Renewable Energy 123 (2018) 204e217

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Optimization under uncertainty of a biomass-integrated renewable

energy microgrid with energy storage
Yingying Zheng a, *, Bryan M. Jenkins a, Kurt Kornbluth a, Chresten Træholt b
a
Department of Biological and Agricultural Engineering, University of California, Davis, One Shields Avenue, Davis, CA 95616, USA
b
Department of Electrical Engineering, Technical University of Denmark, Elektrovej, 2800 Kgs. Lyngby, Denmark

a r t i c l e i n f o a b s t r a c t

Article history: Deterministic constrained optimization and stochastic optimization approaches were used to evaluate
Available online 3 February 2018 uncertainties in biomass-integrated microgrids supplying both electricity and heat. An economic linear
programming model with a sliding time window was developed to assess design and scheduling of
Keywords: biomass combined heat and power (BCHP) based microgrid systems. Other available technologies
Microgrids considered within the microgrid were small-scale wind turbines, photovoltaic modules (PV), producer
Renewables integration
gas storage, battery storage, thermal energy storage and heat-only boilers. As an illustrative example, a
Combined heat and power
case study was examined for a conceptual utility grid-connected microgrid application in Davis, Cali-
Biomass
Modeling
fornia. The results show that for the assumptions used, a BCHP/PV with battery storage combination is
Energy storage the most cost effective design based on the assumed energy load profile, local climate data, utility tariff
Uncertainty structure, and technical and financial performance of the various components of the microgrid. Monte
Stochastic analysis Carlo simulation was used to evaluate uncertainties in weather and economic assumptions, generating a
probability density function for the cost of energy.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction The most common dispatchable units are diesel, natural gas, or
biomass powered engine-generators. Moreover, an energy storage
Microgrids (MG) are smaller distribution networks usually system is adopted in most cases to neutralize mismatch between
installed close to the end users, and frequently contain hybrid en- generation and demand and tackle the uncertainty of demand
ergy resources, storage devices, and controllable loads. The tradi- forecasts. Energy storage provides the necessary means to shift the
tional power grid is generally a large-scale centralized network microgrid supply to a higher market price period based on the time
where power plants generate high voltage electricity that is of use. As an alternative to energy storage, load shifting can be
transferred and distributed to lower voltage end users. A significant applied to match demand with renewable energy generation. Load
fraction of electrical energy is dissipated in delivery due to the long shedding may also be feasible, or other types of generation added to
distances between generator and load. Microgrids have been ensure demand is satisfied [2]. MG can also be operated with
developed around the world as a means to address the high connection to the central power grid, in which case the central grid
penetration level of renewable generation and reduce greenhouse is used as a backup to reduce or eliminate the need for local storage,
gas emissions while attempting to address supply-demand or while completely disconnected from the central grid or islanded
balancing at a more local level [1]. [3]. When connected, the customer sometimes has the option of
The electricity generation of microgrid via solar PV and wind selling surplus electricity back to the utility grid operator under a
turbines depends, of course, on the total solar radiation and the net metering, feed-in, or other power purchase agreement.
wind speed in general. Due to the stochastic nature of these In microgrid applications, both manufacturers and customers
renewable energy resources, load behaviors, and market prices, a are interested to know the optimal capacity of the associated
dispatchable generation unit is frequently included that can be components of the system and the dispatch strategy to use in order
turned on or off or modulated to adjust power output accordingly. to minimize cost and environmental impacts. Due to the compu-
tational complexity, a number of software packages have been
developed to assist in microgrid design and assessment including
* Corresponding author. HOMER [4e8] and DER-CAM [9-11]. Rohit et al. [12] proposed a
E-mail address: [email protected] (Y. Zheng).

https://doi.org/10.1016/j.renene.2018.01.120
0960-1481/© 2018 Elsevier Ltd. All rights reserved.
 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217 205

Nomenclature O&M operation & maintenance cost ($/kWh)

F feedstock or fuel cost ($/kWh)
C hourly capital cost ($/h)
Abbreviations M rated capacity (kW)
MG microgrid N expected life (years)
CHP combined heat and power s economic scaling factor
BCHP biomass-based CHP T length of planning horizon (h)
GS biomass gasifier V set of all system components that contribute to
ICE internal combustion engine capital cost
WT wind turbine U set of all system components that contribute to O&M
PV photovoltaic cost
ES energy storage x decision variable: hourly energy flow (kWh/h)
BT battery
PGS producer gas storage Superscripts and subscripts
TES thermal energy storage t time step (s)
HOB heat-only boiler p actual facility
MCS Monte Carlo simulation o reference facility
COE cost of energy i index of installed units that contribute to capital cost
PDF probability density function j index of installed units that contribute to O&M and
fuel cost
Symbols chr charging
E electricity demand (kWh) dis discharging
H heat demand (kWh) min minimum charging and discharging rate (kWh/h)
Z net acquisition cost ($) max maximum charging and discharging rate (kWh/h)
P purchase and installation (capital) cost ($)

hybrid off-grid system for a rural application with HOMER. Bras- However, where a fuel generation unit, such as a biomass gasifier, is
lavsky et al. [13] presented an economic model of a shopping deployed, producer gas production and electricity generation can
center, developed in DER-CAM, using on-site-specific demand, also be treated as two separate and independent processes. The
tariffs, and performance data for each technology option available. producer gas after biomass gasification can be used directly to fuel
Furthermore, substantial studies on microgrid optimal design an internal combustion engine, microturbine, or another prime
and operation are typically formulated as minimization or maxi- mover for power generation, and also used in a furnace or boiler for
mization problem constrained by energy demand, capacity limits, heat generation to offset utility natural gas demand [56]. Most
ramping rate, and startup or shutdown times [14e21], and most studies considering energy storage include thermal storage or
address electricity only although thermal loads may also exist. Both battery; separate gas storage is typically not considered, instead
thermal and electrical load profiles can fluctuate hourly and relying on pipeline supply as the storage equivalent and thereby
seasonally and utility tariff prices for natural gas and electricity may subject to utility pricing. Biomass integrated models using RHC to
change dynamically as well. In these cases, electricity-led as- schedule combined fuel gas generation and storage, engine
sumptions cannot guarantee an optimal solution overall. A number cogeneration, auxiliary boiler and thermal energy storage opera-
of modeling studies incorporating CHP units in the microgrid have tions have not previously been developed.
considered both electricity and thermal demand [22e29] but few To address the above-mentioned issues, a model was developed
address biomass integration including separation of the fuel pro- to optimize the design and scheduling of an integrated biomass
duction and power generation components. combined heat and power microgrid (BCHP-MG) system. The
Variables that are subject to uncertainty in microgrid design and model combines a deterministic optimization module with a sto-
operation include unscheduled maintenance, climatic conditions chastic module and Monte Carlo simulation. Developing a more
(e.g. wind and cloudiness), and energy market prices and demands general model capable of solving for the optimal configuration and
[30e40]. Model prediction control and receding horizon control dispatch of a renewable energy microgrid with the flexibility of
(RHC) are frequently used to predict and make decisions under biomass integration was the primary objective for this work. Spe-
uncertainty [41e47]. Jiang and Fei addressed the problem of cifically, the objectives include: 1) finding the optimal capacity of
adopting multiple CHPs for cost reduction in microgrid using hi- wind and PV generation in each proposed scenario, 2) developing
erarchical optimization [48]. Xie et al. [49] developed a look-ahead an optimal dispatch strategy between the various BCHP, wind tur-
optimal control algorithm for dispatching the generation units with bine (WT), PV, battery (BT), producer gas storage (PGS), thermal
the objective of minimizing both generation and environmental energy storage (TES), and heat-only boiler (HOB) components of
costs. Silvente et al. [50] used the RHC approach to analyze un- the microgrid and the main utility grid (electricity and natural gas)
certainty in both energy generation and demand. Monte Carlo based on hourly energy demands and tariff rates, 3) estimating the
simulation (MCS) has been widely used to evaluate the reliability of effects of BT capacity on the cost of energy (COE) for different
a microgrid by generating data from fixed probability distributions scenarios, 4) evaluating the influence of tariff rates and demand
of stochastic variables, such as wind speed, solar irradiance, profiles on the COE and unit dispatch strategies, and 5) investi-
customer demands, and others [51e55]. gating the impact of stochastic variables on the final COE proba-
Currently most of the CHP integrated microgrid sizing and bility distribution.
scheduling studies have assumed the CHP system as a single unit.
206 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217

