation directed on the InP cathode. Our experiments show that He causes
lower breakdown voltage, whereas air causes higher at the same conditions. And cathode diameter is very crucial
in plasma current. Plasma current increases with increasing cathode diameter.

REFERENCES

[1] Abroyan I A, Eremeev M A, Petrov N N, Sov. Phys. Usp.10 332–367 (1967) .

[2] V.A. Lisovskiy, P.P. Platonov, S.V. Dudin, Plasma Physics. 25, 156 (2019)
[3] I. H. Hutchinson, Principles of Plasma Diagnostics, 2nd ed. (Cambridge: Cambridge UP, 1987. Print)
[4] Sh. S. Kasymov and L. G. Paritskii, Device for tracking images Russian Authors’ Certificate 1798020/18-10, 1973
[5] Yu.A. Astrov, A.N. Lodygin, and L.M. Portsel, J. Phys. D Appl. Phys. 49, 095202 (2016)
[6] Y. Y. Peter, C. Manuel, Fundamentals of Semiconductors Physics and Materials Properties, ( Fourth Edition, Springer,
2010)
[7] H.H. Kurt, E. Tanrıverdi, Journal of Electronic Materials. 46, 4024 (2017)

804
 THE ELECTRON CURRENT DENSITY PATTERNS IN
SEMICONDUCTOR-GAS DISCHARGE SYSTEM

H. Hilal Kurt
Department of Physics, Faculty of Science, Gazi University, 06500 Teknikokullar, Ankara, Turkey, [email protected],
ORCID: 0000-0002-1277-5204

B.G. Salamov
Department of Physics, Faculty of Science, Gazi University, 06500 Teknikokullar, Ankara, Turkey,
[email protected], ORCID: 0000-0002-9819-8160

E.Tanrıverdi
Department of Physics, Faculty of Science, Gazi University, 06500 Teknikokullar, Ankara, Turkey,
[email protected]

Kurt H. Hilal , Salamov, BG and E.Tanrıverdi, The Electron Current Density Patterns in
Cite this paper as: Semiconductor-Gas Discharge System 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: In this study, electron current densities for different cathode materials are obtained by COMSOL
simulation programme in the semiconductor-gas discharge structure. Interesting results are
obtained with varying gas pressure and inter-electrode distances. Electron current density maps
change between cathode and anode depending on the plasma conditions. Local currents density
domains are observed in the different part of the microplasma system. We concluded that those
regions are very sensitive both for the type of cathode material and plasma parameters.
Keywords: Gas discharge system, COMSOL, current density
© 2020 Published by ECRES

1. INTRODUCTION

Townsend discharge is the gas ionization phase of free electrons, which is accelerated in a sufficiently strong
electric field, giving rise to the electrical conductivity in the gas with increasing avalanche multiplier. The
decrease in the number of free charges or the weakening of the electric field stops Townsend discharge.
Townsend discharge got its name from John Sealy Townsend and is also commonly known as "Townsend
avalanche". Avalanche is a staged reaction in an area containing electrons in a sufficiently high electric field.
This reaction takes place in an ionizable environment, such as air. By ionizing an atom or a molecule in an
applied electric field in the devices, the positive ion is moved towards the cathode while the free electron is
moved towards the anode. If the electric field is strong enough, the electrons gain enough energy to release more
free charge. With the collision of electrons with another atom or molecule, more electrons are released when it
reaches a sufficient energy level. The two free electrons move together after advancing a little in the electric
field and gaining energy before another collision occurs. This is an effective chain reaction of electron formation.
The number of electrons moves towards the anode, increasing to 2n in each collision. Here n is a collision. The
limit of the number of avalanche multipliers is known as RAETHER LIMIT (Kao, Chih-cheng, et Yu-cheng Liu
(2009)).

In Townsend type discharging, a voltage is placed across the two electrodes, making a gas-filled charge-
discharge transition. Electrons are accelerated by the ionizing collisions in the electric field to produce more
electrons, such as high-energy muon interactions, and are found in space due to natural stages.

805
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

dne / dx = α (E, p) ne [1]

ne is the electron density in particles per unit volume. α is the number of ionizations per unit length to which a
single
electron is exposed, E is the electric field and p is the pressure [2].

α = Ape - Bp / E, [2]

where A and B are characteristic of the empirical coefficient of a given gas. The stage of electron production is
shown as the secondary emission coefficient γ. This study compares the electron current densities for different
cathode materials in a wide region plasma parameter interval. This is the first study that shows current
densities in semiconductor gas discharge system by using COMSOL Multiphysics programme.

2. RESULTS AND DISCUSSION

In our system, semiconductor materials are defined as cathodes, while SnO 2 coated glass is defined as the anode.
In the first stage, the geometric structure was created. Then, argon gas and materials were defined from the
material library. COMSOL Multiphysics package has been used for the modelling of semiconductor-plasma
structure. The numerical simulation is an indispensable tool to understand the plasma behaviour inside the micro
discharge cell with InP electrode and it is also vital to improve the knowledge in the field of dc discharges which
is used for deposition or etching [3,4]. The optimal mesh numbers are used as 42560 elements. The system uses
gas discharge formalism and the required material constants to find the steady-state solutions for the defined
simulation cell [5]. It is noted that there exist limited theoretical studies on the transport properties of III-V group
semiconductors.

Figure 1. Current density graphs: a) for GaAs, b) for GaP in the case of 160 Torr gas pressure and 50 m interelectrode
distance.

As seen from the Figure, GaAs gives higher current density compared with GaP cathode. The current density
value is 7.32 x 10 5 (A / m2) for GaAs and 1.85x 10 5 (A / m2) for GaP cathode. Present investigations have
shown that Current density is strongly dependent on the material type and also the mobility of the electrons in
the plasma cell.

Here, secondary electrons oscillate at very high speeds from the cathode, that is, they have high acceleration.
These high energy electrons are at a certain distance from the cathode (which corresponds to the average free
path) and consequently form the cathode glow. The colour of the emitted radiation depends on both the cathode
material and the working gas. The axial length of the cathode glow depends on gas pressure and the nature of
the gas. The cathode glow is followed by the cathode dark space (Crookes or Hittorf), where the electric field is
of medium size, where the space charge is positive and highly high density. In this dark area, positive ions are
accelerated against the cathode. Since the mobility of ions is smaller than that of electrons, this dark region is
mostly made up of ions. The cathode dark zone is followed by the negative glow zone. Negative glow is the
region with the brightest luminous intensity across its entire area. This leads to ionization and excitation

806
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

collisions in the region. Here the electric field is very low and the current is almost entirely produced by the
electron.

3. CONCLUSION

In this study, current densities maps have been obtained by the simulation programme. Current densities show
that current domain regions around electron densities exist in plasma structure. And those domains move from
cathode anode resulting luminous part inside the plasma. The properties of the glow discharge are changeable
with accordance to its discharge parameters such as gas pressure, plasma gas composition, gas pressure and
applied voltage to the electrodes (input electrical power). The plasma gas composition is particularly important
in the excitation and ionization mechanisms including the production of active species [2]. The cathode is an
electrically conductive metal structure. As a result of bombarding the cathode surface with positive ions, the
electron is released from the cathode. This is called secondary electron diffusion. The potential difference
between cathode and anode is generally uneven. The area right next to the cathode is the primary dark zone or
the Aston dark zone. This region has a strong electric field and a negative space charge. With the accumulation
of the secondary electron, Aston dark space or primary dark space is formed in the cathode. When there are
stages of acceleration from the cathode, the accumulation of slow electrons occurs, creating a negative surface
charge and a strong electric field in this region.

REFERENCES

[1] Kao, Chih-cheng, et Yu-cheng Liu. Intense green emission of ZnS:Cu, Al phosphor obtained by using diode structure
of carbon nano-tubes field emission display. Materials Chemistry and Physics 115, no 1 (15 mai 2009): 463-466.
doi:10.1016/j.matchemphys.2009.01.013.
[2] Y.P. ,Gas Discharge Physics. Springer-Verlag, 1997.
[3] Coburn J W and Harrison W W 1981 Plasma sources in analytical mass spectrometry Appl. Spectrosc. Rev. 17 95–164
de la Rue W and Müller H W 1880 Experimental researches on the electric discharge with the chloride of silver battery
Phil. Trans. R. Soc. London 171 65–116
[4] Delcroix J-L and Trindade A R 1974 Gas discharge displays: a critical review Advances in Electronics and Electron
Physics vol 35 ed L Marton (New York: Academic) pp 87–190
[5] Kurt, H.H., Tanrıverdi, E. Electrical Properties of ZnS and ZnSe Semiconductors in a Plasma-Semiconductor
System. Journal of Elec Materi 46, 3965–3975 (2017).

807
 THE CHARACTERISTIC FEATURES OF THE COLD PLASMA
WITH ZNSE ELECTRODE

H. Hilal Kurt
Department of Physics, Faculty of Science, Gazi University, 06500 Teknikokullar, Ankara, Turkey, [email protected],
ORCID: 0000-0002-1277-5204

S.Kızışar
Department of Physics, Faculty of Science, Gazi University, 06500 Teknikokullar, Ankara, Turkey,
[email protected]

Kurt H. Hilal , Kızışar Sereni, The Characteristic Features of The Cold Plasma with ZnSe
Cite this paper as:
Electrode 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Rapid advances have been in Infrared sensitive materials with a direct bandgap. ZnSe is ideal
Infrared detector due to its low absorption at infrared wavelengths and ıts operation at room
temperature. Its operation is not limited in visible but it is used in a wide spectral range up to
Infrared. There is a fundamental relationship between optical properties and band structure of the
material. ZnSe served as a cathode material in our system. ZnSe is very sensitive for IR
stimulations and ZnSe can operate up to atmospheric pressures. ZnSe gave better plasma values at
high gas pressure values.
Keywords: Cold plasma, ZnSe, CVC curves, IR stimulation
© 2020 Published by ECRES

1. INTRODUCTION

Many different materials used in the field of IR conversion. Interest has focused mainly on the wavelengths of
the two atmospheric windows 3–5 and 8–14 m, even though the recent researches are increasing in longer
wavelengths stimulated by space applications. The advance in IR detector technology is connected to the
development of semiconductor infrared detectors. IR systems have found great potential in the enhancement of
the automobile and aircraft safety and automatic supermarket door openers [1-4]. The plasma – semiconductor
structure has an advantage for providing low power consumption, high-speed response and gets high efficiently
discharge light emission from the micro discharge plasma cell where the power consumption of the structure is
lower than the other plasma-cells [1-4]. The semiconductor materials under investigation are attractive because
they can work at room temperature and they have electronic and optical stability up to atmospheric pressure.

2. EXPERIMENTAL

The semiconductor electrode (photodetector) is produced in the form of a disk with a thickness of  1 mm and
a diameter of  30 mm. One of the surfaces of the photodetector faces is in the discharge zone. The transduced
electrical contact is created on the outer surface of the photodetector. A constant or pulsed voltage is applied to
the electrodes of the structure. At an application voltage that exceeds the breakdown voltage of the discharge
range, self-feeding discharge begins. Using a window in the image converter, an IR image is dropped to the
photodetector plane, which changes the local resistance of the photodetector. An increase in local current
density and gas discharge brightness occurs according to the intensity of the incoming IR radiation. Thus, the
IR image can be converted into the visible radiation of the gas; and can be observed through a window in the
converter. Discharge current density can be changed in the area of 10-6 A / cm2 by illuminating the

808
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

semiconductor detector. Thus, images can be recorded in a wide dynamic region of IR radiation intensity [5].
Under these conditions, the current of the system can be controlled either by changing the Uo or by changing
the spatially homogeneous illumination of the semiconductor layer along the transparent electrode on the
surface of the semiconductor. The concentration of freeload carriers as a result of illumination, impurity-band
and band-band electronic transitions.

Figure. 1a. Semiconductor- plasma cell which is the main part of IR image converters. 1. light source, 2. Si filter, 3. IR
light beam, 4. semi transparent Au contact, 5.ZnSe material, 6. gas discharge gap, 7. Mica foil, 8. UV- visible light beam,
9. Transparent conductive SnO2 contact, 10. glass disc.)

3. RESULT AND DISCUSSION

When an electron that gains energy in the electric field reaches slightly larger energy than its ionization
potential, it loses its energy in collisions with gas atoms and ionizes the gas atoms. As a result, two slow
electrons appear. These electrons are accelerated again in the field, ionizing atoms; they produce four
electrons, and this is how it goes.

Gas breakdown is essentially a threshold process. This means that if the electric field exceeds a value that
characterizes a special set of conditions, the breakdown will begin. When the intensity of voltage or
electromagnetic radiation is gradually increased throughout the discharge range, no change in the condition of
the medium can be noticed. At a certain value of voltage or intensity, ionization suddenly increases, the system
detects a current and a glow is observed. Thus, both supply voltage and light intensity can serve as bifurcation
parameters of the physical system under examination. In the experiment, the gas pressure and conductivity of
the semiconductor cathode p are fixed, and the application voltage Uo is gradually increased from 0 V, thereby
increasing the voltage drop in the gas layer. As soon as the Uo reaches the critical voltage for the gas
breakdown; homogeneous ignition of discharge occurs.

Figure 2 ( a, b) shows plasma current curves for 760 Torr and 900 Torr. Interesting results have been observed
for ZnSe electrode. Optimal plasma currents have been seen at higher pressures. Thus , we can conclude that
ZnSe operates better at 900 Torr. The current is recorded around 6x10 -5 A for 760 Torr, and 7x10-5 A for 900
Torr.

Figure 2 (a, b) CVCs for 760 Torr and 900 Torr.

809
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3 shows CVCs for various ZnSe diameter D and also for different discharge gap d. As observed from
the graph, lower electrode distance causes higher current under weak illumination intensity L1. The higher
plasma current is reached for D = 18 mm.

Figure 3. 3D CVC curves in the case of D = 9, 12 and 18 mm as a function of voltage and inter-electrode distance d
According to CVC curves, ZnSe gives higher plasma current for D = 18 mm and d = 50 m because of the high
electron energy in the evolution of the plasma discharge. New plasma trends are focused on stable discharges.
Nevertheless, the optimization of the plasma requires to understand the complex physical and chemical process.
Microdischarge systems with semiconductor electrodes promise the uniform plasma by suppressing the
instabilities in the cell.

3. CONCLUSION

Many semiconductors have been proposed for IR related technology . But, No no one alone met all expectations.
ZnSe is a semiconductor that can overcome most of it. In comparison with GaAs, InP and GaP, ZnSe has a better
IR response. Detectors must be cooled to increase the Infrared detection. But, ZnSe can operate at room
temperatures up to atmospheric pressures even beyond it. New plasma trends are focused on stable discharges.
Nevertheless, the optimization of the plasma requires to understand the complex physical and chemical process.
Microdischarge systems with semiconductor electrodes promise the uniform plasma by suppressing the
instabilities in the cell.
REFERENCES

[1] Peter Y. Yu and M. Cardona, Fundamentals of Semiconductors Physics and Materials Properties (Fourth Edition,
Springer: 2010)
[2] O. Madelung, Semiconductors: Data Handbook. (3rd ed., Springer, 2004, pp. 736-757)
[3] A. Rogalski, Infrared Detectors, 2nd ed (Boca Raton: CRP press, 2010), p.309
[4] Y. Y. Peter, C. Manuel, Fundamentals of Semiconductors Physics and Materials Properties, ( Fourth Edition, Springer,
2010), p.1
[5] Astrov, Yu. A., Beregulin, E. V., Valov, P. M., Paritskii, L. G., Portsel, L. M., S. Ryvkin, M. and Sreseli, O. M.,
“Photographic evice for high-speed measurement of the spatial distribution of the output from a pulsed C02 laser”,
Sov. Phys. Tech. Phys., 23: 231-238 (1978).

810
 STEP BY STEP MODELLING, ANALYSIS AND SIMULATION
OF A HIGH-EFFICIENCY BATTERY CONTROL SYSTEM BY
USING MATLAB ENVIRONMENT PROGRAM

Mohammed Ayad Alkhafaji

Karabuk University, ,Karabuk, Turkey, [email protected], ORCID: 0000-0001-9552-1332

Yunus Uzun
Aksaray University, Aksaray, Turkey, [email protected], ORCID: 0000-0002-1412-8473

Alkhafaji MA, Uzun,Y Step By Step Modelling, Analysis And Simulation of A High-Efficiency
Cite this paper as: Battery Control System by Using Matlab Environment Program. 8th Eur. Conf. Ren. Energy
Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: This paper explains step by step modeling and simulation a full circuits of battery control system
and connected together starting from the AC input source to the battery control and storage system.
The three-phase half controlled rectifier has designed to control and convert the AC power into
DC power. In addition, two types of direct current converters have used in this paper which are a
buck and bidirectional DC\DC converters. These systems regulate the output voltage to lower or
higher levels from the input voltage. There is a main switch that operates on conduction or cut-off
mode in buck converters. The switch is triggered by a pulse-width modulated (PWM) signal. The
duty cycle of the PWM signal is calculated by the ratio of output and input voltage levels. So, the
duty cycle indicates the operation mode of the converter in steady-state operation. In this study,
analysis and control of a buck converter are investigated with the PWM signal. Besides, the
bidirectional DC\DC converter has achieved and optimized by PI control methods to control the
battery charging and discharging modes. The simulation has applied via Matlab/Simulink
environment. Hence, the result will show the activity of each part of the designed circuits starting
from the converters and the battery control system in charge and discharge modes.
Keywords: Three-phase half controlled rectifier, Buck converter, bidirectional DC\DC converters, PI
controller, pulse-width modulation (PWM), Battery controller.
© 2020 Published by ECRES

1. INTRODUCTION

The present electronic gadgets have higher portability and are greener than at any time in recent memory. Battery
headways are energizing this movement in a wide scope of items from convenient force apparatuses to module
half breed electric vehicles and remote speakers With ongoing headways, you can buy a lithium-particle battery
to kick off your vehicle, and it just gauges two or three pounds and is the size of your hand. The continuous
change of battery innovation has incited numerous newcomers to find out about planning battery the board
frameworks. This article gives an apprentice's manual for the battery of the executive's framework (BMS)
engineering, talks about the major useful parts, and clarifies the significance of each part to the battery control
of the executive's framework [1-4].

