See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/322237820

Optimization and energy management of a standalone hybrid microgrid in the

presence of battery storage system

Article  in  Energy · January 2018

DOI: 10.1016/j.energy.2018.01.016

CITATIONS READS

5 237

4 authors, including:

Hadis Moradi Mahdi Esfahanian

Florida Atlantic University Florida Atlantic University
14 PUBLICATIONS   67 CITATIONS    20 PUBLICATIONS   109 CITATIONS   

SEE PROFILE SEE PROFILE

Amir Abtahi
Florida Atlantic University
27 PUBLICATIONS   138 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Electromechanical Activation Time via HemoTag Correlation with Pulmonary Artery Pressure by Right Heart Catheterization: Implications for Diagnosis and Clinical
Practice View project

Modeling a Hybrid Microgrid Using Probabilistic Reconfiguration under System Uncertainties View project

All content following this page was uploaded by Hadis Moradi on 04 February 2018.

The user has requested enhancement of the downloaded file.

 Energy 147 (2018) 226e238

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Optimization and energy management of a standalone hybrid

microgrid in the presence of battery storage system
Hadis Moradi a, *, Mahdi Esfahanian a, Amir Abtahi b, Ali Zilouchian a
a
Department of Computer and Electrical Engineering and Computer Science, Florida Atlantic University, Boca Raton, USA
b
Department of Ocean and Mechanical Engineering, Florida Atlantic University, Boca Raton, USA

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

Article history: In this study, an optimal energy scheduling of a standalone microgrid under system uncertainties is
Received 30 August 2016 investigated and several operational strategies are tested to evaluate the system performance. The main
Received in revised form objective of the proposed optimization energy management system is to improve energy utilization
12 December 2017
efficiency, decrease system fuel cost and gas emissions reduction by planning generations of energy
Accepted 2 January 2018
Available online 3 January 2018
resources hourly for the next day. The system is modeled as a constrained single-objective optimization
problem in order to minimize the operation and generated emission costs. Employing an advanced
dynamic programming method, the optimization problem is solved. In order to achieve optimal utili-
Keywords:
Microgrid
zation of renewable resources in a microgrid environment and system cost-effective operations, the
Distributed energy resources effectiveness of a battery storage unit has to be examined. Two different operational policies are
Optimal operation investigated to examine the system behavior, where the energy production of a microgrid is supplied by
Renewable energy penetration generation units with and without access to a battery storage system. The obtained results show a
Emission reduction considerable reduction in system total cost and produced emissions when the MG has access to battery
storage system in the proposed second policy. Simulation results demonstrate the feasibility and
effectiveness of the proposed technique in microgrid energy planning and implementation.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction
Microgrid (MG) systems have played essential roles in solving
the energy-related issues around the world and have drawn more
attention in recent years [1,2]. According to the definition by the
U.S. Department of Energy, a MG is “a group of interconnected loads
and Distributed Energy Resources (DER) within clearly defined
electrical boundaries which is considered as a single controllable
entity from the main grid point of view” [3]. The MGs are classified
according to their application, loads and ownership [4,5]. The
utility-scale MG delivers electric power for central urban loads. The
MG in commercial or industrial sectors serves large loads that need
premium and highly reliable power quality. An isolated MG is a
practical option to feed the loads that are placed far away from
devoted feeders [6]. Customer-driven MG is a new instance which
presents associated controls and real-time information to the cus-
tomers [7]. In addition, traditional power systems suffer from
* Corresponding author. 777 Glades Road, EG96-EE 409, Boca Raton, FL 33431,
USA.
E-mail address: hmoradi@fau.edu (H. Moradi).
several concerns such as fossil fuel deficiency, low energy efficiency
in power delivery sector, and growing pollution of conventional
energy resources. Furthermore, electric power utilization is
constantly increasing as a result of the industrialization, population
growth, and especially urban expansions. To accomplish this
request, DERs being penetrated at the distribution level to supply
the load locally [8e10]. Based on current studies, energy consumers
would have access to a standalone MG in the near future. This is
comprised of micro-generation systems where their operation and
management can be carried out by Energy Management System
(EMS), depending on real-time electricity pricing [11]. Likewise, the
significance of MG, which primarily corresponds to the coordinated
operation of a series of DERs, loads and ESS, is pretty appealing due
to its flexibility, controllability, and energy planning strengths. MGs
are expected to operate in both grid-connected and standalone
modes in order to provide power supply to the loads without
interruption and fulfill the load demand on a momentary basis
economically [12e14].
The optimal utilization control of such a MG system is a chal-
lenging task due to the complication of coordination among the
generating units, Energy Storage Systems (ESSs) and load

https://doi.org/10.1016/j.energy.2018.01.016
0360-5442/© 2018 Elsevier Ltd. All rights reserved.
 H. Moradi et al. / Energy 147 (2018) 226e238 227

Nomenclature EFjk emission factor of generating unit jth and emission

type k
N total number of generators SOCmax ESS state of charge upper limit
FDG,j DG total consumption function (liter/h) SOCmin ESS state of charge lower limit
CD diesel fuel price ($/liter) vavg average speed of wind
s standard deviation of stochastic variable c Nmax maximum number allowed for a unit to be on/off in a
PFCj FC generated power at interval jth certain time
hFCj cell efficiency of FC dstart/stop number of starts and stops of a component
PMTj MT generated power at interval jth Pmax
ess ESS output power upper limit
hMTj MT output efficiency Pmin
ess ESS output power lower limit
GIIR solar incident irradiance ht battery charging/discharging efficiency
Tc PV cell temperature Cess ESS capacity
Ta ambient temperature Pmax
j unit jth output power upper limit
Pr WT rated power Pmin
j unit jth output power lower limit
Vci cut-in wind speed Toff,j unit shout down duration of jth unit
Vr rated wind speed c a stochastic variable
k shape factor of the Weibull PDF 2 Rayleigh scale index
2 scale factor of the Weibull PDF
Vco cut-out wind speed List of abbreviations
sd mean of the load demand EMS Energy Management System
md standard deviation of the load demand MG Microgrid
Cj fuel cost of the generating unit jth STC Standard Test Condition
Fj(Pj) fuel consumption rate of the unit jth SOC Battery State of Charge
O&Mj(Pj) operation and maintenance cost of the generating DER Distributed Energy resources
unit jth ($/h) PDF Probability Density Function
STCj start-up cost of the unit jth ($/h) MCEMS Microgrid Central Energy Management System
bk externality costs of emission type k PV Photovoltaic
COMj proportionally constant of resource jth DG Diesel Generator
sj hot start-up costs of generating unit jth MT Micro-gas Turbine
dj cold start-up costs of generating unit jth FC Fuel Cells
tj constant of unit cooling time OPF Optimal Power Flow
m mean of stochastic variable c ESS Energy Storage Systems
Cng natural gas price WT Wind Turbine
REBG Renewable Energy Based Generators