2. Model development
PBCHP ðtÞ ¼ f ðloadÞ (1)
2.1. Microgrid system design In the case of BCHP, routine maintenance is required, but there is
also the risk of unscheduled outages due to mechanical and other
For quantifying the analysis, microgrid systems with the failures. In this study both the gasifier and engine are assumed to
following components are considered (Fig. 1): have a certain failure risk for unscheduled maintenance and would
The electricity demand is met by the sum of the BCHP, WT, PV not be available for generation. The gasifier and engine failures are
generation and the BT discharge, within their operating limits and assumed to be independent, so that the gasifier or the engine may
constraints. The electricity generated via the PV array and WT de- be available when the other fails. Both the gasifier and engine
pends on the solar radiation and the wind speed in general. The operation are treated as binary being either on (1) or off (0) with
power from the BCHP, WT and PV modules is allowed to charge the variable capacity when the units are on. The probability density
BT, depending on the operating strategy selected. Producer gas function of the Bernoulli distribution is used to represent the sto-
from the BCHP unit is assumed to be purified and cooled, and can be chastic nature of gasifier and engine.
used directly as fuel for the engine-generator sets and the boiler,


stored in the producer gas storage tank, or simply flared for disposal 1 for Pf  x < 1
fðOÞ ¼ (2)
if no economic demand exists and storage is at full capacity [57]. 0 for 0  x  Pf
The PGS can be charged when the energy demand is low and
discharge during high demand to improve system reliability. It can where f (O) is the operating mode of the gasifier or engine (inde-
also be deployed to increase export electricity under a power pendent); Pf is the failure chance of gasifier or engine.
purchase agreement to raise system revenue when the utility price
is high if on a time of use tariff. For internal combustion
reciprocating-type engines, heat for other uses can typically be 2.2.2. Wind turbine
recovered from the engine cooling jacket, exhaust, and potentially Wind power generation depends on wind speed and the inter-
the engine surfaces in a combined heat and power mode. The ference of the turbine with the wind. The output power of the
recovered heat can be employed for a number of purposes, turbine can be one of these three values [44]:
including direct heat utilization but also chilling, cooling, and 8
additional electricity generation in a combined cycle mode >
> 0 if VðtÞ < Vcutin or VðtÞ > Vcutout
>
>
although the latter is not included here. Similar to PGS, the recov- >
< 1
Pwt ðtÞ ¼ Cp rAwt VðtÞ3 if Vcutin < VðtÞ < Vrated
ered thermal energy can be used immediately or stored in a ther- 2
>
>
mal energy storage system, in this case a warm water tank is >
> 1
>
: Cp rA V3
assumed. The auxiliary HOB operates to make up any heat shortage, wt rated if Vrated < VðtÞ < Vcutout
2
with heat otherwise supplied from producer gas burned directly
(3)
from the gasifier, producer gas taken from storage (PGS), or utility
natural gas. where Pwt is the mechanical output power of the turbine (W), r is
For utility grid-connected scenarios, any electricity or heat air density (kg/m3), Awt is the turbine swept area (m2), V is the
supply deficits from the microgrid are satisfied by purchasing undisturbed wind speed (m/s), Cp is the performance coefficient (or
electricity or natural gas from the utility. In some circumstances power coefficient) of the turbine, and Vcutin, Vrated and Vcutout are
surplus electricity from the microgrid is available for delivery to the the cut in, rated, and cut out wind speed (m/s) of the turbine.
utility under a net metering, feed-in tariff, or other power purchase Probability density functions (PDF) were used to characterize
arrangement generating revenue for the microgrid operation. the stochastic behavior of wind speed. The wind speed over a
predefined time period was estimated using a Weibull PDF [55].
2.2. Microgrid component modeling     k
k V k1 V
fðVÞ ¼ exp  (4)
2.2.1. BCHP c c c
BCHP is assumed to operate as a load following power plant and
 1:086
alter its output to meet varying demands within the capacity limit. s
k¼ (5)
m

m
c¼   (6)
G 1 þ k1

where f(V) is the frequency rate of wind velocity; c is the Weibull

scale parameter, a measure of the characteristic wind speed of the
distribution; k is the Weibull shape parameter and specifies the
shape of a Weibull distribution, taking on a value of between 1 and
3; m is the mean wind speed (m/s) and s is the standard deviation of
the wind speed (m/s). The parameters k and c can be computed
from m and s. A small value for k signifies highly variable winds,
while constant winds are characterized by a larger k.

2.2.3. Photovoltaic module

The output power of the PV is given by the following equation
Fig. 1. Schematic microgrid system. [58]:
 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217 207

Table 1
Ppv ðtÞ ¼ hpv Apv S (7) Decision variables.

Decision variables Energy Flow from Energy Flow in to

where Ppv is the output power of the PV (W), hpv denotes the
BCHP x1 gasifier
conversion efficiency of the PV array (%) including the intrinsic x2 gasifier gas storage
module efficiency and array shading factor as appropriate, Apv is the x3 gas storage
array area (m2), and S is the solar radiation, treated as a random x4 gas storage engine
variable (W/m2). x5 engine
x6 engine electricity demand
Solar irradiation is a stochastic variable that depends on the x7 electricity grid electricity demand
weather conditions and possible changes in shading throughout x8 electricity demand
the day. Local shading or terrain effects that may also influence the x9 gasifier engine
resource availability are highly site-specific and not part of this x10 gasifier boiler
x11 storage boiler
analysis. The probabilistic nature of solar irradiance is considered to
x12 boiler
follow a beta PDF [59]. x13 natural gas grid boiler
x14 boiler
Gða þ bÞ ða1Þ
ð1  SÞb1 ;
x15 engine heat recovery thermal demand
fðSÞ ¼ S a  0; b  0 (8)
GðaÞ þ GðbÞ x16 thermal production
x17 thermal production thermal storage
  x18 thermal production heating demand
mð1 þ mÞ x19 thermal storage heating demand
b ¼ ð1  mÞ  1 (9)
2 s x20 heating demand
x21 gasifier flare
x22 engine electricity grid
mb x23 engine battery
a¼ (10)
1m WT x24 wind turbine electricity demand
x25 wind turbine electricity grid
x26 wind turbine battery
where a and b are the function parameters and S is the horizontal PV x27 solar panel electricity demand
solar irradiance (kW/m2); a and b are calculated from the mean and x28 solar panel electricity grid
standard deviation of solar irradiance m and s. Similar to the wind x29 solar panel battery
speed, an hourly average solar irradiance is used. BT x30 battery
x31 battery electricity demand