Moreover, Three-phase fully controlled converters are exceptionally mainstream in numerous mechanical
applications especially in circumstances where power recovery from the dc side is basic. It can handle sensibly
high force and has satisfactory input and output consonant contortion. The design additionally fits simple series
and parallel association for expanding voltage and current rating or improvement in symphonious conduct. In
any case, this adaptability of a three-phase fully controlled converters is acquired at the expense of expanded
circuit intricacy because of the utilization of six thyristors and their related control circuit [5,6]. This intricacy
can be impressively decreased in applications where power recovery isn't essential. All things considered, three

811
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

thyristors of the top gathering or the base gathering of a three-phase fully controlled converter can be supplanted
by three diodes [4-7]. The subsequent converter is known as a three-phase half controlled converter. Supplanting
three thyristors by three diodes diminishes circuit multifaceted nature and yet forestalls negative voltage showing
up at the output whenever [7,9]. Hence the converter can't work in the rearranging mode. The three-phase half
controlled converter has a few different points of interest over a three-phase fully controlled converter. For a
similar terminating edge, it has a lower input side uprooting factor contrasted with a fully controlled converter.
It additionally broadens the scope of consistent conduction of the converter. It has one genuine detriment, be
that as it may [5-11]. The output voltage is periodic more than 33% of the input cycle as opposed to one-6th just
like the case with fully controlled converters. This suggests both input and output music are of lower recurrence
and require heavier separating. Hence, half controlled three-phase converters are not as famous as their fully
controlled partner [10,12]. Despite the fact that, from the perspective of development and circuit intricacy the
half controlled converter is less difficult contrasted with the fully controlled converter, its examination is
extensively progressively troublesome.
Direct current converters in power electronic circuits are being used in a large application area from home
appliances, automotive, smartphones and tablets to information technologies. Efficiency has become the most
critical parameter in designs according to international quality standards [13]. Depending on these expectations,
converters are shaped through low losses and high efficiency. The idea of the design of a control algorithm is to
keep the output voltage at the desired level in response to load and line variations. Various control actions have
been defining for converters. As a control perspective, control actions that used power converters are classified
as the linear controllers and nonlinear controllers. The linear controller design is depended on frequency
response and root locus analysis. To design the controller, first, the mathematical model of the converter is
obtained then, the linear controller is designed based on the classical control system approach. The linear
controllers are designed based on the small-signal model of the converter [14]. Thus, the linear controllers do
not have a satisfactory response under varying operation points such as load and line variations. To achieve good
performance under the change in the operation points, a more accurate model is necessary. But the model
becomes too complex and another solution is to use a trial and error controller [5-19]. To control of the buck
converters, various nonlinear methods are PI Controller coefficients were determined by the genetic algorithm
by using the closed-loop design on a single switch type in the simulation environment. Hence, this paper includes
multi-design to convert the current and voltage in two parts AC/DC and DC/DC with various strategies. Besides,
the PI controlled has utilized to and connected the bidirectional [8-16].

2. THE THREE PHASE HALF CONTROLLED RECTIFIER

Traditionally, the units of energy conservation in any electric device dealing with battery are equipped with
direct current, and the main sources of equipment are running on alternating current. So it is necessary to transfer
this energy from the main source to the system of energy conservation in electrical electronics devices and here
comes the importance of using the transformer [4,5,12,13]. In this design, the three phase half controlled rectifier
has using to convert the AC into DC source. So the average DC value of the output voltage waveform from the
three phase half controlled rectifier is given in Eq. [1] below:
3√3
𝑉𝐷𝐶 = 𝜋 𝑉𝑠 × 𝐶𝑂𝑆(𝛼) (1)
The design of the three-phase rectifier depends on three thyristors and three diodes are connected as shown in
figure 1. In addition, each of the thyristor and the diodes have the same parameters as shown in table 1 below;

812
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 1. The simulation design of the three phase rectifier.

Table 1. The Thyristors parameter in the three phase rectifier

Parameters value
Resistance Ron (Ohms) 1
Forward voltage Vf (V) 0.8
Snubber resistance Rs (Ohms) 500
Snubber capacitance Cs (F) 5×10-6

Besides, the three Thyristors are supported by three pulse generator systems as shown in figure 1 to control the
pulse width of the output signals of the three phase rectifier and the parameters of the pulse generator are
explained in the Table 2. Moreover, to control the voltage and current of the three-phase rectifier has connected
it to resistor with 10 ohm.

Table 2. The pulses parameters of the three phase rectifier

Parameters Pulse 1 Pulse 2 Pulse3
Amplitude 10 10 10
Period (secs) 0.02 0.02 0.02
Pulse Width 10 33.34 63
Phase delay (secs) 0.0025 0.0025 0.0025

The three-phase circuit has a constant voltage and frequency it can be used through a rectification circuit to
produce and grant a constant DC voltage strength, which can then be filtered ensuing in an output DC voltage
with much less ripple in contrast to a single-phase rectifying circuit as that explained in figure 2. In addition, the
input three-phase power source that feeds three-phase rectifier has simulated as the normal powers that available
in many countries and used to charge the electric vehicles which it consists a 220 voltage of Peak amplitude and
50 Hz of the frequency with a (0 degree) angle for the phase A, (-130 degree) angle for the phase B and (130
degree) angle for the phase C.

Figure 2. Rectification of three phase supply using thyristor.

813
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3. THE BATTERY DESIGN

Generally, one of the most important parts that determine the efficiency of any electric-electronic device is the
cell and its management, it is considered the soul of the device. Besides, the batteries of any electrical device
vary according to the type, capacity and its manufacturer [1-4].
Consequently, the systems controlling the batteries also differ from one device to another, and there are a
number of different strategies and methods used previously to manage the battery and provide the necessary
electric power that needs under different operation conditions [16,17]. This paper has simulated the normal
battery that has used in several fields like electric vehicles, fans, and other electronic devices which is the
Lithium-Ion battery. Besides, to test the functionality and the effectiveness of using the three-phase full
connected rectifier with the buck converter and the bidirectional DC/DC converter that controlled by using PI
controller as shown in figure 3.
In addition, the battery that has used in this paper contains a nominal voltage equal to 24 V and rated capacity
equal to 50 Ah. The efficiency of any batteries could determine by using the state of charge (SOC), which is the
level of charge of an electric battery relative to its capacity. Generally, the units of SOC are percentage points
(0% = empty; 100% = full) as an ideal value. Thus, the simulation of the battery has tried to reach the maximum
point of charging about 100% and show the effectiveness of the voltage and current at the load which acts like
the parts of the electric device. Besides, the parameters of the battery cell that has used in this paper are shown
in Table 3 below;
Table 3. The battery cell parameters
The battery parameters The value
Maximum capacity (Ah) 50
Cut-off Voltage (V) 18
Fully charged voltage (V) 27.9357
Nominal discharge current (A) 21.7391
Internal resistance (Ohms) 0.0048
Capacity (Ah) at nominal voltage 45.2174

4. THE MAIN SYSTEMS SIMULATION AND RESULTS

This part show and explain the component of the battery control circuit system. Besides, this simulation will
explain the work and effective of each part of the system individually. Therefore, show and explain the full
system as shown in figure 3 with the result.

Figure 3. The battery control system simulation design.

814
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

A- The three phase half controlled rectifier

The three-phase rectifier half connected rectifier has fed directly from the input three phase AC source with
220V for each phase as shown in figure 4. The work of the rectifier is depended on the input parameters of it
design as explained previously in the design part [14,15].

Figure 4. The input voltage of the three-phase rectifier.

Figure 5. The output voltage of the three-phase rectifier in two cases.

The simulation result of the three phase rectifier output voltage as shown in figure 5 will be equal to 180 V in
DC. Besides, the results shows there are two cases of the output could be found; normal signal and effected
signal. the first one that started from (0 to 2.5 sec) which is effected by the PWM signal that generated by the
buck converter and this signal has many properties to use in this system in charging mode to supply the battery
with high band of voltage. In addition, the another signal can be found as shown in the simulation figure 5 that
started from (2.5 to 5 sec) is the normal output signal for the three phase half controller rectifier, and this signal
will be less characteristics than the previous one because of the low bandwidth and this signal has founded in
case of turn of the buck convertor.

B- The buck converter

The top-level test bench is created to simulate the control output buck design. By starting to build the test bench,
it is required to choose the output capacitance and the inductance as a first step since they are put into the test
bench as ideal components [8-17]. A capacitor with 300μF, L = 150μH typical capacitance, and inductance value
is chosen.as seen in a figure 6 the PWM voltage signal acts as Continuous Current Mode (CCM) the conduction
loss is reduced and allows the bidirectional inductor current flow. Thus, the buck converter always maintains in
CCM. The buck converter has higher efficiency at full load because of the forward voltage drop of the diode.
The simulation graph is shown two figure, the first is the current that crossing of the buck converter and it's
about 35 A. The second figure is shown the voltage that crossing through the buck converter and it's about 45
V. in addition, the result has shown how the output signal of the three phase rectifier has affected by the PWM
signal through the buck converter and who It has become more bandwidth without losing any of its properties,
and we notice this in the miniature shape of both current and voltage waves inside the scheme.

815
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 6. PWM buck converter capacitance voltage and Inductance current.

C- The Bidirectional DC/DC converter

Generally, there are two cases for any battery status in the electric or electronic device with respect to the load
of the device, which is charging and discharging modes. Technically, when the voltage source is disabled, the
battery will supply the load and when the voltage source is enabled, the battery will charge and the load will be
supplied from the voltage Source [3]. Besides, the power that feeds the battery in charging mode or the power
that feeds the load in discharge mode must be controlled to be suitable with the battery requirements without
damages [13-16]. The DC/DC converter is responsible for control the battery status by connected the converter
with a control system using a PI controller after trial and error searching [8-19]. The aim of using a PI controller
is to control the output and be as constant values for the systems. In addition, the efficiency of any battery control
system is reaching the battery charging from (0 to 100 %) Soc level in a short time. Besides, in the discharge
mode battery drain should be carrying the load requirements of the device for a longer time. Hence, to reach this
aims there are several strategies that have applied to the system to be as efficient as possible.

D- The battery charging and discharging modes

Current control strategy:

Basically, the batteries are charging control in two modes: Constant current control and constant voltage control
methods. Technically, in each battery management system should check the allowed maximum current for the
battery that connected to the system [16-19]. In this design, as shown in Table 3, the allowed maximum current
for the battery is 21.7391 A as the maximum required to charge. Besides, some battery companies separated the
constant current control and constant voltage control sides at 80% SOC level or 100% SOC level that depended
on the battery manufacturing requirements and the scientific strategies that have applied it in the system to get
high efficiency. Therefore, the open circuit has applied to determine the required voltage to the maximum level
of charging and the result is 28.04 volts.
In addition, the current control strategy has utilized to test the efficiency of the battery charging mode [17,18,19].
This strategy has done by using the PI controller with pulse width modulation generator (DC-DC) as shown in
figure 7 that connected to the battery management system as shown in the figure 3 directly with two Mosfets.
Besides, the PWM parameters that used in the simulation are; 10e3 Hz for Switching frequency and 5e-6 for
sample time. Generally, the PI controller is a manual controller and it has used to test the functionality of the
system. Therefore, by using a constant simulation block next to the PI controller and the input values are
enforced as they are respectively (-5, -25, -40, 0, 25) and it has started from minus part because of the battery in
the charging mode.

816
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 7. The simulation of the current control strategy.

Figure 8. The change of the current value during current control strategy.

Figure 9. The change of the voltage value during current control strategy.

Figure 8 explained the current control strategy that has applied to the system. Besides, the values that have
entered to the PI controller are acting the red in the simulation and it's clear to notice that when the red line
change depending on the input parameters, the current value is change also, which mean, the input value change
the current direction will change also depending on the current control strategy.
Moreover, when taking a look at figure 9 it explains the change of voltage values in the charging mode because
the current is inversely proportional to the voltages. Therefore, the voltage values in the charging mode are
decreases and in the discharge mode when the load feed by the battery and the current values are increasing, the
voltage values are decreases.

Voltage based control strategy:

The voltage-based control strategy has applied to the system to determine the reference battery current that needs
to charge the battery and also to show the eligibility for charging the battery SOC level into 100 % [18,19]. In
addition, another PI controller has used and applied to determine the reference battery current and it's connected
with the previous PI controller. Which mean, the first PI will control the current and the second PI will control
the voltage with respect to the maximum charging voltage of the battery as shown in figure 10 the maximum
charging voltage that determined by the open circuit is 28.04 V and it has used as the input parameter for the
voltage control PI.

817
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 10. The simulation of voltage control by using PI controller.

The Load control:
The load in any electrical device is considered as an important part because it's depended on the result of all
electrical-electronic working parts of the device. Besides, the load of any device has a known range of voltage
that should be fed by the converter of the electrical device without losses because of its effect directly on the
device working and efficiency [2-10]. In this simulation, another PI controller has used in the battery control
magnet to control the load and insure to deliver the needed voltage to the load of the device. The minimum
voltage that assumed in this simulation is 48 at the discharge mode. Therefore, the load could be carried a voltage
between (48 to 220) volt in DC, which depended on the type of the device on the load. Moreover, to open and
closed the voltage gates between charge and discharge mods there a switch control has designed and is
considered as a master gets for all parts of the battery control system to control the load voltage. The load control
PI with the switch control is shown in figure 11 below.

Figure 11. The simulation of the load control PI with the switch control.

5. CONCLUSION

In this study, the importance of the efficiency of buck converters and bidirectional DC/DC converter in a portable
system and methods for increasing efficiency is explained in a switching buck converter. Power loss mechanisms
and efficiency increasing methods for light load and heavy. The previous works in the literature to improve
efficiency such as Pulse width Modulation (PWM). The digital controller provides the capability to implement
a flexible control algorithm. The nonlinear PI function with trial and error method power coefficients which can
be easily applied in the digital controller to the perturbations much more quickly than the conventional and has
a less steady-state error. At the charge mode, the battery can be affected by two signals which are the half
controlled rectifier and the PWM signal. Therefore, the battery will be powered up in a short and perfect time.
At the discharge mode, the battery output will support also the rectifier signal by the PWM signal but it will be
lower amplitude and frequency form the charging mode status. Therefore, the power will be discharged in an
ideal long time and the reason for supporting the output signal of the battery via the rectifier signal even there
has no input power source, because the full system working with a close loop hence each circuit support feed
each other.

818
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] Lawder, M. T., Suthar, B., Northrop, P. W., De, S., Hoff, C. M., Leitermann, O., Subramanian, V. R Battery energy
storage system (BESS) and battery management system (BMS) for grid-scale applications. Proceedings of the IEEE,
2014; 102(6), 1014-1030.
[2] Chakroun, J., Hamam, H. General Study for energy recovery from used batteries using fuzzy logic and PI controllers.
International Journal of Intelligent Engineering Informatics, 2018: 6; 308-324 <https:/dl.acm.org/doi/10.555/3271875>.
[3] Hemmati, R., Azizi, N. Advanced control strategy on battery storage system for energy management and bidirectional
power control in electrical networks. Energy, 2017; 138, 520-528 < https:// DOI: 10.1016/j.energy.2017.07.071>.
[4] Zurfi, A., & Zhang, J. Exploitation of Battery Energy Storage in Load Frequency Control -A Literature Survey.
American Journal of Engineering and Applied Sciences 2016: 9; 1173-1188 <https://
DOI:10.3844/ajeassp.2016.1173.1188>.
[5] Singh, A. K., Jeyasankar, E., Das, P., Panda, S. K. A matrix-based nonisolated three-phase AC–DC rectifier with large
step-down voltage gain. IEEE Transactions on Power Electronics 2016: 32; 4796-4811
<https://ieeexplore.ieee.org/document/7527663>
[6] Mallik, A., Ding, W., Shi, C., Khaligh, A. Input voltage sensorless duty compensation control for a three-phase boost
PFC converter. IEEE Transactions on Industry Applications 2016: 53; 1527-1537
<http://ieeexplore.org/document/7738511>.
[7] Alkhafaji MA, Uzun Y. Design and Speed Control of SynRM Using Cascade PID Controller with PSO Algorithm.
International Journal of Renewable Energy Development. 18 February 2020, pp.69-76
[8] AlKhafaji MA, Uzun Y. Simulation and control of an electric vehicle by using PSO and specify driving route topology.
In 2019 International Conference on Power Generation Systems and Renewable Energy Technologies (PGSRET) 26
Augusts 2019: IEEE, pp. 1-6.
[9] Khan, A., Lu, Y. W., Khan, U., Wang, L., Eberle, W., Agamy, M. ovel. Transformer less Buck-Boost Inverters without
Leakage Current. IEEE Transactions on Industrial Electronics 2020 < https://ieeexplore.ieee.org/document/8948311>.
[10] Kobaku, T., Jeyasenthil, R., Sahoo, S., Ramchand, R., Dragicevic, T. Quantitative Feedback Design Based Robust PID
Control of Voltage Mode Controlled DC-DC Boost Converter. IEEE Transactions on Circuits and Systems II 2020
<https://ieee.org/document/9069243>
[11] Rameshbabu, A., V. Sivachidambaranathan. PV-Based Multiple-Input Single-Output DC–DC Luo Converter for
Critical Load Application. Cognitive Informatics and Soft Computing. Springer, Singapore, 2020.pp. 685-701.
[12] Shestopal, S. S., Lyubashits, V. Y., Astakhov, V. V., Pismennaya, E. E., Ryazantsev, S. V., Fedulov, A. M., and
Barsukov, P. V. Features of building control systems sub-locality in modern russia. International Review of Management
and Marketing 2016: 6; 78-83 <https://search.proquest.com/docview/1762628258?accountid=25087>.
[13] Iqbal, M. T., Tariq, M., Ahmad, M. K., Arif, M. S. B. Modeling, analysis and control of buck converter and Z-source
converter for photo voltaic emulator. In 2016 IEEE 1st International Conference on Power Electronics, Intelligent
Control and Energy Systems (ICPEICES); 4-6 July 2016: IEEE, pp. 1-6.
[14] Hernández-Márquez, E., Silva-Ortigoza, R., Dong, S. H., Garcia-Rodriguez, V. H., Saldana-Gonzalez, G., Marcelino-
Aranda, M. A new DC/DC Buck-Boost converter-DC motor system: Modeling and simulation. In 2016 International
Conference on Mechatronics, Electronics and Automotive Engineering (ICMEAE): 22-25 Nov. 2016: IEEE, pp. 101-
106.
[15] Khubalkar, S., Chopade, A., Junghare, A., Aware, M., Das, S. Design and Realization of Stand-Alone Digital
Fractional Order PID Controller for Buck Converter Fed DC Motor. Circuits Syst Signal Process 2016; 35: 2189–221
<https://doi.org/10.1007/s00034-016-0262-2>.
[16] Vigneysh, T., Kumarappan, N. Autonomous operation and control of photovoltaic/solid oxide fuel cell/battery energy
storage based microgrid using fuzzy logic controller. International journal of hydrogen energy 2016: 41(3); 1877-1891
https://doi.org/10.1016/j.ijhydene.2015.11.022.
[17] Ayad, A., Karamanakos, P., Kennel, R. Direct model predictive current control strategy of quasi-Z-source inverters.
IEEE Transactions on Power Electronics 2016: 32(7); 5786-5801< https://ieeexplore.ieee.org/document/7570193>.
[18] LU Gai-feng, DU Shuai, HOU Peng-fei, ZHANG Peng-fei. Adaptive Current Control Strategy for Harmonic
Compensation in Single Phase Solar Inverter. Journal of University of Electronic Science and Technology of China
2018; 47(5): 705-713 < doi: 10.3969/j.issn.1001-0548.2018.05.011>.
[19] Merai, M., Naouar, M. W., Slama-Belkhodja, I. An improved DC-link voltage control strategy for grid connected
converters. IEEE Transactions on Power Electronics 2017: 33(4); 3575-3582 <
https://ieeexplore.ieee.org/documeVOnt/7932921>.

