Energy 117 (2016) 176e189

Contents lists available at ScienceDirect

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

Optimal coordinated scheduling of combined heat and power fuel cell,

wind, and photovoltaic units in micro grids considering uncertainties
Mosayeb Bornapour, Rahmat-Allah Hooshmand*, Amin Khodabakhshian,
Moein Parastegari
Department of Electrical Engineering, University of Isfahan, Isfahan, Iran

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 19 May 2016
Received in revised form
14 October 2016
Accepted 19 October 2016

In this paper, a stochastic model is proposed for coordinated scheduling of combined heat and power
units in micro grid considering wind turbine and photovoltaic units. Uncertainties of electrical market
price; the speed of wind and solar radiation are considered using a scenario-based method. In the
method, scenarios are generated using roulette wheel mechanism based on probability distribution
functions of input random variables. Using this method, the probabilistic specics of the problem are
distributed and the problem is converted to a deterministic one. The type of the objective function,
coordinated scheduling of combined heat and power, wind turbine, and photovoltaic units change this
problem to a mixed integer nonlinear one. Therefore to solve this problem modied particle swarm
optimization algorithm is employed. The mentioned uncertainties lead to an increase in prot. Moreover,
the optimal coordinated scheduling of renewable energy resources and thermal units in micro grids
increase the total prot. In order to evaluate the performance of the proposed method, its performance is
executed on modied 33 bus distributed system as a micro grid.
2016 Elsevier Ltd. All rights reserved.

Keywords:
Micro grid
Optimal scheduling
Deregulated electricity market
Wind turbine
Photovoltaic unit
Combined heat and power

1. Introduction
Nowadays, the majority of the electrical energy is supplied by
nuclear or large fossil fuel power plants around the world and they
have high reliability and capacity. However, regarding the fact that
the investment cost of these power plants is very high, in order to
draw the attention of small capitals and reducing environmental
emissions, Renewable Energy Sources (RESs) at small scales attract
a higher interest [1]. On the other hand, increasing the use of
renewable energy and Distributed Generation (DG) sources of high
efciency in power systems has provided new points of view. This
new viewpoint, i.e. generation units generating power in a smaller
scale, is close to consumers in the Micro Grid (MG). MGs comprise
distribution networks with distributed energy resources: i.e. Wind

List of abbreviations: CHP, Combined Heat and Power; MG, Micro Grid; WT,
Wind Turbine; PV, Photovoltaic; MPSO, Modied Particle Swarm Optimization algorithm; DG, Distributed Generation; TH, Thermal unit; FC, Fuel Cell; RES,
Renewable Energy Source.
* Corresponding author.
E-mail addresses: mbornapour@eng.ui.ac.ir (M. Bornapour), hooshmand_r@eng.
ui.ac.ir (R.-A. Hooshmand), aminkh@eng.ui.ac.ir (A. Khodabakhshian), parastegari@
eng.ui.ac.ir (M. Parastegari).
http://dx.doi.org/10.1016/j.energy.2016.10.072
0360-5442/ 2016 Elsevier Ltd. All rights reserved.

Turbine (WT), Fuel Cells (FCs) based Combined Heat and Power
(CHP), Photovoltaic (PV), etc. [2,3]. MGs can be managed in a nonautonomous mode, if interconnected to the main grid, or in an
autonomous mode, if disconnected from the main grid [4]. The
operation of DGs in the network can provide different benets for
the overall network performance, if optimally scheduled and coordinated [4]. One of the most important issues in MGs is optimal
operation in deregulated markets by determining the best value
generation of each unit to achieve the highest prot. In this regard,
distributed generators that are capable of CHP production are
increasingly used in MGs as well as in RESs such as solar and wind
energy sources which have stochastic nature [5]. Since, the generation of RESs is uncertain, their generation should be forecasted; on
the other hand, the uncertainties of load and RESs are considered in
Ref. [6].
Many studies have been carried out on the performance of MG
and prot maximization objective function [7e18]. The pool electricity market of MG for PV and FC units has also been investigated
in Ref. [7]. In this study, the prot maximization objective function
is considered and genetic optimization algorithm is used to solve
the problem. In Ref. [8], the optimal operation and participation of
MGs in electricity market are proposed by considering energy
sources such as WT, PV, and FC, while adopting prot-maximizing

M. Bornapour et al. / Energy 117 (2016) 176e189

Nomenclature:
CBuy,t,s
cost of buying electrical energy in sth scenario
cvf
coefcient of variation
0
DTs;s0
the distance between scenarios s and s
fTH,h,fFCCHP,l,fWind,j, fPV,q cost functions related to thermal, CHP,
WT, PV units respectively
HBoiler,l,t,s the heat generated by the boiler can be parallel CHP
units in sth scenario
HCHP,l,t,s heat generated by CHP units in sth scenario
HDemand,t,s thermal load demand in sth scenario
Loadt,s
electrical load in time t in sth scenario
Nbus
total number of buses
NFC
total number of FCs
Ns
number of scenarios of each batch
PBuy,t,s
amount of electrical power bought from market in sth
scenario
PenaltyMarket, PenaltyDemand penalty factor for market and
demand
t
PFCj;s
active power generated by FCj during time t in sth
scenario
t
PHj;s
equivalent electric power for hydrogen production
during time t in sth scenario
PLRj
part load ratio of FCj (equal to electrical generated
power/maximum power)
PMax,FC maximum power of FC
min
max
PUnit;i;t;s
, PUnit;i;t;s
minimum and maximum of active power
produced by Uniti during time t in sth scenario
Pnot-supplied,Market,t,s not-supplied power on the market at the
time t in sth scenario
PSell,t,s,PH,Sell,t,s electrical power and heat energy sold to market in
sth scenario