2.2.4. Energy storage

As an energy storage device, the battery storage injects power to
planning horizon (h) and t is the time step.
the microgrid when the local generation is insufficient and absorbs
The net energy supply cost consists of the levelized fixed or
power from the microgrid when the local generation is abundant or
capital costs of the system, the feedstock and fuel supply costs, and
a model decision criterion indicates that saving the electrical en-
all operating and maintenance (O&M) costs, all resolved to a uni-
ergy for the future hours would increase net economic benefit.
form cost of energy considering the time value of money. Capital
Producer gas storage can be charged when the producer gas pro-
and O&M costs in general are subject to economies of scale and
duction is abundant or a model decision criterion decides that
hence influenced by the size of the units included in the system
storing the producer gas for future generation either as electricity
[60]. Feedstock and O&M costs assumed not to be subject to
or heat would improve the value of the objective function. Thermal
economies of scale although in practice pricing may depend on
energy storage was modeled as sensible heat storage, using water
supply quantities.
as the storage medium, and considering only energy flows through
the warm water storage tank. 0 1
X
T X X 
Z¼ @ Cti þ t t tA
O&Mj þ Fj xj (12)
2.3. Mathematical formulation t¼1 i2V j2U

2.3.1. Decision variables

Pi irð1 þ irÞN
Decision variables express the microgrid operating modes and Ci ¼ (13)
the energy flows (kWh/h) between system components. The x 8760 ð1 þ irÞN  1
variables define the energy flows throughout the microgrid system,
each labeled with a subscript denoting the specific energy transfer  s
Pp Mp
(Fig. 1). The variables mentioned, along with other energy transfers, ¼ (14)
Po Mo
are now defined according to this notation. Each hour of the anal-
ysis involves 31 decision variables (Table 1). where V is the set of all installed system components that
contribute to capital cost; i is the index for all the installed units
2.3.2. Objective function that contribute to capital cost; U is the set of all energy flows that
The optimization in this case is developed from the objective to create O&M and fuel cost; j is the index for the energy flows that
minimize the cost of energy of over a particular time horizon (T) create O&M and fuel cost; C is the hourly levelized capital cost ($);
using an hourly time base. The objective function is formulated as: O&M and F are the hourly O&M and fuel cost of energy flow j at
time t ($/kWh); x is the energy flow (kWh); P is the overnight
Z purchase and installation cost that is influenced by an economy of
Min COE ¼ PT   (11)
t¼1 Etload þ Htload scale defined by the value of s (0  s  1) [61]; ir is the interest rate,
and N is expected life time (y). The constant 8760 is the conversion
where Z is the net energy supply cost ($); Eload and Hload are the for the number of hours per year and is uncorrected for leap years.
electricity and heating demand (kWh); T is the length of the For the equation defining the economy of scale, Pp is the capital cost
208 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217

of facility or unit under consideration within the microgrid; Po is

the known capital cost of a reference facility or unit of the same xt1  xt1
1  G ramprate (25)
type, Mp is the rated capacity of the unit under consideration; Mo is
the rated capacity of the reference unit.
xt6  xt1
6  E ramprate (26)

2.3.3. Constraints where x1 is the biomass gasifier production at time t; x6 is the ICE
All energy flows (x1-31) are signed with lower (zero) and upper production at time t; G_ ramprate and E_ ramprate are the ramping
bounds with the latter being the maximum acceptable capacities. rates of the gasifier and the engine-generator, which is related to
The electricity balance constraints the electrical demand to be the capacity and type.
satisfied by BCHP, WT, PV, BT or the grid power. The heat balance Some decision variables are coupling with each other and con-
constraints the heating demand to be satisfied by the producer gas strained by energy balance, for example:
powered boiler, the natural gas powered boiler, the heat recovered
from engine generator set, or some combination of these sources. xt1 ¼ xt2 þ xt9 þ xt10 þ xt21 (27)
Therefore, the energy balances at time t for the microgrid can be
written as follows: where x2, x9, x10, and x21 represents the energy flow out of biomass
gasifier to either gas storage, engine, boiler, or flare, respectively
xt6 þ xt7 þ xt24 þ xt27 þ xt31 ¼ Etload (15) (Table 5). Similar energy balance constraints are shown in Fig. 1.
This mathematical formulation of the system design and unit
commitment problem is a linear convex optimization problem. The
xt20 ¼ Htload (16) model implementation was here solved using MATLAB with its
The BT and PGS storage levels at the current time step t depend optimization toolbox (MATLAB 2016a, Mathworks, Natick,
on the storage level at previous time step (t-1) and the current Massachusetts).1
charging or discharging rate. The BT and PGS energy balances are:
2.4. Solution method
t t1
BT ¼ BT þ xt30  xt31 (17)
The model discussed here chains a deterministic planning
optimization module with a stochastic module (Fig. 2).
PGSt ¼ PGSt1 þ xt2  xt3 (18)
2.4.1. Performance surface method
where BTt and PGSt are the energy storage level at current time step The model is first solved deterministically to derive the optimal
t, BTt-1 and PGSt-1 the amounts of energy stored in BT and PGS at wind and solar capacities for each proposed microgrid scenario.
previous time step t-1. The capacity of WT and PV units is gradually increased from none to
Storage level constraints require that storage levels should be in 250 kW (with 10 kW of increment) in a search for the optimal ca-
the range between the minimum and maximum determined safety pacity yielding minimum cost with the rest of units fixed. The
and economy. The constraints of charging and discharging indicate criterion of selecting the best hybrid energy system combination
the changing rate for BT and PGS should be within the upper and for a proposed site is based on minimizing the cost for different
lower limits. The maximum charge and discharge rate is for the renewable combinations, the output of the optimal sizing and
model developed here assumed to be half of the rated capacity. The operation being the preferred set of WT and PV modules. The entire
charging and discharging efficiencies of BT and PGS are assumed procedure is repeated for all the possible combinations. The com-
small although this is not a general constraint of the model. The bination with the lowest cost overall is selected as optimal design
constraints for TES are as same as BT and PGS. for each scenario.

BTmin  BTt  BTmax (19) 2.4.2. Sliding time window

Once all the unit capacity has been fixed, the dispatch of the
BT chrmin  xt30  BT chrmax (20) available units to meet demand at the lowest cost is required. The
hourly operation strategy of the different hybrid configurations is
determined by using linear constrained optimization. The sizing
BT dismin  xt31  BT dismax (21)
and operating strategies are interdependent so a different set of
component configurations is analyzed in each hybrid combination
PGSmin  PGSt  PGSmax (22) to find the optimal hybrid system.
A sliding time window method is used to first determine the
PGS chrmin  xt2  PGS chrmax (23) optimal operation of all microgrid components [43,50]. For the
examples included here, a 4-h time window is used with known
electrical and heat demand. Each hour has a total of 31 solution
PGS dismin  xt3  PGS dismax (24) variables (Fig. 1), which for a 4-h horizon requires solving for 124
variables. Within each time window, linear optimization is applied
where BTmin and BTmax are the minimum and maximum allowed to obtain the gasifier, gas storage, engine, boiler, thermal storage,
BT energy storage level at any time, and the same goes for PGS; WT, PV and BT operation giving the minimum operating cost. Only
chrmax and dismax are the maximum allowed charging and dis- the solution for the first hour is retained to compute the actual
charging rates; chrmin and dismin are the minimum allowed generation and cost. At the same time, the new initial conditions of
charging and discharging rates.
The power ramping constraint, expressed in kWh per hour, in-
dicates how much a generator can change its output between two 1
Mention of a specific tradename does not constitute an endorsement by the
successive time steps. University of California.
 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217 209