management possibilities [15]. Several real-time optimization and
energy management approaches for a standalone MG, considering
ESS lifetime and using multi-layer ant colony optimization have
been presented in Refs. [16,17]. A multi-agent based distributed
control for handling complex energy management of a MG using
game theory implementation for multi-agent coordination is pro-
posed in Ref. [18]. In Ref. [19], an optimal strategy for supplying
required energy for an autonomous MG is addressed. The proposed
system utilizes a hybrid energy system, including photovoltaic,
wind, battery banks, and diesel generators, based on operational
and financial perspectives. The problem has been solved using
guaranteed convergence Particle Swarm Optimization (PSO) with
the Gaussian mutation method. Although many metaheuristic al-
gorithms implement some form of stochastic optimization, they
commonly do not guarantee an optimal solution to be found in the
global search. In addition, some of these methods rely on initial
guesses to reach the desired results that might not lead to the
optimum solution. Some of these techniques are time-consuming
and inefficient.
In general, uncertainty can be considered as one of the impor-
tant concerns related to the energy management in MGs, especially
for the components with intermittent nature and load demand. As
pointed out in Refs. [20e23], the uncertainty commonly can be
described as the probability of discrepancy between the forecasted
measurements and the actual values.
In this paper, a novel system modeling is proposed to investigate
the MG concerns in terms of accuracy and performance. Firstly, the
MG configuration as well as associated energy components and
loads are introduced and formulated in this paper. Secondly, in
order to develop the system modeling, reviewing minimization of
the total cost of the system has to be accomplished. Therefore, it is
essential to evaluate the problem of costs minimization as well as
fulfilling the load demand. Consequently, an advanced dynamic
programming method as presented in Ref. [24] is applied to the
energy-scheduling problem using MATLAB and GAMS/Cplex12
software on a personal computer (2.6 GHz with 8 GB of RAM). This
method is a numerical algorithm and has an appropriate conver-
gence feature. In the presented model, the dynamic process of the
ESS is considered with the inclusion of load demand, wind and
solar production uncertainties using various probabilistic
modeling. The model is implemented in a typical MG with respect
to two distinctive operational policies: (i) the system operates in
islanded mode without having access to any energy storage sys-
tems. (ii) the system is considered as an off-grid MG with access to a
set of battery storage units. In this research, the proposed algorithm
provides enough flexibility and has adequate rapid response to any
incident occurring in the system. In the presented methodology,
the resources' timing schedules are studied by utilizing day-ahead
time intervals. The energy planning is capable of decreasing opti-
mization timing to intra-day or intra-hour decision-making. Ulti-
mately, the optimal results for generation scheduling will be
presented and analyzed for final conclusions to be drawn.
In short, the main contributions of this paper can be described
as follows:
228 H. Moradi et al. / Energy 147 (2018) 226e238

A robust optimal MG energy-scheduling model is proposed in
islanding mode operation under appropriate technical con-
straints. The proposed procedure is a numerical algorithm,
which has a decent convergence property.

Probabilistic models are applied to reflect the uncertainty of
renewable energy resources and load. The solution guarantees
convergence with all possible realizations of the modeled
operational uncertainties. Considering probabilistic configura-
tion and unit commitment concretely for MG optimal energy
planning, with the integration of renewable energy and load
demand, is a novel operation arrangement to minimize MG's
cost and emission in the presence of uncertainties.

Various policies are applied on a simulated MG and the effec-
tiveness and feasibility of the proposed model are validated. A
comprehensive optimization technique is proposed for day-
ahead MG energy management and unit commitment.
2. System configuration
The configuration of a multi-source MG composed of renewable
and non-renewable resources along with the battery storage sys-
tem is shown in Fig. 1. The configuration includes Micro-gas Tur-
bine (MT), Fuel Cells (FC) and Diesel Generator (DG) as conventional
energy sources, which consume natural gas and diesel as input
fuels. Wind and Photovoltaic (PV) based active generators are
associated with the storage system that consists of a set of batteries
as a long-term storage system and a set of ultracapacitors for short-
term and real-time power balancing. Thus, surplus energy from
generating units can be locally stored in batteries for future use [5].
In order to send power references and exchange data, a
communication network exists among energy resources, local load
demand and the Microgrid Central Energy Management System
(MCEMS). The control structure relies on a MCEMS due to deliver
the energy management and controls the Optimal Power Flow
(OPF) of a local area to meet the local requirements. The distribu-
tion system operator views the MCEMS as a single entity, which is
capable to control a cluster of generators and flexible loads locally
and to allow Renewable Energy Based Generators (REBG) to supply
their full benefits while minimizing the pollution of energy com-
ponents [25e27]. The installed nominated power of each power
generation unit is presented in Table 1.
Descriptions of MG energy components as well as their corre-
sponding characteristics and expressions for various fuel con-
sumption rates are presented as follows:
2.1. Diesel generator
Due to a diesel generator's capability to operate standalone
and remotely, diesel generators typically are applied in power
systems to supply local loads. Also, a DG is usually capable of
meeting peak load on its own at expensive rates without the
need for adding other energy sources. Currently, DGs operate
with robustness, and low maintenance cost, hence DGs are
typically utilized in MG technology for use in supplying isolated
areas [28]. The DG operation is constrained by its power output
boundaries; power ramp-up and -down rates and also start-up
time, when time frames are considered in the order of minutes.
The fuel cost of a power source can be represented and modeled
as a quadratic polynomial function of its real output power [29].
The following equation expresses the total consumption function
of a DG in liter/h:
X
N
FDG;j ðpðtÞÞ ¼ aj þ bj PDG;j þ cj PDG;j ðtÞ2 (1)
j¼1
where aj,bj and cj are associated parameters for each cost function
of a typical generator and PDG,j, j ¼ 1, 2, …, N is the output power of
the jth DG (kW) assumed to be known. Commonly, parameters aj,
bj, and cj are known and given by the manufacturer as component
specifications. These parameters for a typical DG are shown in
Table 2.
The fuel cost of a DG unit (in $/h) can be calculated as a function
of fuel consumption as follows:
X
CDG ¼ CD FDG;j ðpðtÞÞ (2)
j¼1
where Eq. (2) is a function of fuel consumption (in liter/h) and DG's
fuel price (in $/liter) and it varies with the change of diesel fuel
prices.
2.2. Fuel Cells

FC is one of the usual stored-energy DERs employing in MGs. A

FC unit is an electrochemical device capable of converting chemical
energy of a fuel and an oxidant directly into low voltage DC elec-
tricity. By means of an electrochemical reaction the transformation
is done, in which, the oxidant is mainly oxygen or air. On the other
hand, since the output that is oxidized is not part of the FC structure
and both products can be supplied simultaneously, while there are
reagents, the generation of electricity will keep constant [30]. The
efficiency of a FC is the ratio of the electrical output power over
ratio of the input fuel where both of them are in the same units. The
following equation calculates the fuel cost of a typical FC:

Table 1
Installed distributed generations.