819
 PERFORMANCE STUDY OF THE CELL-SET MODEL FOR A
GURNEY FLAP IMPLEMENTATION

Alejandro Ballesteros-Coll
University of the Basque Country (UPV/EHU), Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0002-
6563-2434

Unai Fernandez-Gamiz
University of the Basque Country (UPV/EHU), Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0001-
9194-2009

Iñigo Aramendia
University of the Basque Country (UPV/EHU), Vitoria-Gasteiz, Spain, iñ[email protected], ORCID: 0000-
0002-4960-2729

Ekaitz Zulueta
University of the Basque Country (UPV/EHU), Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0001-
6062-9343

Jose Manuel Lopez-Guede

University of the Basque Country (UPV/EHU), Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0002-5310-
1601

Ballesteros-Coll, Alejandro, Fernandez-Gamiz, Unai, Aramendia, Iñigo, Zulueta, Ekaitz,

Cite this paper as: Lopez-Guede, J.Manuel. Performance study of the Cell-set model for a Gurney Flap
Implementation. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: In recent years, advances in weight and dimensions of wind turbines have led to the optimization
and advancement of flow control devices. The current work presents a novel method for the
modelling of flow control devices, for Gurney Flaps in particular, named as the Cell-set model.
This innovative model is based on the reusability of the cells around the corresponding geometry
to create a new wall region. Therefore, the flexibility of the Cell-set model provides a versatile
way to implement new geometries on an already generated mesh without having to recalculate the
meshing. In order to validate the Cell-set model, a performance study was carried out on a
DU91W250 airfoil for different Gurney Flap lengths (from 0% to 2% of chord length), by
contrasting the results obtained with the ones of a fully-meshed model. CFD simulations were
performed and RANS equations with a Re = 2 ∙ 106 were introduced as the main physics. Major
resulting data shows that a maximum relative error of 3.715% and a global error of 1.13% was
reached. Subsequently, it is regarded that the model is sufficiently accurate for its implementation
on other computational domains.
Keywords: flow control; wind turbine; aerodynamics; Gurney flap; Vortex generators
© 2020 Published by ECRES
Nomenclature
GF Gurney Flap
AoA Angle-of-attack
c Airfoil chord length
VG Vortex Generator
hGF Gurney Flap Length
CFD Computational Fluid Dynamics
HAWT Horizontal Axis Wind Turbine
CL Lift coefficient
CD Drag coefficient

820
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

1. INTRODUCTION

In these times, Computational Fluid Dynamics (CFD) play an essential role in terms of wind turbine
optimization. Cai et al. analyzed the unsteady aerodynamics for a full-scale HAWT by means of CFD [1].
Studies from Saenz-Aguirre et al. [2] show how the implementation of flow control devices on HAWT turbines
enhance their aerodynamic performance significantly. As the dimensions and loads of wind turbines are
increasing [3,4], the aerodynamic loads of the blades require higher optimization. Vortex Generators (VGs) are
typically used passive flow control devices for the improvement of the airfoil’s performance. These are small
vanes, generally triangular, and are located in pairs close to the leading edge of an airfoil [5,6]. Studies show
that the implementation of VGs on 2.5 MW HAWT can enhance the power output to 15.2% [7]. The usefulness
of other flow control devices such as Gurney Flaps (GFs) has also been researched. These are passive L-shaped
devices, usually located perpendicular to the pressure side of an airfoil and collinear with the trailing edge [8].
Alber et al. [9] studied the effects of GFs on different aerodynamic profiles for a range of GF lengths, where the
increase on performance was optimized from 0.8% to 2%.

The principal objective of the present manuscript is to show the resulting data for different GF length
implementation on a DU91W250 airfoil, by means of the Cell-set novel method. Furthermore, comparative
results with a fully-meshed model are displayed. The following layout is introduced: firstly, the Cell-set model
and its main characteristics are explained. Secondly, the formulation of the numerical simulation setups is
described. After this, the main results are represented by graphical and numerical data. Finally, the more
noteworthy conclusions are explained.

2. SECTIONS

Presenting the Cell-set model

This study presents a novel method to model flow control devices named as the Cell-set model. Essentially, this
model reuses the cells (structured elements) which are around the desired geometry but without having to re-
mesh the computational domain. A selection of the required cells is performed and once a set of finite elements
is generated, a new region is assigned to the mentioned structure. This region will act as if it was an additional
construction of the computational domain, following the same equations applied for the rest of the domain. The
Cell-set method provides flexibility and manageability when new geometries have to be added to an already
generated mesh and the smoothness of a surface is not critical for its correct performance. Figure 1 represents
the application of the model for a GF implementation on a 2D mesh.

(a) (b)
Figure 1. Cell-set region selection on a DU91W250 airfoil trailing edge (a) clean profile, (b) GF Cell-set.

821
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Cell-set performance validation setup

In the present work, the Cell-set source-term model has been used for a GF implementation on the pressure side
of a DU91W250 airfoil. This airfoil has been selected since it is part of the most referenced HAWT developed
by NREL, see Jonkman et al. [10]. The effects of the Cell-set model were studied by means of STAR CCM+
v14.02.012 CFD commercial code [11].
The following physics were applied to the numerical setups: A Reynolds number of Re = 2 ∙ 106 and Reynolds-
Averaged Navier-Stokes (RANS) equations were introduced in all the scenarios. The kinematic viscosity of the
air was set in 𝜈 = 1.52 ∙ 10−5 𝑚 2 /𝑠 and a dynamic viscosity of μ = 1.85 ∙ 10−5 Pa·s was introduced. A free
stream velocity value of U∞ = 30 m/s was introduced. The shear stress turbulence model (SST) formulated by
Menter [12] was used. The SST model could be defined as a combination of two models: K-Ω and K-E, where
the calculations performed on refined regions are carried out by using the K-Ω turbulence and as the mesh is
enlarged, a gradual change to the K-E model is applied.
All the meshes used in the numerical simulations were O-shaped and two-dimensional. In order to calculate the
radius of the mesh, as Sørensen et al. [13] suggest in their work, it was set to be 42 times the chord length of the
airfoil. The DU91W250 has a chord of 1 m, hence, the radius of the mesh was of 42 m. The mesh had a
computational domain of 65348 structured cells. Additionally, mesh-refinement was applied to the whole airfoil
area, as well as to the airfoil trailing edge since GFs were to be included there. The normalized height of the first
structured element (divided by the chord length of the airfoil) was detrermined as Δz/c of 1.351 ∙ 10−6 . A
second order linear upwind scheme was defined for the mesh discretization. The surface area of the DU91W250
airfoil was set as a non-slippery boundary. See Figure 2 where the mesh is illustrated, focused on the
aerodynamic profile.

Figure 2. Mesh representation focused on the DU91W250 airfoil.

With the purpose of validating the performance of the Cell-set model, different simulations were carried out: the
GF length (hGF) varied from 0.25% to 2% of the chord length and the angles-of-attack from 0⁰ to 5⁰. This ranges
were selected based on the work made by Aramendia et al. [14] where the resulting data showed that when AoA
values higher than 5⁰ were introduced a GF implementation is adverse to the aerodynamic performance of the
airfoil. Additionally, the same clean airfoil mesh was employed to introduce the model. The mesh was validated
with experimental and CFD data.

Results for Cell-set performance

The main point of this section is to show the performance results for the validation of the Cell-set model on the
implementation of a GF, for a DU91W250 airfoil. In the interest of having comparative data, the values of a
fully-meshed model calculated by Aramendia et al [14] were used. The evaluation has been carried out with lift
and drag coefficient relationship, commonly known as lift-to-drag ratio (CL/CD) or the aerodynamic performance
of an airfoil. The lift and drag coefficients were calculated for 48 different cases.
Firstly, two lines per plot were displayed (see Figure 3): blue lines with triangular marking points show the
calculated CL/CD values for each GF length at the corresponding AoA while the black lines with triangular
markers show the values of a fully-meshed model.

822
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Comparative results between the Cell-set model and a fully-meshed model.

On the one hand, Figure 3 clearly shows how from 0⁰ to 3⁰ of AoA the CL/CD values are higher than the ones
reached with a clean profile. Nevertheless, for AoAs higher than 3⁰, part of the curves are below the clean line.
At 5⁰ of AoA both fully-meshed and Cell-set curve values are lower than the ones generated by an airfoil that
has no flow control devices. On the other hand, it is clearly illustrated which is the trend followed by the curves.
Initially, at 0⁰, the implementation of a GF is appropriate since the CL/CD ratio is enhanced for all the
implemented GF lengths. Afterwards, the curves start to point down and to go through the clean value. At 4⁰ of
AoA, for hGF larger than 1% of c, the implementation of a GF results to be detrimental for the aerodynamic
performance of the airfoil.
Another point to take into consideration is how the values reached by the Cell-set model and the fully-meshed
model follow a similar path. With the purpose of achieving accurate data, error calculations have been carried
out: Firstly, the relative error per case was calculated based on Eq. [1], see Table 1. Secondly, average error
values were determined by means of Eq. [2]. Subsequently, the global error for all the cases was calculated by
reaching a value of 1.13% (see Eq. [3]). Additionally, 3.72% was the maximum error generated by the model.

823
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

CL
min( i)
CD
i
eri = (1 − CL ) ∙ 100 (1)
max( i )
CD
i

∑𝑁=6
𝑖=1 𝑒𝑟 𝑖
𝑒𝑎𝑣𝑔𝑗 = (2)
𝑁

∑𝑁=8
𝑗=1 𝑒𝑎𝑣𝑔 𝑗
e𝑔 = 𝑁
(3)

Table 1. Relative error percentages for the Cell-set model compared with a fully-meshed model.
Gurney Flap Length (% of chord)
AoA (⁰) 0.25 0.5 0.75 1 1.25 1.5 1.75 2
0 0.478 1.391 0.031 0.581 1.816 0.134 1.566 0.082
1 0.325 1.310 0.159 0.913 2.990 0.253 2.248 0.422
2 0.224 2.295 1.561 1.082 3.520 0.392 2.618 0.629
3 0.120 1.293 0.216 1.082 3.720 0.295 2.753 0.687
4 0.680 1.885 0.106 1.053 3.635 0.166 2.664 0.612
5 0.527 1.855 0.016 0.890 0.118 0.095 2.425 0.429
Av. error [%] 0.392 1.672 0.348 0.933 2.632 0.222 2.379 0.477

CONCLUSIONS

The current work studies the performance of the Cell-set model for a GF implementation on the DU91W250
aerodynamic profile. The basis of this model is the reusability of the structured elements of an already generated
mesh. Consequently, there is dependency on the mesh refinement since the same finite elements will be reused.
The model reproduces an approximated geometry to the desired one, without a new calculation of the whole
computational domain. This provides flexibility and versatility to the user as the main dimensions and
geometries can be easily modified.
In order to determine the accuracy of the Cell-set model, CFD numerical simulations were performed and
comparative calculations with a fully-meshed model were carried out. Six different AoAs and a range between
0.25% to 2% of c were established, by providing a total of 48 scenarios according to the mentioned data. The
results showed that the method was precise enough for its implementation in other computational domains since
a global relative error of 1.13% was determined.
Future studies are being prepared on this topic, where the use of the Cell-set model for different flow control
devices implementation is thought to be performed: micro-tabs, Vortex Generators etc. Moreover, three-
dimensional simulations are being prepared with the purpose of analyzing the effects of the method on a 3D
scenario.

ACKNOWLEDGMENT

The present research has been developed under the frame of the Joint Research Laboratory on Offshore
Renewable Energy (JRL-ORE). The authors are grateful to the government of the Basque Country and the
University of the Basque Country UPV/EHU through the SAIOTEK (S-PE11UN112) and EHU12/26 research
programs, respectively.

REFERENCES

[1] Cai, X., Gu, R., Pan, P., Zhu, J., Unsteady aerodynamics simulation of a full-scale horizontal axis wind turbine using
CFD methodology. Energy Conversion and Management 2016, 112, 146–156.
[2] Saenz-Aguirre, A., Fernandez-Gamiz, U., Zulueta, E., Aramendia, I., Teso-Fz-Betono, D., Flow control based 5 MW
wind turbine enhanced energy production for hydrogen generation cost reduction. International Journal of Hydrogen
Energy 2020, S0360319920300811.
[3] Pechlivanoglou, G., Nayeri, C.N., Paschereit, C.O., Performance Optimization of Wind Turbine Rotors with Active

824
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Flow Control (Part 1). Mechanical Engineering 2012, 134, 51–51.

[4] Pechlivanoglou, G., Nayeri, C.N., Paschereit, C.O., Performance Optimization of Wind Turbine Rotors with Active
Flow Control (Part 2). Mechanical Engineering 2012, 134, 55–55.
[5] Velte, C.M., Hansen, M.O.L., Investigation of flow behind vortex generators by stereo particle image velocimetry on a
thick airfoil near stall. Wind Energy -Chichester- 2013, 16, 775–785.
[6] Godard, G., Stanislas, M., Control of a decelerating boundary layer. Part 1: Optimization of passive vortex generators.
Aerospace Science and Technology 2006, 10, 181–191.
[7] Miller, G.E., Comparative performance tests on the Mod-2, 2.5-mW wind turbine with and without vortex generators.;
Cleveland, OH, United States, 1995.
[8] Kumar, P.M., Samad, A., Introducing Gurney flap to Wells turbine blade and performance analysis with OpenFOAM.
Ocean Engineering 2019, 187, 106212.
[9] Alber, J., Pechlivanoglou, G., Paschereit, C.O., Twele, J., Weinzierl, G., Parametric Investigation of Gurney Flaps for
the Use on Wind Turbine Blades. In Proceedings of the Volume 9: Oil and Gas Applications; Supercritical CO2 Power
Cycles; Wind Energy; American Society of Mechanical Engineers: Charlotte, North Carolina, USA, 2017; p.
V009T49A015.
[10] Jonkman, J., Butterfield, S., Musial, W., Scott, G., Definition of a 5-MW Reference Wind Turbine for Offshore System
Development; 2009; p. NREL/TP-500-38060, 947422;
[11] Siemens Star CCM+ Version 14.02.012 Available online: https://www.plm.automation.siemens.com/global/en/
(accessed on Feb 3, 2020).
[12] Menter, F.R., Two-equation eddy-viscosity turbulence models for engineering applications. AIAA Journal 1994, 32,
1598–1605.
[13] Sørensen, N.N., Mendez, B., Munoz, A., Sieros, G., Jost, E., Lutz, T., Papadakis, G., Voutsinas, S., Barakos, G.N.,
Colonia, S., et al., CFD code comparison for 2D airfoil flows. J. Phys.: Conf. Ser. 2016, 753, 082019.
[14] Aramendia, I., Fernandez-Gamiz, U., Zulueta, E., Saenz-Aguirre, A., Teso-Fz-Betoño, D., Parametric Study of a
Gurney Flap Implementation in a DU91W(2)250 Airfoil. Energies 2019, 12, 294.