categories: 1) customer information (load profile and weather

data), 2) technical information (physical characteristics and speci-
fications of all units, efficiency, heat to power ratio, power gener-
ation PDF, etc.), 3) financial information (capital cost, O&M cost,
fuel costs, tariff rate). In the deterministic model, the output of the
model is the optimal microgrid design and dispatch (based on the
COE values). In the stochastic model, the output of the model is the
probability distribution of COE.
For the case study located in Davis, California, a typical winter
daily residential electricity load profile from the local utility was
scaled up and used for analysis. A winter daily thermal demand
profile from the UC Davis campus was scaled down to represent
Fig. 2. Microgrid optimization modules. thermal energy usage (Fig. 4). The studied microgrid scale is around
100 kW. Therefore, the input load profiles are scaled up or down to
the desired range. For the scenario analyses, the peak and base load
all the energy storage devices including the BT, PGS, TES units are demand for electricity and heating were in the range of 72e200 kW
updated. The time window is then incremented by one hour, and and 50e172 kW, respectively. Davis hourly solar and wind data in
the process repeated for the entire time horizon (27 h for a 24 h February were obtained from the California Irrigation Management
period). The sliding time window approach is summarized in the Information System (CIMIS).
following steps (Fig. 3). Technical and economic parameters for the wind turbines and
PV assumed here are based on a 10 kW unit capacity. The efficiency
1 Specify initial conditions of energy storages. and cost of the power converters have been included in the overall
2 Optimize the system operation as outlined above for the period PV and wind turbines' efficiencies and costs. All parameters
from t_initial to t_initial þ T (T: sliding time window width). assumed for BCHP is based on a 100 kW unit capacity. All units are
3. Obtain the optimal operating points of all units. assumed to subject to 20 years life time and 6% of interest rate
4. Set the operating conditions of the first hour of the window to (Table 2).
the optimal conditions.
5 Update energy storage conditions.
6. Slide the window 1 h forward in time.
7 Repeat from step 2.

2.4.3. Monte Carlo simulation

Monte Carlo simulation was conducted to generate a finite
number of possible outcomes based on the probability distributions
of assumed stochastic parameters. A total of 1000 Monte Carol
simulations were used to generate the cost of energy distributions
for each scenario. The distribution of simulated outcomes for the
1000 realizations of the COE provided the risk profile.

3. Model application

3.1. Case study definition

The input data for the model were divided into the following Fig. 4. Model hourly electricity and heat demand for February in Davis, California.

1 hour 4-hour optimization based on

actual energy demand and price
1 hour 4-hour optimization based on
actual energy demand and price
1 hour 4-hour optimization based on
actual energy demand and price
1 hour 4-hour optimization based on
actual energy demand and price
1 hour 4-hour optimization based on
actual energy demand and price
hour 1 hour 2 hour 3 hour 4 hour 5
Fig. 3. Sliding window method (4-h window illustrated over a period of 5 h).
210 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217

Table 2 For TOU rates, the price of electricity changes by time of day
Technical parameters and cost assumptions for components of the microgrid (Table 4). For natural gas, the price is assumed to be constant
[62e65].
throughout the day. The electricity buyback price is assumed to be
Parameters Unit Value $0.04/kWh based on the net surplus compensation rate approved
All units Discount/interest rate % 6 by the California Public Utilities Commission (CPUC) [66].
Economic life years 20 Five design configuration scenarios were selected to investigate
Economic scale factor e 0.9 various aspects of the biomass integrated microgrid optimization
BCHP Rated power kW 100
(Table 5). All are utility grid interconnected, installed with a 200 kW
Capital cost $/kW 4500
O & M cost $/kWh 0.03 HOB and a 200 kWh TES. A net energy metering agreement is
Feedstock cost $/kWh 0.02 included with compensation for surplus electricity delivered from
Electricity efficiency e 0.3 the microgrid to the utility. BCHP, PV and WT are allowed for
Heat recovery factor e 0.6
connection to a utility meter. Scenario 1 includes only wind and PV
WT Reference module rated power kW 10
Reference module rotor diameter m 7
generation with battery storage. This option is a good alternative for
Capital cost $/kW 2154 locations with very limited heat demand but abundant wind and
O & M cost $/kWh 0.005 solar resources, or areas without abundant biomass resources.
Fuel cost $/kWh 0 Scenario 2 includes all the all three renewable sources but without
Cut-in wind speed m/s 2.5
producers gas storage. Scenario 3 also includes all three renewable
Cut-off wind speed m/s 50
Rated wind speed m/s 11 sources with a full complement of producer gas, thermal, and
Air density kg/m^3 1.23 battery storage but the biomass component is insufficient to meet
Betz Coefficient e 0.593 peak load by itself. Scenario 4 is the same as Scenario 3 but with
PV Reference module rated power kW 10 added BCHP capacity, in this case a duplicate unit for a total
Reference module surface area m^2 64
Capital cost $/kW 3463
biomass generation of 200 kW, slightly higher than peak load.
O & M cost $/kWh 0.005 Scenario 5 is the presumed conventional system and supplies en-
Fuel cost $/kWh 0 ergy demands entirely from the utility electricity and natural gas
Electricity efficiency e 0.2 grids.
BT Rated capacity kWh 200
Capital cost $/kWh 255
O & M cost $/kWh 0
Round trip efficiency e 0.9
3.2. Deterministic model results
PGS Rated capacity kWh 200
Capital cost $/kWh 80 3.2.1. Optimal microgrid design
O & M cost $/kWh 0.005 The lowest COE was found among all the possible combinations
Round trip efficiency e 1
of WT and PV modules (WT: 0e250 kW; PV: 0e250 kW) (Table 6).
HOB Rated power kW 150
Capital cost $/kW 120 Figs. 5e8 illustrate the 3D surface of cost response as a function
O & M cost ($/kWh) $/kWh 0.005 of the capacities of WT and PV for scenarios 1 to 4. The optimum
Efficiency e 0.85 WT and PV capacity can be found around the minimum points in
the figures. For scenario 1, when no BCHP is considered, the model
yields the lowest cost with 180 kW of wind capacity and 170 kW of
Both capital and operating costs are also subject to uncertainty. PV capacity. For scenarios 2, 3 and 4, when BCHP is included, no
An assumption was made here that all the capital and O&M costs wind capacity is adopted for the cost structure assumed. The
are uniformly distributed over the range from zero to twice the reference installed capital cost for PV was assumed to be $3165/kW
reference cost, the lower bound representing an extreme incentive with $0.005/kWh for O&M; for wind, the reference capital cost was
case with a high subsidy. The gasifier and engine are both assumed $2175/kW with $0.005/kWh for O&M. Although the wind is
to have 5% of failure risk for unscheduled maintenance and would assumed to have a lower capital cost, the wind speed profile for the
not be available for generation. The shape and scale factors for site selected (Davis, California) has only 9 h of the day with speeds
Weibull and Beta distribution are estimated by the curve fitting above 2.5 m/s, the cut-in wind speed. Hence, generation is low and
function in Matlab based on historical wind speed and solar irra- generation cost exceeds that of PV. From an economic viewpoint,
diance data. Table 3 lists the 13 uncertainty parameters and their PV and BCHP are the most attractive technology for this site under
associated PDFs. these cost assumptions. The optimal outcomes will in general differ
depending on location. Comparing scenarios 2 and 3, no change in

Table 3
Stochastic parameters and assumed PDF.