Unit Type Min Power (kW) Max power (kW)

MT 9 30
FC 1.6 40
DG 15 50
Wind 0 15
PV 0 14
Fig. 1. Integration of multi-source MG with loads and central energy management
ESS 30 30
systems.
 H. Moradi et al. / Energy 147 (2018) 226e238 229

Table 2
Diesel generator corresponding parameters.

Diesel DNAC Coefficient (a) Coefficient (b) Coefficient(c)

0.4333 0.2333 0.0074

Fig. 3. Estimated model of MT efficiency as a function of output power.

percentage. An estimated function of efficiency vs. output power of

a 30 kW gas turbine unit is shown in Fig. 3. The presented efficiency
function model is inserted as one of the inputs in Eq. (4).
Fig. 2. Estimated model of FC efficiency as a function of output power.

2.4. Photovoltaic system

X PFCj
CFC ¼ Cng
j
hFCj
Regarding Eq. (3), the efficiency of the FC is a function of output
power and drops with the increase of the supplied power. FC effi-
ciency curve can be drawn as a function of net power percentage.
An estimated function of efficiency verses power output of a 40 kW
FC system is presented in Fig. 2 and the FC formulated efficiency
model is utilized as one of the inputs in Eq. (3).
(3)
Solar PV resources have played an important role in present
MGs due to the clean and sustainable use features. A PV cell is a
nonlinear electric resource, that has an output characteristic
affected by light intensity, environment temperature, and also load
magnitude [32]. The output power of each PV module can be
calculated as a function of solar incident irradiation, cell tempera-
ture, and physical characteristics of the module. The following
equation is used to calculate the output power generated by PV
systems:

2.3. Micro gas turbine GIIR

As one of the most competitive DESs in MG, a MT is employed
with advantages such as greater stability, strong fuel adaptability,
low fuel consumption rate and low maintenance cost [31]. The cost
function model of a MT is comparable to the FC function as pre-
sented in Eq. (3). The fuel cost of a typical MG can be expressed in
an equation as follows:
X PMTj
CMT ¼ Cng
j
hMTj
The efficiency of the MT is a function of net output power and
increases when the unit output power rate rises. MT efficiency
curve can be presented as a function of net output power
PPV ¼ PSTC ð1 þ kðTc  Ta ÞÞ (5)
GSTC
where PPV is the output power of the module at irradiance GIIR. In
addition, PSTC, k, GSTC represent the module maximum power at
Standard Test Condition (STC), temperature coefficient of the po-
wer, the irradiance at STC, respectively. In this research, these pa-
rameters are assumed 285 W, 0.5, 1000 W/m2, respectively. GIIR and
Ta as two stochastic variables are considered as PV system un-
certainties and can be modeled by employing Beta Probability
(4)
Density Function (PDF) on historical data [33]. According to
approximately regular alterations of these variables during daily
periods, the created PDFs are different at each time in order to
improve the accuracy. Beta PDFs are used to model aforementioned
variables according to Eqs. (6)e(8):

8  
>
> G a ; b
 >
b b
 <    cab 1 ð1  cÞbb 1 ; a ; b  0; c ¼ GIIR ; Ta
b b
PDF c; ab ; bb ¼ G ab ; G bb (6)
>
>
>
:
0; else
230 H. Moradi et al. / Energy 147 (2018) 226e238

the wind speed profiles, as utilized in this study for this purpose,
where G is determined as gamma distribution function, a and b and produce power of the WT at each time [36].
values are parameters of the beta probability distribution function
   2 !
in kW/m2 calculated utilizing the mean and standard deviation of 2v v
PDFr ðvÞ ¼ exp  (12)
the stochastic variable as follows: 22 2
 
mð1  mÞ
a ¼ m$ 1 (7) where Rayleigh scale index can be estimated by an acceptable
s2 approximation as follows:
   
a 2
b ¼ ð1  mÞ (8) 2 ¼ pffiffiffi vave z1:128vave (13)
m p
Therefore, the scale factor can be calculated using average speed
of wind in the typical area and consequently the PDF of the WT
2.5. Wind turbine output power can be obtained.

A widespread sort of wind power penetration can be observed 2.6. Load demand
in present MGs, though only Wind Turbines (WTs) with a consid-
erable share of the peak load are usual to be operated as part of The behavioral patterns of diverse energy consumers lead to
EMS. The power output of a wind turbine can be express as a variation in load demand profile in MGs systems. Thus, the sto-
function of the wind speed, the hub height and the output char- chastic behavior of load demand, which alters continually with a
acteristics of the wind turbine generator. The most common output high level of uncertainty, can be modeled with a normal distribution
power of a wind generator attached to a WT can be presented as function with a mean value and a standard deviation as follows [37]:
follows [34,35]:
" #
8 1 ðP  m Þ2
>
> P ¼0 Vac < Vci PDFðPd Þ ¼ pffiffiffiffiffiffi$exp  d 2 d (14)
< WT 3 sd 2p 2sd
PWT ¼ a:Vac þ b:Pr Vci  Vac < Vr
(9)
>
> PWT;r ¼ Pr Vr  Vac < Vco
: where sd and md represent mean and standard deviation of the load
PWT ¼ 0 Vac < Vco
respectively. Also, Pd is the power demand at each time.
In this paper, the uncertainty is modeled utilizing Monte Carlo
Pr Vci3 (MC) simulation where for uncertain inputs, i.e., solar radiation,
a¼ ; b ¼ (10)
Vr3  Vci3 Vr3  Vci3 temperature, wind speed and load demand several scenarios are
generated. Consequently, the system is considered under attained
where a and b are coefficients depending on the wind turbine types stochastically scenarios as deterministic inputs. Therefore, there are
and can be calculated using Eq. (10). Fig. 4 shows the model of the different statuses, which are analyzed by employing different sce-
actual power curve of a WT provided by the manufacturer where narios. In fact, MC approach is a method that uses stochastic vari-
the values of Vci, Vr, and Vco are given as 3.5, 11 and 14 (m/s), ables and their corresponding PDFs to solve the uncertain problems
respectively. [38,39].
Due to the stochastic behavior of wind speed, the output power
of wind generator has also comparable nature as solar output po-
2.7. Model input data
wer where the wind speed is typically modeled by using Rayleigh
PDF as a particular case of Weibull PDF. The wind speed is estimated
Locally meteorological data including temperature ( C), solar
by Weibull PDF as following:
irradiation data (W/m2) and wind speed (m/s) as well as a day-
  k1   k  ahead load demand of the MG extracted from MC simulation in
k v v
PDFw ðvÞ ¼ exp  (11) 24-h are depicted in Figs. 5e8, respectively.
2 2 2
The Weibull PDF is converted to Rayleigh PDF when k is equal to
2 as presented in Eq. (12). This PDF can be used to model most of

