Sustainable Energy Technologies and Assessments 48 (2021) 101561

Contents lists available at ScienceDirect

Sustainable Energy Technologies and Assessments

journal homepage: www.elsevier.com/locate/seta

Artificial intelligence assisted technoeconomic optimization scenarios of

hybrid energy systems for water management of an isolated community
Rasikh Tariq *, A.J. Cetina-Quiñones , V. Cardoso-Fernández , Hernández-López Daniela-Abigail ,
M. A. Escalante Soberanis *, A. Bassam , M. Vega De Lille
Unidad de Posgrado de la Facultad de Ingeniería, Universidad Autónoma de Yucatán, Av. Industrias No Contaminantes por Anillo Periférico Norte, Apdo. Postal 150,
Cordemex, Mérida, Yucatán, Mexico

A R T I C L E I N F O A B S T R A C T

Keywords: Water is an essential resource demanded worldwide and it is quite debatable owing to the economic, political,
Water-energy nexus and energy characteristics of any region. Off-grid water filtration plants are an alternative for communities where
Hybrid renewable energy systems transportation of freshwater becomes a real challenge due to a lack of infrastructure for the water potabilization
Potable water filtration
processes. For such potable water filtration plants, hybrid renewable energy systems (HRES) can be a viable
Wastewater treatment
solution to meet their energy demand meanwhile providing a sustainable water solution.
Artificial intelligence
Digital twin The main contribution of this work is the unique methodology, which starts with a sizing procedure of various
Optimization hybrid energy systems using a commercial software “Hybrid Optimization of Multiple Energy Resources (HOMER)”
Sensitivity analysis and spreadsheet algorithms, followed by a “Non-dominating Sorting Genetic Algorithm II (NSGA-II)” based multi­
objective optimization. Single-objective optimization scenarios contain photovoltaic installation capacity, wind
turbines, diesel generators, and battery energy storage systems including Pb-acid (Lead-acid), Li-ion (Lithium-ion),
and AGM (Absorbent Glass Mat) technologies as design variables to maximize the cost of electricity or net-present-cost.
Multiobjective optimization also involved environmental (CO2 emissions i.e. carbon dioxide emissions) and water cost
indices as an additional packet to single-objective optimization scenarios. Afterward, a multicriteria decision-making
tool using “The Order of Preference by Similarity to Ideal Solution (TOPSIS)” is applied on the Pareto front to attain the
final optimization results. The analysis is further explored in depth by generating digital twins (surrogate or meta
model) of HRES data using artificial intelligence techniques (artificial neural network and group-method-of-data-
handling). Furthermore, calculus and statistical sensitivity analysis assist in the identification of the significant var­
iables in the design procedure. In summary, the technical contribution of this work can be divided into two sections.
The first one is the design of a hybrid energy system for the water management of an isolated community of the
indigenous Mayan region of Yucatan, Mexico, which has never been considered before. Secondly, the technical
contribution is related to the usage of environmental emissions as an objective function, which is not considered in the
traditional design of hybrid energy systems by the software HOMER. Environmental emission as an objective function
is not considered while designing a hybrid energy system in commerical softwares like HOMER, in fact, HOMER
provides a list of environmental impacts but it is a secondary outcome as a result of technoeconomic optimization.
Analysis of results between HOMER pro and spreadsheet has shown conformity, reporting that the optimal case
consists of a photovoltaic system, diesel generator, and Li-ion technology of battery storage with capacities of ~17 kW,
~5kW, and 44–48 kWh, respectively, corresponding to a net present cost ranging from 70,000 United States Dollars
(USD) to 79,000 USD and a cost of electricity ranging from 0.205 to 0.229 USD/kWh. The achievements obtained with
multiobjective optimization indicate that the cost of electricity and net present cost can be further reduced by 0.86 %
and 0.73 %, respectively, at a decrement of only 0.4% of the renewable fraction as compared to the single objective
optimization scenario. It is concluded that multiobjective optimization provides an add-in feature to HOMER by using
environmental emissions as an objective function.
The design procedure and adapted methodology can be useful to promote sustainable development in the
statewide context and can provide a scientific justification to national energy policymakers.

* Corresponding authors.
E-mail addresses: [email protected], [email protected] (R. Tariq), [email protected] (M.A.E. Soberanis).
URL: https://sites.google.com/view/rasikhtariq (R. Tariq).

https://doi.org/10.1016/j.seta.2021.101561
Received 23 February 2021; Received in revised form 16 August 2021; Accepted 28 August 2021
Available online 22 September 2021
2213-1388/© 2021 Elsevier Ltd. All rights reserved.
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Introduction and services since the growing network density and unprecedented in­
crease in network traffic require intelligent network operations. Sodhro
Water is an essential resource [1], and in recent decades it has et al. [27] provided a comparative analysis of various tools and appli­
become scarce in some parts of the world [2]. The proportion of fresh­ cations in terms of energy efficiency with objective to minimize the
water on the earth’s surface is only 2.5 % but only 1 % is accessible for energy consumption of software and hardware components. Sodhro
use [3]. Moreover, there is a growing water demand [4], and existing et al. [28] also proposed the need for self-adaptive artificial intelligence-
freshwater resources are not sufficient to meet human needs [5]. based strategies to effectively cluster, examine, and interpret the entire
Coupled with this, the effects of climate change, the increase of pollu­ entities in the system. Memon et al. [29] provided an introduction to
tion, and anthropogenic activities around the world have caused a various key concepts of Wireless Local Area Network (WLAN) that are
decrease in water quality, which represents a big global problem, necessary to design, model, and develop a framework while also dis­
especially in marginalized communities with low quality of life [6]. cussing Quality of Service for IEEE 802.11 WLAN and Medium Access
Water quality in the water bodies is greatly dependent on physical, Control protocols for supporting industrial emergency traffic over the
chemical, and biological parameters which must be evaluated before network.
deciding the specific usage of water. Conventional water treatment During the last years, many studies have been conducted in the area
plants ensure the necessary quality of water either for domestic or for of HRES that includes different optimization techniques using machine
drinking use [7]. However, they involve different processes such as learning in these applications. For instance, Sreeraj et al. [30] proposed
filtration, chlorination, coagulation-flocculation, sedimentation, among a method that is based on the design space approach and can be used to
others, which entail great energy consumption. Moreover, there are determine the conditions for which hybridization of the system is cost-
other costs such as transportation and distribution which limit its ben­ effective. The proposed method was implemented to design an iso­
efits to regions [8] that are situated away from the treatment plant [9]. lated power system for an Indian village (Sukhalai, Hoshangabad dis­
For instance, in Mexico, the amount of energy required by a potable trict, Madhya Pradesh) utilizing the wind-solar photovoltaic system. It is
water treatment plant from the ground or surface is between 0.1 and 4.5 reported that the cost of energy from a photovoltaic/battery (PV/bat­
kWh/m3 [10]. This energy is obtained by the burning of fossil fuels, and tery) system was 0.38 USD/kWh (United States Dollar) and the same for
consequently, the problem of environmental pollution and climate the wind-battery system was 0.24 USD/kWh. The wind power gener­
change persists. Therefore, renewable energy sources have emerged as ating system is a cheap source of electric power as compared to solar PV,
an appealing alternative to conventional power generated from fossil but the variability of the power available from the wind is greater than
fuels [11,12]. This has led to increasingly significant levels of distributed photovoltaic solar energy [31]. Hence, if the reliability requirement is
renewable energy generation being installed on existing distribution low, the wind battery system is cost-effective. Dursun et al. [32]
circuits. This brings the research to address the water-energy nexus in developed a computational model that allows calculating the energy
which no compromise should be made to utilize green energy resources supplied by a stand-alone renewable hybrid power system that can be
for the potable water treatment plant. used in an electrolyzer for the production of hydrogen. The results of the
There are different types of conventional and non-conventional en­ proposed model are calculated and compared with experimental data.
ergy sources used to generate electricity. Solar [13] and wind energy According to the obtained results, the annual hydrogen production is
[14] systems are some of the most prominent sources of energy. The 34.3 kg whereas the differences between the real data and model data
utilization of solar and wind energy systems has become increasingly are quantified with Mean Square Error (MSE) with a value of 8.28e-7.
popular due to their modular and environment-friendly nature [15–18]. Rohani et al. [33] designed a hybrid renewable energy system for the
The combination of solar and wind energy resources [19] has exhibited remote area in Ras Musherib located in the western region of Abu Dhabi.
significant effectiveness in the previous two decades because of their The hybrid system, which consisted of a photovoltaic (PV) array, wind
standalone interactive utility. Nevertheless, due to the intermittent na­ turbines, batteries, and diesel generators, was designed to meet three
ture of solar and wind energy, the system based on either solar or wind known electric loads, 500 kW, 1 MW, and 5 MW to be able to fulfill the
energy alone is found unreliable. A system that is based on renewable primary load for 250, 500, and 2500 household, respectively. The ob­
resources but at the same time reliable is necessary and a solar-wind tained results indicated that the hybrid system with 15 % of photovoltaic
hybrid system with battery storage and a small diesel generator as a and 30 % of wind turbine capacity was found to be the optimal system
backup can meet this requirement. These systems may be utilized in for 500 kW average load with an initial cost of 4,040,000 USD and total
grid-connected mode, isolated [20] from the grid and special aims, and net present cost of 14,504,952 USD over 25 years. The reduction in CO2
are called Hybrid Renewable Energy System (HRES) [21]. There are emission achieved in this study for the 500 kW optimal hybrid system
several types of hybrid energy systems [22] such as wind-solar hybrid, was 37 % as compared to the conventional diesel generator case.
solar-diesel, wind-hydro, and wind-diesel, which are among the present Recently, Forough et. al. [34] presented a methodology for energy
in production plants. The design and sizing of a system and the choice of management systems based on the multi-objective receding horizon
energy resources depend on several considerations. It includes the type optimization to find the optimal scheduling of HRES. The proposed
of storage system which is important for both on-grid and off-grid sys­ hybrid system was experimentally installed in an educational building
tems. The motivation to design and size a hybrid system is the following: that comprises PV panels, wind turbine, battery bank, and diesel
reduction in the peak load (load shaving), provide energy security to the generator as the backup system. The results have shown that using a
user application and minimize the effects of uncertainty related to longer time frame (from 6 h to 24 h), the total share of renewable energy
renewable generation systems [23]. The strategic combination that in supplying weekly demand can be improved up to 18.7 %. Hence, the
minimizes the cost and maximizes the energy profit from a hybrid sys­ proposed methodology can manage a system to make better use of re­
tem can be attainable through an optimization study. For these reasons, sources resulting in improved system scheduling. Harajli et. al. [35]
it is important to locate the optimal operation in any energy system [24] presented a study of a hybrid solar PV/diesel system to assess the en­
with a requirement that a system must provide a consistent energy ergetic, environmental, financial, and economic performances of a hy­
meeting the demand [25]. The factors affecting the choice of hybrid bridized solar photovoltaic energy using first-hand data and information
power technology can also tell us why people use hybrids and some of collected from the Palestinian, Lebanese, and Iraqi commercial and/or
the advantages. industrial sectors. Results have shown that hybrid PV/diesel systems
The usage of the techniques of machine learning has benefited and have largely beneficial energy, environmental, and economic perfor­
extended the research circle nearly in every area. Ahmed et al. [26] have mances in all three countries, whereas their financial performance is also
applied machine learning in different types of networks and networking positive for Palestine and Lebanon.
technologies to meet the requirements of future communicating devices Several recent studies have simulated the HRES using commercial