Stochastic parameter PDF PDF specifications

BCHP Capital cost ($/kW) Uniform [0,9000]

WT Capital cost ($/kW) Uniform [0,4308]
PV Capital cost ($/kW) Uniform [0,6926]
BT Capital cost ($/kW) Uniform [0,510]
GS O&M cost ($/kWh) Uniform [0,0.04]
ICE O&M cost ($/kWh) Uniform [0,0.02]
WT O&M cost ($/kWh) Uniform [0,0.01]
PV O&M cost ($/kWh) Uniform [0,0.01]
BT O&M cost ($/kWh) Uniform [0,0.002]
Wind speed (m/s) Weibull Shape factor k ¼ 1.6337; scale factor c ¼ 2.7813
Solar irradiance (w/m^2) Beta Shape factor a ¼ 0.0058; scale factor b ¼ 0.070
GS availability Bernoulli Pf ¼ 0.05
ICE availability Bernoulli Pf ¼ 0.05
 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217 211

Table 4
Electricity tariff rate

Energy Source Category Tariff rate ($/kWh) Time

Electricity Buy off-peak 0.22 00:00e09:59 a.m.; 09:00e11:59 p.m.

partial-peak 0.30 10:00e11:59 a.m.; 07:00e08:59 p.m.
peak 0.40 12:00e06:59 p.m.
Electricity Sell all day 0.04 N/A
Natural gas Buy all day 0.08 N/A

Table 5
System components of the 5 proposed scenarios (C ¼ unit included, x ¼ unit
excluded).

Scenario BCHP_1 BCHP_2 WT PV PGS BT

1 x x C C x C
2 C x C C x C
3 C x C C C C
4 C C C C C C
5 x x x x x x

installed PV capacity is associated with the addition of producer gas

storage. For scenario 4, with an oversized 200 kW BCHP, the
optimal PV capacity declines to 130 kW.
For the assumptions used in these examples, the total cost of the
optimal system configuration varies from a low of $0.1182/kWh for
Scenario 2, the 100 kW BCHP with 200 kWh battery storage sce-
nario, to a high of $0.2029/kWh for Scenario 5, the utility only Fig. 5. COE surface of scenario 1 as a function of WT and PV capacities.
supply scenario. Scenario 4, with a 200 kW BCHP capacity and
130 kW of PV achieves 100% renewable supply through the
microgrid with no utility purchase, but is higher in generation cost
than the hybrid microgrids of Scenarios 2 and 3 relying on both
microgrid and utility generation. Because of the boiler's installa-
tion, there is also installation cost in scenario 5 (Table 7).
Fig. 9 illustrates the economic results for the proposed 5 sce-
narios. Even with additional capital and installation costs, the
introduction of BCHP reduces the need for electricity and natural
gas purchases, and the overall total cost and COE are decreased
through this on-site generation. In addition, the producer gas
storage does not lead to obvious economic gains, due to the low
cost assumed for purchased natural gas.
To evaluate the effect of demand patterns from other times of
the year on the optimal microgrid configuration, the model was
also tested using daily electrical and thermal load data for summer.
The lowest cost, $0.1136/kWh, is also found for scenario 2
employing the 100 kW BCHP with solar PV and battery storage.
Scenario 1, with a 170 kW wind turbine capacity and 130 kW of PV
achieves a COE of $0.1349/kWh, which is lower than the COE of Fig. 6. COE surface of scenario 2 as a function of WT and PV capacities.
scenario 1 in winter. That is because the heating demand is much
lower in summer and the influence from the absence of a heat
source in scenario 1 is minimized. For the same reason, the COE of 3.2.2. Optimal microgrid dispatch
scenario 4, with a 200 kW gasifier, is $0.1390/kWh in summer, Figs. 10e13 illustrate the optimal energy flows from the BCHP,
which is higher compared to the COE in winter. WT, PV, and BT units as well as the grid to the demand during the
selected 24 h period. These graphs show the optimal dynamic
operation based on the cost minimization.

Table 6
Optimal system combinations.

Scenarios System Configuration Optimal WT Installed units Optimal PV Installed units

1 BCHP (0), PGS (0), BT (200) 180 170

2 BCHP (100), PGS (0), BT (200) 0 160
3 BCHP (100), PGS (200), BT (200) 0 160
4 BCHP (200), PGS (200), BT (200) 0 130
5 BCHP (0), PGS (0), BT (0) N/A N/A
212 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217

Fig. 9. COE values and cost composition for various scenarios.

Fig. 7. COE surface of scenario 3 as a function of WT and PV capacities.

Fig. 10. Optimal electricity supply for scenario 1.

not met by PV. Grid electricity is purchased after 20:00 when both
BT reserve is depleted, PV is absent and the residual demand ex-
Fig. 8. COE surface of scenario 4 as a function of WT and PV capacities. ceeds the 100 kW supply from the BCHP unit (Figs. 11e12).
Scenario 4 considers the case of having an oversized BCHP
(200 kW) in the microgrid. In this case the BCHP is large enough to
For scenario 1, when no BCHP is adopted, most of the electricity meet virtually all the nighttime demand when lower cost PV gen-
during midnight to early morning is supplied by purchasing elec- eration is absent with the exception of a small contribution from
tricity from the utility due to the absence of PV generation and low the battery in the later evening. No grid power is purchased and the
wind speed (<2.5 m/s) for the data set selected. The morning peak microgrid independently meets the full system demand (Fig. 13).
load demand occurs between 07:00 to 09:00 a.m. coinciding with
an increasing time of use tariff; therefore, the BT is used to balance
the demand in conjunction with PV. From 10:00 a.m. to 16:00 p.m.,
all of the electricity is generated from PV. After 07:00 p.m., PV 3.3. Sensitivity analyses
generation declines while as does the wind generation and import
of grid power again increases to meet the nighttime demand Sensitivity analyses were performed on BT capacity, electricity
(Fig. 10). and natural gas prices, and energy demands. The analysis was
Scenarios 2 and 3 show similar optimal operating schedules based on the results obtained from the most optimistic WT and PV
except for the addition of the BCHP that carries most of the demand capacities giving the lowest cost.

Table 7
Optimal COE values and cost composition for 5 scenarios.

Composition of total cost ($) scenario 1 scenario 2 scenario 3 scenario 4 scenario 5

Installation 239.75 259.91 263.37 334.94 7.70

O&M þ Fuel 19.44 276.51 281.33 342.99 11.60
Electricity purchase 290.26 55.18 55.18 0.00 804.42
Natural gas purchase 218.44 28.44 28.45 0.00 218.44
Net metering credit 16.48 12.77 10.95 5.01 0.00
Daily total cost 751.40 607.27 617.37 672.91 1042.16
COE ($/kWh) 0.1462 0.1182 0.1202 0.1310 0.2029
 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217 213

Fig. 14. COE as a function of battery capacity.

Fig. 11. Optimal electricity supply for scenario 2.

to increase due to the limitation of the gasifier capacity. With

increasing length of the prediction window, part of the storage
capacity becomes redundant. For example, if with an 80 kW
engine-generator set and a 12 h sliding time window, even if 100%
of the model generation is stored over the first 11 h, the optimal BT
capacity will not be 880 kWh, but instead will be something less to
still meet the demand over the model interval.

3.3.2. Effects of electricity and natural gas price

Figs. 15 and 16 show the results for the cost reduction ratio
obtained by changing the electricity and natural gas price from 60%
below to 60% above the reference prices. The microgrid provides
greater cost savings as the price of purchased electricity increases

Fig. 12. Optimal electricity supply for scenario 3.

Fig. 15. COE reduction ratio as a function of increased electricity price.

Fig. 13. Optimal electricity supply for scenario 4.