825
 LARGE EDDY SIMULATION OF A RECTANGULAR VORTEX
GENERATOR ON A FLAT PLATE WITH AN INCIDENT ANGLE
OF 25º

Iosu Ibarra-Udaeta
University of the Basque Country, UPV/EHU, Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0001-8112-
9030

Koldo Portal-Porras
University of the Basque Country, UPV/EHU, Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0003-4747-
963X

Unai Fernandez-Gamiz
University of the Basque Country, UPV/EHU, Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0001-
9194-2009

Iñigo Aramendia
University of the Basque Country, UPV/EHU, Vitoria-Gasteiz, Spain, [email protected],ORCID: 0000-0002-
4960-2729

Ekaitz Zulueta
University of the Basque Country, UPV/EHU, Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0001-
6062-9343

Javier Sancho
University of the Basque Country, UPV/EHU, Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0002-
9217-2680

Jose Manuel Lopez-Guede

University of the Basque Country, UPV/EHU, Vitoria-Gasteiz, Spain, [email protected],ORCID: 0000-0002-5310-
1601

Ibarra-Udaeta, Iosu, Portal-Porras, Koldo, Fernandez-Gamiz, Unai, Aramendia, Iñigo,

Zulueta, Ekaitz, Sancho, Javier, Lopez-Guede, Jose Manuel. Large Eddy Simulation of a
Cite this paper as:
Rectangular Vortex Generator on a flat plate with an incident angle of 25º. 8th Eur. Conf.
Ren. Energy Sys. 24-25 August 2020, Istanbul, Turkey

Abstract: Passive controls are included in the design of the wind turbine blades in order to obtain better
performance and reduce loads without consuming any external energy. Main passive techniques
include vortex generators, microtabs, serrated trailing edges, fences, spoilers and Gurney flaps.
The main target of a passive Vortex Generator, is to delay the flow separation and increase the
maximum lift coefficient. Vortex generators are small vanes usually triangular or rectangular,
usually displayed in pairs and with an angle of inclination with the inflow. Thanks to these
elements, the energy produced from the outer flow region (higher velocity) is transferred into the
boundary layer region.
Source term models can be used to simulate the side force that would result from a vane-type
vortex generator in the flow field, but they require additional computational uncertainty and
processing times (Bender E.E). Large Eddy based simulations have been performed in order to
reproduce the physics of the wake downstream of a VG on a flat plate in a negligible streamwise
pressure gradient flow.
Keywords: Large Eddy Simulation, Computational Fluid Dynamics, Vortex Generators, coherent structures

826
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

© 2020 Published by ECRES

Nomenclature
CFD Computational Fluid Dynamics
H Vane height
L Vane length
LES Large Eddy Simulation
PIV Particle Image Velocimetry
RANS Reynolds-Averaged Navier-Stokes
Re Reynolds number
SGS sub-grid-scale
TE Trailing Edge
VG Vortex Generator
x/H Normalized distance from the TE
α Incident angle

1. INTRODUCTION

Wind energy has become a key source of electricity generation for the change to a cleaner and sustainable energy
model. In order to compete in energy production and cost against the traditional energies, improvement and
optimization of wind turbines is required.
To improve the efficiency of wind turbine blades both active (such as trailing edge flaps, air jet vortex
generators and synthetic jets) and passive devices (such as vortex generators, microtabs, spoilers, fences,
serrated trailing edge and Gurney flaps) can be added. Aramendia et al. [1] exhaustively reviewed these devices.
Among these two types of devices, passive devices can be highlighted, since they improve performance of the
wind turbine blades without consuming any external energy.
Vortex generators (VG) are passive flow control devices which bring momentum from the outer flow region to
the inner flow region, modifying the boundary layer, and therefore, delaying the boundary layer separation and
increasing the lift coefficient. There are several experimental studies in which turbine blades with and without
VGs are compared. Øye [2] and Miller [3] compared the power curves obtained both with and without VGs
on a 1 MW wind turbine and on a 2.5 MW wind turbine, respectively. Improved results of the output power
were obtained using VGs in both studies.
In previous works, CFD (Computational Fluid Dynamics) simulations were carried out in a rectangular VG on
a flat plate in order to assess the accuracy of CFD simulations. Ibarra-Udaeta et al. [4] and Urkiola et al. [5]
analyzed the streamwise vortexes generated wake downstream of the VG with different vane heights (H) and
different incident angles (α) with a Reynolds number (Re) around 27000, and Fernandez-Gamiz et al. [6] tested
sub-boundary layer rectangular VGs with an α equal to 18º and a Re around 2600. In all the mentioned studies
RANS (Reynolds-Averaged Navier-Stokes) turbulence models were used, and good agreements with the
experimental data were obtained.
Other authors used LES (Large Eddy Simulation) turbulence models to carry out CFD simulations with VGs.
Bjerg et al. [7] performed LES simulations to analyze the flow structures and heat transfer characteristics of
rectangular VG pairs with different heights and distances between VGs, and Colleoni et al. [8] compared the
results obtained in RANS and LES simulations, showing more accurate results in LES simulations.
In the present paper, CFD simulations using LES turbulence models are conducted in order to reproduce the
physics of the wake downstream of a rectangular VG on a flat plate in a negligible streamwise pressure gradient
flow with a H equal to 0.25 m, an α equal to 25º and a Re around 27000.

2. NUMERICAL SETUP

CFD simulations were conducted in order to reproduce the physics of the wake downstream of a VG on a flat
plate in a negligible streamwise pressure gradient flow. To conduct the simulations Star CCM+ [9] solver was
used.
In these simulations a three-dimensional incompressible flow with constant values of 1.225 kg/m3 for density
and 1.511·10-5 m2 /s for kinematic viscosity is considered, and for the modeling of turbulence LES Smagorinsky
[10] SGS (sub-grid-scale) model is selected.

827
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The numerical domain consists of a block with a rectangular VG located on its lower surface. The height of the
VG (H) is equal to 0.25 m, the length (L) is twice its height and the incident angle (α) is equal to 25º. The lower
surface of the block and the surfaces of the VG are considered walls with no-slip conditions. The upstream part
of the domain is set as inlet, and the downstream part of the domain as outlet. The rest of the surfaces are
considered symmetry planes, in order to not affect the flow. The free stream velocity of the flow is set at 20 m/s.
Therefore, the Reynolds number (Re) is around 27000. More information about the numerical domain can be
shown in Figure 1.

Figure 1. Numerical domain and boundary conditions

With the above-mentioned numerical domain, a structured mesh of 11534336 cells was generated. In order to
enhance the accuracy of the results in the area near the VG, the cell density is much higher in this region. Figure
2 shows the refined mesh near the VG.

Figure 2. Meshing near the VG.

3. RESULTS AND DISCUSSIONS

Wall Shear Stress

As mentioned above, the main target of VGs is to delay the flow separation. The wall shear stress is a significant
parameter to understand this phenomenon. Figure 3 shows the wall shear stress values obtained in the VG TE
(Trailing Edge). x/H represents the normalized distance from the TE.

828
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 3. Wall shear stress in the TE of the VG.

The high wall shear stress values along the TE of the VG show a good performance of the VG, since the flow is
attached to the surface for a long distance.

Coherent Structures
To identify the coherent structures of the flow the Q-criterion method is used, represented by an isosurface. This
isosurface can be shown in Figure 4.

Figure 4. Isosurface colored by the instantaneous velocity of the flow.

Velte et al. [11] showed by PIV (Particle Image Velocimetry) measurements that two vortical structures appear
in the wake behind the vane. The primary vortex appears at the top of the VG and follows the direction of the
flow. The secondary vortical structure is a horseshoe vortex which appears at the lower part of the VG. This
secondary vortex is divided into two sides, the pressure side and the suction side. As the suction side of the
horseshoe vortex has the opposite sign of the primary vortex, at the early stage of its generation the primary
vortex sweeps the secondary vortex. Therefore, at the early stage of its generation the secondary vortex does not
follow the direction of the follow. The conducted simulation accurately predicted both vortical structures.

ACKNOWLEDGMENT

The present research has been developed under the frame of the Joint Research Laboratory on Offshore
Renewable Energy (JRL-ORE). The authors are grateful to the government of the Basque Country and the

829
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

University of the Basque Country UPV/EHU through the SAIOTEK (S-PE11UN112) and EHU12/26 research
programs, respectively.

4. CONCLUSIONS

The physics of the wake downstream of a VG on a flat plate in a negligible streamwise pressure gradient flow
were analyzed by LES.
The results show good agreements with the experimental data available in literature in terms of coherent
structures, with accurate predictions of the main vortical structures.
The wall shear stress shows a good performance of the VG, as the flow is attached to the surface for a long
distance.

REFERENCES

[1] I. Aramendia, U. Fernandez-Gamiz, J. A. Ramos-Hernanz, J. Sancho, J. M. Lopez-Guede, and E. Zulueta, ‘Flow

Control Devices for Wind Turbines’, in Energy Harvesting and Energy Efficiency: Technology, Methods, and
Applications, N. Bizon, N. Mahdavi Tabatabaei, F. Blaabjerg, and E. Kurt, Eds. Cham: Springer International
Publishing, 2017, pp. 629–655.
[2] S. Øye, ‘The effect of vortex generators on the performance of the ELKRAFT 1000 kW turbine’, presented at the 9th
IEA symposium on aerodynamics of wind turbines, Stockholm, 1995.
[3] G. Miller, ‘Comparative performance tests on the Mod-2, 2.5-MW wind turbine with and without vortex generators’,
presented at the DOE/NASA workshop on horizontal axis wind turbine technology, Cleveland, 1984.
[4] I. Ibarra-Udaeta, I. Errasti, U. Fernandez-Gamiz, E. Zulueta, and J. Sancho, ‘Computational Characterization of a
Rectangular Vortex Generator on a Flat Plate for Different Vane Heights and Angles’, Applied Sciences, vol. 9, no. 5,
Art. no. 5, Jan. 2019, doi: 10.3390/app9050995.
[5] A. Urkiola, U. Fernandez-Gamiz, I. Errasti, and E. Zulueta, ‘Computational characterization of the vortex generated
by a Vortex Generator on a flat plate for different vane angles’, Aerospace Science and Technology, vol. 65, pp. 18–
25, Jun. 2017, doi: 10.1016/j.ast.2017.02.008.
[6] U. Fernandez-Gamiz, I. Errasti, R. Gutierrez-Amo, A. Boyano, and O. Barambones, ‘Computational Modelling of
Rectangular Sub-Boundary Layer Vortex Generators’, Applied Sciences, vol. 8, no. 1, Art. no. 1, Jan. 2018, doi:
10.3390/app8010138.
[7] A. Bjerg, K. Christoffersen, H. Sørensen, and J. Hærvig, ‘Flow structures and heat transfer in repeating arrangements
of staggered rectangular winglet pairs by Large Eddy Simulations: Effect of winglet height and longitudinal pitch
distance’, International Journal of Heat and Mass Transfer, vol. 131, pp. 654–663, Mar. 2019, doi:
10.1016/j.ijheatmasstransfer.2018.11.015.
[8] A. Colleoni, A. Toutant, and G. Olalde, ‘Simulation of an innovative internal design of a plate solar receiver:
Comparison between RANS and LES results’, Solar Energy, vol. 105, pp. 732–741, Jul. 2014, doi:
10.1016/j.solener.2014.04.017.
[9] ‘STAR-CCM+ v2019.1’. https://www.plm.automation.siemens.com/ (accessed Jun. 02, 2020).
[10] J. Smagorinsky, ‘General circulation experiments with the primitive equations’, Mon. Wea. Rev., vol. 91, no. 3, pp.
99–164, Mar. 1963, doi: 10.1175/1520-0493(1963)091<0099:GCEWTP>2.3.CO;2.
[11] C. M. Velte, M. O. L. Hansen, and V. L. Okulov, ‘Multiple vortex structures in the wake of a rectangular winglet in
ground effect’, Experimental Thermal and Fluid Science, vol. 72, pp. 31–39, Apr. 2016, doi:
10.1016/j.expthermflusci.2015.10.026.

830
 ATOM SEARCH OPTIMIZATION ALGORITHM FOR
OPTIMIZATION OF POWER GENERATION IN MICROGRID

Wafeeqa Abdulrazak Hasan

Southern Technical University, Engineering Technical College ,Department of Electrical Power Engineering ,Basra,
Iraq, wafeeqaalibraheem82@ gmail.com, ORCID: 0000-0002-1332-4632

Issa Ahmed Abed

Southern Technical University, Engineering Technical College, Department of Automation and Control Technologies
Engineering, Basra, Iraq, [email protected], ORCID: 0000-0002-4007-2872

Diyah Kammel Shary

Southern Technical University, Engineering Technical College Department of Electrical Power Engineering, Basra,
Iraq, [email protected], ORCID: 0000-0002-0649-3091

Surname, FN, Surname, FN. Atom Search Optimization Algorithm For Optimization Of
Cite this paper as: Power Generation In Microgrid. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: In this study, a novel atom search optimization (ASO) is used to solve the economic scheduling of
one of the real world problems of the power system. This method of optimization mimic the pattern
of atomic movement in nature, where to the atoms interact by Lennard-Jones potential interaction
forces and restriction forces as a result of the bond-length potential. A new optimization approach
is used to solve the constraint of balanced generation for scheduling energy generation among the
available energy sources in the microgrid system at different value of essential parameters for ASO
algorithm. Eventually, the results of the simulation are compared with the technique of particle
swarm optimization (PSO) available in literature for validating the global capability of new
optimization method.
Keywords: microgrid, optimization algorithm, scheduling, atom search optimization
© 2020 Published by ECRES
Nomenclature
ASO Atom Search Optimization
PSO Particle Swarm Optimization
MGs Microgrids
LDCs Least Developed Countries
HPS Hybrid Power System Problems

1. INTRODUCTION

A key to optimal scheduling of available on dispatch generators is any microgrid operates economically.
Microgrid is confident and optimum solution for small and insulated electrification rural regions in many least
developed countries (LDCs) [1]. There is currently a lot of interest in the use of microgrids (MGs) in power
systems because they are a resilient, intelligent and energetic energy network [2].
Use of artificial intelligence technology to tackle problems with electricity source problems that uses various
sources due to their robustness and capability was an active to handle easy to use with complicated systems [3].
The conventional approaches however could not solve hybrid power system problems (HPS), where
information is always imperfect and naturally qualitative it must be approach with several opposing objectives.
Artificial intelligence in this field will boost performance Of HPS for precision outcome, there are different type
of artificial.

831
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

techniques such as the neural networks, fuzzy logic and algorithms with bio-inspiration [4]. A literature analysis
of the problem with the hybrid power system demonstrates that several methods of optimization solving a
problem of energy management system are they proposed:
Bo Zhao et.al [5] the researchers developed a non-dominated genetic sorting algorithm (NSGA-II) to solve
problems a multi-objective. Shariatzadeh et al. [6] have been proposed a solution to the problem of shipboard
microgrid power system (SMPS) by applied particle swarm optimization (PSO) technique and utilize genetic
algorithm (GA). Rajesh Kumar et.al [7] applied biogeography based optimization algorithm (BBO). Askarzadeh
[8] has been proposed a memory-based genetic algorithm (MGA). Ngoc An and Tuan [9] used dynamic
programming technique to locate a minimum cash flow (CF) to optimize the sources schedule. Basu and
Chowdhury [10] have been presented cuckoo search algorithm (CSA) for solving both convex and no convex
ED (economic dispatch) problems of fossil fuel fired generators. Furthermore, Aghajani and Ghadimi [11] had
been used particle swarm optimization (MOPSO) method for management and optimize distribution of energy
resources in proposed microgrid. Liang Luo et.al [12] presented the modified bat algorithm (MBA) for efficient
scheduling in microgrid. Peng Li et al.[13] are Suggested binary gravitational search algorithm (BGSA) to
improve scheduling of operations of MG and optimize microgrid dynamically.
A new method of global Meta-heuristic optimization called atom search optimization (ASO) algorithm. This
optimization method was suggested by Zhao et al. 2018 on the basis of the atom model in force movement in
dynamics of molecules [14].
ASO is a kind of swarm intelligence algorithm which is inspired by physics, needs algorithm minimal parameters
and well capable of global exploration method optimization. Thus, it was implemented successfully to measure
hydrogeological parameters and groundwater coefficient of dispersion of [14], [15].
ASO is used to manage the scheduling of an output power in a microgrid system in 24 hours for the optimum
cost, and tested it at different values of algorithm parameters. In addition, the effectiveness of ASO algorithm is
evaluated through comparison of results with result of the PSO algorithm.
Organizing the paper includes: problem formulation in the Section 2, Section 3 demonstrations the proposed
optimization method ASO, Next Section 4 Simulation performance and discussion. In the end, the conclusion
was presented in Section 5.

2. PROBLEM FORMULATION

In this study, micro grid contains from two diesel plants, two wind plants and three fuel cell plants. In a diesel
plant the objective function of Generator consists of a cost-function similar for fuel function for traditional
generating fossil-fuel plants. The generator of wind turbines is based on the characteristics of variable output,
since generative wind turbine generation is evaluated from the wind's intensity. Cost of operation in the fuel cell
scheme, cost of fuel, and efficiency are electric fuel production included. Power restrictions include equilibrium
equations as well as the limits of generating power in the microgrid [10], [14].
2.1 Objective Functionality and Restriction
To minimize the cost of generating defines the power generated as well as the objective function (OF) according
to the quadratic cost function for each generator Eq.1 [10].The power generated should be the same as the power
required to satisfy the balance in the microgrid system [10].
𝑁 𝑁 (1)
𝐹𝑡 = ∑ 𝐹𝑖 ( 𝑃𝑖 ) = ∑ 𝑎𝑖 𝑥𝑃𝑖2 + 𝑏𝑖 𝑥𝑃𝑖 + 𝑐𝑖
𝑖=1 𝑖=1

Thus, Fi (Pi) is a function of costs, i denote that ith number of generator, N represented the number of committed
generators; Pi represented the output of the power of i-th generator. ai , bi and ci are ith generator cost
coefficients.
Total cost function Includes of whole distributed generation cost functions and has been used as the objective
function as in Eq.2 [10]
𝐹𝑡𝑜𝑡𝑎𝑙 = 𝐹𝑑𝑖𝑒𝑠𝑒𝑙 + 𝐹𝑤𝑖𝑛𝑑 + 𝐹𝑓𝑢𝑒𝑙𝑙 𝐶𝑒𝑙 (2)

832
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

2.1.2 Real Power Equilibrium Restriction

The load balance equation consists of the requiring load must equal the power generation in a micro grid at
any time period over scheduling period as following in Eq.3 [10].

𝑁
∑ 𝑃𝑖 − 𝑃𝐷 = 0 (3)
𝑖=1

PD : express the power to demand.

2.1.2 Real Power Restrictions on Power Generation

i
The power which is the generation of each unit must be within its lower bound Pmin as well as upper
i
boundPmax , to verify that:

𝑃𝑖𝑚𝑖𝑛 ≤ 𝑃𝑖 ≤ 𝑃𝑖𝑚𝑎𝑥
, ,𝑖 ∈ 𝑁 (4)

3. EXISTING MECHANISM ATOM SEARCH OPTIMIZATION (ASO)

3.1Algorithm
For applied to a real-world engineering problems, metaheuristic optimization algorithms are becoming more
common in and widely used in intelligent computing. Its success comes from subsequent aspects. First of all, all
these methods of optimization of the certain simple theories and mathematics Models proved rational,
originating out of the real world and influenced by all types of physical or biological behaviors [17],[18].
ASO is a community oriented, newly developed physics-inspired Heuristic algorithm mimicking atomic
movement with interaction control forces and design constraints an efficient global optimization search
mechanism problems. [15].
Matter is made up of molecules, connecting atoms through the covalent bonds in molecular shape. Atoms are
made of mass as well as volume. The Interaction forces there are atoms in between indicated like repulsion or
an attraction, depending on distances against atoms. The repulsion powers in the repelling zone among atoms,
when distances reduce sharply. In An attractive area, when the distances are in rise to any extent, the forces of
attraction reach maximum.
The distances keep on increasing including; the forces of attraction slowly decline to 0 if there are two atoms at
a distance of balance, the force of interaction among them is nil. In Atomic molecules the effects of geometric
constraints also need to be considered by the forces and the internal movements of atoms. Geometric limitations
and the internal movements of atoms are called binding force. [15],[16],[19].
All materials are composed of atoms that travel the whole time, and atomic movement takes the classics
engineering [15],[16]. Referring to the second law of Newton, presume the force Fi is an The interaction force
and Gi act together as a constraint force on the i-th atom in an atom The system, then the acceleration associated
with its mass mi a be given as follows in Eq.(5)[15],[20].