2
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

software and comparing the results with another method. Generally, water-energy nexus [42]. The feasibility of such hybrid energy systems
most of these methods are computer programs. One example was pre­ is extremely practical for isolated indigenous communities, like the one
sented by Eltamaly et al. [36], where the authors proposed a new presented here as a case study (indigenous community of San Jose
methodology that uses the following input data: hourly wind speed, Tipech, Yucatan). Another gap identified in the literature is that the
hourly radiation, and hourly load power with many different types of researchers have not used environmental emissions as an objective
wind turbines and PV module types, which was used to optimize a PV/ function to design a hybrid energy system especially for the water
wind/battery energy system for ten locations in Saudi Arabia through a management of remote communities. Some authors have provided in­
computer program which was compared with “Hybrid Optimization of formation about environmental emissions like the Ramesh et al. [43],
Multiple Energy Resources (HOMER)”. The obtained results revealed Soberanis et al. [44], Sarker et al. [45], Shezan et al. [46], and Upadhyay
that the cost per kWh of HRES was found between 0.3 and 11.8 cents/ et al. [47]. However, they have not used those emissions for the initial
kWh and that obtained by HOMER between 10.6 and 12.8 cents/kWh, design. Moreover, the environmental emissions are presented merely as
which indicated that the program developed was ideal for the optimi­ an outcome of the primary design. Thus, with this background, it is
zation of hybrid systems and the cost range is within a reasonable important to contribute further in this area by using environmental
margin for HRES investment by the private and government sectors. One features of a hybrid energy system for the initial design of water filtra­
year later, Eltamaly et al. [37] presented a design and simulation study tion plants for remote communities
of a wind system for five sites in Saudi Arabia through a computer This is also aligned with the sustainable development goal no. 10,
program developed using Visual Fortran which was verified through an reduced inequality, of the United Nations. This research also addresses
Excel spreadsheet. The obtained results indicated that the best site was other sustainable development goals like 7, 8, 9, 11, 12, and 13 of the
Dhahran with a cost of ~5.84 cents/kWh using a KMW-ERNO turbine, United Nations. On the other hand, there is a country-wise sustainable
while the worst site was Riyadh with a minimum kWh cost of ~12.81 development goal to align with the goals of the United Nations. In this
cents/kWh using a GE (General Electric) Energy 2 turbine. That same case, the national sustainable development goals of Mexico are also
year, Eltamaly et al. [38] developed a computer simulation program to addressed which are a part of the national agenda 2030 [48]. Further­
design and optimize a self-contained battery/wind/PV hybrid power more, the state of Yucatan has a statewide agenda too. Through the help
system for the city of Dhahran, Saudi Arabia. The main objective of this of quantitative indicators, it is reported that the state of Yucatan is still
study was to determine the optimal size of each component of the energy struggling to meet the statewide goals, for example, the goal of clean
system that would allow obtaining the lowest cost of kWh, which was water and sanitation, affordable and clean energy, and reduced
compared with HOMER. The results revealed that the initial costs of the inequality, are only attained by 37 %, 58 %, and 55 %, respectively [49].
system, replacement, operation, and maintenance and the salvage price This is a worrisome situation in which the state must put more effort and
obtained with HOMER were $46,780,000, $20,161,322, $18,644,542, take diligent steps to address these issues. In this context, the current
and $-6,920,542, respectively. The developed program by the authors work can be a beacon to lift statewide sustainability indicators,
were $49,126,780, $18,089,010, $8,809,040 and $-2,011,168 respec­ contributing towards the national agenda 2030 with a projection to­
tively. Therefore, the computer program is able to report reliable pre­ wards the global sustainable development goals. Another justification of
dictions and it is a useful tool for sizing and optimizing HRES. the work is related to the scientific development projects to address the
Suresh et. al. [39] presented a modeling and optimization study of an national problems of Mexico specifically directly targeting (a) compre­
off-grid HRES where the main objective was to reduce the Total System hensive water management, water security, and water rights, (b) miti­
Net Preset Cost (TNPC), Cost of Energy (COE), unmet load, and CO2 gation and adaptation to climate change, (c) sustainable energy
emissions using a Genetic Algorithm (GA) and HOMER Pro software for consumption, (d) development and use of clean renewable energies, and
three un-electrified village hamlets in Chamarajanagar district, Karna­ indirectly targeting, (a) fight poverty and food security.
taka state (India). The results of the two methods were compared with Apart from addressing the sustainable development goals, techni­
four combinations (biogas-biomass/solar/wind/fuel cell with battery) cally, this study also bridges other gaps likesuch as the limitations of the
of hybrid renewable energy systems. The results indicated that usage of the software HOMER for hybrid energy systems. The optimi­
compared with HOMER, GA was found to be the optimal solution sup­ zation scenario considered in HOMER is based on a single objective
plying energy with 0 % unmet load providing a cost of energy of 0.163 optimization, considering linearly dependent net present cost and cost of
USD/kWh which is more cost-effective than HOMER. Mokhtara et al. electricity. However, in real-world conditions, the problem is multidi­
[40] presented a novel approach by integrating demand–supply man­ mensional and requires extra indicators, specifically environmental
agement with particle swarm optimization and applying it to optimally emissions, contemplating the context of global warming challenges.
designed off-grid hybrid PV/diesel/battery system for the electrification Even though a renewable energy system does not have carbon dioxide
of residential buildings in an arid environment, using a typical dwelling emissions during the operation, it is marked with environmental emis­
in Adrar, Algeria, as a case study. The proposed HRES was first modeled sions during the life cycle. Hence, even a renewable hybrid energy
by an in-house MATLAB code and then validated with the HOMER system should have a carbon dioxide accounting before implementation.
software. Moreover, techno-economic analyses including sensitivity In summary, the novelty of this work is to fill various gaps in the existing
study were carried out considering different battery technologies. The body of knowledge by addressing water-energy nexus through the
results have shown that the energy demand and total net present cost integration of wastewater treatment and potable water filtration plant
(TNPC) were reduced by 7 % and 18 % as compared to the case of using with the hybrid energy system. Moreover, it, addressesing the statewide,
solely supply-side management. Also, it was found that PV/Li-ion rep­ national, and global sustainable development goals, and highlightsing
resented the best configuration, with TNPC of 23,427 USD and Cost of additional environmental add-in feature as an objective function for
Electricity (COE) of 0.23 USD/kWh. optimization purposes in the HOMER software.
In the context of the problem formulation, some hybrid technologies Thus, the motivation of this work is to provide an integrated design
can provide a completely different approach to water treatment, espe­ of a hybrid energy system for the water management of an entire
cially for remote communities, where the cost of power transmission indigenous Mayan community located in the south-east of Mexico. The
lines is expensive. Therefore, the implementation of HRES is highly proposed design would help to mitigate various environmental, health,
recommended to ensure cost-effective, reliable power delivery, and and social problems of the community and would significantly
consequently, potable water [41]. The existing literature reports a contribute to the sustainable development goals of the state. Another
versatility coupling case studies with a considerable depth of analysis. motivation of this work is to use environmental emissions as an objective
However, the technical feasibility of a hybrid energy system should be function for the design of the hybrid energy system, which has a sig­
evaluated further in the case of water filtration plants to address the nificant importance, considering the global environmental degradation

3
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Fig. 1. Representation of the processes involved in the wastewater treatment plant and potable water filtration plant, considered for the sizing of the hybrid
renewable energy systems.

problems. highlighting each process involved in both water treatment plants, from
The objectives of this research work is to evaluate the technical and its entry to the wastewater treatment plant to its final disposal in the
economic indicators of different hybrid energy systems proposed to community. The development of a similar model, where industrial
satisfy the energy demand of a suggested Wastewater Treatment Plant wastewater and potable water treatment plant was analyzed, is reported
(WWTP) and Potable Water Filtration Plant (PWFP) for a remote com­ by Gude [50]. Moreover, Fig. 1 illustrates that from the integration of
munity in the state of Yucatan. The authors used a spreadsheet whose both plants, a load curve is determined which describes the energy de­
results were compared with the software HOMER Pro, which led to a mand throughout the day. This load curve is used to size the hybrid
multiobjective analysis by using artificial intelligence techniques to find renewable energy systems by different techniques.
the optimal case and the different economic indicators that allowed to
conclude the viability of the project. The model developed in the present
work consists of a HOMER pro and developed spreadsheet techniques for Wastewater treatment plant
the sizing of the HRES capable of supplying the energy demanded from
the water treatment and filtration plants, followed by a multiobjective The wastewater treatment process is divided generally into four
optimization process developed to further enhance the performance of steps. The first one consists of a preliminary treatment in which the
both techniques. Furthermore, in relation to some previous studies re­ objective is to remove the coarse solids and other large materials that are
ported in the literature [23,36–38], the methodology presented in this often found in raw wastewater. Preliminary treatment is carried out in
work consists of a complete study of an HRES with sizing, optimization the proposed treatment plant in the bar screen and aerated grit units.
scenarios, and sensitivity analysis. Also, there are few optimization Subsequently, a primary treatment is carried out in the primary settling
studies reported in the literature where different scenarios are consid­ to reduce the suspended and floating solids in the wastewater by gravity
ered with different energy storage technologies. The developed action. Later, secondary treatment is carried out in the activated sludge
spreadsheet allows to find the optimal combination of HRES that min­ unit, which is a biological process, since this consists of biological
imizes the COE and maximizes the Net Present Cost (NPC), and mainly degradation of dissolved degradable organic matter and remaining
contemplates three different storage technologies considering the initial suspended solids by microorganisms, reducing the organic loading in
costs and the lifetime of each one. Finally, it is also pertinent to mention terms of COD (Chemical Oxygen Demand) and BOD (Biochemical Ox­
that no research has targeted the usage of the hybrid energy system with ygen Demand), as well as nutrients like total nitrogen. It should be noted
a potable water filtration plant for an indigenous community of the that during this treatment some nutrients or resources from the residual
Mayan region with the consideration of optimization scenarios built on sludge can be recovered since they contain a considerable amount of
HOMER, spreadsheet, and MATRIX Laboratory (MATLAB). nutrients, particularly phosphorus in the form of proteinaceous material
[51]. Afterward, a secondary settling is used to remove the sludge
System description (biological flocs) from the treated wastewater. Finally, in tertiary
treatment, the purpose is to enhance the quality of the treated water to
In this section, the processes that take place for the water to be fulfill regulations for its discharge into the environment or reuse as a
potable for its usage in the community as well as the ones involved in the water source. This is done in the chlorination unit, which main purpose
wastewater treatment once the resource has been used are described. is the removal of pathogens. The final disposal of the water at the end of
Fig. 1 presents the schematic diagram of the complete water cycle this process is in an aquifer reservoir, where it will be taken to be treated
in the water purification plant described in the next section.

4
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Table 1 filtration process to make sure that every solid particle has been elimi­
Estimated energy consumption of the processes involved in the wastewater nated and finally remains stored in a well until it’s demanded by the
treatment plant and potable water filtration plant [50]. subject community. The unwanted particles obtained in the last filtra­
Loop of water Process Specific energy Total energy tion process and some detected while in storage, are redirected to a
plant consumption consumption waste backwash process where it reverses the flow of water lifting it and
(kWh/m3) (kWh) expelling the dirty water to a drying bed, or to the environment in case
Wastewater Pumping 0.0317 6.791 the pollution levels are low enough. Drying beds receive the particles
treatment Screening 0.0081 1.743 obtained as a residue from the sedimentation process and along with the
Aerated grit 0.0035 0.758
dirty water that comes out of the waste backwash, it forms a sludge that
removal
Primary 0.0041 0.877 is dried by percolation and evaporation; the substance obtained from
sedimentation this process is often used as a biosolid in agriculture.
Activated sludge 0.1406 30.107
Secondary 0.0041 0.877
Methodology
sedimentation
Chlorination 0.0007 0.153
Gravity 0.0483 10.356 Description of the study site
thickening
Anaerobic 0.0370 7.923 This research work evaluates energy and economic indicators of
digestion
Dewatering 0.0101 2.173
different hybrid energy systems proposed to satisfy the energy demand
Potable water Pumping 0.0317 6.791 of a wastewater and potabilization plant suggested for the San José
filtration Alum + Polymer 0.0015 0.323 Tipceh community, Muna municipality (20.48 ◦ N, 89.72 ◦ W), Yucatán,
treatment Rapid mix 0.0081 1.743 México. This town has a population of 700 inhabitants [52], distributed
Flocculation 0.0024 0.509
in 148 homes, of which 100 % have electricity, 97.71 % have drinking
tanks
Sedimentation 0.0023 0.498 water and 88.55 % have a toilet [53]. Moreover, the average daily water
Drying beds 0.0011 0.226 consumption per capita is estimated at 200 L [54]. The climate of this
Waste backwash 0.0053 1.132 municipality is tropical savanna type, it has an average annual solar
Clear well 0.0053 1.132 radiation of 388.84 W/m2 during sunny hours, an ambient temperature,
storage
Total 74.11
and wind speed of 25.18 ◦ C and 5.21 m/s respectively.

Estimation of energy consumption of water treatment plants and power

demand curve

To develop the sizing of the treatment plants, it is necessary to know

the daily water consumption by the entire community. The parameters
necessary to carry out this estimation are the number of inhabitants of
the community, as well as for each house, the daily water consumption
of a person, the peak and return factors. This parameter is estimated as
[55]:
q = k1 .k2 .p.w.N/86400 (1)

where q is the water flow to be treated (L/s), k1 is the peak factor (1.8 for
simplified sewerage [56]), k2 is the return factor (wastewater flow/
water consumption), p is the average number of inhabitants per dwelling
and N is the total number of dwellings in the community. The calcula­
tion of Eq. (1) resulted in a value of 2.479 L/s, which is equal to 214.2
m3/day of water to be treated. This result was used to estimate the en­
Fig. 2. Hourly variation curve of the power demand for water treatment plants.
ergy consumption of the respective water plants that allow the sizing of
the hybrid renewable energy system. Therefore, it is necessary to know
Additionally, large suspended solids and sludge coming from the pri­
the energy demand curve of the wastewater treatment plants and the
mary and secondary settling are thickened (concentrated) in the gravity
water purification plant. In this study, a single energy consumption
thickening and afterward stabilized in the anaerobic digestion unit.
curve is considered, which corresponds to both water plants and consists
Finally, the stabilized sludge is passed through a belt press dewatering to
of the average behavior of energy for each hour of the day. This allows to
remove excess water before disposal. The resulting sludge can be
identify the periods with the highest consumption, as well as those in
incinerated or used as a biosolid in agriculture.
which the energy demanded from the plant is minimal.
Each water treatment plant has different processes. Table 1 lists each
Potable water filtration plant process involved in both water treatment plants illustrated in Fig. 1, as
well as the unit energy consumption and the total consumption. These
The process begins by taking the water demanded by the community values were calculated considering the result obtained from Eq. (1) with
daily from the aquifer reservoir. First, it goes through a rapid mix which the context of the wastewater and potable water filtration plant as
consists of identifying the solid suspended particles existing in the water presented in Fig. 1 [50].
and mixing them with the help of aluminum and polymers to form even Regarding the hourly power demand curve, a single curve is pro­
bigger particles and making it easier for them to be extracted from the posed for both plants, as mentioned above, since it is considered an ideal
water. The flocculation tanks are the space where the water with the case of energy demand for a community. The behavior of the hourly
now visible suspended particles is received before entering the sedi­ power demand was estimated taking into account the variation of the
mentation process, where the heaviest components that remain in the hourly water flow, which are modeled through stochastic methods,
fluid fall to the bottom of the container. The cleaned water goes to a final considering comprehensive data from the Mexican National Water

5
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Fig. 3. Schematic diagram of the different cases proposed in this work.

optimization toolboxes applied using HOMER energy, spreadsheet, and

Table 2
MATLAB.
Technical specifications of photovoltaic modules [61].
Regarding the battery storage system, three different technologies
Concept Value were used: Pb-acid, Li-ion, and AGM, to obtain the most optimal tech­
Brand and model Sharp NU-JB395L nology from a technical and economic point of view. Finally, the
Maximum power 395 W photovoltaic and wind systems were modeled through different equa­
Nominal Operating Cell Temperature 45 ◦ C tions that allow estimating energy generation, considering different
Temperature coefficient 0.353 %/◦ C
factors such as solar radiation, wind resource, ambient temperature,
among others.
Commission (CONAGUA, by its acronym in Spanish) [57]. Fig. 2 illus­
trates the hourly power demand of the water treatment plants for the Photovoltaic solar system
present case study. As seen in Fig. 2, the peak energy consumption oc­ The electrical energy production of a photovoltaic panel (PV) is
curs at 9:00 a.m., in which most people consume a greater amount of given in terms of its temperature and the reference temperature for the
water since they take morning baths before leaving for work, and this design conditions [59].
trend continues throughout the afternoon, where housewives use a large [ ( )]
PPV = Gt ηPV,ref 1 − β Tc − Tref (2)
amount of water for housework. However, during the night and in the
early morning hours of the following day, water consumption is
where Gt is the incident solar radiation on the PV module, ηPV,ref is its
reduced, so that this behavior is reflected in the energy consumption of
efficiency under the reference temperature Tref which is 20 ◦ C, β is the
water treatment plants.
temperature coefficient specified by the manufacturer and Tc is the
temperature of the panel cell according to the NOCT criteria, which is
Sizing of the hybrid energy system calculated as follows [60]:
( )
NOCT − 20
In this section, the sizing of the different combinations of the hybrid Tc = Tamb + Gt (3)
800
energy system consisting of a photovoltaic system (PV), a wind gener­
ation system (W), a battery energy storage system (BESS), and a diesel Here Tamb is the ambient temperature. The specifications of the
generator (DG) is proposed as a backup power system. For this purpose, photovoltaic modules proposed in this work are presented in Table 2.
different equations were considered that allow obtaining the average The energy losses for the photovoltaic system and inverter were
power and energy generated by each system throughout the year and considered to be 3 % [62].
estimating the number of elements by each system. For the case of a
diesel generator, an HYUNDAI diesel generator model HHY5500 [58] Wind generation system
with a nominal power of 5 kW was used, which has a fixed power, which To calculate the power generated by the wind turbine, the power
does not depend on any renewable resource, but only on the volume of curve from the datasheet of the wind generator to be implemented can
fuel to be used. Hence, the diesel generator is chosen based on the be used so that this power is given as a function of the wind speed at the
maximum power consumed by the study community to satisfy the height of the wind turbine [63].
maximum annual energy demand. The different proposed cases of the The proposed wind turbine, as well as its technical specifications, are
hybrid power system are illustrated in Fig. 3 along with the usage of detailed in Table 3.