3.3.1. Effects of battery capacity

To study the effect of battery capacity on the COE, the battery
energy storage was varied from 0 to 300 kWh. The upper limit
corresponds to the storage size of charging the battery with
100 kWh/h for 3 h. Because the sliding time window width is 4 h,
the BT should be able to store at least 3 h of production. Fig. 14 il-
lustrates COE as a function of the battery size for the 4 scenarios
with the optimal capacity of WT and PV. For scenarios 1, 2 and 3, the
COE decreases as the storage size increases. The larger the storage,
the less purchased electricity is required by the customer during
the higher tariff period. At approximately 125 kWh, the COE begins Fig. 16. COE reduction ratio as a function of increased natural gas price.
214 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217

(Fig. 15). However, the marginal benefit of having the microgrid installed, the cost reduction ratio decreases slightly as the heat
declines with increasing grid price. A breakeven point is found for demand increases (Fig. 18). The overall impact on the COE is minor
scenarios 1 and 4 at grid prices that are 55 and 45% lower than the for scenarios 2 and 3, however, because the only heat resource in
assumed base case or reference price (negative cost reduction ra- the microgrid is from BCHP (no electric resistance heating), which
tios indicate a preference for utility grid purchase). Moreover, is absent in scenario 1, the COE becomes more sensitive to heat
although scenario 4 has the lowest cost savings at lower grid prices, demand. For scenario 4, with 200 kW BCHP capacity, even when
as the grid price increases this scenario eventually achieves the the heat demand is increased by 60% and the peak load is slightly
same savings as scenarios 2 and 3 and breaks even with scenario over 200 kW, the microgrid still supplies most of the thermal en-
1 at a grid price about 20% lower than the reference price. If the ergy from engine heat recovery and the boiler. Therefore, increasing
electricity price is reduced more than 20%, the no-BCHP case, sce- the heat demand does not require much additional natural gas and
nario 1, is preferred. As the purchased electricity price continues to a positive relationship results in contrast to that for scenario 1.
increase, the larger BCHP capacity becomes more attractive.
For scenario 1, as natural gas price changes from 60% below to
60% above the reference price, the COE cost reduction ratio de- 3.4. Stochastic model results
creases from 32% to 25% due to the lack of heat recovery from a
BCHP unit (Fig. 16). For scenarios 2, 3 and 4, the BCHP heat recovery Many of the technical and economic assumptions used in the
and producer gas can almost meet the full heat demand, therefore, model are subject to uncertainty as well as variability. To assess the
increasing natural gas price does not influence the COE reduction potential risk associated with decisions around a particular
ratio and a nearly positive linear relationship develops over the microgrid design, a stochastic model was developed employing
remainder of the cost range. Monte Carlo simulation for COE. Histograms from the Monte Carlo
simulations are presented in Fig. 19. The BCHP scenario distribu-
tions (scenarios 2e4) are centered at lower COE even with a 5%
3.3.3. Effects of demand changes possibility of the gasifier and/or engine failure than for the micro-
Figs. 17 and 18 show the results for the cost reduction ratio grid without BCHP (scenario 1). The 200 kW BCHP case (scenario 4)
obtained by changing the electricity and heat demand from 60% shows the widest variation in COE with cost mostly ranging be-
below to 60% above the reference values. The model yields a tween 0.01 and 0.26 $/kWh, a width of $0.25/kWh while the sce-
maximum cost reduction at around a 20% increase in the base nario 2 and 3 span about $0.17/kWh. The no BCHP case (scenario 1)
electricity demand. At this demand level, the microgrid capacity is shows the narrowest variation in COE around the mean of $0.13/
fully utilized for an overall improvement in cost of generation kWh.
(Fig. 17). For the case of no BCHP or only a single BCHP unit is Table 8 shows statistics for these and other important results
from the Monte Carlo simulations including descriptive statistics of
the scenario distributions, the probabilities associated with any
microgrid scenario reaching the optimal situation (deterministic
COE), and the probability that any microgrid scenario will be
preferred over the conventional utility grid supply (scenario 5).
Note that scenario 4 has the highest probability, about 46%, of
reaching the optimal COE. There is a better than 28 and 31% chance
that the COE of scenarios 2 and 3 are less than or equal to the
optimal value. For scenario 1, the probability of achieving the
optimal COE is only about 20%., however, the distribution of COE is
narrower around the mean. Even with uncertainty in the renewable
generation, the microgrid options still yield odds of having lower
COE than the utility supply only option (scenario 5) under the

Fig. 17. COE reduction ratio as a function of increased electricity demand.

Fig. 18. COE reduction ratio as a function of increased heat demand. Fig. 19. COE probability distribution of scenario 1e4.
 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217 215

Table 8
COE results from MCS.

Scenario COE values from MCS ($/kWh) P (optimum COE) P (COE ofScenario 5)

Min Max Mean Range

1 0.0911 0.2217 0.1653 0.1306 19.80% 96.00%

2 0.0482 0.2278 0.1402 0.1796 28.00% >99.9%
3 0.0570 0.2263 0.1391 0.1693 30.60% >99.9%
4 0.0123 0.2570 0.1427 0.2447 45.89% >99.9%

assumptions used. 4. Summary and conclusions

From the deterministic optimization analysis, scenario 2 has the
lowest COE, nevertheless, if all the uncertainty factors are consid- A sliding time window optimization modeling approach was
ered, scenario 3 provides the opportunity to achieve the lowest COE applied to the optimal design and dispatch scheduling of a
overall, albeit with reasonably low probability. With a 4-h sliding renewable microgrid supplying both heat and electricity. A model
time window, and the large BCHP capacity, any gasifier or engine microgrid was evaluated with biomass combined heat and power,
failure in any but the first hour while the utility TOU electricity wind and solar electricity generation, gas-fired boiler, and battery
price is high, allows the BT and PGS storage to accommodate the electric, producer gas, and thermal energy storage included.
lack of BCHP generation. For scenario 2, with the smaller 100 kW For the economic assumptions employed, BCHP can significantly
BCHP unit and no PGS, accommodation cannot be fully provided by improve the cost-efficiency of such a microgrid when compared
the BT. Scenario 2 is also more dependent on solar energy than with utility-supplied grid electricity and natural gas to meet the
scenario 3 and 4, consequently, the uncertainty from solar radiation electrical and heat demands. While for the example location used
is reflected in the higher minimum cost for scenario 2. Scenario 4 the mean wind speed was low and wind generation was not
also shows a much wider range of possible outcomes. To illustrate, selected as optimal, other locations may show superior wind per-
the expected COE from scenario 2 is lower than scenario 4, but the formance. In the optimal scheduling, the inclusion of batteries al-
odds of achieving that cost are lower. The microgrid design should lows storing electrical energy when utility time of use rates are low
consider the odds of obtaining performance below expectations and electricity purchase is acceptable, and satisfying demand from
and understand the tolerable risk. storage when the utility rates are high.
Sensitivity analyses show how battery capacity can be opti-
mized. Utility electricity and natural gas prices, as well as energy
3.5. Sensitivity analysis of stochastic variables demand levels all have a significant impact on microgrid design
decisions. Cost breakeven points exist for the microgrid against the
Sensitivities of the results are presented for the three main more conventional utility supply scenario depend on utility pricing,
power generation units, BCHP, WT and PV (Table 9). Changing the demand, and supply capacities.
BCHP capital cost by 15% from the reference value results in average The optimum dispatch was evaluated under uncertainty using
changes of 2.62, 2.00 and 9.87% in the mean COE for scenarios 2e4. the probability density functions anticipated for the primary pa-
When the BCHP O&M cost was changed by 15%, the average rameters of concern. Monte Carlo simulation was then used to
changes were 2.82, 4.89 and 6.96% in mean COE for these same generate the probability distribution of COE as an indicator of risk.
scenarios 2e4. It is apparent that the BCHP investment and O&M The lowest cost option may also have a higher risk of failing to reach
cost uncertainty is more sensitive in the COE of scenario 4. With the expected design performance. Sensitivity analysis indicated a
respect to the O&M cost, for each scenario, the COE varies by less greater sensitivity to capital cost than the O&M cost for the range of
than 5% indicating that COE is less sensitive to O&M cost category, assumptions evaluated. The model provides a means to determine
although BCHP O&M cost is more sensitive than the WT and PV the major risk factors in the microgrid design and weigh the various
O&M costs. advantages and disadvantages of each microgrid configuration.
The variation in wind speed and solar radiation yield greater Further work will compare optimized scenarios based on both
than 5% change in the final cost for scenario 1, however, solar ra- short-term and annual performance and include the uncertainties
diation does not show significant influence on the other scenarios. arising from demand-side management to alter both electricity and
Therefore, in terms of total cost, the variation in market price is heat demand in a combined optimization.
more important than the variations in wind speed and solar irra-
diance at the site if there is a BCHP on site.
Acknowledgments