strategy. In this article, uncertainties related to the market price

and electrical load are also considered. In Ref. [9], the problem of
coordination between PV and electrical energy storage batteries is
proposed and their participation in the electricity market is solved.
The energy generated by WT and PV units as well as the energy
stored by means of hydrogen and their participation in short-term
electricity market are investigated in Ref. [10]. In this study, the
day-ahead electricity market has been simulated and the uncertainty of WT production is considered to reach the maximum prot.
In Refs. [11,12], the electricity market for distribution companies
with DG units and interruptible loads has been solved. Prot
maximization in both the energy and reserve markets is considered
as the objective function. The coordinated scheduling of WT, PV,
energy storage, and pumped storage to reach the maximum prot
in the power market and ancillary services have been investigated
in Refs. [13,14]. The uncertainties of WT, PV generation, and market
price are considered. In Ref. [15], the problem of coordinated
schedule of WT and thermal units (TH) to participate in the energy
and reserve market has been modeled by multi-objective optimization problem. The objective function of this problem is the prot
maximization and emission minimization. Coordinated scheduling
of PV and energy storage device units with CHP units in an industrial MG is described in Ref. [16]. In this paper, the prot maximization objective function is considered. In Ref. [17], the
participation of MG in the day-ahead electricity market with the
goal of maximizing prots has been studied and the scenario-based

177

t
Pthi;s

produced thermal by FC in bus i if there is a FCPP in this

bus during time t in sth scenario
PTH,h,t,s, PFCCHP,l,t,s, PWT,j,t,s, PPV,q,t,s generated power related to
thermal, CHP, WT, PV units in
sth scenario respectively
VFC
output voltage of FC
Vit
voltage magnitude of the bus i during time t
VMin,VMax lowest and highest voltages of each bus
Vref
nominal voltage
k
Xmut;j
mutant vector of particle j at iteration k
k
Xnew;j
the trial vector
rand (.) random number between 0 and 1
rTEj
thermal energy to electrical energy ratio
Smax
maximum power exchanged with the upstream
exch
network
T
the total time
Vcon
over voltage due to concentration
Velkj , Velk1
current velocity of particle j at iterationk and k1
j
Xjk
current position of particle j at iteration k
rs
probability of each scenario
hact
activation polarization
htj
efciency of FCj during time t
hohm
ohmic polarization
hst
hydrogen storage efciency
ppenalti,Market,t,s penalty factor for failing to meet the power of
market at the time t in sth scenario
psell,t,s
tariff of electrical energy selling to market in sth
scenario
pTariff,s,pH,sell,t,s tariff of electrical and heat energy selling to
customer in sth scenario
mf
mean value of the output variable f
sf
standard deviation value of the output variable f

method is used for modeling the uncertainties. In Ref. [18], MG was

scheduled for participation in day-ahead electricity market in a way
to reach the maximum prot and minimum cost while considering
heat consumers. Moreover, CHP plants have been investigated from
diverse and different mechanical engineering points of view such
as energy balance, fueled microturbine, and thermo-economic
analysis [19e21].
The performance of MG and prot maximization objective
function is studied in Refs. [7e18]. Furthermore, many different
types of DGs such as WT unit [7,8,10,13e15]; PV unit [7e10,13,14
and 16]; FC unit [7,8 and 10]; CHP unit [16,18]; and some DGs
[11,12,17 and 18] have been applied to MGs. The uncertainties of
uncertain variables of MGs have been considered in Refs. [8,10,13,14
and 17] by different methods. In the above-mentioned studies, the
coordinated scheduling of WT, PV, TH, and FC-CHP units are not
simultaneously considered for participation in an electricity market. Furthermore, in some of the studies mentioned above, the
essential uncertainties of MGs such as electrical market price, the
speed of wind, and solar radiation are not considered.
In this paper, a scenario-based stochastic programming
framework is used for the coordinated scheduling of CHP, PV, WT,
and TH DGs by considering uncertainties of wind speed, solar
radiation, and the electrical market price. The normal probability
density function is used for modeling the variation of electrical
and thermal loads forecasting, solar radiation, and the electrical
market price. The Weibull distribution function is used for

178

M. Bornapour et al. / Energy 117 (2016) 176e189

modeling the variation of wind speed. Since the proposed problem

is a mixed integer nonlinear problem, an evolutionary algorithm
named Modied Particle Swarm Optimization (MPSO) is used to
solve this problem.
The contributions of this paper are as follows: (i) coordinated
scheduling of CHP units in MG; (ii) coordinated scheduling among
CHP, WT, and, PV units; (iii) the coordinated scheduling problem
planned by a static model; (iv) the uncertainty of electrical market
price taken into consideration by a scenario based method; (v)
uncertainties of wind speed and solar radiation; (vi) the computation time reduced by a backward scenario reduction method; (vii)
the performance and robustness of this algorithm improved by the
proposed technique; and (viii) the favorable and benecial performance of the proposed method conrmed by the simulation
results.

Costs

24
X

CBuy;t;s  PBuy;t;s

t1

N
CHP
X

NTH
24 X
X

24

 X
fTH;h PTH;h;t;s

t1 h1

t1



fFCCHP;l PFCCHP;l;t;s ; HFCCHP;l;t;s ; HBoiler;l;t;s

l1

24 NX
Wind
X

NPV
24 X

 X


fWT;j PWT;j;t;s
fPV;q PPV;q;t;s

t1 j1

t1 q1

PenaltyMarket s PenaltyDemand s
(5)
where each term of cost function is calculated as follows:



fTH;h PTH;h;t;s CTH;s $PTH;h;t;s

(6)



fFCCHP;l PFCCHP;l;t;s ; HFCCHP;l;t;s ; HBoiler;l;t;s

2. Problem formulation

CFCCHP;s $PCHP;l;t;s CBoiler;s $HBoiler;l;t;s

In this paper, the optimal coordinated scheduling of CHP units in

MG considering WT and PV units are proposed. In this section, the
market prot objective function and the related constrains are
determined.



fWT;j PWT;j;t;s CWT;s $PWT;j;t;s

(8)

2.1. Decision variables



fPV;q PPV;q;t;s CPV;s $PPV;q;t;s

(9)