6
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Table 3 economic data of the hybrid system such as the capital cost of each
Wind turbine technical specifications [63]. system, the cost of the storage system capacity, operation and mainte­
Concept Value nance (O&M) costs, among others. These parameters are presented in
Table 5.
Brand and model Enair E200
Maximum power 10 kW To carry out the economic study to find the best combination of the
Diameter 9.8 m system, two economic indicators were used: Net Present Cost (NPC) and
Hub eight 40 m Cost of Energy (COE). The NPC (USD) of a project is defined as the
Power present value of all the costs of installation and operation of the system
characteristic
curve
during the useful life of the project minus the present value of all the
income obtained during the useful life of the project and is calculated by
the Eq. (4) [33]:
Cann,tot
NPC = ( ) (4)
CRF i, Rproj

where Cann,tot is the total annuity and CRF(I,Rproj) is the capital recovery
factor used to calculate the present value of an annuity, and it can be
calculated using Eq. (5) [33]
( ) [i(1 + i)n ]
CRF i, Rproj = (5)
Table 4 [(1 + i)n − 1]
Technical specifications of storage batteries [64].
Technology
where i represents the interest rate, of which 10 % per year was
considered and n is the number of years, considering the 20-year dura­
Pb-acid Li-ion AGM
tion for this project.
Number of cycles (80% discharge) 1500 3500 2500 Finally, the COE (USD/kWh) is an indicator that represents the
Allowable discharge depth 40 % 15 % 20 %
current cost to buy energy, so the objective is to minimize this value
Energy losses (charge–discharge) 10 % 10 % 10 %
Lifetime 5 years 10 years 8 years
[68], which can be calculated as Eq. (6):
NPC
COE = (6)
Ec
Table 5
Economic data used in this work. where Ec is the total load of the community over 20 years.
System Concept Value

Photovoltaic Capital cost 1,000 USD/kW [65] Supplementary initial data

O&M 20 USD/kW [65]
Wind Capital cost 5,000 USD/kW [65]
Details of some other initial data are given as follows:
O&M 30 USD/kW [65]
Power converter Capital cost 1,000 USD/kW + 500 USD/kW additional
[66] 1. Variable and hour-dependent elevation and declination angle are
Diesel generator Unit cost 754 USD [58] considered for the sizing of the photovoltaic system [69].
O&M 0.025 USD/kWh [67] 2. The efficiency used for the inverter is 97 % [70].
Fuel cost 2 USD/L [44]
Battery (Pb- Storage 200 USD/kW [64]
3. For the wind generation system, power obtained at different altitudes
acid) capacity is analyzed through the use of wind shear [71].
Battery (Li-ion) 500 USD/kW [64] 4. In the case of the diesel generator, it is turned on and off following
Battery (AGM) 400 USD/kW [64] the recommendable percentages of minimum and maximum state of
charge of the battery technology [64].
5. The utilized economical parameters are as follows: a minimum rate
Energy losses in the power converter are estimated to be 3% [62].
of return on investment of 10 %, annual increase on the costs of O&M
of 5 % [72], annual inflation of 5 % [73] is considered.
Battery energy storage system
6. Regarding the environmental parameters, both sizing techniques
The battery storage system comprises three different technologies:
utilized the solar radiation, ambient temperature, and wind speed
Pb-acid, Li-ion, and AGM. Each technology has different specifications
database which is obtained from NASA (National Aeronautics and
regarding the number of charge–discharge cycles, energy losses, charge
Space Administration) Power Data Access [74].
and discharge depth, and lifetime. These parameters are specified in
7. Energy tariffs [75] and the cost of potable water [54] are taken from
Table 4.
the Mexican electricity and water main suppliers, respectively.
The nominal capacity of each battery bank varies to find the optimal
value that presents the lowest possible cost integrated with the hybrid
renewable energy system. For the economic analysis, different costs Results and discussion of single-objective optimization using
were considered in each system, which is described in the following HOMER and spreadsheet
section.
The sizing of the hybrid systems was carried out through the HOMER
Pro software, as well as using a developed algorithm developed in an
Economic study of the hybrid energy system Excel spreadsheet. There are two optimization algorithms within the
interface of HOMER Pro software. The original grid search algorithm
This section describes the procedure used to calculate the economic simulates all the feasible configurations as defined by the search space.
indicators proposed in this work that allows decision-making of the The other algorithm is related to a proprietary derivative-free algorithm
system that presents the lowest net present cost and the lowest energy that searches the least costly system and finally, the software displays
cost. To carry out the economic study, it is necessary to know certain the list of configurations sorted by the net present cost which is used for

7
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

the comparison purposes of different system designs [67].

HOMER
There is a dedicated algorithm which defines the operational
Power converter (kW) behavior in various cases of the supplied loads by PV, diesel generators,

6.8
7.4
7.2
7.3
7.3
7.3
7.3

6.8

6.5
4.4
4.9
5.1
9.1
9.1
6.1
-
-
-
wind turbines, and batteries. It also includes the usage of the charging
and discharging cycles of the batteries. Halabi et al. [76] and Alsham­
Spreadsheet

mari et al. [77] have described this energy management strategy in

detail. The system operates in different modes according to the condi­

5.3
5.3
5.3
5.3
5.3
5.3
5.3

5.3

5.3
5.3
5.3
5.3
5.3
5.3
5.3
tions of the surrounding atmosphere. During the normal and standard

-
-
-
operating conditions when the sun is available, the control system gives
HOMER

the highest priority to the arrays of solar panels to supply the demand of
BESS capacity (kWh)

124

the load. At this duration, the excess energy of the system is used to

48
87
55
55

88
77

75

49
28
34
35
62
44
43
6
3
4 charge the battery until it reaches its maximum state of charge position.
Spreadsheet

If there is a sufficient amount of solar radiation which generates enough

photovoltaic electricity to meet the load and also charges the batteries to
the maximum state of charge, then the excess energy is used by the
130

44
85
48
54

80
86

65

45
77
36
25
57
38
39
6
3
4

dumped loads. In another case, when the solar panels can not provide
sufficient electricity to handle the variation in the load, the batteries
HOMER

start operation and handle the load until reaching their minimum state
DG capacity (kW)

of charge. If the renewable energy potential of some particular hours of

5
5
5
5
5

5
5
5
5
5
5
5
-
-
-
-
-
-

the year is not sufficient enough to carry the supply load, and the bat­
Spreadsheet

* The annual behavior of total load and total power output of the optimization configuration (case 4 with Li-ion battery) is displayed in Appendix A. teries are also reaching their minimum state of charge, then the diesel
generator is turned on and supplies the entire load. This control man­
agement system is sensitive to hourly calculations
5
5
5
5
5

5
5
5
5
5
5
5
-
-
-
-
-
-

Table 6 reports the comparison of the six cases illustrated in Fig. 3

HOMER

with the three different battery technologies, contrasting the results of

the spreadsheet and HOMER. Various technical parameters of each case
30
30
30

10
10
10
10
10
10
-
-
-
-

-
-
-
-

-
W capacity (kW)

are reported, as well as economic indicators. It is possible to observe that

the results obtained from the spreadsheet are well correlated with the
Spreadsheet

results obtained from HOMER. Considering the minimum values ob­

tained by each technique used for sizing the hybrid systems, the best
30
30
30

10
10
10
10
10
10

results regarding the cost of energy are obtained in the following order:
-
-
-
-

-
-
-
-

case 4, case 1, case 5, case 6, case 3, and case 2.

HOMER

There are different parameters hidden in each case, leading them to

17.0
19.9
18.3
19.9

18.3

18.1

12.5
12.5
10.5

different COE scenarios. For instance, the cost of installation for 1 kW of

PV capacity (kW)

-
-
-
-
-
-

-
-
-

wind turbines is much more expensive than for photovoltaic modules,

Spreadsheet

with the hub being one of the most expensive parts of this technology.
This is due to the three cases with wind turbines are the worst COE
17.7
25.7
21.3
23.7

19.7

19.7

12.2
12.2
11.4

scenarios, just after the diesel generator with a battery storage system,
-
-
-
-
-
-

-
-
-

which is affected considerably by the high cost of the fuel. Moreover,

Results were obtained from the spreadsheet and HOMER for the three battery technologies.

cases with photovoltaic modules resulted in two of the three best COE
HOMER

0.205
0.240
0.217
0.218
0.670
0.667
0.635
0.695
0.688
0.694
0.226

0.207
0.326
0.295
0.294
0.323
0.323
0.302

scenarios, since this renewable technology has been developing with

success worldwide, making its cost of installation way lower than the
COE (USD/kWh)

one for wind turbines. The climatic conditions of the region are also a
Spreadsheet

key point for analyzing the obtained results since the Yucatan peninsula
is well known for its solar resource with solar daily radiation above the
0.229
0.286
0.254
0.268
0.754
0.722
0.749
0.722
0.692
0.653
0.236

0.245
0.328
0.299
0.296
0.339
0.332
0.337

national average.
Furthermore, important features were discovered while comparing
230,298
236,589
238,612
238,903
236,589
238,612

111,989
101,359
100,294
111,244
110,940
104,005
HOMER

70,415

the results obtained between the spreadsheet and HOMER. From the
82,703
74,521
74,976

77,826

71,207

results obtained with HOMER for case 1, the capacity proposed for the
photovoltaic system is 21.3 kW for the Li-ion battery technology, while
the capacity for the power converter is 7.2 kW, only 34 % of the total
Spreadsheet
NPC (USD)

capacity of the photovoltaic system. From this percentage, it is possible

259,027
248,275
257,607
248,123
238,047
239,039

112,920
102,862
101,885
116,693
114,229
115,781
78,732
98,264
87,336
92,029

81,216

84,271

to ensure that HOMER uses a charge controller since the power con­
verter does not have the capacity enough to convert all the energy
generated from the photovoltaic system, from DC (direct current) to AC
Technology

(alternating current). Due to this observation, a charge controller was

Pb-acid

Pb-acid

Pb-acid

Pb-acid

Pb-acid

Pb-acid
Li-ion

Li-ion

Li-ion

Li-ion

Li-ion

Li-ion

added to the calculations developed in the spreadsheet. This led to a

AGM

AGM

AGM

AGM

AGM

AGM

decrease in the capacity of the power converter, as well as a modification

in the COE regarding the cost of the charge controller.
Case 6: PV/W/DG/ BESS

Another difference discovered between the spreadsheet and HOMER

Case 4: PV/DG/BESS*

Case 5: W/DG/BESS

was the batteries’ limit level of charge. The only percentage that
Case 3: DG/BESS
Case 1: PV/BESS

HOMER reports is the limit level of discharge, whereas, in the spread­

Case 2: W/BESS

sheet, both charge and discharge limit levels are modified according to
the technology used. This led to reducing the number of batteries in the
Table 6

spreadsheet calculations regardless of the technology used for each case.