(5)
Fi +Gi =mi a

3.2 Principle of ASO Theory and Mathematics

Every atom in ASO performs a viable solving in the seeking area. The weight of an atom is the strength of an
atom practicable Solving. The better, represented higher the solution. That will be the atomic mass, and vice
versa. Both the atoms inside the population attracts one another or fends one another as per theirs distance that
results in atoms that are lighter moving to the more heavy atoms. Less acceleration of heavier atoms, making
they are great at leveraging local region.
Therefore, much smaller accelerations to the atoms that Let them go exploring a more extensive search area.
Might be the mass of its atom i on the iteration t just tested with fitness function. Expressing the 𝐦𝐢 𝐭 is indicated
in Eq.7 [15-16]:
𝑀 (𝑡) = 𝑒 −(𝐹𝑖𝑡𝑖 (𝑡)−𝐹𝑖𝑡𝑏𝑒𝑠𝑡 (𝑡)⁄𝐹𝑖𝑡𝑤𝑜𝑟𝑠𝑡(𝑡)−𝐹𝑖𝑡𝑏𝑒𝑠𝑡 (𝑡)) ( 6)
𝑖

833
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

𝑀𝑖 (𝑡)
𝑚𝑖 𝑡 = 𝑁 ( 7)
∑𝑗=1 𝑀𝑗 (𝑡)

mi Where Fit i(t) represented the fitness, N denotes the total number of atoms, atomic value of function i during
iteration t, Fit worst (t) and Fit best (t) denotes the worst and most values of the fitness function respectively on
iteration t at best atoms, and defined as respectively.

𝐹𝑖𝑡𝑏𝑒𝑠𝑡 (𝑡) = 𝑚𝑖𝑛 𝐹𝑖𝑡𝑖 (𝑡) (8)

𝑖𝜖{1,2,….,𝑁}

𝐹𝑖𝑡𝑤𝑜𝑟𝑠𝑡 (𝑡) = 𝑚𝑎𝑥 𝐹𝑖𝑡𝑖 (𝑡) (9)

𝑖𝜖{1,2,….,𝑁}

The overall interaction forces to act at the i-th atom in the d-th dimension that simply Is the total value of that
vector the disharmony and attraction from dynamically exerted. The change of adjoining atoms on i-th is given
in Eq.5 [15], [16]

𝐹𝑖𝑑 (𝑡) = ∑ 𝑟𝑎𝑛𝑑𝑗 𝐹𝑖𝑗𝑑 (𝑡) (10)

𝑗∈𝐾𝑏𝑒𝑠𝑡

Thus, randj is the random number within [0, 1] range and K best is a sub-population of the population of atoms,
that composed of atoms first K with the highest values of function [15], [16].

Making ASO doing the early iterations more exploration, every atom wants to communicate With as many of
them suitable atoms as possible fitness values like its neighbors K, and make algorithms do further exploitation
at the ultimately iterations , every atom requires interaction with little, atoms have values of better fitness than
their K neighbors. The value of K therefore the need is increasingly reduced With the Iteration process.
Hence the number of iterations rises by K lowering. When T is the cumulative number of iterations, then K is
computed from Eq.11 [15-16]:
𝑡
𝐾(𝑡) = 𝑁 − (𝑛 − 2)𝑥√ (11)
𝑇

So, N accounts for the complete number of atoms in the atomic system, the initial iteration is t, and the
cumulative number of iterations is T.
As ASO wants to increment power of attraction and decrease repellent force by increasing the times of iteration.
Then, Zhao et al. [15] suggested simplified design modelFij interaction, Fij the description is as shown in Eq.12
[15-16]:

13 7
𝜎 𝜎 𝑟𝑖𝑗→
𝐹𝑖𝑗𝑑 = − 𝜂 [2 ( ) −( ) ] 𝑑 (12)
𝑟𝑖𝑗 𝑟𝑖𝑗 𝑟𝑖𝑗

𝑡−1 3 −20𝑡
𝜂(𝑡) = 𝛼 (1 − ) 𝑒 𝑇 (13)
𝑇

𝑟𝑖𝑗 (𝑡)
ℎ𝑚𝑖𝑛 , < ℎ𝑚𝑖𝑛 ,
𝜎(𝑡)
𝑟𝑖𝑗 (𝑡) 𝑟𝑖𝑗 (𝑡)
ℎ𝑖𝑗 (𝑡) = ℎ𝑚𝑖𝑛 ≤ ≤ ℎ𝑚𝑎𝑥 , (14)
𝜎(𝑡) 𝜎(𝑡)
𝑟𝑖𝑗 (𝑡)
ℎ𝑚𝑎𝑥 , ,
{ 𝜎(𝑡) > ℎ𝑚𝑎𝑥

So, define η(t) is the repulsion adjustment depth function or the attractive regions, the feature depth is described
in (13). Where α is the weight for depth. Pursuant to equation

834
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

(14), the hij (t) distance function is modified, in which hmin = 1.1 , and the features are defined as following
hmin = 1.4 represented the maximum, and minimize boundary, of h, respectively. The Euclidean from the atom
i to the atom j represents in rij t in t in the iteration, σ(t) is thus the length that is set as in the Eq.15 [15-16]:
∑∈𝐾𝑏𝑒𝑠𝑡 𝑥𝑖𝑗 (𝑡)
𝜎(𝑡) = ‖𝑥𝑖𝑗 (𝑡), ‖ (15)
𝐾(𝑡) 2

xij , is appeared the component atom i position in dimension the search area j. The hmin changed dynamically
in the following Eq. 14, and hmax are defined:

ℎ = 𝑔0 + 𝑔(𝑡)
{ 𝑚𝑖𝑛 (16)
ℎ𝑚𝑎𝑥 = 𝑢

And g is just the represented of drift ingredient and it is described as below:

𝜋 𝑡
𝑔(𝑡) = 0.1𝑥 𝑠𝑖𝑛 ( 𝑥 ) (17)
2 𝑇
A geometrical theory constraint in molecular dynamics plays a key atomic Roll movement. Assume that each
atom comprises covalent bonds in K best with every atom.
Is the difference as a weighted position between the atoms and better atom described in the following Eq.18
[15-16]:

𝐺𝑖𝑑 (𝑡) = 𝜆(𝑡)(𝑋𝑏𝑒𝑠𝑡

𝑑
− 𝑋𝑖𝑑 (𝑡) ), (18)

20𝑡
𝜆(𝑡) = 𝛽𝑒 −( 𝑇
) (19)

d
So, λ is represented of the Lagrange multiplier, and β is define the weight of the multiplier, and Xbest is better
atom position at iteration t in the dth dimension of the better atom. By equations (5), (13), and (18), the modified
acceleration equation is extracted as then follows in Eq.20[15-16]:

𝐹𝑖𝑑 (𝑡) 𝐺𝑖𝑑 (𝑡)

𝑎𝑖𝑑 (𝑡) = +
𝑚𝑖𝑑 (𝑡) 𝑚𝑖𝑑 (𝑡)
13 7
𝑟𝑎𝑛𝑑𝑗 [2𝑥(ℎ𝑖𝑗 (𝑡)) −(ℎ𝑖𝑗 ) ]
20𝑡
−( ) ∑𝑗=𝐾𝑏𝑒𝑠𝑡 ( )𝑥((𝑥𝑗𝑑 (𝑡)−𝑥𝑖𝑑 (𝑡))/‖𝑥𝑗(𝑡)−𝑥𝑖 (𝑡)‖ ) (20)
𝑡−1 3 𝑇 𝑚𝑖 (𝑡) 2
= 𝛼 (1 − ) 𝑒
𝑇
20𝑡 𝑑
(𝑡)−𝑋𝑖𝑑 (𝑡)/𝑚𝑖 (𝑡))
+𝛽𝑒 −( 𝑇 )(𝑋𝑏𝑒𝑠𝑡

The update for equations, of speed and place, from the previous equations found in this atom i in iteration (t+1)
described in the Eqs.21-22 [15-16], respectively:

𝑣𝑖𝑑 (𝑡 + 1) = 𝑟𝑎𝑛𝑑𝑖𝑑 𝑣𝑖𝑑 (𝑡) + 𝑎𝑖𝑑 (𝑡) (21)

(22)
𝑥𝑖𝑑 (𝑡 + 1) = 𝑥𝑖𝑑 (𝑡) + 𝑣𝑖𝑑 (𝑡 + 1)

835
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

3.3 The Pseudo Code of ASO Algorithm

Figure 1Pseudo code of the ASO algorithm [15]

4. SIMULATION AND NUMERICAL RESULTS

The performance of ASO will be verified by proved that ASO has ability to schedule the optimal generation of
the microgrid system. In this study, the microgrid system is consisting of several traditional and renewable
energy power plants such two diesel generators, two wind turbine and three fuel - cell plants .The system data
comprehensive cost coefficients, power of load and wind velocity as described in [10]. The micorgrid system
data as described in [10], which comprise cost coefficients, power demand and wind velocity. The parameters
setting of the ASO algorithm are listed in Table1. Since the ASO algorithm is randomly initialized, The number
of trials shall be 30 in every hour within 24 hour for minimum and average cost, to see how effective and
notability the proposal algorithm to schedule the optimal generation with minimum cost of the proposed
microgrid system at different case of change number of maximum iteration and number of atom (population
size). Additionally, the best ASO algorithm results are compared to PSO algorithm to determine effectiveness
of the algorithm proposed at same algorithm parameters [10]. The results of the simulation are realized using
MATLAB2017b with PC (Intel(R) core(TM) i7-75000 CPU@ 2.70 GB, 2.90 GHZ) and 12.00 GB RAM.

Table 1. ASO algorithm parameters

Parameter Value
Number of atom(population) 100
Maximum iteration number 500
alpha(α) 50
beta(β) 0.2

4.1 Analysis of Tests Results ASO Algorithm

To validate the proposed ASO algorithm based approach, it has the ability to schedule the optimal generation
of the microgrid system. To solve the microgrid optimization problem in 24 hours, the proposed algorithm aims
to provide minimum cost of the necessary power at any hour.
ASO algorithm is tested for different values for the parameters as the number of atom (population size) and
number of iterations. And, the simulation research was performed over thirty consecutive trials and best
outcomes tabulated at every test case as shown below:

Test Case 1

By variation, the atom numbers (population size) within the same number of maximum iteration (500) as well
as the atom values were chosen 20, 50 and 100 respectively. The results of test for the microgrid generation
obtained from ASO have been shown in Table 2.And the schedule of power generation for the microgrid for this
test in each case shown in figures (2, 3, and 4).

836
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 2. Results of Test for Different Iterations

Population Sizes Minimum Cost Maximum Cost Average Cost
20 32675.3782 35635.1724 34311.362
50 32625.2676 35558.3276 33906.1323
100 32236.9526 35595.6446 33911.6515

Pdiesel1 Pdiesel2 Pwind1 Pwind2 Pfuel1 Pfuel2 Pfuel3

1000
Power Generation (Kw)

800
600
400
200
0
1 5 9 13 17 21
Hour

Fig. 2 Optimal power scheduling obtained by ASO at atom population (20).

Pdiesel1 Pdiesel2 Pwind1 Pwind2 Pfuel1 Pfuel2 Pfuel3

1000
Power Generation (Kw)

800
600
400
200
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

Figure 3 optimal power scheduling obtained by ASO at atom population (50).

Pdiesel1 Pdiesel2 Pwind1 Pwind2 Pfuel1 Pfuel2 #REF!

1000
Power Generation (Kw)

800
600
400
200
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

Figure 4 optimal power scheduling obtained by ASO at atom population (100).

Add that, Figure (5) illustration the compares the generation power of the microgrid with that of the requested
power in every hour of a day. The results were obtained through ASO. The power balance equation is seen to
have been fulfilled because the generation is in correspondence with the power demanded.

837
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Since ASO is an optimization algorithm, its basis is population-based, where the forces of interaction comprise
attraction and repulsion. Therefore, constrained force is an essential medium for communicating
knowledge among all population [15].
Accordingly, ASO algorithm can resolve the issue of optimization immediately and evaluate the power schedule
optimum for microgrid.

PD(Kw) PGeneration(kW)

2000
Power (kW) )

1500
1000
500
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

Figure 5 Comparison the power generated by microgrid and the demand power

Test case II

By change the iteration number within the same number of atom (population size) (100) and the iteration values
were chosen 100, 250 and 500 respectively. The results of experiment for the microgrid generation obtained
from ASO have been shown in Table 3. And the convergence curves for this test in each case atom population
shown in figures (6, 7, and 8). ASO has faster speed of convergence in early iterations. We can see that there
are path curves even the early iterations large-scale variations are frequent. And it tends to be stabilized
gradually in later iterations, and then stabilized at the time of iteration.
The results show that ASO shows a strong equilibrium for both exploration and exploitation. Additionally, the
global optimum search efficiency is also satisfactory [16].

Table 3 Results of Test for different Iterations

Population Sizes Minimum Cost Maximum Cost Average Cost
100 33074.5344 35316.194 34088.5315
250 32841.2057 35501.0316 34012.2771
500 32236.9526 35595.6446 33911.6515

Figure 6 Convergence curve for using ASO for atom population (100)

838
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Figure 7 Convergence curve for using ASO for atom population(250)

Figure 8 Convergence curve for using ASO for atom population (500)

4.2 Comparison Between ASO And PSO For Optimal Performance

The ASO is a kind of algorithm for swarm intelligence based on physics; the algorithm will use less parameters
and a better global exploration capability the process of optimization. So, the optimal cost result gained from
the proposed ASO method has been compared with the result of optimum cost obtained from the particle swarm
optimization (PSO) method at the same standard parameters of algorithm (population size (100) and
iteration(500).
The algorithm is performed 30 times to provide a fair comparison, and executed for each hour through in 24
hour to reach at optimal cost result, which have (32236.9526$) is consider less than the cost compare with result
obtained from proposed PSO algorithm (38189.31$) [10].
Experimental results demonstrated that ASO is massively best as comparison with particulate swarm
optimization (PSO) and also, genetic algorithm (GA) [16]. It represents a promising replacement
to the other algorithms particulate swarm optimization (PSO), genetic algorithm (GA) and bacterial foraging
optimization (BFO) to apply widely in several practical problems [16].

5. CONCLUSIONS

In the present study, metaheuristic algorithm inspired by physics named atom search optimization (ASO)
algorithm being applied for the optimum power scheduling generation in the microgrid, which the load demand
must equal the generation of power in a microgrid at each time interval over the scheduling time. The
performance of ASO was analyzed on scheduling problem for reducing the cost of microgrid generation by
varying the parameters of ASO algorithm. According to results obtained, ASO obviously does have global
search ability. And the result obtained from it is also confirmed best than the PSO algorithm result mentioned
in recent literature. In addition, we can suggest that ASO could be considered as an alternate optimization method
for energy management system optimization scheduling solution as promising replacement to PSO algorithm.

ACKNOWLEDGMENT

We would like to express our acknowledgment to the Southern Technical University, Engineering Technical
College of Basra and Department of Electrical Power Engineering for supports the work in this paper.

839
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

REFERENCES

[1] L. L. GARVER . Power Generation Scheduling by Integer Programming-Development of Theory, Transactions of the
American Institute of Electrical Engineers, Part III: Power Apparatus and Systems. 1962, 81(3),pp. 730–734.
[2] Moataz E , Amrane O, Hamid G & Ottavio L. Optimal economic and environment operation of micro-grid power
systems, Energy Conversion and Management 2016;122: pp. 182-194.
[3] J. A. Laghari, H. Mokhlis, A. H. A. Bakar & Hasmaini M. Application of computational intelligence techniques for
load shedding in power systems: A review. Journal of Energy Conversion and Management 2013; 75, pp.130–140.
[4] S. Berrazouane, & K. Mohammedi .Parameter optimization via cuckoo optimization algorithm of fuzzy controller for
energy management of a hybrid power system .Journal of Energy Conversion and Management.2014; 78, pp.652–
660,> http://dx.doi.org/10.1016/j.enconman.2013.11.018<.
[5] Bo Z , Xuesong Z , Jian C , Caisheng W & Li G . Operation Optimization of Standalone Microgrids Considering
Lifetime Characteristics of Battery Energy Storage System .In: IEEE Transactions on Sustainable Energy 2013; 4(4),
pp.934–943.> http://ieeexplore.ieee.org. DOI:10.1109/TSTE.2013.2248400>.
[6] Farshid S , Nikhil K & Anurag K S . Optimal Control Algorithms for Reconfiguration of Shipboard Microgrid
Distribution System Using Intelligent Techniques. IEEE Transactions on Industry Applications 2017; 53(1):pp. 474-
482.
[7] Rajesh K, R. A. Gupta &Ajay K B . Economic analysis and power management of a stand-alone wind/photovoltaic
hybrid energy system using biogeography based optimization algorithm, Swarm and Evolutionary Computation .In:
journal.homepage:www.elsevier.com/swevo 2013, 8: p. 33-43.
[8] Alireza A. Memory-Based Genetic Algorithm for Optimization of Power Generation in a Microgrid. In: IEEE
Transactions on Sustainable Energy 2018; 9(3), 1081–1089.> DOI 10.1109/TSTE.2017.2765483<
[9] LUU Ngoc A & TRAN Q .Optimal energy management for grid connected microgrid by using dynamic
programming method. In: IEEE Power & Energy Society General Meeting 2015.
[10] M. Basu & A. Chowdhury. Cuckoo search algorithm for economic dispatch .In: Energy 2013;vol. 60: pp. 99-108.
<http://dx.doi.org/10.1016/j.energy.2013.07.011>
[11] Gholamreza A & Noradin G.Multi-objective energy management in a micro-grid. Journal of Energy Reports 2.18; 4:
pp. 218-225. < (http://creativecommons.org/licenses/by-ncnd/4.0/)>.
[12] Liang L, Sarkew S A , Alireza R , Mohammad R M , Sarminah S , Nahla A & Mehdi P. Optimal scheduling of a
renewable based microgrid considering photovoltaic system and battery energy storage under uncertainty. Journal of
Energy Storage, 17 February 2020,28,101306, 2020. <https://doi.org/10.1016/j.est.2020.101306>
[13] Peng L ,Zeyuan Z, Xiaopeng L, Xavier Y& Xingyan N. Dynamic Optimal Operation Scheduling of Microgrid
Using Binary Gravitational Search Algorithm. In: International Conference on Power System Technology 2014,
pp.3175-3180.
[14] S. Krishnamurthy, T. M. Jahns & R. H. Lasseter. The operation of diesel gensets in a CERTS microgrid. In: IEEE
Power and Energy Society General Meeting - Conversion and Delivery of Electrical Energy in the 21st Century,
2008.
[15] Weiguo Z, Liying W & Zhenxing Z. Atom search optimization and its application to solve a hydrogeologic
parameter estimation problem,”Knowledge-Based Systems 2.18; vol.163, pp.283-304<
https://doi.org/10.1016/j.knosys.2018.08.030>
[16] Weiguo Z, Liying W & Zhenxing Z. A novel atom search optimization for dispersion coefficient estimation in
groundwater, Journal of Future Generation Computer Systems 2018, vol. 91, pp. 601-610 <
https://doi.org/10.1016/j.future.2018.05.037>.
[17] S. Kirkpatrick, C.D. Gelatt, M.P. Vecchi. Optimization by simulated annealing, In: Science, 13May 1983, vol.
220(4598), pp.671-680.
[18] James K & Russell E .Particle swarm optimization. In: IEEE International Conference on Neural Networks 1995, pp.
1942-1948.
[19] Yanming F, Zhuohang L , Chiwen Q & Haiqiang C. Modified Atom Search Optimization Based on Immunologic
Mechanism and Reinforcement Learning. Journal of Mathematical Problems in Engineering 2020; vol.2020, Article
ID 4568906, 22 pages,2020.
[20] Jean –Paul R, Giovanni C , Herman J C B .Numerical integration of the cartesian equations of motion of a system
with constraints: molecular dynamics of n-alkanes. Journal of Computational Physics 1977; vol.23, pp.321-341.