X is the vector of all decision variables of the problem. This

vector includes the sub-vectors of the generated active power of
each FC units in sth scenario (PFC;NFC ;t;s ) and the generated active
power of each TH units in sth scenario (PTH;NTH ;t;s ). By the selection
of optimal values of the variables in the search space, the applied
optimization algorithm will nd the optimal solution to the problem. The vector X is presented as follows:

:::

X PFC;1;t;s

PFC;NFC ;t;s

PTH;1;t;s

:::

PTH;NTH ;t;s

PenaltyMarket s

24
X

(7)

Pnotsupplied;Market;t;s  ppenalti;Market;t;s

t1

(10)
(

ppenalti;Market;t;s

1 b$pSell;t;s
1  b$pSell;t;s

~
PSell;t;s > P
Sell;t;s
~
PSell;t;s < P
Sell;t;s

(11)

NTH NFC T

(1)
2.3. Constraints
2.2. Objective function of market prot
Maximizing prots (expected prot-EPF) for the participation of
an MG in the electrical market is considered as the objective
function of this problem. EPF is dened as follows:

MaxEPF

rs $PFs

(2)

s2S

 Demand and generation equality

PBuy;t;s

NX
Thermal

PTH;h;t;s

h1

PFs Max$Revenues  Costs

(3)

The proposed revenue function includes three terms which are

the selling of electrical energy to customers and the market, and
selling heat energy to customers. Eq. (4) represents these three
terms:

Revenues

The constraints of above mentioned optimization problem are

as follows:

24
X

pTariff ;s  Loadt;s

t1

24
X

24
X

psell;t;s  PSell;t;s

NPV
X

N
CHP
X

PCHP;l;t;s

l1

NX
Wind

PWind;j;t;s

j1

PPV;q;t;s

q1

Loadt;s Losst;s

ct 1 : 24

(12)

 Supply of thermal load

N
CHP 
X


HCHP;l;t;s HBoiler;l;t;s HDemand;t;s

t1

(13)

l1

pH;sell;t;s  PH;Sell;t;s

(4)

t1

The proposed cost function includes seven terms which are the
costs of buying electrical energy from market; costs of generated
electrical energy by TH, FC-CHP, WT, and PV units; and penalties for
failure to supply the market and customers. The cost term of prot
function is as follows:

 The limits of bus voltage

 
VMin  Vit   VMax
 Output active power of DG units

(14)

M. Bornapour et al. / Energy 117 (2016) 176e189

min
max
PUnit;i;t;s
 PUnit;i;t;s It  PUnit;i;t;s

ci2SUnit

t 1 : 24

(15)

 Steady state security constraints of MG

Sij;t;s  Smax
ij
Et;s  Smax
exch

ci  j2Sb

t 1 : 24

ct 1 : 24

(16)
(17)

179

3.1.2. Heat generation of FC

Generating both electrical and thermal energy concurrently by
FCs increases the efciencies to the desirable ratio. Also, here in this
paper, FCs are used for the generation of electrical and thermal
energies simultaneously. Fig. 1 depicts the efciency and ratio of
thermal to electrical energy with varying art load ratio (PLR). These
curves can be tted by the mathematical expression in Ref. [23].
The following relation efciency and the ratio of electrical energy to
thermal energy depend on part load ratio (PLR) as follows [22]:

PLRj
3. Steady state model for RES in micro grid

For

t
PFCj

PMax;FC

PLRj < 0:05

(20)

hFCj 0:2716 rTEj 0:6801

3.1. Fuel cell based combined heat and power (FCCHP)

3.1.1. FC model
Fuel Cell (FC) as the name suggests, is one of the latest generator
models which generates either electrical energy or thermal energy

For

where PLR is the part load ratio of FC (equal to electrical generated

power/maximum power).
The produced thermal energy in FCs depends on the electrical

PLR > 0:05

hFCj 0:9033PLR5  2:999PLR4 3:6503PLR3  2:0794PLR2 0:4623PLR 0:3747

(21)

rTEj 1:0785PLR4j  1:9739PLR3j 1:5005PLR2j  0:2817PLRj 0:6938

by converting chemical energy. That's why they are mentioned as

the electro-thermal devices. This chemical energy is the consequence of the combination of two fed gases, i.e. oxygen and
hydrogen with the efcient production of water, heat, and the
required electricity. The produced electricity and heat can be used
as the power supply for the distribution system. The generated
voltage range of an FC is just about 0.5e0.9 V; therefore, by
adjoining specied number of single cells in series, the resultant FC
stack can be applied to generation systems.
The producing voltage of a single cell can be obtained by the
following equation [22]:



t
t
t
Pthi
rTEj  PFCj
PHj

(22)

t
t
PMax;FC PFCj
PHj

(23)

VFC ENernst  hact  hohm  Vcon

htoverall

(18)

energy produced according to the following equation [22]:

Without considering FCs as CHP, their efciency is approximately in the range of 30e40%. Employing FCs as CHP noticeably
improves their efciency. The overall efciency of FCs could be
calculated as follows [22]:



t h  P t min P t ; Lt
PFC
st
H
th th

(24)

t
PFC
PHt

htFCj

Of which ENernst is the thermodynamic potential of the cell (Volt)

and represents the reversible voltage. The related equations were
previously described [22]. Considering the unity power factor, the
output active power of a fuel cell is determined as follows:

3.2. Wind turbines (WT)

PFC iFC  VFC

(19)
The electrical energy generated by variable speed wind turbine
is done in two steps: rst, a synchronous or a double-fed generator
applied to their structure converts the kinetic and mechanical

Pr

Vc i
Fig. 1. Performance cure of the FC.

Vr

Vc o

Fig. 2. Nonlinear function of power of wind turbine.