Case

Otherwise, analyzing the development that photovoltaic modules have

8
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Fig. 4. Scatter matrix plot providing information on the generation of the numerical experiments for surrogate modeling.

had in the last 20 years, as well as results obtained from other authors on Sequential procedure of multi-objective optimization
the sizing of hybrid systems, it is justified that a hybrid system integrated
photovoltaic modules with a diesel generator and battery storage system Optimization involves a stepwise procedure that starts from the
is the optimal solution since wind turbines are still way too expensive for generation of a database, otherwise called ‘numerical experiments’,
them to be used in a small demand project, comparing it with photo­ followed by the development of the digital twin, and finally, the actual
voltaic modules. The combination of wind turbines with diesel genera­ optimization is accomplished.
tors gives this renewable energy a chance to decrease its COE and have a
better-looking scenario for future applications.
In general, it can be seen from Table 6 that the economic indicators Generation of numerical experiments
obtained by the spreadsheet are higher than the ones presented by
HOMER. This is due to the fact that the best scenarios calculated with The first step in the development of a digital twin is correlated with
the spreadsheet were obtained through a manual search, by changing the generation of a set of ‘numerical experiments’ which mimic a
unit by unit the nominal capacity of the elements that integrate each working database having all the required information of the input var­
case. As for HOMER, it is software that calculates automatically the iables and the output indicators. In this case, such a database is gener­
economic indicators taking into consideration the elements provided: ated using HOMER [78] having the following characteristics of the
photovoltaic modules, wind turbine, power converter, diesel generator, tuning parameters: (a) the PV capacity is varied from 15 kW to 20 kW,
and energy storage systems. As a consequence, the computational time (b) the generator capacity is varied from 1 kW to 8 kW, (c) the battery
required from HOMER to present its results is much higher than the one storage capacity is varied from 40 kWh to 55 kWh, and (c) the inverter
associated with the spreadsheet. capacity is varied from 5 kW to 8 kW. The variation of this set of input
Finally, from these results, it can be concluded that the best result variables is noted on the indicators like a renewable fraction, cost of
obtained by these techniques was case 4 with Li-ion batteries technology potable water, and the CO2 emissions which are depicted in Fig. 4. Fig. 4
since the NPC and the COE were minimized. These results were con­ also presents the sensitivity analysis of the input variables on the output
ducted to develop an optimization process to maximize the renewable variables. It can be noted that the renewable fraction increases stepwise
fraction, minimize the cost of potable water, and reducing CO2 with the increase in PV capacity and battery capacity, however, it in­
emissions. creases with the decrease in CO2 emissions. Also, the relationship be­
tween renewable fraction and the cost of potable water is not well
defined. Therefore, it appears to be a multi-objective problem where it is
desired to have high renewable energy penetration while maintaining a
low cost of potable water [79].

9
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Table 7
Selection of best architecture for the surrogate modeling of the performance indicators in the multi-objective optimization scenario.
Renewable fraction (%)

Sample Training algorithm Number of hidden layers R2 Epoch

Training Validation Testing Total

1 Levenberg–Marquardt algorithm1 1 0.82 0.82 0.77 0.81 16

2 Levenberg–Marquardt algorithm 4 0.92 0.86 0.84 0.90 29
3 Levenberg–Marquardt algorithm 6 0.92 0.92 0.92 0.92 32
4 Levenberg–Marquardt algorithm 8 0.94 0.94 0.95 0.94 16
5 Levenberg–Marquardt algorithm 10 0.94 0.94 0.95 0.94 17
6 Bayesian regularization algorithm2 8 0.93 0.95 0.92 0.94 176
7 Scaled conjugate gradient algorithm3 8 0.92 0.93 0.92 0.92 41

Cost of potable water (USD/m3)

Sample Maximum number of input neurons4 Input source4 Maximum number of neurons4 Polynomial power4 Total SSE Total MSE Total R2

1 2 0 1 2 0.000326 4.96E-07 0.689437

2 2 0 1 3 0.000325 4.96E-07 0.689649
3 2 0 2 2 0.000231 3.52E-07 0.779515
4 2 0 8 2 9.13E-05 1.39E-07 0.912947
5 3 0 5 3 7.56E-05 1.15E-07 0.927917
6 3 0 6 2 8.82E-05 1.34E-07 0.915908
7 3 1 5 2 3.97E-05 6.06E-08 0.962085
8 3 1 5 3 5.38E-06 8.20E-09 0.994867
9 3 1 6 2 4.98E-05 7.59E-08 0.952509
10 3 1 10 3 8.45E-06 1.29E-08 0.991942

CO2 emissions (kgCO2)

Sample Training algorithm Number of hidden layers R2 Epoch

Training Validation Testing Total

1 Levenberg–Marquardt algorithm 4 0.99 0.99 0.99 0.99 45

2 Levenberg–Marquardt algorithm 6 0.99 0.99 0.99 0.99 73
3 Levenberg–Marquardt algorithm 8 0.99 0.99 0.99 0.99 94
4 Bayesian regularization algorithm 8 0.99 0.99 0.99 0.99 145
5 Scaled conjugate gradient algorithm 8 0.99 0.99 0.99 0.99 104
1
Levenberg–Marquardt algorithm typically required more memory but less time. Training automatically stops when generalization stops improving as indicated by an
increase in the mean square error of the validation samples.
2
Bayesian regularization algorithms typically require more time but can result in good generalization for difficult, small, or noisy datasets. Training stops according to
adaptive weight minimization.
3
Scaled conjugate gradient algorithm typically required less memory. Training automatically stops when generalization stops improving as indicated by an increase in
the mean square error of the validation samples.
Bold rows show the best architecture.
4
Referenced from the work [92].

The cost of potable water (USD/m3) and the CO2 emissions (kgCO2/ Development of a digital twin through the techniques of artificial
kWh) are calculated using the following equations: intelligence

Digital twins (also called metamodels or surrogate models), which is

a digital duplicate of a physical model, are gaining popularity in energy

NPC
Cost of potable water = (7)
Project life time (years) × Hours in a day × daily water demand

systems [88,89] owing to their easiness of usage in the optimization

number of process. The prime purpose of the generation of such a metamodel is
power sources correlated with complicated mathematical models and is associated
∑
CO2 emissions = (CO2 emission factor × Capacity of power source)i with a high computational burden. Consequently, such physical models
i=1 are not user-friendly for the multi-objective optimization process where
(8) the model is evaluated various times during the mutation, selection, and
In Eq. (7), the project lifetime has 20 years, and the daily water crossover operations to find the best fitness function. In such cases, a
demand is 214.2 m3/day (see section 3.2), yielding the units USD/m3 of simplified digital twin can significantly reduce the optimization
the cost of potable water. In Eq. (8), the CO2 emission factor of photo­ computational times because it just consists of a black box with a set of
voltaic panels, batteries, generator, and inverter are 0.032 kgCO2/kWh input variables and one or more output variables, whereas, the physical
[80,81], 0.11 kgCO2/kWh [82–84], 0.778 kgCO2/kWh [81], and 0.641 model consists of energy balances, control strategies, parameter calcu­
kgCO2/kWh [85–87], respectively. lations, and verification processes of assumptions. The digital twin is
generated using the data-driven fitting techniques of artificial

10
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

intelligence including the application of artificial neural network (ANN), Problem formulation of multi-objective optimization
and group-method-of-data-handling (GMDH).
ANN consists of a feed-forward multilayered perceptron neural Multi-objective optimization is applicable in the situation where
network which is applied for the metamodeling of renewable fraction there is a trade-off (or conflict) between the objective functions.
and CO2 emissions. For this process, the database is divided into three Therefore, the prime objective is to provide an upgrade to the baseline
parts corresponding to training, validation, and testing sets making case conditions and the single objective optimization results. The
percentages of 70, 15, and 15, respectively. The training dataset trains formulation of the multi-objective optimization problem involves the
the neural network for an appropriate fit, the validation assures that the presentation of the objective function, design variables, and the search
training dataset does not overfit the database, and finally, the testing space. It should be noted that the current problem is an unconstrainted
database is used to verify the regression fit. ANN consists of a particular multiobjective optimization problem, and it is described as follows:
architecture1 in which the training algorithm and the number of hidden 1. Objective functions for multiobjective optimization: In mul­
layers are the tuning parameters. An appropriate combination of these tiobjective optimization, it is desired to maximize the renewable frac­
two tuning parameters must be evaluated for the development of the tion, and minimize the cost of potable water and CO2 emissions.
metamodel. MATLAB neural network toolbox is utilized for this method. Mathematically, it can be written as:
It can be seen in Table 7 that the most suitable training algorithm for the ⎧
renewable fraction is Levenberg–Marquardt having a neural network of ⎨ obj1 : − Renewable fraction
8 number of hidden layers corresponding to an R2 (Coefficient of Min(obj) = obj2 : Cost of potable water (9)
⎩
obj3 : CO2 emissions
determination, see Appendix B) value of 0.94. Similarly, the most suit­
able training algorithm for CO2 emissions is also Levenberg–Marquardt Note: The renewable fraction is multiplied by a negative one to
having a neural network of 6 number of hidden layers corresponding to accommodate the syntax of the optimization problem and the algorithm,
an R2 value of 0.99. The regression fit of the selected architectures of therefore, the minimization of obj1 : − Renewable fraction yields its
renewable fraction and CO2 emissions are presented in Fig. 5(a) and maximum value. obj refers to the objective function.
Fig. 5(c). It can be noted from Fig. 5(a1 and c1) that datapoints follow 2. Design or tuning variables: The optimization is carried out based
the linear pattern between the target and the output variables having on the following design variables: PV capacity, generator capacity,
acceptable error performance indicators (RMSE(Root-Mean-Square- inverter capacity, and battery capacity.
Error) = 0.460, SI(Scatter Index) = 0.46 %, MAPE(Mean Absolute Per­ 3. Search space for optimization: The optimization problem is
centage Error) = 0.0029 for renewable fraction, and RMSE = 184.7, SI
constrained with a set of limits on the design variables which defines the
= 0.23 %, MAPE = 0.0035 for CO2 emissions). It can also be noted from
search space or solution zone for the multiobjective algorithm. This
Fig. 5(a2 and c2) that there is no visible pattern in the error distribution
space is expressed mathematically in the following equation:
that guarantees the independence and homoscedasticity of errors.
Similarly, it can be noted from Fig. 5(a3 and c3) that the error distri­ 15 kW ≤ PV system ≤ 20 kW
bution is also normal which guarantees satisfaction towards the neural 1 kW ≤ Diesel generator ≤ 8 kW (10)
regression model. 40 kWh ≤ Batteries capacity ≤ 54 kWh
The cost of potable water is digitally modeled using the GMDH The constrained search space on the optimization problem is esti­
technique. It belongs to the family of polynomial neural network2. mated from the findings of the single objective optimization (section 4).
However, this technique is a self-organizer that performs an automatic Here, it can be considered as a preliminary study to estimate the search
evaluation of the number of neurons, layers, and transfer function. This space for multiobjective optimization. Another justification to use this
technique is only adapted for the digital twin generation of cost of constrained search space also comes up from the limited resources of the
potable water because ANN could not produce a satisfactory model local context in terms of the available space, and capital.
owing to the manual evaluation of the training algorithm and the
number of hidden layers. Nevertheless, GMDH comes up with other
types of tuning parameters which are (a) the maximum number of input Theory and implementation of the NSGA-II (Non-dominating sorting
neurons, input source, the maximum number of neurons, and the genetic algorithm II)
polynomial power. It is implemented in MATLAB [90] using the algo­
rithm published by Sohani et al. [91]. The dataset of the numerical ex­ Non-dominating sorting genetic algorithm II (NSGA-II) is applied for
periments is only divided into two parts corresponding to the testing and the multi-objective optimization problem and as suggested by Sohani
training phases scoring 85 % and 15 %, respectively. Through a set of et al. [93] that NSGA-II has several technical advantages over other
iterations, it can be noted from Table 7 that the most suitable archi­ techniques. The technique works on the process of natural selection in
tecture is sample no. 8 yielding an R2 of 0.994. The regression fit which which the final breed is the best fit as compared to its ancestors. The
can be seen in Fig. 5(b1) shows the suitability of fit in which RMSE = problem is initiated based on the range of the problem along with the set
0.00013, SI = 0.23 %, and MAPE = 0.0019. Similarly, the test on the of constraints. The initialized population is sorted based on the non-
errors can be verified from Fig. 5(b2 and b3). domination characteristics which is a measure of an individual to
It is extremely vital to verify and validate the generation of the dominate another if its objective function is not worse than the other
digital twins because the quality of the optimization process completely members and at least one of its objective functions is better than the
depends upon it. Therefore, extensive external validation of these digital other members. After the non-dominated sorting process, the crowding
twins is carried out which is attached in Appendix C. This gives sufficient distance is assigned which is a measure of closeness of one member to its
confidence to the authors to claim the reliability of the digital twins and neighbor. The concept behind the assignation of crowing distance is to
can be employed in the optimization process. evaluate the Euclidian distance between the individuals and larger
values of crowing distance will yield a better diversity in the population.
Afterwards, the selection of the individuals is considered using a binary
tournament selection with a crowed-comparison-operator. The best-fit
solutions are passed on to the new generation and the worst-case sce­
1
Theoretical details of the ANN architecture can be reviewed in the previous narios are eliminated. Afterwards, the processes of mutation, cross-over,
work of authors [92]. and selection are carried out and this new generation is replaced with
2
Theoretical details of the GMDH architecture can be reviewed in the pre­ the previously rejected members. Since all the previous and the current
vious work of authors [92] and the one published by Sohani et al. [91]. best individuals are added to the population, therefore, elitism is

11
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Fig. 5. (a1, b1, and c1) Comparison between the target and output performance indicator, (a2, b2, and c2) verification of independence and homoscedasticity of
errors, and (c1, c2, and c3) verification of normal distribution of errors.

12
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

evaluated using a suitable decision-making method. In this case, The

Order of Preference by Similarity to Ideal Solution (TOPSIS) is adapted
in which the relative distance (d+ i , di ) of each optimized performance
−

indicator (PI) is calculated with respect to the ideal and non-ideal so­
lution, given by the equation:
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√ 3 [ ]2
√∑
di = √
+
PIij − PIideal j (11)
j=1

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√ 3 [ ]2
√∑
d−i =√ PIij − PInon− j
ideal
(12)
j=1

In the next step, the relative closeness of each alternative with

respect to the ideal solution (d+
i ) is calculated, using the equation:

d−i
RCi = (13)
d−i + d+i

The matrix of relative closeness is sorted in ascending order and a

single solution with its maximum value is chosen.