840
 ARTIFICIAL NEURAL NETWORK-BASED STOP
CRITERIA FOR A GENETIC ALGORITHM

Ekaitz Zulueta Guerrero

University of the Basque Country, UPV-EHU, Vitoria-Gasteiz, Spain, [email protected],
ORCID: 0000-0001-6062-9343

Ander Sánchez Chica

University of the Basque Country, UPV-EHU, Vitoria-Gasteiz, Spain, [email protected],
ORCID: 0000-0002-0271-6825

Daniel Teso-Fz-Betoño
University of the Basque Country, UPV-EHU, Vitoria-Gasteiz, Spain, [email protected],
ORCID: 0000-0001-6062-9343

Unai Fernández-Gámiz
University of the Basque Country, UPV-EHU, Vitoria-Gasteiz, Spain, [email protected],
ORCID: 0000-0001-9194-2009

Jose Antonio Ramos-Hernanz

University of the Basque Country, UPV-EHU, Vitoria-Gasteiz, Spain, [email protected],
ORCID: 0000-0001-9706-4016

Jose Manuel Lopez-Guede

University of the Basque Country, UPV-EHU, Vitoria-Gasteiz, Spain, [email protected],
ORCID: 0000-0002-5310-1601

Ekaitz, Zulueta Guerrero, Ander, Sánchez Chica, Daniel, Teso-Fz-Betoño, Unai Fernández-
Gámiz, Jose Antonio Ramos-Hernanz, Jose Manuel Lopez-Guede. Artificial neural network-
Cite this paper as:
based stop criteria for a Genetic algorithm. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020,
Istanbul, Turkey

Abstract: In this work, a new relationship between Artificial neural networks (ANN) and optimization
algorithms is present. ANN’s are largely used in forecasting and modeling applications. A typical
application of optimization algorithms is optimizing the cost function of the ANN error. In the
present work, the research team proposes a new approach to use these techniques together. A
genetic algorithm is developed in order to optimize a selection of test functions and finally the
design of an air impingement system. The points and the cost calculated by the optimization
algorithm are used as inputs and targets of ANN training. The use of the performance of the ANN
training as a stop condition of the optimizing process. The main objectives are to reduce the number
of iterations of the optimization process and, at the same time, obtain a neural model of the cost
function
Keywords: Artificial neural networks, Genetic algorithms, Performance, Training, Optimization, Stop
condition, Modelling
© 2020 Published by ECRES

1. INTRODUCTION

Artificial neural networks have become very important in research over the last few years. They have been
widely used in modeling, forecasting and control tasks such as photovoltaic modeling [1]. There are different
ways to approach the training of artificial neural networks (ANN). In this article, we talk specifically about
supervised training. The error is defined as the difference between the expected value and the value obtained by

841
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

the network for the same inputs. The supervised training algorithms are based on this error to modify the synaptic
weights of the net and try to improve its performance.
The most common training algorithm in this type of application is the backpropagation algorithm. However,
some authors have employed optimization algorithms such as Particle swarm algorithms (PSO) and Genetic
algorithms (Ga) to reduce network error by taking the synaptic weight and bias of the network as particles or
individuals in each case. In the works of Karaboga [2] and Shafaei [3] examples of both techniques can be found.
Optimization algorithms aim to adjust the variables of a function so the best response is obtained. Throughout
the last decades, special emphasis has been placed on the generation of new algorithms based on natural
phenomena, especially on the behavior of living beings. In [4] Mohit, Vijander and Asha made a review of many
of the algorithms developed in recent times such as Ant colony optimization [5], Flower Pollination algorithm
[6] and the Multi-Verse optimizer [7]. These researchers analyze the behavior of living beings and translate it,
in mathematical expressions, to implement the processes of exploration and exploitation. These inspirations seek
to find new algorithms that quickly converge to a solution without falling into the local optimum.
In this paper, it is intended to use these two techniques together. The idea proposed by the research team is to
use the genetic algorithm to optimize a series of test functions and finally a real design problem. At the same
time that the optimization process is executed, the points found by the algorithm are used to train the neural
network to obtain a model of the cost function. The fundamental idea of the work is to use the training error of
the network as a criterion to stop the algorithm. Therefore, once an acceptable error is reached, the optimization
process is stopped. It is intended to demonstrate that this new technique reduces the number of iterations of the
optimization algorithm as well as allows to obtain a model of the cost function.

2. Neural Network

The type of neural network chosen depends on the application. For example, in numerical regression
applications, the use of feed-forward networks is very widespread, in [8] Ansari, Zarei Sabbaghi and
Keshavars made a model for relative viscosity of nanofluids based on feed-forward neural networks. For
image processing and recognition, there are many examples of the use of convolutional networks such as
ImageNet [9]. Finally, in applications where time is a key factor, long/short term memory networks are used.
For example, the lithium-ion batteries life prediction [10].

Since one of the main objectives is to obtain a numerical model based on known data, the feed-forward network
architecture is selected.

Defining the architecture is a challenging task. There is no technique that defines the number of layers and the
number of neurons suitable for a specific application. The most common method to determine the network
structure is to perform a test with different configurations and select the one with fewer errors. In this work, this
technique has been applied and it has been determined that the best configuration is 2 hidden layers with 15
neurons in each one. The activation function of each hidden layer is sigmoid. In the output layer, a linear activation
function is used. Table 1 shows an overview of the parameters of the ANN.
For multilayer feedforward neural networks there are different methods of training. Backpropagation (BP)
algorithms based on the Jacobian derivative such as Levenberg-Marquardt or based in the gradient derivate such
as gradient descendent backpropagation are two examples. As previously mentioned, heuristic algorithms such
as optimization algorithms can also be applied. In this work, the Levenberg-Marquardt BP algorithm is used
because is the default algorithm defined by Matlab.

842
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 1. ANN parameters Summary

3. GENETIC ALGORITHM

As mentioned in the introduction, there are a variety of optimization algorithms. In recent years research has
focused especially on algorithms based on nature's behavior. In this work, because the points obtained by the
algorithm are going to be used as the training data, it is necessary to choose an algorithm that generates a
population of great diversity. Though the algorithm converges more slowly, a population as heterogeneous as
possible is necessary for the correct training of the network.

The research team decides to use genetic (Ga) algorithms. The mutation capacity of Ga allows for obtaining
diverse populations. Although more advanced alternatives to Ga have emerged in recent years, they are well
known by the research team. Because of this, the modifications of the code needed to implement the training of
the network along with the optimization process will be faster. These are the reasons that motivate the team to
choose the genetic algorithms.

A Ga is composed of agents. The population is defined as a group of agents. In each generation (iteration) several
processes are applied to the agents: crossover, mutation and elite count [11]. In the same way that the structure of
the neuronal network has been decided in the previous section, the size of the population and the percentages that
are destined to each process have been decided after testing with different values.

The configuration chosen allows us to obtain a diverse population. At the same time, the algorithm converges
fast. Table 2 shows an overview of the parameters of the Ga.

Table 2. Ga Parameters summary

In this work half of the population goes through the process of mutation, while the ten agents with the lowest
cost survive and pass to the next generation. The crossover process is applied to the rest of the population. The
total size of the population is 500 agents. Finally, each agent will be composed of two variables.

Finally, a maximum of 500 generations is set as a stop criterion. The other stop criterion will be the network
training error. The following section details how this new stop criterion is implemented.

843
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

4. STOP CONDITION
The network training is performed every 50 iterations of the optimization algorithm. In each iteration, the points
obtained by Ga are stored. In this way, enough data is obtained to perform a correct training. The training
validation error is used as a stop criterion. If it is lower than 2 the optimization stops.
These values are decided by the research team. As in the case of Ga and ANN parameters, the stop criterion was
tested with different combinations. The parameters presented in this paper are the ones that obtained the best
results. Fig. 1 shows the flow chart of the optimization process:

Fig.1.Optimization process flow chart

5. APPLICATION TO BENCHMARK-FUNCTIONS
Four functions are optimized by the algorithm with the new stop criterion before applying it to a real case. Table
3 shows these functions and their characteristics. Four functions are optimized by two methods. First by the
genetic algorithm without the new stop criterion. Second, the stop criterion is added, and the optimization process
is repeated. Each function is optimized 20 times. Finally, the results of both methodologies are compared.
Table 3. Benchmark functions for the testing process

Table 4 shows the results of the first method. It can be seen that the algorithm converges to the local minimum
in all cases. Furthermore, the standard deviation of the test is small.

Table 4. Results of the optimization without the new stop criterion

844
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 5 shows the results of the second method. It can be seen that the algorithm still reaches the minimum.
Despite the value reached in the first method is closer to the minimum, the result obtained by the second method
is acceptable. The standard deviation increases with respect to the first method, nevertheless, it is an acceptable
result.
Table 5. Results of the optimization with the new stop criterion

Two columns summarizing the result of the training are added. A net is obtained in each of the twenty repetitions.
The added columns show the validation error of the best obtained network and the median of the error of the
twenty networks.
Different results can be observed. Precise models are achieved for soft contouring functions such as Sphere or
Rosenbrock. By contrast, for irregular shape functions such as Shubert and Ackley, the models obtained are
inconsistent.

In the first method, all the executions of the algorithm reached 500 iterations. In contrast, in the second method,
the number of iterations is reduced. Thanks to the new stop criterion, more than half of the runs are reduced in
the case of smooth shape functions.
In Table 6. the iteration reduction is shown. In the left column, the iteration reached in the best case is presented.
In the second column, the iteration reached in the worst case.

Table 6. Iteration reached with the second method

In the case of sharp shape functions, the results are different. On one hand, a good enough model of the Shubert
function was not achieved during the optimization. So, the process has not stopped and the limit of iterations has
been reached. The results of the optimization of the Ackley function are acceptable. They are similar to those
obtained in the optimization of the Sphere and Rosenbrock functions.
In Fig. 2 it can be seen in which iteration the optimization process was stopped for the Sphere function.
Rosenbrock and Ackley functions have similar results. Shubert function reached the last iteration in all the runs
of the algorithm.
The new stop criterion reduces the number of iterations of the optimization algorithm. The computational cost of
the optimization is reduced. However, network training also has a high computational cost. The algorithm with
the new stop criterion takes longer to optimize the function.
However, this new proposal has the advantage that the optimization stops when the function to be optimized is
known sufficiently well. At the end of the process, the optimal values of the function and the model of that
function are obtained.

845
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

7. CONCLUSIONS

The new stop criterion allows saving iterations when optimizing smooth shape functions. The model obtained,
in these types of functions, during the optimization process has a small error. In the future, the work can advance
improving the training of the network in such a way that the algorithm is able to optimize and learn sharp
functions. The algorithm is also capable of optimizing and learn complex functions such as the impingement
plate design. However, the accuracy of the model can be improved.
In conclusion, an enchantment of the training process would improve the performance and the applicability of
the new approach.

Number of times the algorithm stops in a given iteration when optimizing the Sphere function

ACKNOWLEDGMENT

The Regional Development Agency of the Basque Country (SPRI) is gratefully acknowledged for economic
support through the research project KK-2018/00109 (ELKARTEK Program).

REFERENCES

[1] J. Manuel Lopez-Guede, J. Antonio Ramos-Hernanz, E. Zulueta, U. Fernandez-Gamiz and F. Oterino, '"Systematic
modeling of photovoltaic modules based on artificial neural networks," Int J Hydrogen Energy, vol. 41, no. 29, pp.
12672-12687.
[2] D. Karaboga and E. Kaya, '"Adaptive network based fuzzy inference system (ANFIS) training approaches: a
comprehensive survey," Artif.Intell.Rev., vol. 52, no. 4, DEC, pp. 2263-2293.
[3] A. Shafaei and G.R. Khayati, '"A predictive model on size of silver nanoparticles prepared by green synthesis method
using hybrid artificial neural network-particle swarm optimization algorithm," Measurement, vol. 151, FEB, pp.
107199.
[4] M. Jain, V. Singh and A. Rani, '"A novel nature-inspired algorithm for optimization: Squirrel search algorithm," Swarm
Evol.Comput., vol. 44, FEB, pp. 148-175.
[5] M. Dorigo, M. Birattari and T. Stuetzle, '"Ant colony optimization - Artificial ants as a computational intelligence
technique," IEEE Comput.Intell.Mag., vol. 1, no. 4, NOV, pp. 28-39.
[6] M. Abdel-Basset and L.A. Shawky, '"Flower pollination algorithm: a comprehensive review," Artif.Intell.Rev., vol. 52,
no. 4, DEC, pp. 2533-2557.
[7] S. Mirjalili, S.M. Mirjalili and A. Hatamlou, '"Multi-Verse Optimizer: a nature-inspired algorithm for global
optimization," Neural Comput.Appl., vol. 27, no. 2, FEB, pp. 495-513.
[8] H.R. Ansari, M.J. Zarei, S. Sabbaghi and P. Keshavarz, '"A new comprehensive model for relative viscosity of various
nanofluids using feed-forward back-propagation MLP neural networks," Int.Commun.Heat Mass Transfer, vol. 91,
FEB, pp. 158-164.
[9] A. Krizhevsky, I. Sutskever and G.E. Hinton, '"ImageNet Classification with Deep Convolutional Neural Networks,"
Commun ACM, vol. 60, no. 6, JUN, pp. 84-90.
[10] Y. Zhang, R. Xiong, H. He and M.G. Pecht, '"Long Short-Term Memory Recurrent Neural Network for Remaining
Useful Life Prediction of Lithium-Ion Batteries," IEEE Trans.Veh.Technol., vol. 67, no. 7, JUL, pp. 5695-5705.
[11] D. Beasley, D.R. Bull and R.R. Martin, '"An overview of genetic algorithms: Part 1, fundamentals," Univ.Comput.,
vol. 15, no. 2, pp. 56-59.

846
 HIGH TEMPERATURE LEAD ACID BATTERY SOC AND SOH
CHARACTERIZATION BASED ON ELECTROCHEMICAL
IMPEDANCE SPECTROSCOPY DATA

Javier Olarte
Center for Cooperative Research on Alternative Energies (CIC energiGUNE), Basque Research and Technology
Alliance (BRTA), Parque Tecnológico de Alava, Albert Einstein 48, 01510 Vitoria-Gasteiz, Spain. Bcare, Vitoria,
Spain University of the Basque Country, Vitoria, Spain, [email protected], ORCID: 0000-0002-4050-6845

Ekaitz Zulueta
University of the Basque Country, Vitoria, Spain, [email protected], ORCID: 0000-0001-6062-9343

Raquel Ferret
Centro de Investigación Cooperativa de Energías Alternativas (CIC energiGUNE), Basque Research and Technology
Alliance (BRTA), Vitoria, Spain, [email protected], ORCID: 0000-0002-7552-6467

Erol Kurt
Gazi University, Technology Faculty, Department of Electrical and Electronics Engineering, Ankara, Turkey,
[email protected], ORCID: 0000-0002-3615-6926

Jaione Martínez de Ilarduya

Bcare, Vitoria, Spain, [email protected], ORCID: 0000-0001-8304-3388

Jose Manuel Lopez-Guede

University of the Basque Country (UPV/EHU), Vitoria-Gasteiz, Spain, [email protected], ORCID: 0000-0002-5310-
1601

Olarte, J, Zulueta, E, Ferret, R, Kurt, E, Martínez de Ilarduya, J, Lopez-Guede, JM. High

temperature lead acid battery SOC and SOH characterization based on Electrochemical
Cite this paper as:
Impedance Spectroscopy data. 8th Eur. Conf. Ren. Energy Sys. 24-25 August 2020, Istanbul,
Turkey

Abstract: The development of tools for examining and predicting battery performance represents a
significant challenge for the field of monitoring, as it is dependent on nondestructive evaluation
techniques to provide key behavior descriptors. At the research level, an innovative path is marked
in the development of sensors that allow monitoring at cell level and undoubtedly, Electrochemical
Impedance Spectrometry (EIS) is one of the most promising tools in parallel with the development
of advance multivariable models that integrate environment and internal battery information. In
this article, we present some promising in operando experimental results of EIS measurements for
a set of high temperature lead acid battery that will allow further development of electrochemical
models to estimate the battery state of health (SOH) and state of charge (SOC) based on the
impedance data monitoring. By the present approach, it is observed that the combination of
different impedance modulus measurements at different frequencies, at 0.1 Hz and 31 Hz can be
used for the implementation in a smart electrochemical model to detect the charge and health status
of a high temperature lead acid battery. Further extraction and analysis of impedance data variables
will be very valued for the advance in predictive modelling of high temperature lead acid batteries.
Keywords: Lead-Acid batteries, Electrochemical Impedance Spectroscopy, Electrochemical model
© 2020 Published by ECRES

847
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Nomenclature
EIS Electrochemical Impedance Spectroscopy
SOH State of Health
SOC State of Charge

1. INTRODUCTION

The biggest challenges from the point of view of lead-acid battery operation and maintenance is to determine
the health status throughout its service life with the best availability and reliability rates, as well as to minimize
incidents and optimize costs during operation As the requirements faced by the implementation of the batteries
increase, both in terms of demand and complexity, the ability to understand, control, and predict battery
performance becomes ever more urgent. It seems axiomatic that the identification and characterization of battery
electrochemical models is a crucial part of battery life prediction, and further control and understanding of the
battery itself. These electrochemical models are based on the interpretation of battery electrochemical behavior
with respect to a wide range of battery properties, including performance, life span, and safety, in which
Electrochemical Impedance Spectroscopy (EIS) plays a remarkable role [1,2,3,4].