180

M. Bornapour et al. / Energy 117 (2016) 176e189

energy of the wind into electrical energy. Next, a power electronic

device conforms the electricity generated in the rst part with the
AC network power as shown in Fig. 2 [24]. The following relation
represents the power output of WT according to the speed of wind
[24]:
t
PWT

8
< 0 
j wt
: max
Pw

wt  vci orwt  vco

vci  wt  vr
t 1; 2; :::; T
vr  wt  vco

(25)

Eq. (25) is not used for calculating because it is non-linear.

Therefore, according to the linear approximation of this equation
in small pieces, Eq. (26) can be replaced. Here, the relation between
the varieties of wind speeds and wind power is observable.

PtWT

8
r wt  vci ; vci  w  v1
>
>
> 1
>
r v  vci r2 wt  v1 ; v1  w  v2
>
>
< 1 1
r1 v
 1t vci  r2 v2  v1
Pmax

w
>
w  v2 ; v2  w  vr
r
3
>
>
>
>
1 vr  w  vco
>
:
0 otherwise
(26)

In Eq. (26) vci4 m/s, vr14 m/s, vco25 m/s, v17 m/s, v212 m/
s, r10.2/(v1vci), r2(0.960.2)/(v2v1), r3(10.96)/(vrv2).
Also, ri and vi refer to the slope and the break-point of the ith part of
the wind power generation curve, respectively [24]. Alsovci, vco and
vr are the cut-in, cut-off, and rated speed of the wind turbine,
respectively. For characterizing the proper position of the wind
turbines, it is necessary to predict the wind speed hour by hour in
the next day. By using Eq. (26), the wind power can be obtained
from the estimated information. These results can be used as the
upper bond of the associated decision variable.
3.3. Photovoltaic systems (PV)
PV is arrays of cells containing a material that converts solar
radiation into direct current electricity. PV electric power generation is a promising clean technology with enormous potential [25].
Orbiting satellites were the rst practical application of PVs, but
nowadays it gains much attention to be used as a source of power in
micro grids. There are three available factors inuencing the generation of electric power by PV module including, i.e. solar radiance,
ambient temperature, and module characteristics. The following
relation expresses the conversion function of solar radiance to
electric power [26]:

PPV

 t
8
G
>
>
P
>
sn
<
Gstd Rc

t
>
>
>
: Psn G
Gstd

0 < Gt < Rc
(27)
Gt > Rc

where G is the forecast solar radiation, Gstd is the solar radiation in

the standard radiation (1000W/m2), Psn is the nominal power for
the PV systems, and Rc is the certain radiation point equal to 150W/
m2.

4. Stochastic model description

Monte Carlo is the most accurate method for stochastic programming [27]. Monte Carlo method tries to solve the problem by
considering the involved uncertainties through random sampling
of the distribution function of random variables. The very heavy
computational burden is the main bottleneck of this method.
Monte Carlo simulation accuracy depends on the selected number
of examples of distribution functions. The more the number of
samples increases, the higher the accuracy and computational
burden will be. In this way, by specifying the input distribution
functions, the behavior of the outputs will be achieved. In solving
the stochastic programming, the problem should be divided into a
number of deterministic problems. Thus, the answer to stochastic
problems will be achieved by this approach using the results of
deterministic problems. In the scenario-based method, a scenario
refers to a deterministic problem. Scenario method can be used for
modeling uncertainties [28]. The following items describe the
scenario generation and reduction method.
4.1. Scenario generation
The following steps describe how to generate scenario [28]:
Step 1: As is visible in Fig. 3, the continuous distribution functions
should be discrete and proper occurrence probabilities
should be selected for them, i.e. each of the last operations
brings about errors and it is necessary to specify them.
Step 2: According to the resultant equation of the previous step, a
cumulative distribution function should be computed so
that by analyzing Figs. 3 and 4 it should be inferable that
each piece of the graphs is respondent to the other. These
pieces are organized by chance. As a problem may have
several uncertain variables, this operation should be
repeated for all of them.
Step 3: The process of generating scenarios for each uncertain
variable in this step is almost like the operation of roulette
wheels. The purpose of operating this method is to determine the percent of errors according to each of the uncertain variables and their occurrence probabilities. The above
method should be repeated for all the uncertain variables
one by one and independently.
Step 4: In this step, new variables should be determined for each
scenario. For this purpose, new values are computed for
uncertain variables and the error of each variable should be

Probability
Density
Level 1

3.4. Thermal units

4, 5

A thermal generation unit converts heat energy to electric power. In this paper, an important producer of electricity in the MG is
considered alongside other DG units. Micro gas turbines that can be
synchronized with network are used as thermal units in this paper.
Relations related to its cost and revenue (prot objective function)
are brought up in Section 2.

6, 7

Level 2

Level 3

2, 3

Leve 7

Level 4

Level 5

-3

Leve 6

-2

Forecast Error
Fig. 3. Typical discretization of the probability distribution functions of the load level
forecast error.

M. Bornapour et al. / Energy 117 (2016) 176e189

S S  fdg; DS DS fdg; rr rr rd

Accumulated

181

(32)

Probability

Step 5: By cycling steps 2 to 4, the ideal number of mitigated scenarios can be visible.
This scenario-reduction method is performed on basic scenarios; now as a result, the reduced scenarios can be operated for
solving uncertain problems.
Error level
Fig. 4. Accumulated normalized probabilities of the forecast error levels.

added to its computed value. By this operation, the new

ideal variables for scenarios are determined.
Step 5: The last process of generating scenario is ended by multiplying the probabilities of all variables to achieve the
probability of the scenario.
Above-mentioned steps, i.e. 1 to 5 should be cycled orderly until
a proper number of scenarios are generated.