Fig. 6. Pareto front along with TOPSIS best selection of the multi-objective Optimization results (Pareto front)
optimization scenario.
The outcome of the application of multi-objective optimization and
ensured. This process is repeated in such a way the best individuals the TOPSIS solution are presented in Fig. 6. It can be noted that the
could pass their best genes in the upcoming generations until reaching a results of the multi-objective optimization are a set of solutions enclosed
convergence criterion. The multi-objective optimization is implemented in the area of the performance indicators with the following range:
in MATLAB global optimization toolbox [90] using the default charac­ renewable fraction ranges from 98 % to 99 %, cost of potable water
teristics and the flow diagram highlighting the steps for the imple­ ranges from ~ 0.0446 USD/m3 to ~ 0.0456 USD/m3, and the CO2
mentation of the code is presented in Appendix D. emissions range from 36,970 kg to 37,210 kg. The best selection opted
by TOPSIS is the following point on the Pareto front: the renewable
fraction of 98.5 %, the cost of potable water of 0.0447 USD/m3, and the
Multicriteria decision analysis using The Order of Preference by Similarity CO2 emissions of 36,970 kg.
to Ideal Solution (TOPSIS)
Comparison of optimization scenarios
Unlike the single-objective optimization, the results of multi-
objective optimization are a set of non-dominating candidate solutions In this section, the results of the spreadsheet and HOMER are
which is represented in a Pareto front (see Fig. 6). The final multi- compared with the multi-objective optimization scenario. The results of
objective optimization point among the Pareto front should be HOMER and spreadsheet consider the optimization scenario in which

Table 8
Concluding a comparison of the optimized performance indicators and the system capacity for various optimization scenarios.
Optimization scenario Optimized performance indicators System capacity

Cost of Net Renewable Cost of CO2 PV, Generator, Batteries, Inverter,

electricity, present fraction, % potable emissions, kW kW kWh kW
USD/kWh cost, water, kgCO2/year
USD USD/m3

Optimization scenario no. 1: Min(obj) = 0.205 $70,415 98.9 0.0450 – 17 5 48 6.8

⎧
⎨ obj : Net present cost (NPC)
↕ *Interpretation:
⎩
obj : Cost of electricity (COE)
Minimize net-present-cost (NPC) and cost of
electricity (COE) through HOMER.
Optimization scenario no. 2: Min(obj) = 0.229 $78,732 95.9 0.0503 – 17.7 5 44 5.3
⎧
⎨ obj : Net present cost (NPC)
↕ *Interpretation:
⎩
obj : Cost of electricity (COE)
Minimize net-present-cost (NPC) and cost of
electricity (COE) through a spreadsheet.
Optimization scenario no. 3: Min(obj) = 0.203 $69,895 98.5 0.0447 36,970 16.5 4 50 6.5
⎧
⎨ obj1 : − Renewable fraction
obj2 : Cost of potable water Interpretation:
⎩
obj3 : CO2 emissions
Multiobjective optimization to maximize renewable
fraction, minimize the cost of potable water, and
minimize CO2 emissions using surrogate (digital
twin) model implemented on MATLAB.

* Optimization scenario no. 1 and no. 2 are single-objective optimization because net present cost and the cost of electricity are linearly correlated.

13
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Table 9
Results of sensitivity analysis. Note: R is the Pearson correlation coefficient. A p-value greater than 0.05 indicates statistical insignificance for a confidence interval of
95%. The table is symmetric at y = -x.

the decision is carried out using only a single-objective optimization, i. based sensitivity analysis is displayed in Table 9, which highlights the
e., net-present-cost and the cost of electricity are linearly correlated. correlation of the input variables (PV capacity, inverter capacity,
However, the multi-objective optimization considered the energetic, generator capacity, and battery storage capacity) with the output vari­
economical, and environmental performance indicators in the form of a ables (net present cost, cost of electricity, renewable fraction, cost of
renewable fraction, cost of potable water, and yearly CO2 emissions, potable water, and CO2 emissions). Statistical-based sensitivity analysis
respectively. Consequently, the perspective of the multi-objective opti­ is implemented on the software STATGRAPHICS [95]. This analysis also
mization is much broader than the optimization proposed by HOMER or assists to understand the correlation and interaction between the input
the one in the spreadsheet. variables and the output variables. It can be noted that a positively
It can be seen in Table 8 that the performance indicators through the correlated and statistically significant strongest correlation is noted
multi-objective optimization are more refined as compared to the other between the cost of electricity and cost of potable water since they are
methods. It is reported that through the multi-objective optimization, linearly dependent while being a function only of the net present cost. It
the cost of electricity and the net present cost is further reduced by 0.86 is followed by a renewable fraction with the cost of potable water which
% and 0.73 %, respectively at a decrement of only 0.4 % of the renew­ is rational because a very high renewable fraction can also correspond to
able fraction. It makes sense because even a high renewable fraction can the high cost of renewable energy equipment which yields a rise in the
also contribute to carbon emissions because of the emission factor of a cost of potable water. Similarly, there is a positive correlation between
photovoltaic system, battery storage, and inverters. In other words, a the generator and inverter capacity with the CO2 emissions, and it cor­
higher renewable energy fraction can move the system to an overdesign responds to the fact that the increment in the size of the generator and
point where the emissions are high. Therefore, the results of multi- inverter would contribute towards the life cycle CO2 emissions. The
objective optimization are significant because they are based on the statistically significant negatively correlated is noted between the net
minimization of CO2 emissions. Consequently, a lower renewable frac­ present cost and cost of electricity with the renewable fraction signifying
tion is justified to constraint the emissions of CO2. that a very high net present cost and cost of electricity corresponds to
high usage of diesel generators, resulting in a lower renewable fraction.
Sensitivity of design variables on the objective functions However, there is one drawback of this statistical-based sensitivity
analysis: although it exhibits the relationship between the variables, it
Sensitivity analysis is an assessment of the uncertainty of the design does not exhibit the sensitivity of the model with respect to any uncer­
variables on the objective functions. It is used to test the robustness of tainty in the design variables. Therefore, calculus-based sensitivity
the results of a model in the presence of uncertainty, development of analysis is reported here to further extend the analysis.
better robust models, calibration of the theoretical model, identifying
the desired optimization space, to enhance the communication from the
modeling hand to the decision-makers, to simplify the model, to look for Calculus-based (partial derivative-based) sensitivity analysis
errors in the model, to reduce the uncertainty, and to understand the
nature of the relationship between the input and the output variables of The calculus-based sensitivity method considers the summation of
a system. This tool also helps to identify more responsive variables in the each design variable and its respective uncertainty, which can be
system. expressed as follows [96]:
̂x
X=X±U (14)
Statistics based sensitivity analysis
In eq. (14), X is the design variable, which can be PV capacity,
In this work, a calculus [94] and statistics-based sensitivity analysis, inverter capacity, generator capacity, or battery capacity. X is its nom­
which is reliable and fundamental in any area is carried out. Statistical inal value, and ± U
̂ x is the possible uncertainty in the nominal value of

14
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Fig. 7. Results of sensitivity analysis including normalized sensitivity coefficient and relative sensitivity coefficient for (a) net present cost, (b) cost of electricity, (c)
Renewable fraction, (d) cost of potable water, and (e) CO2 emissions.

the design variable [97]. This method expresses the corresponding un­ ̂ Y represents the uncertainty in any Y objective function,
In eq. (15), U
certainty in the objective function owing to the uncertainty in the design which can be net present cost, cost of electricity, renewable fraction,
variable, which can be written as: cost of potable water, and/or CO2 emissions. The term dY represents the
dX
change in the objective function Y with respect to the set of design
̂ Y = dY U
U ̂x (15)
dX variables X. Therefore, in such multivariate problems, the uncertainty in

15
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Table 10
Comparison of different results about hybrid renewable energy systems in the literature.
Research Hybrid system Load detail Sizing Approach Results of sizing Overview
study component
Technical Economic Environmental
indicators indicators indicators

[43] PV, Energy Storage 24.4 kW HOMER PV: 332 kW NPC: 465,790 CO2 emissions: This research study is carried out to
(ES), DG, Hydro, BESS total ES: 3 kW USD 20,868 kg/yr analyze the provide power supply
BESS: 1100 COE: 0.10 using an HRES for a remote rural area
kWh USD/kWh in the state of Karnataka (India) with a
combined dispatch strategy.
[103] PV, ES, Pumped hydro 255.6 kWh/ Particle Swarm PV: 84 kW COE: 0.196 Unspecified This study investigates and compares
storage (PHS) day total Optimization ES: 75 kW USD/kWh the various combinations of
PHS: 320 kWh renewable energies (solar, wind) and
storage technologies (battery,
pumped hydro storage, hybrid
storage) for an off-grid power supply
system.
[39] PV, ES, Biogas 724.83 GA PV: 100 kW NPC: 890,013 CO2 emissions: This work compares two methods to
generator (BGG), kWh/day HOMER ES: 50 unities USD 3,842 kg/yr find the best optimal solution of NPC,
biomass generator total BESS: 200 COE: 0.163 COE, and CO2 emissions of an HRES.
(BMG), Battery Energy 149.21 kW unities USD/kWh
Storage System (BESS), peak BGG: 60 kW
and fuel cell BMG: 50 kW
Fuel cell: 57
kW
Electrolyzer:
50 kW
[104] PV, ES, DG, BESS 3000 kW Multi-Objective Self- PV: 40.4 kW COE: 0.050 RF: 95.18 % This paper conducts the optimal sizing
peak/yr Adaptive ES: 10 kW USD/kWh of a PV/ES/DG with battery storage
Differential using the Multi-Objective Self-
Evolution Adaptive Differential Evolution
(MOSaDE) (MOSaDE) algorithm for the city of
Yanbu, Saudi Arabia.
[44] PV, ES, DG, BESS 200 kW HOMER PV: 200 kW NPC: CO2 emissions: This research study develops an
average ES: 200 kW 3,677,573 314.6 Tons/yr economical and technical feasibility
DG: 365 kW USD method to determine the best
BESS: 150 COE: 0.751 implementation of a novel energy
kWh USD/kWh storage system in an HRES.

[105] PV, DG, Grid 30 kW base HOMER PV: 200 kW NPC: RF: 18 % This work investigates the
758 kW Grid: 850 kW 3,336,403 optimization of several hybrid energy
peak USD system models consisting of solar PV,
COE: 0.130 diesel generators, and grid.
USD/kWh
[106] PV, ES; BMG, BGG, 2,434,500 Genetic algorithm Hydro: 16 kW NPC: 605,376 Unspecified The present work focuses on the
Hydro, BESS kWh/yr (GA) BMG: 77 kW USD optimal sizing of an integrated
BGG: 50 kW COE: 0.087 renewable energy system considering
ES: 49 kW USD/kWh locally available different renewable
BESS: 936 kVA energy sources for electrification of a
remote area in Karnataka state in
India.

[45] PV, ES, BESS, BGG 54 kWh/ HOMER PV: 1 kW NPC: 72,232 CO2 emissions: This paper presents an economic
day average ES: 1 kW USD 26,894 kg/yr feasibility analysis of a single
BGG: 2 kW RF: 100 % standalone house operating with a
BESS: 2.16 COE: 0.306 hybrid power plant consisting of a
kWh USD/kWh fixed capacity producer gas generator
and other renewable energy sources
(PV and ES).

[46] PV, ES, DG, BESS 33 kWh/ HOMER PV: 18 kW NPC: 288,194 CO2 emissions: This study is carried out to analyze the
day ES: 20 kW USD 198,347.9 kTons/ performance of an off-grid/PV/ES/
DG: 3 kW COE: 1.877 yr DG/BESS hybrid energy system for a
3.9 kW peak BESS: 25 USD/kWh RF: 66.3 % remote area in Malaysia.
unities
[47] PV, BESS, DG, BMG, 100 kW Particle Swarm PV: 56 kW COE: 0.077 CO2 emissions: The present work includes sizing of
BGG, Micro-hydro peak Optimization BESS: 110 kVA USD/kWh 14,614 kg/yr hybrid energy systems using various
generator (MHG) GA BMG: 30 kW RF: 96 % energy management strategies which
Biogeography BGG: 35 kW are the cycle charging strategy, load
techniques DG: 30 kW following strategy, and peak shaving
MHG: 15 kW strategy.
Present PV, ES, DG, BESS 4.25 kW Spreadsheet PV: 16.5 kW NPC: 69,895.6 CO2 emissions: This research study aims to evaluate
study maximum ANN DG: 4 kW USD 36,970 kg/yr the technical and economic indicators
74.09 kWh/ HOMER BESS: 50 kWh COE: 0.2032 RF: 98.5 % of different HRES to satisfy the energy
day Power USD/kWh demand of two water treatment plants
Converter: 6.5 Cost of in a remote community in the state of
kW potable water:
(continued on next page)

16
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Table 10 (continued )
Research Hybrid system Load detail Sizing Approach Results of sizing Overview
study component
Technical Economic Environmental
indicators indicators indicators

0.0447 USD/ Yucatan by using artificial

m3 intelligence and HOMER.

Abbreviations and/or acronyms: PV: photovoltaic panels, ES: energy storage, DG: diesel generator, Hydro: Hydropower, BESS: battery energy storage system, HOMER:
Hybrid Optimization of Multiple Energy Resources, NPC: net present cost, COE: cost of energy, HRES: hybrid renewable energy systems, PHS: pumped hydro storage,
BGG: biogas generator, BMG: biomass generator, GA: genetic algorithm, MOSaDE: Multi-Objective Self-Adaptive Differential Evolution, RF: renewable factor, MHG:
micro-hydro generator.