Fig. 4. The power curves of a typical wind turbine. Fig. 5. 24-hour ahead temperature forecast.
 H. Moradi et al. / Energy 147 (2018) 226e238 231

3. Optimization model formulation

The operation of the MG is presented under two different pol-

icies as follows:

3.1. Policy 1: standalone system without access to energy storage

In this policy, the local generating units but ESS satisfy the load
demand of a MG at each time. The objective function that mini-
mizes operation and emission costs can be modeled as follows:

X
N       XN
minFðPÞ ¼ Cj  Fj Pj ðtÞ þ O&Mj Pj ðtÞ þ STCj þ
j¼1 j¼1

X
M  
 bk EFjk Pj ðtÞ
Fig. 7. 24-hour ahead wind speed forecast.
k¼1
(15)

where Pj is a decision variable defining the power output from

generating unit jth with P ¼ (P1, P2, …, PN), which represents the
decision variable vector. Numerical values of externality cost,
emission factor of type k emission are given in Table 3. N and M are
variables representing the generating unit j and emission types
(CO2 or NOx or SO2).

3.1.1. Operating and maintenance cost

The operating and maintenance costs depend on the energy
produced by the generating resources.

X
N   XN
O&M ¼ O&Mj Pj ðtÞ ¼ COMj Pj ðtÞ (16)
j¼1 j¼1

where it is function of proportionally constant for each generating Fig. 8. 24-hour ahead load demand forecast.
unit which is related to the types of energy resources [40]. COMj
values have been used in this study are provided in Table 4. " !#
Toff ;j
STCj ¼ sj þ dj 1  exp (17)
tj
3.1.2. Start-up cost
Regarding the time period that an energy resource has been off which is a function of hot and cold start-up costs, the constant of
before its start-up, the generator start-up cost can be calculated. unit cooling time, and unit shutdown duration.
This cost for unit j at any given time interval can be measured by an
exponential function as:
3.2. Policy 2: standalone system with battery storage

The lithium-ion battery storage system is one of the core parts of

the windesolaredieselebattery MG system. Lithium-ion batteries
have been widely used in modern electrified vehicles, which led to
a good understanding of the technology. The reliable, efficient, and
safe operation of lithium-ion batteries requires monitoring, control,
and management. The strategy of managing the lithium-ion bat-
teries considerably impacts the performance of the overall system.
Renewable energy sources are depending on weather conditions
and have intermittent behavior. Therefore, battery storage is used
to store surplus power to maintain the system frequency, system
voltage and supply the power to loads in case of low solar radiation
or wind speed or load changes in the MG. Additionally, storage
systems can smooth the fluctuation of solar and enhance the load
availability [41,42]. In this policy, the MG is studied while it is
operating with access to the battery storage system where ESS can
be charged and discharge whenever is required. Although, the
objective function and related constraints of the first policy are
Fig. 6. 24-hour ahead solar irradiation forecast.
applicable in the seconded scenario modeling, more constraints
232 H. Moradi et al. / Energy 147 (2018) 226e238

Table 3
Externality costs and emission factors of the system.

Emission Type Externality costs $/lb Emission factors for DG lb/MWh Emission factors for FC lb/MWh Emission factors for MT lb/MWh

NOx 4.2 21.8 0.03 0.44

SO2 0.99 0.454 0.006 0.008
CO2 0.014 1.432 1.078 1.596

Table 4
Proportionally constant of components.

COMj Diesel Generator ($/kWh) Microgas Turbine ($/kWh) Fuel Cells ($/kWh)

0.01258 0.00419 0.00587

will be introduced and applied for ESS.

dstart=stop  Nmax (22)
4. System constraints
where Eq. (22) is a function of unit's number of starts and stops and
4.1. Constrains in policy1: Nmax is a specific number given by the manufactures.

Applied constraints in the first policy are described as following 4.2. Constraints in policy 2:

Since the BESS is added to the generating sector, some con-

4.1.1. Power balancing of supply and load demand constraints
straints are required to be changed, and some new constraints need
to be considered as follows
X
N
Pj ðtÞ ¼ PL ðtÞ  PPV ðtÞ  PWT ðtÞ (18) 4.2.1. Power balancing constraints
j¼1

where Pj(t) denotes power generation of jth unit at time t in kW, X

N

PL(t) is power demand at time t in kW and PPV(t) and PWT(t) are Pj ðtÞ ¼ PL ðtÞ  PPV ðtÞ  PWT ðtÞ  Pess ðtÞ (23)
j¼1
output power of PV and wind at time t respectively.
where Pess(t) is the output power of the battery storage at time t in
4.1.2. Generation capacity constraints kW and other variables are similar to the equality of Eq. (18).
Output power of the jth dispatchable unit at time t should fall
between two maximum and minimum values in order to stabilize 4.2.2. ESS power output constraints
system operation. The output power of the battery at time t is restricted between
two limits as follows:
Pjmin  Pj ðtÞ  Pjmax ; cj ¼ 1; …; N (19)
min max
Pess  Pess ðtÞ  Pess (24)

where in the above inequality relation, ESS is in discharging mode

4.1.3. Run and start time constrains
when Pess is positive and is in charging mode when Pess is negative.
Also, ESS has no generation when Pess is equal to zero.
  
on
Tt1;j  MUTj ut1;j  ut;j  0 (20) 4.2.3. ESS state of charge constraints
The battery's State of Charge (SOC) prevents overcharging or
  
off undercharging of the battery. Hence, adding constraints to the
Tt1;j  MDTj ut;j  ut1;j  0 (21)
battery SOC would help the ESS perform appropriately. The SOC
off
value at time t is determined by the SOC value at time t-1 and the
where Tt1;j on are the jth unit off and on time, respectively.
and Tt1;j battery output power during the period as described below:
ut,j represents the unit off/on status which is either 0 or 1. Also,
MUTj and MDTj are minimum up and down time limit of jth unit. ht Pess ðtÞ
Constraints in Eqs. (20) and (21) assure that once a microsource SOCðtÞ ¼ SOCðt  1Þ  (25)
Cess
turns on, it will operate uninterruptedly for a specific time before it
turns off. The corresponding constraints can be formulated so that the
Lieion battery SOC will be within a certain range at any hour
[16,43]:
4.1.4. Number of the starts and stops constraints
This constraint illustrates that the number of starts and stops of
SOC min  SOCðtÞ  SOC max (26)
generating units should be less that a certain value:
 H. Moradi et al. / Energy 147 (2018) 226e238 233