Several authors review the importance of identifying the correct parameters and models that allow the
management of lead acid batteries, and other technologies, in the most appropriate, accurate and added low-cost
way. In the review presented by Unguren et al. [5] a valuable updated comparison of the different types of
models that have been used, mainly, for electric mobility applications is included. This work focuses on dynamic
identification of electrochemical model parameters, as suggested by Kwiecien et al. [4]. In the work of Lin et al.
[6] is included an overview of the latest work being done in the world of electric mobility in which despite the
enormous efforts made, there is still room for improvement in order to optimize the algorithms of battery
management. Possibly one of the positive messages of this report is the importance of gaining knowledge
collecting as much experimental data as possible in order to implement incremental improvements of the models.
Indeed, more relevant models, which range from simple one-dimensional models to highly complex multi-
dimensional coupled ones, as suggested by Olarte et al. [7] will be made by in operando electrochemical data,
as in the case of the present work.

Several authors have reported different ways to estimate this SOC and SOH variables of the batteries, which can
be divided in different categories as direct measurements, electrical and electrochemical models and adaptative
and machine learning methods [8, 9, 10, 11] independently of the target technology. For example, Kumar et al.
mentioned the problematic of State of Charge (SOC) and State of Health (SOH) identification in batteries
oriented to its implementation in vehicles, which is still not accurate enough. They reviewed different SOC and
SOH indication algorithms and commercially available Battery Management System [8].

Among direct measurements, discharge test capacities can be found to detect capacity at the beginning of life of
the battery, or intermediate measurements to determine State of Health. However, this type of method, even
being precise, cannot be incorporated into an intelligent diagnostic detection system by itself [12].

Some electrochemical models can be based on coulomb counting which can be very accurate but only if the
initial input is valid, high accuracy current sensors are required [13, 14]. OCV measurement is used in
technologies as Lead-Acid, Lithium-ion (depending on which one) or Zn/Br, which is based on the relation of
the OCV and SOC. This Open Circuit Voltage has to be measured in off-line conditions usually, but it can be
online if OCV is inferred from terminal voltage measurements or suitable models.

Electrochemical Impedance Spectroscopy is other of the methods that can be used for real time prediction by
interpreting parameters from the spectra. Although, it is necessary to take into account that the electrochemical
model is difficult to be implemented and it is very specific for each technology [13], so it is necessary to have
an appropriate protocol for extraction of experimental data in each electrochemical energy storage case.

848
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Regarding to the EIS impedance measurement, some comments are done by different authors. According to
Baccouche et al. this method is very reliable as it presents high accuracy, however it could be very time
consuming and hard to implement [14]. According to Ribera-Barrera et al., this estimation method is suitable
only for identical charging conditions, as for example the case of stationary cases [10]. According to Chang,
impedance measurements are very versatile from the point of view that all the parameters about which it can
give information, as state of charge, degradation of the battery, failure modes, etc. [9].

The results presented in the paper correspond to a more ambitious project that has identified the possibility of
improving the certainty of the estimation of the health status, as well as the capacity of the battery, with a new
model, with special emphasis on the correlation of the EIS data and the different failure modes. The process of
identification and use of parameters will be interactive and will be refined in future works, allowing the
integration of more experimental variables in the model. Preliminary, it is presented the evolution of impedance
modulus at different frequencies as one of the first variables to be included in the smart electrochemical model.
This model will allow further development of electrochemical models to estimate the battery state of health
(SOH) and state of charge (SOC) based on the impedance data and equivalent circuit parameters identification.

3. EXPERIMENTAL SETUP

Several electrochemical experiments have been carried out, including measurements of charge current, no-load
voltages, electrochemical impedance and capacity of a set of high temperature lead acid batteries of the same
model (80 Ah/12 V).

Firstly, standard characterization is performed in order to extract electrical parameters as inputs for the
electrochemical model in terms of SOC characterization. Secondly, ageing tests are developed, which include,
electrochemical characterization based on load profile of stationary use case application inducing accelerating
ageing by means of different temperature profiles and load frequencies. From these accelerated aged tests,
suitable extraction of parameters/signals from the electrochemical system is extracted to be integrated in the
modelling for the determination of SOH and failure modes of the high temperature lead acid battery. This testing
protocol includes voltage, current, temperature (experimental temperatures from 25 ºC to 50 ºC) and EIS spectra
identification (measured frequencies from 10 kHz to 0.01 Hz) at different operational conditions. Using the
gathered data, tendencies are analyzed, and electrochemical complete models are developed.

4. RESULT AND DISCUSSION

In this section, from the full set of experiments only some representative data is presented corresponding to one
out of the variables from impedance measurements from the high temperature lead acid battery model measured.
The data presented corresponds to the impedance modulus (Modulus IZI) measured at 25 ºC. The evaluation of
this variable is done at 0.01, 0.1, 31, 100 and 282 Hz, in order to look for the optimum frequency at which SOC
and SOH can be evaluated. Figure 1 illustrates the evolution of the impedance modulus at different health status
(Pristine, SOH 1, SOH 2, SOH 2, SOH 3, SOH 4, SOH 5, SOH 6) and SOC, at the mentioned frequencies.

For the sake of simplicity, only data corresponding to one high temperature lead acid battery reference is
presented although the number of batteries aged during the experiments were higher.

From Fig. 1(a) and Fig. 1(b), corresponding to 0.01 Hz and 0.1 Hz frequencies, it is observed that impedance
modulus values along 0-100% of SOC present similar values, except from 80% to 100%. Range in which there
is an evident increment of the variable analyzed. Values of impedance modulus higher than 0.05 and 0.025 in
the case of 0.01 Hz and 0.1 Hz, respectively, notes the fully charged state of the battery. At 31, 100 and 282 Hz,
the impedance modulus evolution is clearer. By increasing the SOC, there is a slight decrease of the impedance
modulus, reaching values close to 0.03 in all the frequencies mentioned, which could be evaluable.

849
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Regarding the SOH estimation, it is observed that, independently of the frequency analyzed the evolution of the
impedance modulus is similar for the SOC range. However, there are evaluable differences at 100% SOC at 0.1
Hz. As the degradation of the lead acid battery increments, the impedance modulus value is increasing almost
linearly.

It is observed that the combination of different impedance modulus measurements at different frequencies, at
0.1 Hz and 31 Hz, for example, can be used for the implementation in a smart electrochemical model to detect
the charge and health status of a high temperature lead acid battery. The charge state of the battery can be
detected by the evaluation of impedance modulus at 31 Hz, while the value of this variable at 0.1 Hz can be used
to correctly limit the value of the state of charge if it is between 80 and 100%.

Figure 1. Impedance modulus for different SOC (%) and SOH (%) measured at (a) 0.01Hz, (b) 0.1 Hz, (c) 31 Hz, (d) 100
Hz and (e) 282 Hz.

Respect to the identification of SOH, as presented in Fig. 2, if the impedance modulus at 0.1 Hz is higher than
0.030 Ω, the battery will be close to its End of Life. The fact that this variable can be evaluated in states of

850
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

charge close to the fully charged state of the battery could help it to be implemented, in a remote-control model,
in target applications based in floating mode.

90
80
70
60

SOH (%)
50
40
30
20
10
0
0.020 0.025 0.030 0.035 0.040
Modulus IZI

Figure 2. Impedance modulus values at 0.1 Hz and 100% SOC for different SOH (%)

5. CONCLUSIONS

In summary, we have identified some promising correlation data in order to build up a robust estimator of the
SOC and SOH as a function of a set of EIS measurement. Meanwhile, by demonstrating clearly that the results
are robust enough, there is still room for improvement based on experimental data to be gathered from
electrochemical experiments and Post-Mortem analysis. Total identification of EIS parameters evolution for
high temperature lead acid batteries would not only be a significant advance for the field in and of itself, but
also will be key to the development of remote monitoring of batteries and automated validation—particularly
with respect to predicative modelling of performance.

ACKNOWLEDGMENT

The authors would like to express their gratitude to CIC energiGUNE for the facilities and to Joseba Segura and
Gorka Ramírez for their technical and scientific contributions to this work.

REFERENCES

[1] Aksakal C. and Sisman A. On the Compatibility of Electric Equivalent Circuit Models for Enhanced Flooded Lead Acid
Batteries Based on Electrochemical Impedance Spectroscopy. Energies, 2018, vol. 11, no. 1, p. 118. DOI:
10.3390/en11010118.
[2] Andre D., Appel C., Soczka-Guth T., and Sauer D. U. Advanced mathematical methods of SOC and SOH estimation
for lithium-ion batteries. J. Power Sources, 2013, vol. 224, pp. 20–27. DOI: 10.1016/j.jpowsour.2012.10.001.
[3] Badeda J., Kwiecien M., Schulte D., and Sauer D. Battery State Estimation for Lead-Acid Batteries under Float Charge
Conditions by Impedance: Benchmark of Common Detection Methods. Appl. Sci. 2018, vol. 8, no. 8, p. 1308. DOI:
10.3390/app8081308.
[4] Kwiecien M., Badeda J., Huck M., Komut K., Duman D., and Sauer D. Determination of SoH of Lead-Acid Batteries
by Electrochemical Impedance Spectroscopy. Appl. Sci. 2018, vol. 8, no. 6, p. 873. DOI: 10.3390/app8060873.
[5] Ungurean L., Cârstoiu G., Micea M. V., and Groza V. Battery state of health estimation: a structured review of models,
methods and commercial devices: Battery State of Health Estimation: a Structured Review. Int. J. Energy Res. 2017,
vol. 41, no. 2, pp. 151–181, DOI: 10.1002/er.3598.
[6] Lin Q., Wang J., Xiong R., Shen W., and He H. Towards a smarter battery management system: A critical review on
optimal charging methods of lithium ion batteries. Energy, 2019, vol. 183, pp. 220–234. DOI:
10.1016/j.energy.2019.06.128.
[7] Olarte J., Dauvergne J. L., Herrán A., Drewett N. E., Bekaert E., Zulueta, Ferret R. Validation of thermal imaging as a
tool for failure mode detection development. AIMS Energy, 2019, vol. 7, no. 5, pp. 646–659, DOI:
10.3934/energy.2019.5.646.

851
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

[8] Kumar, B., Khare, N., & Chaturvedi, P.K. Fpga Design Scheme For Battery Soc and Soh Algorithms For Advanced
BMS. International journal of engineering sciences & research technology, 2017, 6 pp. 263-279 (7). DOI:
10.5281/ZENODO.827494.
[9] Chang W. Y. The State of Charge Estimating Methods for Battery: A Review. ISRN Appl. Math., 2013, vol. 2013, pp.
1–7. DOI: 10.1155/2013/953792.
[10] Rivera-Barrera J., Muñoz-Galeano N., and Sarmiento-Maldonado H. SoC Estimation for Lithium-ion Batteries:
Review and Future Challenges. Electronics, 2017, vol. 6, no. 4, p. 102. DOI: 10.3390/electronics6040102.
[11] Zhang R., Xia, B., Li B., Cao L., Lai Y., Zheng W., Wang H. and Wang W. State of the Art of Lithium-Ion Battery
SOC Estimation for Electrical Vehicles. Energies, 2018, vol. 11, no. 7, p. 1820. DOI: 10.3390/en11071820.
[12] Srdjan M. L., S. M., Cao J., Bansal R. C., Rodríguez F., and Emadi A. Energy storage systems for automotive
applications. Industrial electronics, IEEE Transactions on, 2008, 55(6):2258-2267. DOI: 10.1109/TIE.2008.918390
[13] Piller S., Perrin M., and Jossen A. Methods for state-of-charge determination and their applications. J. Power Sources
2001, vol. 96, no. 1, pp. 113–120. DOI: 10.1016/S0378-7753(01)00560-2.
[14] Baccouche I., Jemmali S., Mlayah A., Manai B., and Amara N. E. B., Implementation of an Improved Coulomb-
Counting Algorithm Based on a Piecewise SOC-OCV Relationship for SOC Estimation of Li-Ion Battery. International
Journal of Renewable Energy Research-IJRER, 2018, v. 8, N. 1. 178-187.

852
 EFFECTS OF SUBSTRATES ON THE GROWTH OF BETA
VULGARIS SUBSP. VULGARIS IN HYDROPONIC SYSTEMS
Tran Trong Nhan
Institute of Research and Development,Duy Tan University, Danang 550000, Vietnam, Faculty of Civil Engineering,
Duy Tan University, Danang 550000, Vietnam, Email: [email protected], https://orcid.org/0000-0002-
6872-5521

Ngo Van Tuan

3Hong Bang International University, Ho Chi Minh City, Vietnam, email:[email protected],
https://orcid.org/0000-0002-4590-6944

Nguyen Phuc Thien

HoChiMinh City University of Technology (HCMUT), Ho Chi Minh City, Vietnam, Vietnam National University, Ho
Chi Minh City, Vietnam,email:[email protected], https://orcid.org/0000-0002-4639-1639

Nhan, TT, Tuan, NV, Thien, NP. Effects of substrates on the growth of BETA VULGARIS
Cite this paper as: SUBSP. VULGARIS in hydroponic systems. 8th Eur. Conf. Ren. Energy Sys. 24-25 August
2020, Istanbul, Turkey

Abstract: This research focuses on types of substrate materials (rice husk, coconut fiber, sand…), the rate of
media mix between coconut fiber, rice, and concentrations of Thiamine HCl (vitamin B1) on the
growth of Rainbow Vinegar (Beta vulgaris subsp. Vulgaris). In the studies of nutritional absorption
and metal toxicity in the roof, it is important to grow plants without technical damage. The results
showed that Hoagland solution combined coconut fiber, sand, and Thiamine HCl (vitamin B1)
suitable for Beta vulgaris subsp. Vulgaris. The Rainbow grows remarkably in the period from 20
days to 30 days for faster, cleaner, and when cultivating in the soil environment. The method
presented here provides an additional model system for the studies on nutrient growth interaction
in Beta vulgaris subsp. Vulgaris and a simple hydroponic culture system for Beta vulgaris subsp.
Vulgaris to obtain a healthy plant having a well-developed root system with many lateral roots.
Keywords: Substrate, Hoagland solution, Beta vulgaris subsp. vulgaris, hydroponic system
© 2020 Published by ECRES

1. INTRODUCTION

BEETROOT is a very popular vegetable in eastern and central Europe but is much less important in Western
Europe and the USA, where it is known as garden beet. It is grown for its swollen root and is the horticultural
form of Beta vulgaris subsp. vulgaris (2n =18) which also includes the agricultural types sugar beet, mangold,
and fodder beet. The species also includes chard (subsp. Cicla) and sea beet (subsp. Maritima), which is thought
to be the ancestor of most cultivars [1].

The earliest form of domesticated beet was leaf beet. The leaves were eaten and the roots were used for medical
purposes only although much of this research applies to beetroot breeding, the minor status of the horticultural
crop has resulted in much less effort being expended on breeding research. Consequently, there is often an
appreciable delay before new developments in sugar-beet breeding are applied to beetroot. This chapter
concentrates upon research that has been carried out directly upon beetroot [2].

For the molecular biological studies in higher plants, beta vulgaris possesses many suitable properties, such as
a short life cycle and high amenability to gene manipulation.

853
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

The hydroponic system is also important for the studies of growth nutrient interactions and several hydroponic
culture systems have been established for fully developed plants of beta Vulgaris. Following some reports, we
developed a hydroponic apparatus for young seedlings of beta Vulgaris [3].

In this report, we describe a new simple method of hydroponic culture that enables beta Vulgaris to develop
many lateral roots with high viability. Using the growth of the seedlings with this apparatus, the effects of the
short time response to toxic metals were examined.

Water is, of course, always the main component of growing plants. But the major portion, usually about 90
percent, of the dry matter of most plants is made up of three chemical elements: carbon, oxygen, and hydrogen.
Carbon comes from the air, oxygen from the air and water, and hydrogen from water. In addition to these three,
plants contain other elements, such as nitrogen, phosphorous, potassium, and calcium, which they obtain from
the soil. The soil then supplies a large number of chemical elements, but they constitute only a very small portion
of the plant. Yet the various elements that occur in plants in comparatively small amounts are just as essential
to growth as those which compose the bulk of plant tissues.

2. MATERIALS AND METHODS

Culture of Beta vulgaris subsp.vulgaris. Beta vulgaris subsp. vulgaris seeds were labeled and washed with
sterile water then germinated for 24 h on coir.

Hydroponic system. For hydroponic growth, germinating sterilized seeds were planted (3 seeds .pot ') to a depth
of 2 cm in plastic pots (I0cnl top diameter, 5 cm bottom diameter, and I0 cm height). During the 1 week, the
seedlings were irrigated daily with distilled water and each pot was placed in a nutrition container (15 cm height,
12cm diameter pot) with 200 mL of a modified Hoagland solution.

The solution contained the following: KNO3: 3.9 g.L-1; KH2PO4: 26.9 g.L-1; K2SO4: 4,3 g.L-1; MgSO4 : 30.8
g.L-1; ZnSO4 : 0.015 g/L-1; H3BO3: 0.02 g.L-1; MnSO4: 0.115 g.L-1; CuSO4 : 0.01 g/L-1; NH4Mo7O24:
0.003 g.L-1; FeSO4: 0.64 g.L-1; Na-EDTA. : 0.86 g.L-1

The solution in the container was changed every 2nd day to keep the roots under a constant mineral
concentration.