Funding for this research was provided by the National Science

Table 9 Foundation through a Partnership for International Research and
Sensitivity analysis of the MCS results.
Education (PIRE) grant. We also gratefully acknowledge the finan-
Case Mean COE increase/decrease vs. baseline cial support of the China Scholarship Council (CSC).
1 2 3 4

þ15% BCHP Capital cost e 2.62% 2.00% 9.87% References

þ15% WT Capital cost 2.38% e e e
þ15% PV Capital cost 2.28% 3.93% 2.34% 9.56% [1] R.H. Lasseter, Microgrids, in: Power Engineering Society Winter Meeting, IEEE,
þ15% BCHP O&M cost e 2.82% 4.89% 6.96% 2002, p. 305.
þ15% WT O&M cost 0.00% e e e [2] P. Palensky, D. Dietrich, Demand side management: demand response, intel-
þ15% PV O&M cost 0.63% 1.93% 0.07% 0.92% ligent energy systems, and smart loads, IEEE Trans. Ind. Inf. 7 (2011) 381.
þ15% Wind speed 6.36% e e e [3] N. Hatziargyriou, H. Asano, R. Iravani, C. Marnay, Microgrids, IEEE Power
Energy 5 (2007) 78.
þ15% Solar irradiance 7.03% 3.44% 0.36% 3.06% 
[4] L. Montuori, M. Alca zar-Ortega, C. Alvarez-Bel, A. Domijan, Integration of
216 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217