4.2.2. Stopping constraint

In above method, in order to realize the satisfactory limitation of
the estimation accuracy to terminate a constraint, a number of
samples are selected with the help of a criterion called coefcient of
variation determined by Eq. (33). In this method, the stopping rule
is not checked for all scenarios one by one, but also the simulated
results are divided into a number of groups that each one refers to a
specic number of scenarios.

cvf

sf
p
mf Ns

(33)

If cvf is less than a pre-determined tolerance, the outcomes are

considered acceptable and the simulation process is terminated.
5. MPSO algorithm

4.2. Scenario reduction

5.1. Original PSO algorithm
4.2.1. Method of reduction
In the scenario generation method, many scenarios are generated so that the existence of many scenarios increases simulation
time to overcome this; then, scenario reduction method should be
used. Therefore, reducing the number of scenarios by applying
backward method, the resultant program burden can noticeably be
reduced. The operation of this method is illustrated as follows
[28,29]:
Consider r(s)as the probability and DTs;s0 as the distance of
0
scenario pair (s,s )of each xs(s1,...,Ns) as differentNs. The following
steps present the procedure of using the simultaneous backward
method:
Step 1: ConsiderS as an inception set of scenarios. Ds is the scenarios which should be deleted and the inception Ds should
be considered null, so the following relation computes the
distances of all the scenario pairs:

DT

s;s0

r

Xd 
s  xs0
x
DTxs ; xs0
i
i
i1

s; s0 2S

(28)

Step 2: The minimum distance of scenario k to scenario r for

different k can be determined by the following equation:

DTk;r minDTk;s0 k; s0 2S; ssk

(29)

Step 3: In this step, the following amounts should be calculated:

PDk;r rk  DTk;r k2S

(30)

PDd minPDk k2S

(31)

Step 4: By performing the following relations, the new scenario

sets are achieved:

James Kennedy was the rst who discovered a global method for
optimization by analyzing the social life of animals that lived
collectively and achieved their purposes like feeding and traveling
by communication [30]. This optimization method that is based on
population is called particle swarm optimization algorithm. In this
algorithm rst of all a set of particles should be characterized. In
fact, these particles are the considered occurrence solutions. Then,
this algorithm produces the modied particles in their new situations by operating the equations of velocity update (34) and position update (35) until the best solution is reached.



k
Velk1
u  Velkj C1  rand$  PBest;j
 Xjk C2  rand
j


 GkBest  Xjk
(34)
Xjk1 Xjk Velk1
j

(35)

where Velkj , Velk1

and Xjk are related to the current velocity, the
j
modied velocity and the current position of particle j. Another
unidentied component in the above equations is rand (.) that
produces a positive occurrence number smaller than unity. The
particles obtained in each iteration should be stored in a memory
because each particle follows its best position among the last itk ). These positions are also called tness positions. The
erations (PBest
k
result of these processes is produced by PBest
according to each
particle of the population. Thus, among these local tness positions,
a global tness position is available for a particle shown with
(GkBest ). C1 and C2 in Eq. (34) that are always equal to 2.0 have a
controlling effect for the particles [24]. The operated u in Eq. (34) is
the inertia weight that inuences the local and global convergence
characteristics. The value of this parameter is determined according
to PSO's operation during the training process. This value is reduced
from unity to almost zero. This reduction conforms to the linear
function. It is therefore possible to determine the xed values for

182

M. Bornapour et al. / Energy 117 (2016) 176e189

convergence even if this algorithm is implemented perfectly [30].

Since in the PSO algorithm the initial population is generated
randomly similar to that of other optimization methods, so the
program may not be robust. Furthermore, the rate of obtaining
convergence in this method is slower than that of others such as the
classical and gradient-based methods which is a disadvantage in
this method. It can be a great and effective weakness for this
algorithm.

follow: three different particles are selected from the initial population and a mutation operator will be applied to them. Furthermore, other particles are muted by the selected ones to constitute
the population. Thus, in each iteration, an Xj that is unequal to the
three selected particles is selected so as to be proper for all
searching domains. This operation is cycled for all particles of the
population. The following equation shows how to produce a
k
mutant vector ( Xmut;j
):



k
k
k
k
Xmut;j
Xm1
rand$  Xm2
 Xm3

5.2. Modication of PSO algorithm

k
xmut1 ; xmut2 ; :::; xmutD 
Xmut;j

The PSO algorithm under special situations loses its efciency;

for example, it may stop working with high population diversity.
Also, sometimes the algorithm may produce premature convergence; for instance, when a local convergence is determined or
when there is no diversity in population and when the processing
rate of algorithm is slow. Therefore, this algorithm is independently
useless and needs to be modied. The PSO algorithms are modied
for special purposes. For example, using some technics and equations to make the PSO algorithms diversify to prevent the generation of premature convergence from local minimal mutation is as

According to the following equation, the mutant vector should

be combined with each swarm to determine trial vector:

i
h
k
Xnew;j
xknew;1 ; xknew;2 ; :::; xknew;D
(
xkmut;z if Cr  rand$
k
xnew;z
xkswarm;z otherwise

Read input data

Generate an initial population and an initial velocity

Calculate the objective function for each individual
Select the global position based on the objective function values
Select the ith individual

Select the local position for the ith individual

Update the MPSO parameters
Calculate the next position for each individual MPSO parameter
Check the new position with its limits

No

(37)

where Cr is the occurrence number within (0.1, 0.9) range. By

k
k
comprising Xnew;j
and Xswarm;z
, it is inferable that the cost function

start

i=i+1

(36)

Are all individuals

selected?
Yes
Is convergence
condition satisfied?
Yes
Stop and print the results.
Fig. 5. Flowchart of the MPSO algorithm.

No

M. Bornapour et al. / Energy 117 (2016) 176e189

183

Fig. 6. Single line diagram of modied 33-bus MG test system.

values following relations are truthful:

k1
Xswarm;j

Table 1
Primary data related to prot objective function and network [22,33].
Parameter

Value ($/kWh)

Cost of WT active power,CWT

Cost of PV active power,CPV
Cost of FC active power,CFCCHP
Cost of TH active power,CTH
Tariff of electrical energy selling to customer, pTariff
Tariff of heat energy selling to customer, pH,sell,t
Price of natural gas for parallel boiler,CBoiler

0.005
0.005
0.039
0.039
0.045
0.045
0.05

8
< Xk

new;j if
: Xk
swarm;j





k
k
 f Xswarm;j
f Xnew;j
otherwise

(38)

6. Implementation of MPSO for solving coordinated

scheduling problem
In this paper, in order to solve the coordinated scheduling
problem, MPSO algorithm is used as follows:
Step 1: The initial data should be determined in this stage of the
proposed algorithm, i.e. the cost coefcients of DGs,
electrical market price, network data, electrical loads, the
number of population, the initial value of the inertia
weight, and the value of the learning factor.