In one set of experiments, in hydroponics (two experiments, each treatment in 20 replicates), the plant shoots
were grown under normal winter environmental conditions in Rehovot during November, February, for 110
days (10-11h light, 10 25°C). The nutrient containers were placed in a water bath to keep the roots under a
constant temperature of 22°C. In the second set of experiments (two experiments each treatment in 20 replicates),
the plants were grown for 45 days in a phytotron at the Faculty of Agriculture, Rehovot, at a constant temperature
of 22°C (day and night) with a photoperiod of 9 h light: 15 h dark

3. RESULTS

In this experiment, the primary roots grew for a longer period than those cultured by the methods developed for
monitoring the growth of the primary root. The viability of the root is important for the study of root function in
nutrient uptake and to examine the damage symptoms of the root under stress conditions [4]

854
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table1. Effects of nutrient medium on the growth and development of Rainbow (Beta vulgaris subsp. Vulgaris) during 30
days
Number of Plant height (cm) Leaf length (cm) Leaf width (cm) Root length (cm)
NT
10 Leaf/plant
20 30 10 20 30 10 20 30 10 20 30 10 20 30
days day 6,00daysc days days days daysc day 11,1
day 3,12day 3,97 days
day 6,03 dayd day days
A0 3,80 4,50 9,58 14,0 20,30 4,62 9,01 - - 20,2
ab sc b
A1 4,03 6,00 7,10b 11,5 8c
19,7 d
29,03 7,85a 12,8sd 7sd
17,8 sc
4,35 sd
6,10 10,1 s- s- 5c
25,1
a a
A2 3,67 6,00 7,83a 9,50 2a 2a
19,9 b
35,23 0b
6,87b 13,8 2b
21,8
a
3,36 7,37 b
2b
13,7 - - 3b
27,1
b a b b c a a d a a b a a c a
A3 3,67 5,50 7,00 8,60 0
17,8 25,67 5,80 10,4 0 7
15,3 2,85 5,27 7,92 5 - - 24,7
b
Note: in the sameb column, the average8values
b c
with 7c
the characters 0cb, c, d.)
(a, d c
followed 8b
the same with no statistical
difference (P <0.05) (-) : No figures available

Thus, this experiment showed that using the nutritive medium proposed by Hoagland will help Rainbow
Rainbow (Beta vulgaris subsp. Vulgaris) grow and develop better than when using water and lips. the nutritional
school proposed by Knop or Alan Cooper. This can be explained by Hoagland's proposed nutritional
environment, which contains all the macronutrients and micronutrients needed for the Rainbow mustard (Beta
vulgaris subsp. Vulgaris) to grow and develop, in When Knop proposed the nutritional environment in addition
to macronutrients, the micronutrients needed for plants also contain too many types of vitamin that can hinder
the growth and development of Rainbow Chard (Beta vulgaris subsp vulgaris). According to [5], when studying
the effects of several nutritional environments on the growth and development of broccoli grown by hydroponics
technique, in the treatment using mixed nutrition solution according to the formula Hoagland gives the best
results.

Figure 1. Effect of Hoagland hydroponic nutrition on the growth and development of Rainbow mustard (Beta vulgaris
subsp. Vulgaris after 10; 20; 30 days (left to right respectively).

Effect of media on the growth and development of Rainbow mustard (Beta vulgaris subsp. Vulgaris) after 30
days

The analysis results in Table 2 and Figure 2 show that different types of media affect the growth of Rainbow
Chard.

Table 2: Effect of media on the growth of Rainbow Chard (Beta vulgaris subsp. Vulgaris) after 30 days
N Number of Plant height (cm) Leaf length (cm) Leaf width (cm) Root length (cm)
T 10 Leaf/plant
20 30 10 20 30 10 20 30 10 20 30 10 20 30
day day
B0 3,80 4,5 days day days
6,00 9,85 14,0 days
20,3 days
6,42 day9,0 day11, day 3,9 days
3,1 day 6,03 da - day
- day
20,2
B1 3,50 sa 0sb
3,8
c
7,10 sa
6,27 8b
11,5 4d
21,8 b
3,78 5,7 2sd s d 1,9
17
11, 2sb 7sb
2,6 c
5,15 ys
- s- 5sb
17,9
b
B2 3,67 3c
6,0 b
7,83 9,50c
5c
19,9 5c
35,2
d
6,87 8a
13, 9c
21, 7d
3,3 3d
7,3 d
13,2 - - 4d
27,1
ab a a b a b a a ac a a a a a
B3 3,67 4,0 0 7,00 8,60 13,5 0 3
23,8 4,69 7,4 8 87
14, 2,46 7
3,5 5
6,45 - - 19,5
ab
0c b c
8b c
52c 8c 7c b
5c

855
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Note: in the same column, the average values with the characters (a, b, c, d) followed the same with no
statistical difference (P <0.05) (-) : No figures available

The results of the study showed that in 3 types of media: rice husk, coir, sand, which use soil as a control,
coconut fiber is the most suitable medium for Rainbow Vulgaris (Beta vulgaris subsp. Vulgaris) to grow and
grow in rooted hydroponic systems. The coir substrate can retain good moisture, have high ventilation, and is
capable of decomposing, these are favorable factors for good plant growth and development. This result is
similar [6]when evaluating the effectiveness of this experiment is similar [7]when evaluating the effectiveness
of Types of growing medium for hydroponic cultivation of white Hue (Polianthes tuberosa L.) tissue culture

Figure 2. Effect of coconut fiber media on the growth and development of Rainbow mustard (Beta vulgaris subsp. Vulgaris)
after 10; 20; 30 days (from left to right respectively).

Effect of the incidence of mixing medium between coconut fiber and rice husk on the growth and development
of Rainbow chard (Beta vulgaris subsp. Vulgaris) after 30 days

The analytical results in Table 3 show that the mixing ratio of medium type between coir and rice husk influence
not much different on the growth of Rainbow chard.

Table 3. The effect of the mash mix of coconut fiber and rice husk on the growth and development of Rainbow Rainbow
(Beta vulgaris subsp. Vulgaris) after 30 days
N Number of Plant height (cm)Leaf length (cm) Leaf width (cm) Root length (cm)
T 10 Leaf/plant
20 30 10 20 30
10 20 30 10 20 30 10 20 30
day day
C0 3,67 6,0 days daya 19,9
7,83 9,5 daysa 35,23
days
day day days
6,8 13,8 21,8 3,63 day 7,37 day days
13,2 day - days
- days
27,1
C1 3,83 sa 0sa
4,1 b a
8,5 s
8,10 17,6 30,65
a
7sa
5,8 sa
8,94 7a
20,9 sa
2,87 sa
4,75 5a
10,3 s- - 0a
20,1
b
C2 3,17 3,1 7b b
5,5d 5,88 13,22b b
21,93 4,75b b
7,75 12,85b c b
2,33 4,22 6,82 2b c - - 7b
17,5
b
C3 3,67 7d
3,8 c
6,00 8,00 3d c 23,17
15,3 d
0c
4,9 c
8,67 8c
13,2 3,00d c
4,68 6,23d - - 8c
17,3
a
3 column, the average values with5the characters2 (a, b, c, ...) followed the same with no 3c
Note: in the same
c c b c c b c b b

statistical difference (P <0.05), (- ): No figures available

Thus, this experiment has not seen the effectiveness of the mixing ratio of 70:30, 50:50, 30:70 media between
coconut fiber and rice husk compared to using coconut fiber medium. This result is different [6] when evaluating
the effectiveness of vegetable lettuce growing medium. This result is also different from [8] when studying the
effect of growing media on the growth and off-season green vegetables yield using circulating hydroponic
technology. This explanation can be explained by the high husk content of potassium husk and SiO2, which is
not suitable for the development of Rainbow chard (Beta vulgaris subsp. Vulgaris). Besides, the husk also
absorbs heat due to its high carbon content, which causes the growing medium and nutrient environment to
affect root growth and absorption. Besides, the structure of the coir medium can hold moisture well, is capable

856
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

of decomposing, and still tighter than rice husks, thereby helping the tree stand and absorb nutrients easily, These
may be the factors that give more advantages to the coir body compared to when mixing between coir and rice
husk.

The effect of the media mix ratio between coir and sand on the growth and development of Rainbow chard (Beta
vulgaris subsp. Vulgaris) after 30 days

The analysis results in Table 4 and Figure 3 show that the mixing ratio of coconut fiber and sand has different
effects on the growth of the Rainbow chard.

Table 4. The effect of the medium mixing ratio between coir and sand on the growth and development of Rainbow mustard
(Beta vulgaris subsp. Vulgaris) after 30 days
Number of Plant height (cm) Leaf length (cm) Leaf width (cm) Root length (cm)
NT
10 Leaf/plant
20 30 10 20 30 10 20 30 10 20 30 10 20 30
day 6,00day 7,83day 9,50 days days
day 19,90 35,23 6,87 day 13,8 days
day 21,87 3,63 day 7,37 days
day 13,25 a days day day
D0 3,67 - - 27,1
D1 4,33 sc sa
6,00 sa
6,83 sa
8,45
a b
17,37 22,13 sa
6,00 sa
10,6 a
13,88 sa
2,80 sa
5,73 7,03c - s- sa
21,1
a a b b c c c
D2 3,83 5,17 5,83 8,47 14,88 19,88 6,558b
9,25
b c c
13,67 3,10 4,08 7,20c - - 8b
18,2
bc b c b d d b d b b d
D3 4,00 5,33 8,00 9,67 18,47 37,68 10,1 22,27 3,58 6,25 12,73b
7,10 - - 7c
27,5
b b a a b a a a a
3c
Note: in the same column, the average values with the characters
a b
(a, b, c, ...) followed the same with no 8
statistical difference (P <0.05), (- ): No figures available

This experiment showed that the efficiency of the 30:70 media mix between coconut fiber and sand media was
better than using coconut fiber media or mixing coconut fiber media and sand media. at 70:30 and 50:50 ratios.
This result shows that the substrate is firmer, able to support the tree, has good ventilation, and moderate
humidity when combining coconut fiber and sand at a ratio of 30:70.

Figure 3. The effect of the 30:70 media mix ratio on coir and sand on the growth and development of the Rainbow mustard
(Beta vulgaris subsp. Vulgaris) after 10; 20; 30 days (from left to right respectively).

Effect of thiamine HCl (vitamin B1) concentration on the growth and development of Rainbow mustard (Beta
vulgaris subsp. Vulgaris) after 30 days.

The analytical results in Table 4 and Figure 3 show that the concentrations of thiamine HCl (vitamin B1) have
different effects on the growth of Rainbow Chard.

857
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

Table 5. Effect of thiamine HCl concentration (vitamin B1) on the growth and development of Rainbow mustard (Beta
vulgaris subsp. Vulgaris) after 30 days
Number of Plant height (cm) Leaf length (cm) Leaf width (cm) Root length (cm)
NT
10 Leaf/plant
20 30 10 20 30 10 20 30 10 20 30 10 20 30
E0 3,67 6,00 7,83
day days day days9,50 19,90
days days35,2 6,87 daysa 21,87
day 13,8 days 3,63
day 7,37 13,2 day
day days - day- 27,1
days
a
E1 3,17 sb 3,96 sb
7,92 b
11,2 16,50
a 35,4
3c 5,42
as b
10,3 21,95 sa
2,52 sa
4,27 5b
13,3 s- s- 0c
28,6
c c b
E2 3,50 4,08 8,17 2a
9,53 c
15,68 5bc
36,2 c
5,30 3c
10,8 b
22,02 c
2,20 c
4,58 5b
13,4 - - 7b
21,5
b c b b d b c b b d b b
E3 4,00 4,42
b 8,83 11,4 17,58 39,8 7 6,25 10,03 24,52 2,93 5,42 14,0 5 - - 0d
30,3
a b a a b a b c d a b c a c a
E4 3,17 4,17 6,33 9,85 3 16,25 30,67 5,47 7,423 18,72 2,22 4,33 8,23 7 - - 22,6
Note: in the same column, the average values3 with the characters (a, b, c, d) followed the same with no 7b
c bc c b c d c c d c

statistical difference (P <0.05), (- ): No figures available

The results of this experiment showed that when thiamine HCl (vitamin B1) was added at a concentration of 0.6
g / l to the Hoagland nutrient medium, the Rainbow mustard (Beta vulgaris subsp. Vulgaris) in the soaked
hydroponic system. Roots will grow and develop better than without thiamine HCl (vitamin B1).

Figure 4. Effect of concentration of 0.6 g / l thiamine HCl (vitamin B1) on the growth and development of Rainbow mustard
(Beta vulgaris subsp. Vulgaris) after 10; 20; 30 days (from left to right respectively)

4. CONCLUSIONS

Three types of nutrient medium proposed by Knop, Hoagland, and Alan Cooper in the root system of Rainbow
Root (Beta vulgaris subsp. Vulgaris), the nutrient medium proposed by Hoagland is suitable for growing
vegetables. Rainbow. Of the 3 types of coir, husk, and sand media used in hydroponic systems soaked in the
roots of the Rainbow mustard (Beta vulgaris subsp. Vulgaris), the coir media are the best. When mixing coir
media with rice husk media at mixing ratios of 70:30, 50:50, 30:70, it is not suitable for rainbow chard. When
mixing coir media with sand media at mixing ratios of 70:30, 50:50, 30:70, the 30:70 mixing ratio is most
appropriate. Of the 4 concentrations of thiamine HCl (vitamin B1) are: 0.2 g / l, 0.4 g / l, 0.6 g / l and 0.8 g / l
used in capillary hydroponic systems for growing mustard Rainbow (Beta vulgaris subsp. Vulgaris) has a
concentration of 0.6 g / l in the most appropriate Hoagland field. Rainbow mustard (Beta vulgaris subsp.
Vulgaris) grown in root-grown hydroponics systems thrives at a period of 20 days to 30 days.

5. FUTURE WORKS

The most successful strategy of using bio-control agents against root pathogens in soilless systems may be as
protectants or prophylactic treatments. If the bio-control agent can be established before the pathogen
introduction, can it prevent infection or interfere with the production of pathogen inoculum that could spread
through the system? These are the questions that should be addressed experimentally. But, so far, one of the
drawbacks of inoculation experiments in greenhouses is that they do not reflect exactly what happens in a

858
 8th EUROPEAN CONFERENCE ON RENEWABLE ENERGY SYSTEMS ISTANBUL/TURKEY 24-25 August 2020

commercial greenhouse. Most systems start clean, but the pathogen is introduced sometime later, probably in
low population densities, in a random fashion and possibly many times during the cropping cycle. But in most
experiments, the bio-controls are added to the soilless system at the start, before adding one massive dose of
pathogen inoculum. Under these conditions, the bio-control agents cannot completely control disease, so most
bio-control treatments have disease severities midway between the inoculated non-protected control, and the
non-inoculated, healthy control. Growers expect near-complete disease control and may not accept simple
disease reduction. Experiments should be performed with many introductions of low population densities of the
pathogen. Under these conditions, the bio-control agent may prevent the explosive build-up in pathogen
populations. The bio-control agents may delay or reduce the establishment of the pathogen, keeping the pathogen
population densities below the economic threshold. We are also unclear about how the initial infection is
influenced by the stage of crop development. Are early infestations more important than later infestations, in
terms of disease? We also know very little about how environmental and plant factors such as temperature,
electrical conductivity, transpiration, and fruit set interact with the bio-control agents and pathogens to influence
disease. Finally, more work should be done on the use of booster applications of the bio-control agents to the
system over the whole cropping cycle. This may overcome the natural decline of some bio-control agents over
time and may maintain a suppressive state in the system. Given the ease of introducing nutrients into a fertigation
system, formulations of bacteria may also be developed that could be added to the injector system for rapid
dispersal in rock wool slabs.

In conclusion, biological control may offer another tool for disease management in soilless systems. Given the
lack of chemical alternatives and the innovative technology of these systems, many of the future applications of
biological control will be in these intensively managed greenhouse systems.

ACKNOWLEDGEMENT

We would like to thank Ho Chi Minh City University of Technology (HCMUT), VNU-HCM for the support
of time and facilities for this study.

REFERENCE

1.PINK, D.A.C., Beetroot Beta vulgaris subsp. vulgaris Genetic Improvement of Vegetable Crops, ed. B.O.B. G.
KALLOO. 1992, Great Britain: Pergamon Press Ltd
2.G. Kalloo, B.O.B., Genetic Improvement of Vegetable Crops. 2012, ENGLAND: PERGAMON PRESS.
3.Tomomi TODA, H.K.T.H., A Simple Hydroponic Culture Method for the Development of a Highly Viable Root System
in Arabidopsis thaliana. Bioscience, Biotechnology, and Biochemistry, 2014. 63(1): p. 210-212. DOI:
org/10.1271/bbb.63.210
4.Koyama, H., Toda, T., Kojima, H., and Hara, T, Direct obser¬vation of root-elongation of Arabidopsis thaliana seedlings
grown in hydroponic culture. Soil Sci. Plant Nutr, 1995. 41: p. 173-176. DOI: 10.1080/00380768.1995.10419573
5.Hà, N.T., Nghiên cứu một số giá thể phối trộn trồng cây cải bắp, cà rốt, dưa chuột trong nhà lưới tại Gia Lâm-Hà Nội,
in FACULTY OF AGRONOMY. 2010, Vietnam National University of Agriculture. p. 151-154.
6.Trần Thị Ba, B.V.T.v.T.N.L., Effects of different substrates, varieties, nutrient solutions on growth and yield of
hydroponic lettuce, Winter-spring 2007-2008, in Journal of Science. 2008. p. 339-346.
7.Nguyễn Hồng Ửng, N.B.T., Effects of substrates and hydroponic nutrient media on Polianthes tuberose from tissue
culture, in Journal of Science. 2009. p. 146-156.
8.Sơn, N.K.T., Nghiên cứu ảnh hưởng của một số loại dung dịch khác nhau đến sự sinh trưởng và phát triển của một số
cây rau - quả trong ký thuật thủy canh. 1996, Thai Nguyen University of Agriculture and Forestry.

859
 8. European Conference on Renewable Energy Systems Istanbul 24-25 August 2020

WE WISH TO SEE YOU IN NEXT

EVENT.

FOLLOW
www.ecres.net

860