In this research, the SOCmax, SOCmin and ht are assumed to be 0.3 2 3

p.u and 0.95 p.u and 85%, respectively [5]. 1 0 0 0 0 0 0 0
6 7
6 1 0 0 0 0 0 0 07
6 7
6 7
6 0 1 0 0 0 0 0 07
5. Implementation of the algorithm 6 7
6 7
6 :: :: :: :: :: :: :: :: 7
6 7
In this section, the formulation of the two policies, as described   6 7
in Section 3, is converted into a more suitable form for the proposed cjB xj ; uj ¼ 6
6 0 0 0 0 0 0 1 07 7xj
6 7
optimization technique in Section 6. The power output of dis- 6 0 0 0 0 0 0 1 07
6 7
patchable units at time j is considered to be a state vector (xj) and 6 hchj 7
6 0 17
the planning generation adjustment of each dispatchable unit is 6 0 0 0 0 0 7
6 Cess 7
considered to be a decision vector (uj). CFj is the cost function at 6 7
4 hchj 5
time j and the function fj parameter represents the relation be- 0 0 0 0 0 0 1
tween adjacent states in Eq. (28). The following formula is
Cess
2 3
expressed to minimize the total cost: P1max
6 P min 7
6 1 7
X
n1   6 P max 7
CF ¼ CFj xj ; uj þ CFn ðxn Þ (27) 6 2 7
6 :: 7
j þ6
6 P max
7
7 (33)
The system's constraints can be described as follows: 6 ess 7
6 P min 7
  6 ess 7
4 SOC max 5
xjþ1 ¼ fj xj ; uj
  SOC min
cjA xj ; uj ¼ 0; cnA ðxn Þ ¼ 0 j ¼ 1; 2; …; n  1 (28)
 
cjB xj ; uj  0; cnB ðxn Þ  0 j ¼ 1; 2; …; n  1 In above matrices, the state vector dimension is h þ 2 and the
decision vector dimension is h þ 1, where h represents the number
In Eqs. (27) and (28), CFj, fj, cjA and cjB are all twice differentiable
of generating units in the MG.
functions. The optimization task is to find control variables u1, u2,
…, un1 and the initial state variable x1 in this discrete-time system,
6. Methodology in solving models
which minimize the cost function. Having the initial value of x1 at
time j and adjustment of x1 at time jþ1, the value of x1 at time jþ1
An Advanced Dynamic Programming (ADP) method with con-
can be found. In general case, the transformations of formulations
strained discrete-time optimal control is employed in this study to
are presented as:
solve MG's energy scheduling problems [24]. Based on the formu-
lations in Section 5, the equivalent formulations can be written as
xj ¼ ½P1 ðjÞP2 ðjÞ…Pm ðjÞPess ðjÞSOCðjÞT
(29) below:
uj ¼ ½DP1 ðjÞDP2 ðjÞ…DPm ðjÞDPess ðjÞT
X
n1h  . i
where xj is state vector of units power output and uj stands for minCF ¼ yj þ xTj yj þ uTj sj þ xTj Gj xj þ 2xTj Lj uj þ uTj Sj uj 2
decision vector of power production adjustments for each j
dispatchable unit. By using (28) cost function can be transformed þ CFn þ xTn yn þ xTn Gn xn =2
to:
(34)
 
CFj xj ; uj ¼ xTj yj þ xTj Gj xj þ yj Subject to the associated constraints as follows:
  (30)
CFn xj ¼ xTn yn þ xTn Gn xn þ yn
xjþ1 ¼ Aj xj þ Bj uj
where, the cost function CFj and the vector yj and matrix Gj and also AjA xj þ BjA uj þ b
c jA ¼ 0; AnA xn þ b
c nA ¼ 0 (35)
constant yj are pertaining to the cost function of DERs utilized in AjB xj þ BjB uj þ b
c jB  0; AnB xn þ b
c nB  0
MG. Function fj can be written in matrix form as a function of state
By applying quadratic programming method to the system
and decision vectors as follows:
formulation, the model is divided into only linear and quadratic
2 3 2 3 terms. The Lagrangian multipliers can be applied to the model and
1 0 0 0 0 0 1 0 0 0 0
60 1 0 0 0 07 60 1 0 0 07 also slack variables can be added into the inequalities in constraints
6 7 6 7
  6 :: :: :: :: :: :: 7 6 :: 07 and a new formulation can be acquired as presented in Eq. (36). The
fj xj ; uj ¼ 6 7 xj þ ¼ 6 0 0 0 7uj
60 0 0 1 0 07 60 0 0 :: 07 goal is minimizing the obtained function at iteration k.
6 7 6 7
40 0 0 0 1 05 40 0 0 0 15
0 0 0 0 0 0 0 0 0 0 0 Xh
n1  . i
min Fqi ¼ yj þ ~xTj þ u~ Tj sj þ ~xTj Gj ~xj þ 2~xTj Lj u~ j þ u~ Tj Sj uj 2
(31) j
h   
Constraints in Eq. (28) can be rewritten in a matrix form as  AjA ~ ~ j þ cjA T a
xj þ BjA u ~ iA þ AjB ~ ~ j þ cjB T a
xj þ BjB u ~ jB
follow:
ih
  a ~ jB 2 ~
~ TjB Qj a xTn yn þ ~xTn Gn ~xn þ ðAnA ~ xn þ cnA ÞT a
~ nA
cjA xj ; uj ¼ ½ 1 1 :: :: 1 0 xj þ ½Pwind ðjÞ þ PPV ðjÞ  Pload ðjÞ i
(32) þ ðAnB ~ xn þ cnB ÞT a ~ TnB Q n a
~ nB  a ~ nB =2

(36)
234 H. Moradi et al. / Energy 147 (2018) 226e238

The matrices expressed in Eq. (37) are defined which need to be 2 3 2 3

T
solved as: H2j h5j
6 7 4 cj 5
2 3 F2j ¼4 Aj 5; f3j ¼ (47)
~1
u 2 3 2 3 2 3 c
a1A a1B ~z1B Ajþ1;n Aj jþ1;n
6 :: 7 ~
6
~s ¼ 4 7 ;z ¼ 4 :: a5 ~¼ 4 5
:: ; ~z ¼ 4 :: 5 (37)
~ n1 5
u Matrix F1j can be factored as:
a a ~z
~
x1 nA nB nB

F1j ¼ uj Sj uTj (48)