Y with respect to the perturbations in X are calculated as the root sum summation of RSC’s for a particular objective function for all the design
square product of each individual perturbation [98]. It is expressed in variables is equal to 1 (or 100%).
the following equation: Calculus-based sensitivity analysis is applied on Engineering Equa­
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ tion Solver (EES) using the toolbox “Uncertainty Propagation” [101].
√ last objective function [( )2 ]
√ ∑ dY The results of the derivate-based sensitivity analysis are reported in
̂Y = √
U Û x,objective function (16)
first objective function
dXobjective function Fig. 7. It can be noted from Fig. 7(a1, b1, and d1) that the uncertainty of
change in the PV capacity, inverter capacity, and battery capacity would
Another suitable option is to report the results of the sensitivity bring down the net present cost, cost of electricity, and cost of potable
analysis through Normalized Sensitivity Coefficients (NSC) which is water. Practically, it is also reasonable, besides, it also complies with the
attained through the normalization of the uncertainty U
̂ Y by dividing it findings of Table 9, where the energy load is constant. Therefore, an
with the nominal value of the objective function [99]. Thus, it can be increase in these design variables would techno-economically justify the
written as: renewable energy components in this hybrid energy system. However, it
can also be noted from the same figures that the uncertainty of change in

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
⎡⎛ ⎞2 ⎛ ⎞2 ⎤
√
√ last objective function
̂Y √
U ∑ ⎢ ⎜ dY X ⎟ ⎜ Û ⎟ ⎥
=√ (17)
objective function objective function x,objective function
√ ⎣⎝ × ⎠⎝ ⎠ ⎦
Y first objective function
dX objective function Y objective function X objective function

Summarizing all the information, the normalized value of the the generator capacity would increase the net present cost, cost of
sensitivity analysis of just the Net present cost (NPC) can be written as: electricity, and the cost of potable water since the diesel generator
would be using an expensive resource (diesel) instead of exploiting

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√( )2 ⎛ ⎞2 ( )2 ⎛ ⎞2
√
̂
U NPC √ dNPC PV capacity Û PV capacity dNPC Generator capacity Û Generator capacity
=√ × ⎝ ⎠ + × ⎝ ⎠
NPC d(PV capacity) NPC PV capacity d(Generator capacity) NPC Generator capacity
( )2 ⎛ ⎞2 ( )2 ⎛ ⎞2
dNPC Battery capacity ̂ Battery capacity
U dNPC Inverter capacity ̂ Inverter capacity
U
+ × ⎝ ⎠+ × ⎝ ⎠ (18)
d(Battery capacity) NPC Battery capacity d(Inverter capacity) NPC Inverter capacity

Once these parameters are evaluated, a relative sensitivity coeffi­ freely available solar energy. The findings of Fig. 7(c1) illustrate that the
cient can be estimated to identify the dominant contributors among the renewable energy components of this hybrid system including solar
design variables towards the objective functions. Mathematically, it is panels, inverters, and batteries are contributing positively towards the
calculated by multiplying the sensitivity coefficient and uncertainty renewable fraction, whereas, the generator pulls down this renewable
normalized by the uncertainty in the objective function accompanied by fraction. The interpretation of Fig. 7(e1) and the last column of Table 9
the square of the whole equation [100]. Thus, it can be written as: are somewhat challenging where apparently it seems like only the
⎛⎞2 inverter and the generator are contributing positively towards the
dY ̂ environmental emissions, whereas incrementing solar panels and bat­
⎜dX U x⎟
RSC = ⎝ ⎠ (19) teries lowers the environmental emissions. It is explained through the
̂Y
U concept that the design variables are also correlated. Therefore, an in­
crease in the PV capacity would yield a decrease in the generator ca­
In eq. (19), RSC is the relative sensitivity coefficient and the pacity to meet the same energy load. Since the CO2 emissions of batteries

17
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

and solar panels are lower than the other components [82–84]. There­ the fact that the distribution comes from a water treatment plant which
fore, it seems that the inverter and generator are contributing positively can be considered private and does not involve other costs that JAPAY
in this case. In summary, the net present cost, renewable fraction, and considers, such as operation, pumping, expansion of the distribution
cost of potable water are strongly correlated with the battery size (Fig. 7 system, among others. Moreover, this water does not go through the
(a2, c2, and d2)). Similarly, the most sensitive parameter for the cost of processes needed for the water to be potable, as it just serves for cleaning
electricity and the CO2 emissions are solar panels and diesel generators, purposes.
respectively (Fig. 7(b2, and e2)).
Conclusion
Comparison of results with external literature
In the present work, different configurations were analyzed for the
It is possible to find different works related to this research study in sizing of hybrid renewable energy systems proposed for supplying the
the current literature, which tend to optimize energetic, economic, and energy demanded from a water purification system suggested for
environmental indicators from the hybrid renewable energy system installation in the indigenous Mayan community of San Jose Tipceh
proposed for the corresponding application. In Table 10, a summary of located in Yucatan, Mexico [107]. The technologies considered for the
the results obtained by different authors is presented. different configurations of the hybrid systems were photovoltaic mod­
As seen from Table 10, COE results are reported between 0.050 USD/ ules, wind turbines, diesel generators, and a storage energy system with
kWh and 1.877 USD/kWh, and NPC between 72,232 USD and 3,677,573 Pb-acid, Li-on, and AGM as alternatives for the battery selection. The
USD. In this study, an optimized COE of 0.2032 USD/kWh was obtained, sizing process was performed using a spreadsheet developed by the
which is within the range of values reported in Table 8. However, the authors of this work, as well as with the software HOMER pro. The main
value of the NPC is lower than the results from the literature. This is criteria considered for the selection of the hybrid systems were the
attributed to the daily energy demand of this study, which was 79.06 economic indicators of net present cost and cost of energy. For both
kWh/day, with a maximum demand of 4.25 kW. When comparing this techniques used for the sizing process, the results indicated that the best
value with the one reported by [39], there is an energy demand of hybrid system comprised of photovoltaic technology with a diesel
724.83 kWh/day and an NPC of 890,013 USD. Thus, it is observed from generator and a storage energy system with Li-ion batteries. The values
Table 10 that the installed capacity is much higher than the reported in obtained with HOMER for the NPC and COE were 70,415 USD and 0.205
this work, which would lead to an increase in capital costs, O&M, and USD/kWh, while those obtained with the spreadsheet were 78,732 USD
replacement costs during the entire useful life of the project. and 0.229 USD/kWh, respectively. Moreover, the renewable fraction
Similarly, two important environmental indicators are CO2 emis­ was 98.5 % obtained from HOMER, while from the spreadsheet it was
sions and renewable fraction. The value of CO2 emissions obtained in 95.9 %. Additionally, the potable water costs were 0.0450 USD/m3 and
this work was 36,970 kg/yr, and when comparing this result with 0.0503 USD/m3 for HOMER and the spreadsheet, respectively.
Table 10, it is observed that there are several similar values reported in Regarding the results obtained from the multi-objective optimization, it
the literature [48,50]. Moreover, the renewable fraction of this study was found that the NPC and the COE were 69,895 USD and 0.203 USD/
was 98.5 %, and according to Table 10 values up to 100 %, a renewable kWh, respectively, which represents a lower value than the results from
fraction are reported. It should be clarified that these values depend on HOMER with a difference of 500 USD for the NPC and 0.002 USD/kWh
the percentage of participation of renewable sources within the hybrid for the COE. Moreover, the cost of potable water was optimized and a
system. Finally, the present study reported a cost for potable water and value of 0.0447 USD/m3 was obtained, while the renewable fraction was
energy obtained by the optimized hybrid renewable energy systems. 98.5 % and the total CO2 emission was 36,970 kgCO2/year. Finally, these
However, the indigenous Mayan community already has an established results were compared with results obtained by different authors, con­
cost for these resources, obtained by conventional methods, so it is firming that the present study lies in the context of recent research
important to compare the costs obtained in this work with the ones that literature and makes it suitable for the sizing and optimization of hybrid
already exist. renewable energy systems.
Regarding the cost of energy, the Federal Electricity Commission There are certain advantages and disadvantages to the adapted
(CFE, by its acronym in Spanish) is the Mexican entity in charge of method. One of the advantages is that the usage of spreadsheet algo­
supplying energy to the country. Considering the region where San Jose rithms offers computationally cheap, and reliable results as compared to
Tipceh has located as well as the industrial use that will be given to the the usage of licensed HOMER pro which is equally reliable for single-
energy, the corresponding energy cost is labeled as high demand in or­ objective optimization scenarios. Another advantage of this method is
dinary medium voltage. For this energy tariff and considering the costs that the multiobjective optimization adds certain features to the HOMER
published by CFE for January 2021, the COE is 0.07245 USD/kWh, with pro and spreadsheet algorithm by including the environmental and
a monthly fixed charge of 27.18 USD and distribution and capacity water cost. One of the drawbacks of the usage of multiobjective opti­
charges of 4.66 USD/kW and 13.56 USD/kW, respectively [102]. This mization is that it’s computationally expensive as compared to spread­
COE from CFE is better in comparison with the results obtained from the sheet algorithms since it involves an intermittent step of the generation
hybrid renewable energy systems. However, the Mayan community does of the digital twin. As a precautionary measure, extensive validation
not have the electrical infrastructure needed to supply the energy for should be accomplished before using any digital twin for the multi­
high demand in ordinary medium voltage consumers, which in this case objective optimization because they are vulnerable to several numerical
would be the water treatment plants. errors. Additionally, the operational range of the digital twin should be
Likewise, regarding the cost for potable water, the Yucatan Board of big enough to cover the search space of the optimization results.
Water Supply and Sewerage (JAPAY, by its acronym in Spanish) is the Otherwise, the optimization procedure can also extrapolate the surro­
public organization dedicated to the administration, operation, conser­ gate model which would lead to uninterpretable findings.
vation, expansion, and construction of potable water systems in the State Finally, the lack of proper water and wastewater treatment infra­
of Yucatan, and the cost of water is 0.36 USD/m3 [54]. When comparing structure in isolated rural areas like the one studied must be addressed in
this value with the one obtained in this work, there is a difference of order to guarantee water safety for the future. Hence, it is highly desired
0.3153 USD/m3 favoring the hybrid energy system. This is attributed to to incorporate economically feasible technologies from an energetic

18
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

point of view such as HRES. The design, techno-economic optimization, Methodology, Validation, Resources, Writing – review & editing, Su­
and sensitivity analysis of such a water management plant are reported pervision, Project administration, Funding acquisition. M. Vega De
in this article. At the next step, a feasibility study should also consider Lille: Methodology, Validation, Resources, Writing – review & editing,
the societal and governmental constraints. For instance, social rejection Supervision, Project administration, Funding acquisition.
is a pivotal factor to acknowledge since resistance towards renewable
energy megaprojects in Yucatan, Mexico, by the indigenous community Declaration of Competing Interest
has been previously demonstrated by El-Mekaoui et al. [107]. Conse­
quently, it is important to consider that the implementation of tech­ The authors declare that they have no known competing financial
nologies such as the proposed water management system should be interests or personal relationships that could have appeared to influence
carried out in harmony with the beliefs, culture, and lifestyle of the the work reported in this paper.
targeted communities.
Acknowledgment
CRediT authorship contribution statement
The authors, Rasikh Tariq, A.J. Cetina-Quiñones, V. Cardoso
Rasikh Tariq: Conceptualization, Methodology, Software, Valida­ Fernández, and Daniela Abigail Hernandez Lopez, are grateful to the
tion, Formal analysis, Investigation, Resources, Data curation, Writing – financial support of CONACYT (Consejo Nacional de Ciencia y Tecnología)
original draft, Writing – review & editing, Visualization, Supervision, to pursue a postgraduate degree in Facultad de Ingeniería, Universidad
Project administration. A.J. Cetina-Quiñones: Conceptualization, Autónoma de Yucatán under the following details: (a) Rasikh Tariq, CVU:
Methodology, Software, Validation, Formal analysis, Investigation, Re­ 949314, scholarship no: 784785; (b) A.J. Cetina-Quiñones, CVU:
sources, Data curation, Writing – original draft, Writing – review & 861995, scholarship no: 784880; (c) V. Cardoso Fernández, CVU:
editing, Visualization, Project administration. V. Cardoso-Fernández: 1006703, scholarship no: 752330, and (d) Daniela Abigail Hernández
Conceptualization, Methodology, Software, Formal analysis, Investiga­ López, CVU: 1007727, scholarship no: 752354.
tion, Resources, Data curation, Writing – original draft, Writing – review
& editing, Visualization, Project administration. Hernández-López Appendix A. Annual total load and total power output of
Daniela-Abigail: Methodology, Software, Formal analysis, Resources, optimized hybrid energy system
Data curation, Writing – original draft, Writing – review & editing. M. A.
Escalante Soberanis: Conceptualization, Methodology, Software, See Fig. A1
Validation, Resources, Writing – review & editing, Visualization, Su­
pervision, Project administration, Funding acquisition. A. Bassam:

Fig. A1. Annual total load and total power output of the optimized hybrid energy system (case 4: PV/DG/BESS with Li-ion battery storage, see Table 6) displaying
the pattern for each hour of the whole year. Note: Each individual graph in this figure exhibits a time frame of 3 months, thus, four figures correspond to the
simulation of the whole year. It can be observed that the peaks of total output is higher than the total load and their difference corresponds to the excess en­
ergy production.

19
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Appendix B. Performance indicators of the regression fit

See Table B1.

Table B1
Performance indicators of regression fit.
Indicator Formula Description Significance

Coefficient of R2 = The coefficient of determination refers to the closeness of the fitted data. An R2 value close to unity refers that the
determination (R2) ∑N ( )2 PIactual corresponds to the actual performance index, PIpredicted refers to the regression quality is high.
i=1 PIactual − PIpredicted
1− ( )2 predicted performance index from the regression method, and PIactual is the
∑N
i=1 PIactual − PIactual average value of the actual performance index.
Mean Squared Error This coefficient is a measure of the average of the square of the errors. The values of mean square error close to
MSE =
(MSE) zero are suitable for the regression
1 ∑N ( )2 analysis.
PIactual − PIpredicted
N i=1
Root Mean Square It is the standard deviation of the residuals; thus, indicating the spread of Like the MSE, a root-mean-square error
Error (RMSE) RMSE = the residuals far from the regression line. near zero corresponds to the perfect
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ regression analysis.
∑N ( )2
i=1 PIactual − PIpredicted
N
Scatter Index (SI) RMSE It presents the percentage of root-mean-square difference with respect to A scatter index of less than 5% is
SI = 100 × [∑ ]/
N mean observation or it gives the percentage of expected error for the considered as a strong relation.
i=1 PIpredicted N
parameter [108].

Appendix C. External validation of digital twins

This procedure further extends the reliability of the meta-model. The

motivation to carry out this validation is that a multivariate regression
can often yield a biased high R2 owing to the availability of many tuning
parameters. Additionally, several researchers consider that R2 is not
sufficient proof of high predictive properties [109] of the neural fit.
Consequently, external validation is conducted using the method pre­
sented by Tariq et al. [92] and Tzuc et al. [72]. It is determined by the
correlation fit between the experimental and the predicted data which
should range between 0.85 and 1.15. Similarly, the external predict­
ability indicator should be less than 0.5. The correlations with the
criteria implemented on the digital twins are shown in Table C1. I. It can
be noted that the developed digital twins satisfy all the conditions of
external validation.