renewable energy in microgrids coordinated with demand response re- Energy 59 (2013) 158.
sources: economic evaluation of a biomass gasification plant by Homer [34] M.A. Abdullah, K.M. Muttaqi, A.P. Agalgaonkar, Sustainable energy system
Simulator, Appl. Energy 132 (2014) 15. design with distributed renewable resources considering economic, envi-
[5] A. Acakpovi, E.B. Hagan, M.B. Michael, Cost benefit analysis of self-optimized ronmental and uncertainty aspects, Renew. Energy 78 (2015) 165.
hybrid solar-wind-hydro electrical energy supply as compared to HOMER [35] D. Neves, M.C. Brito, C.A. Silva, Impact of solar and wind forecast uncertainties
optimization, Int. J. Comput. Appl. (2015) 114. on demand response of isolated microgrids, Renew. Energy 87 (2016) 1003.
[6] A.H. Mamaghani, S.A.A. Escandon, B. Najafi, A. Shirazi, F. Rinaldi, Techno- [36] X. Yan, D. Abbes, B. Francois, Uncertainty analysis for day ahead power reserve
economic feasibility of photovoltaic, wind, diesel and hybrid electrification quantification in an urban microgrid including PV generators, Renew. Energy
systems for off-grid rural electrification in Colombia, Renew. Energy 97 (2016) 106 (2017) 288.
293. [37] A. Samimi, M. Nikzad, P. Siano, Scenario-based stochastic framework for
[7] O. Hafez, K. Bhattacharya, Optimal planning and design of a renewable energy coupled active and reactive power market in smart distribution systems with
based supply system for microgrids, Renew. Energy 45 (2012) 7. demand response programs, Renew. Energy 109 (2017) 22.
[8] J.G. Castellanos, M. Walker, D. Poggio, M. Pourkashanian, W. Nimmo, Model- [38] A. Narayan, K. Ponnambalam, Risk-averse stochastic programming approach
ling an off-grid integrated renewable energy system for rural electrification in for microgrid planning under uncertainty, Renew. Energy 101 (2017) 399.
India using photovoltaics and anaerobic digestion, Renew. Energy 74 (2015) [39] S. Mandelli, M. Merlo, E. Colombo, Novel procedure to formulate load profiles
390. for off-grid rural areas, Energy Sustain. Dev. 31 (2016) 130.
[9] C. Marnay, G. Venkataramanan, M. Stadler, A.S. Siddiqui, R. Firestone, [40] M. Chaudry, J. Wu, N. Jenkins, A sequential Monte Carlo model of the com-
B. Chandran, Optimal technology selection and operation of commercial- bined GB gas and electricity network, Energy Policy 62 (2013) 473.
building microgrids, IEEE Trans. Power Syst. 23 (2008) 975. [41] R. Palma-Behnke, C. Benavides, F. Lanas, B. Severino, L. Reyes, J. Llanos, et al.,
[10] A.S. Siddiqui, C. Marnay, J.L. Edwards, R. Firestone, S. Ghosh, M. Stadler, Effects A microgrid energy management system based on the rolling horizon strat-
of carbon tax on microgrid combined heat and power adoption, J. Energy Eng. egy, IEEE Trans. Smart Grid 4 (2013) 996.
131 (2005) 2. [42] M.P. Marietta, M. Graells, J.M. Guerrero, A rolling horizon rescheduling
[11] M. Stadler, A. Siddiqui, C. Marnay, H. Aki, J. Lai, Control of greenhouse gas strategy for flexible energy in a microgrid, in: IEEE Energycon, IEEE, 2014,
emissions by optimal DER technology investment and energy management in p. 1297.
zero-net-energy buildings, Eur. Trans. Electr. Power 21 (2011) 1291. [43] T. Fang, R. Lahdelma, Optimization of combined heat and power production
[12] R. Sen, S.C. Bhattacharyya, Off-grid electricity generation with renewable with heat storage based on sliding time window method, Appl. Energy 162
energy technologies in India: an application of HOMER, Renew. Energy 62 (2016) 723.
(2014) 388. [44] X. Wang, A. Palazoglu, N.H. El-Farra, Operational optimization and demand
[13] J.H. Braslavsky, J.R. Wall, L.J. Reedman, Optimal distributed energy resources response of hybrid renewable energy systems, Appl. Energy 143 (2015) 324.
and the cost of reduced greenhouse gas emissions in a large retail shopping [45] C. Siqi, L. JunYong, Y. Jiaqi, N. Yaqi, X. Yue, Z. Xin, et al., Optimal coordinated
centre, Appl. Energy 155 (2015) 120. operation for microgrid with hybrid energy storage and diesel generator, in:
[14] A. Hawkes, M. Leach, Modelling high level system design and unit commit- Power System Technology (POWERCON), 2014 International Conference on:
ment for a microgrid, Appl. Energy 86 (2009) 1253. IEEE, 2014, p. 3207.
[15] H. Ren, W. Gao, A MILP model for integrated plan and evaluation of distrib- [46] M. Pereira, D.M. de la Pen ~ a, D. Limon, Robust economic model predictive
uted energy systems, Appl. Energy 87 (2010) 1001. control of a community micro-grid, Renew. Energy 100 (2017) 3.
[16] H. Ren, W. Zhou, Nakagami Ki, W. Gao, Q. Wu, Multi-objective optimization [47] C. Wang, Y. Liu, X. Li, L. Guo, L. Qiao, H. Lu, Energy management system for
for the operation of distributed energy systems considering economic and stand-alone diesel-wind-biomass microgrid with energy storage system, En-
environmental aspects, Appl. Energy 87 (2010) 3642. ergy 97 (2016) 90.
[17] W. Kellogg, M. Nehrir, G. Venkataramanan, V. Gerez, Generation unit sizing [48] B. Jiang, Y. Fei, Smart home in smart microgrid: a cost-effective energy
and cost analysis for stand-alone wind, photovoltaic, and hybrid wind/PV ecosystem with intelligent hierarchical agents, IEEE Trans. Smart Grid 6
systems, IEEE Trans. Energy Convers. 13 (1998) 70. (2015) 3.
[18] D. Zhang, N. Shah, L.G. Papageorgiou, Efficient energy consumption and [49] L. Xie, M.D. Ilic, Model predictive economic/environmental dispatch of power
operation management in a smart building with microgrid, Energy Convers. systems with intermittent resources, in: Power & Energy Society General
Manag. 74 (2013) 209. Meeting, 2009 PES'09 IEEE, IEEE, 2009, p. 1.
[19] A. Omu, R. Choudhary, A. Boies, Distributed energy resource system optimi- [50] J. Silvente, G.M. Kopanos, E.N. Pistikopoulos, A. Espun ~ a, A rolling horizon
sation using mixed integer linear programming, Energy Policy 61 (2013) 249. optimization framework for the simultaneous energy supply and demand
[20] M. Marzband, M. Ghadimi, A. Sumper, J.L. Domínguez-García, Experimental planning in microgrids, Appl. Energy 155 (2015) 485.
validation of a real-time energy management system using multi-period [51] B.A. Berg, A. Billoire, Markov Chain Monte Carlo Simulations, Wiley Online
gravitational search algorithm for microgrids in islanded mode, Appl. En- Library, 2008.
ergy 128 (2014) 164. [52] M. Bashir, J. Sadeh, Optimal sizing of hybrid wind/photovoltaic/battery
[21] N.I. Nwulu, X. Xia, Optimal dispatch for a microgrid incorporating renewables considering the uncertainty of wind and photovoltaic power using Monte
and demand response, Renew. Energy 101 (2017) 16. Carlo, in: Environment and Electrical Engineering (EEEIC), 2012 11th Inter-
[22] D. Zhang, S. Evangelisti, P. Lettieri, L.G. Papageorgiou, Optimal design of CHP- national Conference on: IEEE, 2012, p. 1081.
based microgrids: multiobjective optimisation and life cycle assessment, En- [53] S.S. Reddy, Optimal scheduling of thermal-wind-solar power system with
ergy 85 (2015) 181. storage, Renew. Energy 101 (2017) 1357.
[23] H. Wang, W. Yin, E. Abdollahi, R. Lahdelma, W. Jiao, Modelling and optimi- [54] R. Dufo-Lo pez, I.R. Cristo
bal-Monreal, J.M. Yusta, Stochastic-heuristic meth-
zation of CHP based district heating system with renewable energy produc- odology for the optimisation of components and control variables of PV-wind-
tion and energy storage, Appl. Energy 159 (2015) 401. diesel-battery stand-alone systems, Renew. Energy 99 (2016) 919.
[24] M. Motevasel, A.R. Seifi, T. Niknam, Multi-objective energy management of [55] H. Jahangir, A. Ahmadian, M.A. Golkar, Optimal design of stand-alone micro-
CHP (combined heat and power)-based micro-grid, Energy 51 (2013) 123. grid resources based on proposed Monte-Carlo simulation, in: Innovative
[25] M.H. Moradi, M. Hajinazari, S. Jamasb, M. Paripour, An energy management Smart Grid Technologies-Asia, IEEE, 2015, p. 1.
system (EMS) strategy for combined heat and power (CHP) systems based on [56] B.M. Jenkins, L.L. Baxter, J. Koppejan, Biomass Combustion. Thermochemical
a hybrid optimization method employing fuzzy programming, Energy 49 Processing of Biomass: Conversion into Fuels, Chemicals and Power, 2011,
(2013) 86. p. 13.
[26] C. Brandoni, M. Renzi, Optimal sizing of hybrid solar micro-CHP systems for [57] B. Jenkins, L. Baxter, T. Miles, Combustion properties of biomass, Fuel Process.
the household sector, Appl. Therm. Eng. 75 (2015) 896. Technol. 54 (1998) 17.
[27] A.K. Basu, A. Bhattacharya, S. Chowdhury, S. Chowdhury, Planned scheduling [58] D.T. Nguyen, L.B. Le, Optimal energy management for cooperative microgrids
for economic power sharing in a CHP-based micro-grid, IEEE Trans. Power with renewable energy resources, in: Smart Grid Communications (Smart-
Syst. 27 (2012) 30. GridComm), 2013 IEEE International Conference on, IEEE, 2013, p. 678.
[28] A. Basu, S. Chowdhury, S. Chowdhury, Operational management of CHP-based [59] S. Talari, M. Yazdaninejad, M.-R. Haghifam, Stochastic-based scheduling of the
microgrid, in: Power System Technology (POWERCON), 2010 International microgrid operation including wind turbines, photovoltaic cells, energy
Conference on: IEEE, 2010, p. 1. storages and responsive loads, IET Gener. Transm. Distrib. 9 (2015) 1498.
[29] X. Xu, H. Jia, D. Wang, C.Y. David, H.-D. Chiang, Hierarchical energy man- [60] M. Brum, P. Erickson, B. Jenkins, K. Kornbluth, A comparative study of district
agement system for multi-source multi-product microgrids, Renew. Energy and individual energy systems providing electrical-based heating, cooling,
78 (2015) 621. and domestic hot water to a low-energy use residential community, Energy
[30] M.H. Moradi, M. Eskandari, A hybrid method for simultaneous optimization of Build. 92 (2015) 306.
DG capacity and operational strategy in microgrids considering uncertainty in [61] B.M. Jenkins, A comment on the optimal sizing of a biomass utilization facility
electricity price forecasting, Renew. Energy 68 (2014) 697. under constant and variable cost scaling, Biomass Bioenergy 13 (1997) 1.
[31] V.N. Coelho, I.M. Coelho, B.N. Coelho, M.W. Cohen, A.J. Reis, S.M. Silva, et al., [62] J. Silvente, A. Aguirre, G. Crexells, M. Zamarripa, C. Me ndez, M. Graells, et al.,
Multi-objective energy storage power dispatching using plug-in vehicles in a Hybrid time representation for the scheduling of energy supply and demand
smart-microgrid, Renew. Energy 89 (2016) 730. in smart grids, Comput Aided Chem. Eng. 32 (2013) 553.
[32] H. Haddadian, R. Noroozian, Optimal operation of active distribution systems [63] S. Wickwire, Washington, DC, Biomass Combined Heat and Power Catalog of
based on microgrid structure, Renew. Energy 104 (2017) 197. Technologies, 1, 2007, p. 10.
[33] A. Baziar, A. Kavousi-Fard, Considering uncertainty in the optimal energy [64] N. DiOrio, A. Dobos, S. Janzou, Economic Analysis Case Studies of Battery
management of renewable micro-grids including storage devices, Renew. Energy Storage with SAM, National Renewable Energy Lab. (NREL), Golden,
 Y. Zheng et al. / Renewable Energy 123 (2018) 204e217 217

CO, 2015. International, Fairfax, VA, 2010, p. 275.

[65] R. Tidball, J. Bluestein, N. Rodriguez, S. Knoke, Cost and Performance As- [66] Energy Metering (NEM) Available: http://www.cpuc.ca.gov/General.aspx?
sumptions for Modeling Electricity Generation Technologies, ICF id¼3800Net [Accessed: October 2016].