The method to solve these matrices at each iteration is given as
follows:
where uj is an orthogonal matrix of eigenvectors, and Sj can be
k k written as follows:
ykj ¼ yj þ Qj xkj þ Rj ukj þ ATjA ljA þ ATjB ljB
T k T k
(38)  
zkj ¼ zj þ RTj xkj þ k
Sj uj þ BjA ljA þ BjB ljB Sj ¼ diag a1 ; …; anjf ; jai j  jaiþ1 j; j ¼ 1; …; njf  1 (49)

and associated constraints can be rewritten as: Then,

 
ckjA ¼ AjA xkj þ BjA ukj þ b
c jA njr ¼ max maxfjjai =a1 j  1=cmax g; njf  n
 1 (50)
k
(39) njn ¼ njf  njr
ckjB ¼ AjB xkj þ BjB ukj þ b
c jB þ skjB þ diag ljB dik
where cmax is a positive scalar and
The matrix Gki can be defined as below:
   
 1   Sjr ¼ diag a1 ; …; anjr ; Sjn ¼ diag anjrþ1 ; …; anjf
Gkj ¼ diag lkjB diag skjB (40)
 (51)
uj ¼ ujr ; ujn ; ujr 2Rnjf njr ; ujn 2Rnjf njr
The above matrix can also be expressed as:
 " T
#

 Gjb 0 Pjb
Gkj ¼ Pjb Pjf (41) 6.1. Algorithm procedure
0 Gjf PjfT
Given ε a very small positive number, k ¼ 0 with initial condition
where the diagonal elements in the second matrix on the right-side ðz0 ; z0 ; l0 ; s0 Þ, where z0 > 0 and l0 > 0, the residual terms are
is sorted in ascending order. In addition, the number of columns in derived as below:
Pjb equals to the number of diagonal elements that less than one.
Using Eqs. (38)e(41), the following equations can be obtained:  2 X
n 1  . T  . 
Ri1 ¼ ~j
vFqi vu ~j
vFqi vu
yj ¼ y þ ATjB P j ckjB ; zj ¼ zkj þ BTjB P j ckjB ; P j ¼ Pjf G1 T
jf Pjf
j¼1
(42)  2 n 
X  
Q j ¼ Qj þ ATjB P j AjB ; Rj ¼ Rj þ ATjB P j BjB ; Sj ¼ Sj þ BTjB Pj BjB Ri2 ¼ AjA xij þ BjA uij þ b
c jA AjA xij þ BjA uij þ b
c jA
j¼1
where below matrices are derived:
 2 n 
X T  
Ri3 ¼ AjB xij þ BjB uij þ b
c jB þ zijB AjB xij þ BjB uij þ b
c jB þ zijB
" # " # " # j¼1
AjA BjA ckjA 3
Aj ¼ ; Bj ¼ ; cj ¼  2 n h
X  iT h  
P T AjB T
Pjb BjB T k
Pjb cjB Ri4 ¼ i i
min ajB ; zjB i i 5
min aiB ; ziB
" jb #  (43)
ljA 0 0 j¼1
lj ¼ ; Gj ¼
0 Gjb
ljb (52)
At the end of the procedure, the summation of all residential
 
ljB ¼ Pjb ljb þ Pjf G1 T k terms should approach near zero when solution set approaches on
jf Pjf AjB xj þ BjB uj þ cjB
     (44) optimal point. In other words, the optimum result can be obtained
sjB ¼  AjB xkj þ xj þ BjB ukj þ uj þ b
c jB þ skj by calculating Eq. (53), when Ri is close to zero. The following steps
are carried out in order to implement the algorithm:

and,
Step 1) Calculate the residual value defined as Ri:
H1j ¼ Q j þ ATj Wjþ1 Bj ; H2j ¼ Rj þ ATj Wjþ1 Bj rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2  2  2
H3j ¼ Sj þ BTj Wjþ1 Bj ; h4j ¼ yj þ ATj yjþ1 (45) Ri ¼ Ri1 þ Ri2 þ Ri3 þ Ri4 (53)
h5j ¼ zj þ BTj yjþ1

The matrices F1j, F2j and f3j can be extracted as: If Ri < ε, proceed to Step 5.
2 3 Step 2) Solve the affine scaling predictor direction ð~sa ; ~ za ; a~ a ; ~za Þ.
T
H3j Bj BTj ATjþ1;n Step 3) Solve the Combined Predictor-Centering-Corrector
6 7 (CPCC) direction ð~sc ; ~ zc ; a~ c ; ~zc Þ.
F1j ¼4 Bj Gj 0 5 (46)
Ajþ1;n Bj 0 Bjþ1;n Step 4) Calculate the matrixes presented in (37) for the next
iteration ðsiþ1 ; ziþ1 ; aiþ1 ; ziþ1 Þ and proceed to Step 1.
Step 5) Calculate the optimal solutions and end.
 H. Moradi et al. / Energy 147 (2018) 226e238 235

The procedure of optimization algorithm and aforementioned

steps are demonstrated as a flowchart format in Fig. 9.

7. Simulation results

In order to carry out optimal scheduling of a MG to fulfill a load

demand varying between around 30 kW and 110 kW in 24 h, as
shown in Fig. 8, a developed optimization scheme was applied.
First, the available power from the PV and the wind generator was
used and the remaining loads were satisfied by the DG, MT, and FC
in policy 1. In addition, an ESS was added to the available resources
in policy 2 for better support and flexibility. The produced power
from the PV and wind turbine using Eqs. (5)e(9) is shown in Fig. 10.
The PV system agents were in “active” mode during the solar ra-
diation availability and stayed “inactive” for the rest of the day with
Fig. 10. PV and wind turbine output power.

no solar radiation. In contrast, the 15 kW wind generator mode is

“active” throughout the day. Based on the results, the net produced
energies from both PV and wind systems are very close and the PV
had higher output in the middle of the day due to higher solar
radiation, while the wind generator had a uniform production
during the day. Furthermore, a comparison between the pre-
liminary load demand and the remaining load that has not been
supported by the PV and wind generator is shown in Fig. 11. It can
be observed from Fig. 11 that the difference between two graphs is
higher in midday when both the PV and wind resources had power
generation. Optimization techniques are utilized in order to opti-
mally distribute the residual load among the DG, MT, and FC units
and also the ESS in the second policy in order to minimize the cost
function presented in Section 3.

7.1. Policy 1 results

The optimal output power of the system generated from several

resources during a normal operation throughout a day is shown in
Fig. 12. It is observed from the results that the load demand is
satisfied with three available generating units and the DG is the
least desirable production option since it has the highest operating
and emission costs. Due to the low operating cost of the MT, it
supplies a significant part of the load and it operates near its rated
power for 15 h in a day. The FC also works for 13 h in a day and it is
considered another appropriate source of supplying energy. Fig. 13
shows each component's contribution in total power generation on
a daily basis.

Fig. 11. Initial load demand vs. residual load that has not been fulfilled by the PV and
Fig. 9. Algorithm flowchart. wind turbine.
236 H. Moradi et al. / Energy 147 (2018) 226e238

Fig. 12. Energy scheduling in island mode in policy 1.