Fig. 7. a) Daily load prole; b) daily market price; c) daily wind speed; d) daily solar radiation during the next day.

184

M. Bornapour et al. / Energy 117 (2016) 176e189

Step 12: If this algorithm is run, less than the criteria is the determined value, so go to step 6; otherwise, criteria are
stopped. While the iteration counter value is lower than
the considered maximum number refer to step 6; otherwise, the algorithm is ended.

Table 2
Comparison of cost between the results of the proposed method and other ones.
Method

Total power loss (kW)

Total cost ($)

Without optimization [31]

MINLP [31]
GA and PSCAD [31]
Proposed method

3664
369.76
552.3
358.55

5.1695
3.8323
4.1881
3.1349






104
104
104
104

Fig. 5 illustrates the owchart of the MPSO to solve the coordinated scheduling problem.

Table 3
The values of the prot objective function in different cases.

7. Simulation results

Problem

Case

Prot ($)

Stochastic

Considering CHP
Without considering CHP
Considering CHP
Without considering CHP

871.7179
631.1588
819.6278
534.4823

Deterministic

7.1. Input data

Step 2: The initial population (Xj) and an initial velocity (Velj) need
to be generated randomly.
Step 3: In this step, the value of the objective function (Eq. (2))
should be determined for each particle by using the distribution load ow.
Step 4: The positions of each particle represent real power output
k )
for all FCs and THs and the best position of particles (PBest
should be determined.
Step 5: The best position among the selected positions in the
previous step, i.e. (GkBest ) should be recognized.
Step 6: Set i 1.
Step 7: Updating the velocity and position based on Eqs. (34) and
(35).
Step 8: Executing of mutation (presented in Section 5.2).
Step 9: If all the individuals are selected, go to step 10; otherwise,
set up i i 1and return to step 7.
k
Step 10: Update PBest
using the method presented in step 4.
Step 11: Update GkBest using the method presented in step 5.

In this section, the results of the coordinated scheduling of CHP

units in MG with regard to WT and PV units are presented for a
modied 33-bus MG network, whose single line diagram is shown
in Fig. 6 [31]. The total nominal load of the system is 5084.26 kW
and 2547.32 kVAr. Also, Sb and Vb are considered to be 10 MVA and
12.66 kV as the system base values for the calculation of power ow
and objective function, respectively [32]. The network primary
data, i.e. data related to the gas-purchasing price, energy price, and
the parameters of prot objective function are presented in Table 1
[22,33]. In each feeder, 0.4 time of the electrical load is considered
as its thermal load [22,29]. The prole of the electrical load, electrical market price, solar radiation, and wind speed variation curve
are demonstrated in Fig. 7 for the next day [33].
The normal distribution function is used for modeling the uncertainties of solar radiation and electrical market price. The Weibull distribution function is used for modeling the uncertainties of
wind speed. These uncertainties are considered by the scenariobased technique and the problem is solved as a probabilistic one
by using the MPSO method. In this regard, 500 scenarios are
generated using the technique described in Section 5.1. Since
solving the problem by taking all the 500 scenarios into consideration is very complicated and computationally heavy, these scenarios are reduced to 10 scenarios by using the technique described

Table 4
Result of deterministic problem without considering CHP.
Time (h)

DG Source (kW)
WT

PV

FC

TH

Substation

Heat of FC-CHP

Heat of natural gas

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

160
160
160
160
160
160
160
462.1638
462.1638
462.1638
462.1638
583.7517
583.7517
583.7517
583.7517
583.7517
705.3517
705.3517
705.3517
399.5155
399.5155
399.5155
399.5155
160

0
0
0
0
0
0
0
317.592
317.592
317.592
317.592
528
528
528
528
528
212.8896
212.8896
212.8896
0
0
0
0
0

224.5823
136.8596
104.9418
60.3242
133.7998
168.9676
96.1852
148.0003
133.6091
74.8991
211.0335
157.5988
139.6033
182.3453
171.1952
117.6049
122.1727
159.2811
10.2055
160.7015
195.4638
160.157
219.1467
219.9871

1196.1055
976.9995
973.112
1042.2979
946.2623
937.9398
1173.9553
872.4261
856.7697
799.0612
1179.1892
944.2178
686.7447
783.1635
990.1472
779.2042
1022.0498
627.4338
897.4739
1150.1086
1147.4562
1197.2123
1229.1108
1249.233

1448.5284
1763.461
1800.1889
1776.0632
1798.9991
1770.2354
1603.528
2009.9074
2040.4523
2164.4311
1629.0833
2772.5082
3056.387
2908.824
2706.2602
2987.3543
2537.9819
2912.0096
2787.0759
3700.9256
3664.902
3650.5944
3552.0702
1806.8473

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488

1000

PV Generation (kW)

WT Generation (kW)

M. Bornapour et al. / Energy 117 (2016) 176e189

500

10 12

14

16 18

20

22 24

185

800
600
400
200
0

300
200
100
0

10 12

14

10 12

14

16 18

20

22 24

16 18

20

22 24

Time (h)

TH Generation (kW)

FC Generation (kW)

Time (h)

16 18

20

22 24

1500
1000
500
0

Time (h)

10 12

14

Time (h)

Fig. 8. Generated power by different RESs of the MG in deterministic case with neglecting the CHP effect.