Table C1
External validation statistical criteria of the digital twin.
Statistical criteria Condition The digital twin of:

Renewable fraction Cost of potable water CO2 emissions

R2 R2 ≥ 0.64 0.94 0.994 0.99

∑no.ofdata
PIiactual PIipredicted 0.85 ≤ a ≤ 1.15 0.9998 1.0000 1.0003
i=1
a = ∑ ( )2
no.ofdata
i=1 PIipredicted
∑no.ofdata i
PIactual PIipredicted 0.85 ≤ a’ ≤ 1.15 1.0001 1.0000 0.9997
i=1
a’ = ∑no.ofdata ( )2
i=1 PIiactual
∑no.ofdata ( i )2
– 0.9999 1.0000 1.0000
i=1 PIpredicted − aPIipredicted
R2o = 1 − ( )2
∑no.ofdata
i=1 PIipredicted − PIpredicted
∑no.ofdata ( i
PIactual − a’PIiactual
)2 – 0.9999 1.0000 1.0000
i=1
R2u = 1 − ( )2
∑no.ofdata
i=1 PIiactual − PIactual
( √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ⃒)
⃒ Rm ≥ 0.50 0.7099 0.7097 0.7098
2
Rm = R 1 − ⃒R − R2o ⃒
2

2
R − R2o |m| < 0.10 − 0.0637 − 0.0638 − 0.0638
m =
R2
R2 − R2u |m’| < 0.10 − 0.0637 − 0.0638 − 0.0638
m’ =
R2
All the conditions are satisfied? Yes Yes Yes

20
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

Appendix D:. Genetic algorithm-based multi-objective

optimization

See Fig. D.1.

Fig. D.1. The methodology diagram opted for multi-objective optimization using a genetic algorithm.

References [3] Vasistha P, Ganguly R. Water quality assessment of natural lakes and its
importance: an overview. Mater Today Proc 2020;9. https://doi.org/10.1016/j.
matpr.2020.02.092.
[1] Bahar R, Ng KC. Fresh water production by membrane distillation (MD) using
[4] Tariq R, Sheikh NA, Xamán J, Bassam A. An innovative air saturator for
marine engine’s waste heat. Sustain Energy Technol Assessments 2020;42:
humidification-dehumidification desalination application. Appl Energy 2018;
100860. https://doi.org/10.1016/j.seta.2020.100860.
228:789–807. https://doi.org/10.1016/j.apenergy.2018.06.135.
[2] Tariq R, Sheikh NA, Livas-García A, Xamán J, Bassam A, Maisotsenko V.
[5] Pourmoosavi MA, Amraee T, Firuzabad MF. Expansion planning of generation
Projecting global water footprints diminution of a dew-point cooling system:
technologies in electric energy systems under water use constraints with
sustainability approach assisted with energetic and economic assessment. Renew
Sustain Energy Rev 2021;140. https://doi.org/10.1016/j.rser.2021.110741.

21
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

renewable resources. Sustain Energy Technol Assessments 2021;43:100828. [29] Memon SK, Nisar K, Hijazi MHA, Chowdhry BS, Sodhro AH, Pirbhulal S, et al. A
https://doi.org/10.1016/j.seta.2020.100828. survey on 802.11 MAC industrial standards, architecture, security & supporting
[6] Ogbolumani OA, Nwulu NI. Multi-objective optimisation of constrained food- emergency traffic: Future directions. J Ind Inf Integr 2021;24:100225. 10.1016/J.
energy-water-nexus systems for sustainable resource allocation. Sustain Energy JII.2021.100225.
Technol Assessments 2021;44:100967. https://doi.org/10.1016/j. [30] Sreeraj ES, Chatterjee K, Bandyopadhyay S. Design of isolated renewable hybrid
seta.2020.100967. power systems. Sol Energy 2010;84:1124–36. https://doi.org/10.1016/j.
[7] Sharon H. Energy, exergy and enviro-economic assessment of productivity solener.2010.03.017.
enhanced passive double sided vertical convection solar distiller for fresh water [31] Kiwan S, Al-Nimr M, Salim I. A hybrid solar chimney/photovoltaic thermal
production. Sustain Energy Technol Assessments 2020;42:100846. https://doi. system for direct electric power production and water distillation. Sustain Energy
org/10.1016/j.seta.2020.100846. Technol Assessments 2020;38:100680. https://doi.org/10.1016/j.
[8] Essa FA, Elsheikh AH, Sathyamurthy R, Muthu Manokar A, Kandeal AW, seta.2020.100680.
Shanmugan S, et al. Extracting water content from the ambient air in a double- [32] Dursun E, Acarkan B, Kilic O. Modeling of hydrogen production with a stand-
slope half-cylindrical basin solar still using silica gel under Egyptian conditions. alone renewable hybrid power system. Int J Hydrogen Energy 2011;37:
Sustain Energy Technol Assessments 2020;39:100712. https://doi.org/10.1016/ 3098–107. https://doi.org/10.1016/j.ijhydene.2011.11.029.
j.seta.2020.100712. [33] Rohani G, Nour M. Techno-economical analysis of stand-alone hybrid renewable
[9] Zehtabiyan-Rezaie N, Alvandifar N, Saffaraval F, Makkiabadi M, Rahmati N, power system for Ras Musherib in United Arab Emirates. Energy 2014;64:828–41.
Saffar-Avval M. A solar-powered solution for water shortage problem in arid and https://doi.org/10.1016/j.energy.2013.10.065.
semi-arid regions in coastal countries. Sustain Energy Technol Assessments 2019; [34] Forough AB, Roshandel R. Multi objective receding horizon optimization for
35:1–11. https://doi.org/10.1016/j.seta.2019.05.015. optimal scheduling of hybrid renewable energy system. Energy Build 2017.
[10] Vilanova MRN, Balestieri JAP. Energy and hydraulic efficiency in conventional https://doi.org/10.1016/j.enbuild.2017.06.031.
water supply systems. Renew Sustain Energy Rev 2014;30:701–14. [35] Harajli H, Kabakian V, El-Baba J, Diab A, Nassab C. Commercial-scale hybrid
[11] Tariq R, Sheikh NANA, Bassam A, Xamán J. Analysis of Maisotsenko humid air solar photovoltaic – diesel systems in select Arab countries with weak grids: An
bottoming cycle employing mixed flow air saturator. Heat Mass Transf 2018;55. integrated appraisal. Energy Policy 2020;137:111190. https://doi.org/10.1016/j.
https://doi.org/10.1007/s00231-018-2531-z. enpol.2019.111190.
[12] Tariq R, Sheikh NA. Numerical heat transfer analysis of Maisotsenko Humid Air [36] Eltamaly AM, Addoweesh KE, Bawah U, Mohamed MA. New software for hybrid
Bottoming Cycle – a study towards the optimization of the air-water mixture at renewable energy assessment for ten locations in Saudi Arabia. J Renew Sustain
bottoming turbine inlet. Appl Therm Eng 2018;133:49–60. https://doi.org/ Energy 2013;5:033126. https://doi.org/10.1063/1.4809791.
10.1016/j.applthermaleng.2018.01.024. [37] Eltamaly AM, Addoweesh KE, Bawa U, Mohamed MA. Economic modeling of
[13] Temiz M, Dincer I. Enhancement of solar energy use by an integrated system for hybrid renewable energy system: a case study in Saudi Arabia. Arab J Sci Eng
five useful outputs: System assessment. Sustain Energy Technol Assessments 2014;39:3827–39. https://doi.org/10.1007/s13369-014-0945-6.
2021;43:100952. https://doi.org/10.1016/j.seta.2020.100952. [38] Eltamaly AM, Mohamed MA. A novel design and optimization software for
[14] Noorollahi Y, Khatibi A, Eslami S. Replacing natural gas with solar and wind autonomous PV/wind/battery hybrid power systems. Math Probl Eng 2014;2014.
energy to supply the thermal demand of buildings in Iran: A simulation approach. https://doi.org/10.1155/2014/637174.
Sustain Energy Technol Assessments 2021;44:101047. https://doi.org/10.1016/j. [39] Suresh V, Muralidhar M, Kiranmayi R. Modelling and optimization of an off-grid
seta.2021.101047. hybrid renewable energy system for electrification in a rural areas. Energy Rep
[15] Agyekum EB, Velkin VI, Hossain I. Sustainable energy: is it nuclear or solar for 2020. https://doi.org/10.1016/j.egyr.2020.01.013.
African Countries? Case study on Ghana. Sustain Energy Technol Assessments [40] Mokhtara C, Negrou B, Bouferrouk A, Yao Y, Settou N, Ramadan M. Integrated
2020;37:100630. https://doi.org/10.1016/j.seta.2020.100630. supply–demand energy management for optimal design of off-grid hybrid
[16] Tariq R, Zhan C, Zhao X, Sheikh NA. Numerical study of a regenerative counter renewable energy systems for residential electrification in arid climates. Energy
flow evaporative cooler using alumina nanoparticles in wet channel. Energy Build Convers Manag 2020;221:113192. https://doi.org/10.1016/j.
2018;169:430–43. https://doi.org/10.1016/j.enbuild.2018.03.086. enconman.2020.113192.
[17] Tariq R, Sheikh NA, Xamán J, Bassam A. Recovering waste energy in an indirect [41] Patel SK, Singh D, Devnani GL, Sinha S, Singh D. Potable water production via
evaporative cooler – a case for combined space air conditioning for human desalination technique using solar still integrated with partial cooling coil
occupants and produce commodities. Build Environ 2019;152:105–21. https:// condenser. Sustain Energy Technol Assessments 2021;43:100927. https://doi.
doi.org/10.1016/j.buildenv.2019.01.038. org/10.1016/j.seta.2020.100927.
[18] Tariq R, Zhan C, Ahmed Sheikh N, Zhao X. Thermal performance enhancement of [42] Fetanat A, Tayebi M, Mofid H. Water-energy-food security nexus based selection
a cross-flow-type maisotsenko heat and mass exchanger using various nanofluids. of energy recovery from wastewater treatment technologies: An extended
Energies 2018;11:2656. https://doi.org/10.3390/en11102656. decision making framework under intuitionistic fuzzy environment. Sustain
[19] Carvajal-Romo G, Valderrama-Mendoza M, Rodríguez-Urrego D, Rodríguez- Energy Technol Assessments 2021;43:100937. https://doi.org/10.1016/j.
Urrego L. Assessment of solar and wind energy potential in La Guajira, Colombia: seta.2020.100937.
Current status, and future prospects. Sustain Energy Technol Assessments 2019; [43] Ramesh M, Saini RP. Dispatch strategies based performance analysis of a hybrid
36:100531. https://doi.org/10.1016/j.seta.2019.100531. renewable energy system for a remote rural area in India. J Clean Prod 2020;259:
[20] Raghuwanshi SS, Arya R. Reliability evaluation of stand-alone hybrid 120697. https://doi.org/10.1016/j.jclepro.2020.120697.
photovoltaic energy system for rural healthcare centre. Sustain Energy Technol [44] Soberanis MAE, Mithrush T, Bassam A. A sensitivity analysis to determine
Assessments 2020;37:100624. https://doi.org/10.1016/j.seta.2019.100624. technical and economic feasibility of energy storage systems implementation : A
[21] Ben Ali I, Turki M, Belhadj J, Roboam X. Systemic design and energy fl ow battery case study. Renew. Energy 2018;115. https://doi.org/10.1016/j.
management of a standalone battery-less PV/Wind driven brackish water reverse renene.2017.08.082.
osmosis desalination system. Sustain Energy Technol Assessments 2020;42: [45] Sarker S. Feasibility analysis of a renewable hybrid energy system with producer
100884. https://doi.org/10.1016/j.seta.2020.100884. gas generator ful fi lling remote household electricity demand in Southern
[22] Jahangir MH, Cheraghi R. Economic and environmental assessment of solar- Norway. Renew Energy 2016;87:772–81. https://doi.org/10.1016/j.
wind-biomass hybrid renewable energy system supplying rural settlement load. renene.2015.11.013.
Sustain Energy Technol Assessments 2020;42:100895. https://doi.org/10.1016/j. [46] Shezan SKA, Julai S, Kibria MA, Ullah KR, Saidur R, Chong WT, et al.
seta.2020.100895. Performance Analysis of an off-grid Wind-PV-Diesel-Battery Hybrid Energy
[23] Mohamed MA, Jin T, Su W. An effective stochastic framework for smart System Feasible for Remote Areas. J Clean Prod 2016. https://doi.org/10.1016/j.
coordinated operation of wind park and energy storage unit. Appl Energy 2020; jclepro.2016.03.014.
272:115228. https://doi.org/10.1016/j.apenergy.2020.115228. [47] Upadhyay S, Sharma MP. Selection of a suitable energy management strategy for
[24] Gong X, Dong F, Mohamed MA, Awwad EM, Abdullah HM, Ali ZM. Towards a hybrid energy system in a remote rural area of India. Energy 2016;94:352–66.
distributed based energy transaction in a clean smart island. J Clean Prod 2020; https://doi.org/10.1016/j.energy.2015.10.134.
273:122768. https://doi.org/10.1016/j.jclepro.2020.122768. [48] Meixco Agenda 2030. Gov Mex Mex Agenda 2030 2021. https://www.gob.mx/
[25] Mohamed MA, Jin T, Su W. Multi-agent energy management of smart islands agenda2030 (accessed January 30, 2021).
using primal-dual method of multipliers. Energy 2020;208:118306. https://doi. [49] Yucatan Agenda 2030. Gov Yucatan, Agenda 2030 2021. http://www.seplan.
org/10.1016/j.energy.2020.118306. yucatan.gob.mx/agenda2030/ (accessed January 30, 2021).
[26] Ahmad I, Shahabuddin S, Malik H, Harjula E, Leppanen T, Loven L, et al. Machine [50] Gude VG. Energy and water autarky of wastewater treatment and power
Learning Meets Communication Networks: Current Trends and Future Challenges. generation systems. Renew Sustain Energy Rev 2015;45:52–68. https://doi.org/
IEEE Access 2020;8:223418–60. https://doi.org/10.1109/ 10.1016/j.rser.2015.01.055.
ACCESS.2020.3041765. [51] Raheem A, Sikarwar VS, He J, Dastyar W, Dionysiou DD, Wang W, et al.
[27] Sodhro AH, Sangaiah AK, Sodhro GH, Sekhari A, Ouzrout Y, Pirbhulal S. Energy- Opportunities and challenges in sustainable treatment and resource reuse of
Efficiency of Tools and Applications on Internet. In: Sangaiah AK, Zhang Z, Sheng sewage sludge: a review. Chem Eng J 2018;337:616–41.
M, editors. Comput. Intell. Multimed. Big Data Cloud with Eng. Appl., Academic [52] INEGI C de P. vivienda, Instituto Nacional de Geografía y Estadística 2014.
Press; 2018, p. 297–318. 10.1016/B978-0-12-813314-9.00014-1. [53] Gobierno del Estado de Yucatán. San José Tip – Ceh 2020.
[28] Sodhro AH, Obaidat MS, Pirbhulal S, Sodhro GH, Zahid N, Rawat A. A novel [54] JAPAY. Consumo de agua potable. (In Spanish) Los Desafíos Del Agua Http//
energy optimization approach for artificial intelligence-enabled massive internet JapayYucatanGobMx/Noticias/VerarticuloPhp?IdArticulo=358 2019.
of things. Proc. 2019 Int. Symp. Perform. Eval. Comput. Telecommun. Syst. [55] Mara DD, Guimarães ASP. Simplified sewerage: Potential applicability in
SPECTS 2019 - Part SummerSim 2019 Multiconference, Institute of Electrical and industrialized countries. Urban Water 1999;1:257–9. https://doi.org/10.1016/
Electronics Engineers Inc.; 2019. 10.23919/SPECTS.2019.8823317. s1462-0758(99)00015-1.