Fig. 14. Energy scheduling in island mode in policy 2.

By comparing the simulation results, it can be observed that

mainly the MT and FC met the load demand throughout the day and
the required power was mostly generated by the FC in two smaller
load peaks around 8 a.m. and 3 p.m. While in the evening, when the
load was at its highest peak, both the FC and MT operated at their
nominated powers and the DG was turned on to supply the residual
power consumption.

7.2. Policy 2 results

In this scenario, battery storage was added to the microsources

available in the MG in order to help the energy production sector to
be more flexible. The optimal power production from multi-
resources during the day is presented in Fig. 14. Excluding the DG
unit, other generation systems' agents are in “active” mode
throughout the day. It can be seen in Fig. 14 that whenever ESS is in
discharging mode (supplying power) the output of either the MG or
FC or both will decrease. In contrast, when ESS is in charging mode
(demanding power) the output of the MG or FC or both will in-
crease to cover the load and charge the battery. As aforementioned
Fig. 15. Battery charging and discharging.
in the policy 1 result section, the DG is turned on in the evening for
a limited period of time to supply the load residual since the gen-
eration of the MG, FC and ESS is insufficient to supply the whole power output varies between 10 kW and 14 kW. In addition,
load demand. The charging and discharging status of the ESS is Fig. 16 shows the battery SOC, which is limited between the range
indicated in Fig. 15. Based on the optimum results, the battery 30%e95% due to the battery SOC constraints previously mentioned.
storage unit is operated in charging and discharging modes and its