in Section 5.2.
Simulation results are divided into two parts of deterministic
and stochastic. In each part, the simulation is done with and
without considering the CHP effect, separately. In case of with and
without considering CHP effect for supplying the thermal loads, the
shortage of thermal energy is supplied by burning natural gas in
parallel boilers. It should be noted that all DGs generate only active
power and FCs have been considered as CHP.
7.2. Evaluation of proposed method

1000

PV Generation (kW)

WT Generation (kW)

At rst, the proposed method has been applied to a test network

that was examined in the [31]. A comparison made between the
results of the proposed method and other methods in Ref. [31] is
presented in Table 2. In this table, the values of the total cost and

500

total power loss are compared for different methods. As shown in

this table, the cost of using the proposed method is less than that of
other methods. For example, the total cost of the proposed method
is 3.1349  104 $ and for the mixed integer nonlinear programming
(MINLP) and genetic algorithm with PSCAD (GA-PSCAD), it is
3.8323  104 $ and 4.1881  104 $, respectively. Also, the total
power loss by using the proposed method is located in the lower
level than that of other methods such as MINLP and GA-PSCAD. The
total power loss of the test network without optimization is
3664 kW, while for the proposed method, MNILP, and GA-PSCAD it
is 358.55 kW, 369.76 kW, and 552.3 kW, respectively. These comparisons show the superiority of the proposed method to other
methods.

10 12

14

16 18

20

22 24

800
600
400
200
0

300
200
100
0

10 12

14

Time (h)

10 12

14

16 18

20

22 24

16 18

20

22 24

Time (h)

TH Generation (kW)

FC Generation (kW)

Time (h)

16 18

20

22 24

1500
1000
500
0

10 12

14

Time (h)

Fig. 9. Generated power by different RESs of the MG in deterministic case with considering the CHP effect.

186

M. Bornapour et al. / Energy 117 (2016) 176e189

Table 5
Result of deterministic problem considering CHP.
Time (h)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

DG Source (kW)
WT

PV

FC

TH

Substation

Heat of FC-CHP

Heat of natural gas

160
160
160
160
160
160
160
462.16384
462.16384
462.16384
462.16384
583.7517
583.7517
583.7517
583.7517
583.7517
705.3517
705.3517
705.3517
399.5155
399.5155
399.5155
399.5155
160

0
0
0
0
0
0
0
317.592
317.592
317.592
317.592
528
528
528
528
528
212.8896
212.8896
212.8896
0
0
0
0
0

234.967
146.2088
127.3397
131.2292
119.4482
203.6735
205.9898
207.0843
116.0736
159.0779
175.3288
141.5329
126.7371
76.7161
166.6388
123.8297
144.4376
38.5084
125.0347
186.258
207.9963
192.9157
212.9669
224.0838

1253.4201
1003.678
927.011
980.5864
910.8367
997.1158
1069.5104
793.7987
891.9162
718.6969
1219.6333
920.2638
683.6919
828.3859
993.201
773.7523
1015.6939
812.8775
783.8224
1130.6957
1022.7328
1140.8748
1247.5935
1261.9408

1379.0614
1726.2518
1823.7408
1764.8834
1848.8138
1675.2433
1596.3423
2029.8966
2022.8616
2158.7501
1624.3452
2811.171
3075.6087
2980.1301
2710.1312
2983.5348
2520.2596
2842.1882
2781.3019
3694.2167
3783.7317
3674.4298
3539.3743
1789.3543

236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221
236.6221

1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779
1251.3779

deterministic problem. The generation values for each DG at each

hour are illustrated in Fig. 9. The RESs of the MG and thermal energy produced by FC and boilers in this new case are shown in
Table 5. As shown in this table, the total prot has increased
signicantly in the new case, because a part of the required thermal
energy is supplied by CHP.
In case the CHP effect is considered for FCs, the electrical energy
generated by the FCs is increased and the need to buy electrical
power from the market decreases, so prot increases. The comparison made between Tables 4 and 5 shows that the electrical
energy generated by FCs has been increased if FCs generate thermal

7.3. Results of proposed method for solving deterministic problem

In this subsection, at rst the deterministic problem without
considering the CHP effect is considered as a case. The results of this
case are shown in Table 3. The generation values for each DG at each
hour are shown in Table 4 and illustrated in Fig. 8. Also, the
amounts of thermal energy produced by FC and natural gas are
shown in Table 4. As shown in this table, the total thermal load is
supplied by natural gas which increases the operation cost and
decreases the total prot.
The effect of CHP is considered as a new case in the case study of

Table 6
Result of stochastic problem without considering CHP.
Time (h)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

DG Source (kW)
WT

PV

FC

TH

Substation

Heat of FC-CHP

Heat of natural gas

265.8027
651.2537
600.7695
119.6954
185.526
393.3251
174.3731
796.7623
479.4441
413.4723
800
782.5996
800
774.4829
489.2117
465.7514
672.0219
776.1898
800
441.9131
504.5326
800
465.2652
448.3933

0
0
0
0
0
0
0
264.66
396.99
264.66
317.592
660
528
660
396
484
230.6304
266.112
230.6304
0
0
0
0
0

138.7152
189.3716
70.6446
64.172
64.9105
225.9985
160.5914
66.9819
50.3208
75.3329
208.835
167.9726
106.042
155.18758
115.2461
113.921
163.9259
26.3899
110.5426
223.3801
213.13969
126.0244
133.5643
185.302

706.2041
812.7886
805.9647
872.1938
388.4285
437.7965
622.706
699.0134
717.4546
773.2855
698.2022
660.0107
704.6313
614.9319
644.2205
869.3716
726.1966
820.7325
673.419
965.4507
903.823
979.804
804.4591
989.7741

1929.285
1373.1104
1559.198
1987.6467
2423.2714
1983.009
2087.267
1979.4948
2173.1504
2292.5944
1777.4957
2712.9126
2843.6717
2783.2103
3367.8779
3065.6583
2821.1473
2724.849
2793.6015
3785.4644
3788.7085
3486.6386
4024.7494
1813.2615

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488
1488

M. Bornapour et al. / Energy 117 (2016) 176e189

187

Fig. 10. Generated power by different RESs of the MG in stochastic case with neglecting the CHP effect.

energy. Furthermore, less natural gas is needed to supply the

thermal load.
WT and PV units have many benets. Scheduling program of
these RESs can be determined by forecasting the wind speed and
solar radiation. This leads to increased prots. As it follows from the
comparison of the results in Tables 4 and 5, if the CHP impact is not
considered, the electrical energy supply from the market is relatively higher. This issue shows that the electrical energy supply
obtained from the DGs especially from FCs and THs has increased
and has led to higher prots if the CHP impact has been taken into
consideration.