22
R. Tariq et al. Sustainable Energy Technologies and Assessments 48 (2021) 101561

[56] Tchobanoglus G, Burton F, Stensel HD. Wastewater engineering: Treatment and [85] Hawkins TR, Singh B, Majeau-Bettez G, Strømman AH. Comparative
reuse. Am Water Work Assoc J 2003;95:201. environmental life cycle assessment of conventional and electric vehicles. J Ind
[57] Velitchko GT, Víctor HA-Y. Modelación de la variación del consumo de agua Ecol 2013;17:53–64. https://doi.org/10.1111/j.1530-9290.2012.00532.x.
potable con métodos estocásticos. Tecnol Cienc Agua 2016;7. [86] Hu A, Huang L, Lou S, Kuo C-H, Huang C-Y, Chian K-J, et al. Assessment of the
[58] HYUNDAI. [In Spanish] Generador Hyundai Portatil 5.0-5.5 KW C/MOTOR 13 carbon footprint, social benefit of carbon reduction, and energy payback time of a
HP - HHY5500. Hyundai Power Prod 2021. https://www.hyundaipower.com. high-concentration photovoltaic system. Sustainability 2016;9:27. https://doi.
mx/producto/generador-hyundai-portatil-5-0-5-5-kw-c-motor-13-hp-hhy5500/ org/10.3390/su9010027.
(accessed May 9, 2021). [87] Kawamoto R, Mochizuki H, Moriguchi Y, Nakano T, Motohashi M, Sakai Y, et al.
[59] Tariq R, Xamán J, Bassam A, Ricalde LJ, Soberanis MAE. Multidimensional Estimation of CO2 emissions of internal combustion engine vehicle and battery
assessment of a photovoltaic air collector integrated phase changing material electric vehicle using LCA. Sustainability 2019;11:2690. https://doi.org/
considering Mexican Climatic conditions. Energy 2020;209:118304. https://doi. 10.3390/su11092690.
org/10.1016/j.energy.2020.118304. [88] Tariq R, Sohani A, Xamán J, Sayyaadi H, Bassam A, Tzuc OM. Multi-objective
[60] Messenger RA, Abtahi A. Photovoltaic systems engineering. CRC Press; 2017. optimization for the best possible thermal, electrical and overall energy
[61] Sharp Solar. Módulo Fotovoltaico de Silicio Monocristalino NU-JB395L (144 performance of a novel perforated-type regenerative evaporative humidifier.
células) 2020. Energy Convers Manag 2019;198:111802. https://doi.org/10.1016/j.
[62] Kosmadakis IE, Elmasides C, Eleftheriou D, Tsagarakis KP. A techno-economic enconman.2019.111802.
analysis of a PV-battery system in Greece. Energies 2019. https://doi.org/ [89] Golizadeh Akhlaghi Y, Badiei A, Zhao X, Aslansefat K, Xiao X, Shittu S, et al.
10.3390/en12071357. A constraint multi-objective evolutionary optimization of a state-of-the-art dew
[63] enair. E200L Ficha Técnica. 2021. point cooler using digital twins. Energy Convers Manag 2020. https://doi.org/
[64] Mongird K, Viswanathan V, Balducci P, Alam J, Fotedar V, Koritarov V, et al. 10.1016/j.enconman.2020.112772.
Energy Storage Technology and Cost Characterization Report. 2019. [90] MathWorks. The Language of Technical Computing 2017. https://www.
[65] NREL. Distributed Generation Renewable Energy Estimate of Costs 2016. mathworks.com/products/matlab.html.
[66] Mohamed MA, Eltamaly AM. Sizing and Techno-Economic Analysis of Stand- [91] Sohani A, Sayyaadi H, Hoseinpoori S. Modeling and multi-objective optimization
Alone Hybrid Photovoltaic/Wind/Diesel/Battery Energy Systems. Model. Simul. of an M-cycle cross-flow indirect evaporative cooler using the GMDH type neural
Smart Grid Integr. with Hybrid Renew. Energy Syst., Springer; 2018, p. 23–38. network Modélisation et optimisation à objectifs multiples d ’ un refroidisseur
[67] Homer Energy. Homer Pro. Man Homer Energy 2019:1–241. évaporatif indirect à écoulements croisés à cycle M e. Int J Refrig 2016;69:
[68] Bruck M. A Levelized Cost of Energy (LCOE) model for wind farms that includes 186–204. https://doi.org/10.1016/j.ijrefrig.2016.05.011.
Power Purchase Agreement (PPA) energy delivery limits availability-based real [92] Tariq R, Hussain Y, Sheikh NA, Afaq K, Ali HM. Regression-Based Empirical
options approach for determining cost and pricing of performance-based logistics Modeling of Thermal Conductivity of CuO-Water Nanofluid using Data-Driven
contracts view project. Renew Energy 2016;122:131–9. Techniques. Int J Thermophys 2020;41. https://doi.org/10.1007/s10765-020-
[69] Elevation Angle, PVEducation 2021. https://www.pveducation.org/pvcdrom/ 2619-9.
properties-of-sunlight/elevation-angle (accessed May 8, 2021). [93] Sohani A, Sayyaadi H. Design and retrofit optimization of the cellulose
[70] Park CY, Hong SH, Lim SC, Song BS, Park SW, Huh JH, et al. Inverter efficiency evaporative cooling pad systems at diverse climatic conditions. Appl Therm Eng
analysis model based on solar power estimation using solar radiation. Processes 2017;123:1396–418. https://doi.org/10.1016/j.applthermaleng.2017.05.120.
2020;8:1–19. https://doi.org/10.3390/pr8101225. [94] Jamil MA, Goraya TS, Shahzad MW, Zubair SM. Exergoeconomic optimization of
[71] Rehman S, Al-Abbadi NM. Wind shear coefficients and their effect on energy a shell-and-tube heat exchanger. Energy Convers Manag 2020;226:113462.
production. Energy Convers Manag 2005;46:2578–91. https://doi.org/10.1016/j. https://doi.org/10.1016/j.enconman.2020.113462.
enconman.2004.12.005. [95] STATGRAPHICS Centurion 2021. https://www.statgraphics.com/ (accessed May
[72] May Tzuc O, Bassam A, Ricalde LJ, Jaramillo OA, Flota-Bañuelos M, Escalante 11, 2021).
Soberanis MA. Environmental-economic optimization for implementation of [96] Kim JH, Simon TW. Journal of heat transfer policy on reporting uncertainties in
parabolic collectors in the industrial process heat generation: Case study of experimental measurements and results. J Heat Transfer 1993;115:5–6. https://
Mexico. J Clean Prod 2020;242:118538. https://doi.org/10.1016/j. doi.org/10.1115/1.2910670.
jclepro.2019.118538. [97] Hussaini IS, Zubair SM, Antar MA. Area allocation in multi-zone feedwater
[73] [In Spanish] Índice Nacional de Precios al Consumidor (INPC), Instituto Nacional heaters. Energy Convers Manag 2007;48:568–75. https://doi.org/10.1016/j.
de Estadística y Geografía (INEGI) 2021. https://www.inegi.org.mx/temas/inpc/ enconman.2006.06.003.
(accessed May 9, 2021). [98] Qureshi BA, Zubair SM. A comprehensive design and rating study of evaporative
[74] NASA. NASA POWER Data Access Viewer 2021. https://power.larc.nasa.gov/ coolers and condensers. Part II. Sensitivity analysis. Int J Refrig 2006;29:659–68.
data-access-viewer/ (accessed May 8, 2021). https://doi.org/10.1016/j.ijrefrig.2005.09.015.
[75] CFE (Comision Federal de Electricidad). Consulta tu tarifa. 2017 2018. https:// [99] Jamil MA, Ud Din Z, Goraya TS, Yaqoob H, Zubair SM. Thermal-hydraulic
app.cfe.mx/Aplicaciones/CCFE/Tarifas/TarifasCRECasa/Casa.aspx. characteristics of gasketed plate heat exchangers as a preheater for thermal
[76] Halabi LM, Mekhilef S, Olatomiwa L, Hazelton J. Performance analysis of hybrid desalination systems. Energy Convers Manag 2020;205:112425. https://doi.org/
PV/diesel/battery system using HOMER: A case study Sabah. Malaysia. Energy 10.1016/j.enconman.2019.112425.
Convers Manag 2017;144:322–39. https://doi.org/10.1016/J. [100] James CA, Taylor RP, Hodge BK. The application of uncertainty analysis to cross-
ENCONMAN.2017.04.070. flow heat exchanger performance predictions. Heat Transf Eng 1995;16:50–62.
[77] Alshammari N, Asumadu J. Optimum unit sizing of hybrid renewable energy https://doi.org/10.1080/01457639508939863.
system utilizing harmony search, Jaya and particle swarm optimization [101] Software F-Chart. EES: Engineering Equation Solver: Engineering Software. F-
algorithms. Sustain Cities Soc 2020;60:102255. https://doi.org/10.1016/J. Chart Softw 2015;2012:6–8. http://www.fchart.com/ees/ (accessed September
SCS.2020.102255. 28, 2018).
[78] Homer Energy. HOMER – Hybrid Renewable and Distributed Generation System [102] CFE. Tarifas-GranDemandaMTO 2021.
Design Software. Http://WwwHomerenergyCom 2016. [103] Shahzad M, Ma T, Jurasz J, Canales FA, Lin S, Ahmed S, et al. Economic analysis
[79] Tariq R, Jimenez JT, Sheikh NA, Khan S. Mathematical approach to improve the and optimization of a renewable energy based power supply system with different
thermoeconomics of a humidification dehumidification solar desalination system. energy storages for a remote island. Renew Energy 2020;164:1376–94. https://
Mathematics 2021;9:1–31. https://doi.org/10.3390/math9010033. doi.org/10.1016/j.renene.2020.10.063.
[80] Caliskan H. Novel approaches to exergy and economy based enhanced [104] Ramli MAM, Bouchekara HREH, Alghamdi AS. Optimal Sizing of PV/wind/diesel
environmental analyses for energy systems. Energy Convers Manag 2015;89: hybrid microgrid system using multi-objective self-adaptive differential evolution
156–61. https://doi.org/10.1016/j.enconman.2014.09.067. algorithm. Renew Energy 2018. https://doi.org/10.1016/j.renene.2018.01.058.
[81] Sovacool BK. Valuing the greenhouse gas emissions from nuclear power: a critical [105] Usman M, Khan MT, Rana AS, Ali S. Techno-economic analysis of hybrid solar-
survey. Energy Policy 2008;36:2950–63. https://doi.org/10.1016/j. diesel-grid connected power generation system. J Electr Syst Inf Technol 2018;5:
enpol.2008.04.017. 653–62. https://doi.org/10.1016/j.jesit.2017.06.002.
[82] Dai Q, Kelly JC, Gaines L, Wang M. Life cycle analysis of lithium-ion batteries for [106] Rajanna S, Saini RP. Development of optimal integrated renewable energy model
automotive applications. Batteries 2019;5:48. https://doi.org/10.3390/ with battery storage for a remote Indian area. Energy 2016;111:803–17. https://
batteries5020048. doi.org/10.1016/j.energy.2016.06.005.
[83] Gaines L, Sullivan J, Burnham A, Belharouak I. Life-cycle analysis of production [107] El Mekaoui A, Tariq R, Ramírez OB, Méndez-Monroy PE. Sustainability,
and recycling of lithium ion batteries. Transp Res Rec J Transp Res Board 2011; sociocultural challenges, and new power of capitalism for renewable energy
2252:57–65. https://doi.org/10.3141/2252-08. megaprojects in an indigenous Mayan Community of Mexico. Sustain 2020;12:
[84] Peters JF, Baumann M, Zimmermann B, Braun J, Weil M. The environmental 7432. https://doi.org/10.3390/SU12187432.
impact of Li-Ion batteries and the role of key parameters – a review. Renew [108] Objective Scatter Index. Indian Natl Cent Ocean Inf Serv 2021.
Sustain Energy Rev 2017;67:491–506. https://doi.org/10.1016/j. [109] Golbraikh A, Tropsha A. Beware of q2! J Mol Graph Model 2002;20:269–76.
rser.2016.08.039. https://doi.org/10.1016/S1093-3263(01)00123-1.

23