Fig. 13. Optimal power allocation of DERs in policy 1. Fig. 16. Battery SOC.
 H. Moradi et al. / Energy 147 (2018) 226e238
Fig. 17. Optimal power allocation of the DERs in policy 2.
From the obtained results, it can be seen that the battery capacity
was large enough to supply the load for the studied period of time.
It can be concluded that when the load is low the ESS stores energy,
most of which comes from the renewable energy resources. When
the load demand is high, the ESS sends energy back to the MG and
thus lowers the costs of dispatchable resources.
The generation contribution of each unit to the total production
to meet the load demand in policy 2 is shown in Fig. 17. It can be
concluded from Fig. 17 that the FC and MG are competing
throughout the day to fulfill the load except in the evening when
the peak load occurs and both units must operate at their
maximum outputs. Not only is the power production portion pro-
duced by the DG in policy 2 is less than in policy 1 since some part
of the load is covered by ESS, it also leads to a significant reduction
in the system's total cost. A comparison between total costs carried
out for the simultaneous minimization of operational costs and
emissions as two consistent functions for both policies on a daily
basis is shown in Fig. 18.
It can be seen from the financial results in Fig. 18 that the overall
cost of the system in policy 2 is less than in policy 1 when there is
no access to a ESS unit. The difference of the two graphs is directly
related to the battery functionality throughout the day. The results
indicate that at load peak times from 6 p.m. to 10 p.m. the ESS
supplies some part of the load and it leads to a decrease in system
total cost. The MG operational and pollution costs vary between $11
Fig. 18. System total cost in policy 1 vs. policy 2.
237
and $38 in a day in policy 1 and varies between $3 and $31 in a day
in policy 2. Also, it can be concluded from Figs. 15e18 that the
system cost in policy 2 increases when ESS is in its off peak time (at
4 a.m. and at 2 p.m.) and decreases when it is in its peak time (at 10
a.m.).
8. Conclusions
In this paper, a novel optimization model was proposed in order
to simultaneously minimize MG's operational, maintenance and
pollution cost. In addition, a dynamic programming method with
discrete-time optimal control constraints was introduced and
implemented to evaluate the system performance. The simulation
results have shown the effectiveness as well as timely convergence
of the proposed optimization power management system. Two
strategies of power scheduling in a standalone MG were considered
in this research. In order to implement more precise modeling, a
probabilistic programming technique was employed to simulate
the stochastic nature of the solar and wind generator power gen-
eration as well as the local load demand. The results show that the
presence of ESS in the power generation sector can assist the MG
operator to reduce production costs, improve energy utilization
efficiency, and maximize profits. The MG total energy cost is
calculated as $352.10 in policy 1 and as $322.04 in policy 2 per day,
which brings 8.5% cost savings for the entire system. Although the
system responses are affected by several variables including
weather conditions, emissions, operation, and maintenance costs
and, of course, the actual power demand, the proposed optimiza-
tion method can be updated and adjusted accordingly. It is clearly
observed from the achieved results that the functionality of a MG's
optimizer has a satisfactory performance in terms of minimizing
operation, maintenance, and emission costs. The MG optimizer unit
is able to deliver the optimal power flow in the entire system by
taking into account the cost function of each generating unit.
References
[1] Wang L, Li Q, Ding R, Sun M, Wang G. Integrated scheduling of energy supply
and demand in microgrids under uncertainty: a robust multi-objective opti-
mization approach. Energy 2017;130:1e14.
[2] Chen Y, He L, Li J. Stochastic dominant-subordinate-interactive scheduling
optimization for interconnected microgrids with considering wind-
photovoltaic-based distributed generations under uncertainty. Energy
2017;130:581e98.
[3] Mariam L, Basu M, Conlon MF. A review of existing microgrid architectures.
J Eng 2013;2013.
[4] Zhang F, Zhao H, Hong M. Operation of networked microgrids in a distribution
system. CSEE J Power Energy Syst 2015;1:12e21.
[5] Moradi H, Abtahi A, Esfahanian M. Optimal operation of a multi-source
microgrid to achieve cost and emission targets. In: 2016 IEEE power energy
conf. Illinois. PECI 2016; 2016.
[6] Prinsloo G, Mammoli A, Dobson R. Customer domain supply and load coor-
dination: a case for smart villages and transactive control in rural off-grid
microgrids. Energy 2017;135:430e41.
[7] Vasquez JC, Guerrero JM, Miret J, Castilla M, De Vicun ~ a LG. Hierarchical control
of intelligent microgrids. IEEE Ind Electron Mag 2010:23e9.
[8] Athari MH, Ardehali MM. Operational performance of energy storage as
function of electricity prices for on-grid hybrid renewable energy system by
optimized fuzzy logic controller. Renew Energy 2016;85:890e902.
[9] Jin D, Li Z, Hannon C, Chen C, Wang J, Shahidehpour M, et al. Toward a cyber
resilient and secure microgrid using software-defined networking. IEEE Trans
Smart Grid 2017;8:2494e504.
[10] Sadeghian HR, Ardehali MM. A novel approach for optimal economic dispatch
scheduling of integrated combined heat and power systems for maximum
economic profit and minimum environmental emissions based on Benders
decomposition. Energy 2016;102:10e23.
[11] Peerapong P, Limmeechokchai B. Optimal electricity development by
increasing solar resources in diesel-based micro grid of island society in
Thailand. Energy Rep 2017;3:1e13.
[12] Bifaretti S, Cordiner S, Mulone V, Rocco V, Rossi JL, Spagnolo F. Grid-connected
microgrids to support renewable energy sources penetration. Energy Procedia
2017;105:2910e5.
[13] Shaban Boloukat MH, Akbari Foroud A. Stochastic-based resource expansion
238 H. Moradi et al. / Energy 147 (2018) 226e238
planning for a grid-connected microgrid using interval linear programming.
Energy 2016;113:776e87.
[14] Li B, Roche R, Miraoui A. Sizing of a stand-alone microgrid considering electric
power, cooling/heating and hydrogen. In: 2017 IEEE Manchester powertech.
IEEE; 2017. p. 1e6.
[15] Ahmad Khan A, Naeem M, Iqbal M, Qaisar S, Anpalagan A. A compendium of
optimization objectives, constraints, tools and algorithms for energy man-
agement in microgrids. Renew Sustain Energy Rev 2016;58:1664e83.
[16] Zhao B, Zhang X, Chen J, Wang C, Guo L. Operation optimization of standalone
microgrids considering lifetime characteristics of battery energy storage
system. IEEE Trans Software Eng 2013;4:934e43.
[17] Marzband M, Yousefnejad E, Sumper A, Domínguez-García JL. Real time
experimental implementation of optimum energy management system in
standalone Microgrid by using multi-layer ant colony optimization. Int J Electr
Power Energy Syst 2016;75:265e74.
[18] Basir Khan MR, Jidin R, Pasupuleti J. Multi-agent based distributed control
architecture for microgrid energy management and optimization. Energy
Convers Manag 2016;112:288e307.
[19] Abedini M, Moradi MH, Hosseinian SM. Optimal management of microgrids
including renewable energy sources using GPSO-GM algorithm. Renew En-
ergy 2016;90:430e9.
[20] Li H, Zang C, Zeng P, Yu H, Li Z. A stochastic programming strategy in
microgrid cyber physical energy system for energy optimal operation. IEEE/
CAA J Autom Sin 2015;2:296e303.
[21] Hu M-C, Lu S-Y, Chen Y-H. Stochastic programming and market equilibrium
analysis of microgrids energy management systems. Energy 2016;113:
662e70.
[22] Shaban Boloukat MH, Akbari Foroud A. Stochastic-based resource expansion
planning for a grid-connected microgrid using interval linear programming.
Energy 2016;113:776e87.
[23] Niknam T, Golestaneh F, Malekpour A. Probabilistic energy and operation
management of a microgrid containing wind/photovoltaic/fuel cell generation
and energy storage devices based on point estimate method and self-adaptive
gravitational search algorithm. Energy 2012;43:427e37.
[24] Dohrmann CR, Robinett ARD. Dynamic programming method for constrained
discrete-time optimal control 1. J Optim Theor Appl 1999;101:259e83.
[25] Padayattil GM, Thobias T, Sebastian J, Thomas M, Pathirikkat G. Hybrid ring
microgrid with coordinated energy management scheme. Procedia Technol
2016;25:793e800.
[26] Kanchev H, Lu D, Colas F, Lazarov V, Francois B. Energy management and
operational planning of a microgrid with a PV-based active generator for
smart grid applications. IEEE Trans Ind Electron 2011;58:4583e92.
[27] Almada JB, Le~ ao RPS, Sampaio RF, Barroso GC. A centralized and heuristic
approach for energy management of an AC microgrid. Renew Sustain Energy
Rev 2016;60:1396e404.
View publication stats
[28] Theubou Tameghe TA, Wamkeue R, Kamwa I. Diesel generator modelling for
microgrid power plant parameters assessment. In: EIC clim. chang. technol.
conf., Montre al, QC, Canada; 2015. p. 1e12.
[29] Wood AJ, Wollenberg BF. Power generation, operation, and control. second ed.
New York, NY: John Wiley & Sons, Inc; 1996.
[30] Martín JIS, Zamora I, Martín JJS, Aperribay V, Eguia P. Hybrid fuel cells tech-
nologies for electrical microgrids. Elec Power Syst Res 2010;80:993e1005.
[31] Jun Jun L. Modeling and simulation of micro gas turbine generation system for
grid connected operation. In: Asia-Pacific power energy eng. conf. APPEEC;
2010.
[32] Wang J, Li X, Yang H, Kong S. Design and realization of microgrid composing of
photovoltaic and energy storage system. Energy Procedia 2011;12:1008e14.
[33] Aghajani GR, Shayanfar HA, Shayeghi H. Demand side management in a smart
micro-grid in the presence of renewable generation and demand response.
Energy 2017;126:622e37.
[34] Shin J, Lee JH, Realff MJ. Operational planning and optimal sizing of microgrid
considering multi-scale wind uncertainty. Appl Energy 2017;195:616e33.
[35] Craparo E, Karatas M, Singham DI. A robust optimization approach to hybrid
microgrid operation using ensemble weather forecasts. Appl Energy
2017;201:135e47.
[36] Atwa YM, El-Saadany EF, Salama MMA, Seethapathy R. Optimal renewable
resources mix for distribution system energy loss minimization. Power Syst
IEEE Trans 2010;25:360e70.
[37] Jabbari-Sabet R, Moghaddas-Tafreshi S-M, Mirhoseini S-S. Microgrid opera-
tion and management using probabilistic reconfiguration and unit commit-
ment. Int J Electr Power Energy Syst 2016;75:328e36.
[38] Soroudi A, Aien M, Ehsan M. A probabilistic modeling of photovoltaic modules
and wind power generation impact on distribution networks. IEEE Syst J
2012;6:254e9.
[39] Moradi H, Esfahanian M, Abtahi A, Zilouchian A. Modeling a hybrid microgrid
using probabilistic reconfiguration under system uncertainties. Energies
2017;10:1430.
[40] Mohamed FA, Koivo HN. System modelling and online optimal management
of MicroGrid using mesh adaptive Direct search. Int J Electr Power Energy Syst
2010;32:398e407.
[41] Serban I, Marinescu C. Battery energy storage system for frequency support in
microgrids and with enhanced control features for uninterruptible supply of
local loads. Int J Electr Power Energy Syst 2014;54:432e41.
[42] Fedjaev J, Amamra SA, Francois B. Linear programming based optimization
tool for day ahead energy management of a lithium-ion battery for an in-
dustrial microgrid. In: Proc. - 2016 IEEE int. power electron. motion control
conf. PEMC 2016; 2016. p. 406e11.
[43] Guo L, Liu W, Li X, Liu Y, Jiao B, Wang W, et al. Energy management system for
stand-alone wind-powered-Desalination microgrid. IEEE Trans Smart Grid
2016;7:1079e87.