7.4. Results of proposed method for solving stochastic problem

In this substation, uncertainties are taken into account. For this
purpose, the uncertainties of the wind speed, solar radiation, and
electrical market price are considered. The results of this case are
presented in Table 3. As shown in this table, by considering stochastic nature of uncertain parameters, the total benets increase
in comparison with the deterministic problem. Participation in the
electricity market and the amount of unit's production can be
determined realistically. On the other hand, it can be seen that
when there are CHPs in MG, the prot from the electrical market
increases.

Table 7
Result of stochastic problem considering CHP.
Time (h)

DG Source (kW)
WT

PV

FC

TH

Substation

Heat of FC-CHP

Heat of natural gas

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

265.8027
651.2537
600.7695
119.6954
185.526
393.3251
174.3731
796.7623
479.4441
413.4723
800
782.5996
800
774.4829
489.2117
465.7514
672.0219
776.1898
800
441.9131
504.5326
800
465.2652
448.3933

0
0
0
0
0
0
0
264.66
396.99
264.66
317.592
660
528
660
396
484
230.6304
266.112
230.6304
0
0
0
0
0

235
235
126
111
235
2
126
2
235
111
2
235
2
126
116.7159
2
111
111
235
235
2
2
85.6404
235

1021
921.4
433.4
921.4
921.4
537.811
523
413.4
623.0372
772
523
921.4
889.2654
762
386.6184
672.4
862
1021
433.4
593
583.52615
1021
782
862

1507.4203
1219.6692
1879.8401
1897.0193
1697.9495
2120.453
2224.6563
2345.9424
2077.7702
2258.137
2168.5025
2371.9012
2762.6941
2648.9582
3647.2867
3398.8138
2739.171
2419.3692
2919.2302
4167.4399
4385.5277
3571.7038
4101.9749
1894.7997

236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692
236.692

1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308
1251.308

188

M. Bornapour et al. / Energy 117 (2016) 176e189

Fig. 11. Generated power by different RESs of the MG in stochastic case with considering the CHP effect.

Table 8
Comparison of the results obtained by different algorithms considering CHP in case of probabilistic problem for 100 trials.
Objective function

Algorithm

Best solution

Mean solution

Worst solution

SD

No. of global solution

CPU time (s)

Prot ($)

Proposed MPSO
PSO
GA

871.7179
857.8112
846.5346

871.5927
856.9606
843.1765

869.6314
851.7352
834.9548

0.4955
2.1083
5.2545

94
86
71

245.6
272.5
283.7

By neglecting the CHP effect in a stochastic case, the amounts of

electrical energy generated by different RESs of the MG are shown
in Table 6 and illustrated in Fig. 10. In a stochastic case, the production of WT and PV units are increased and are closer to reality
and has helped to increase prots. The increase in the production of
WT and PV units led to reduction in buying electrical power from
the electrical market as shown in Fig. 10. Besides enjoying more
prot, it also enjoys other advantages such as the improvement of
voltage prole, loss reduction, etc.
By considering the CHP effect in a stochastic case, the amount of
electrical energy generated by different DG units of the MG and the
thermal energy generated by FC and natural gas are shown in
Table 7 and Fig. 11. In a stochastic case, when the CHP units are
considered, the amount of prot in power market increases from
631.1588 $ to 871.7179 $. Furthermore, the electrical power obtained from the electrical market is decreased and the generation of
FC and TH units is increased leading to an increase in local
production.
As the results of Tables 6 and 7 show, if CHP exists, the need to
buy natural gas for supplying thermal load is decreased and leads to
a decrease in the cost and an increase in prot. The amount of the
thermal load is considerable, given that the only power source
available to supply the heat load is FC-CHP; MG uses the maximum
heat production of FC-CHP. Moreover, FC-CHP production capacity
is limited. The rest of the unmet thermal load should be supplied
with natural gas.
The results of the optimal coordinated scheduling while disregarding the CHP effect in the case of deterministic and stochastic
problems are presented in Tables 4 and 6, respectively. As these two
tables show, the production of WT and PV units is more in the
stochastic case. The results of the optimal coordinated scheduling

while considering the CHP effect in the case of deterministic and

stochastic problems are presented in Tables 5 and 7, respectively. In
this case, the results show that the renewable energy resource
production of stochastic programming is higher than that of the
deterministic problem. Therefore, this leads to an increase in the
prot. Moreover, Table 3 conrms the mentioned statement.
In Fig. 6 by using Eq. (24), the overall calculated efciency of FCs
that are connected to buses 2, 15 and 32 in stochastic case without
considering CHP effect are 35%, 38%, and 40% respectively; while in
stochastic case with considering CHP effect the overall calculated
efciency of FCs that are connected to buses 2, 15 and 32 are 72%,
75%, and 79% respectively. Therefore, considering the CHP effect on
FC leads to a signicant increment of the overall efciency.
The results of different algorithms for the probabilistic problems
are compared in Table 8. According to the results, the value of the
objective function is higher than that of the other algorithm.
Furthermore, MPSO converges to the global optimal solution more
out of 100 times of the program execution conrming the robustness of the proposed algorithm. Also, the CPU time of MPSO is less
than that of PSO, since it converges in a smaller number of iterations and needs a smaller initial population.

8. Conclusion
In this paper, the stochastic problem of coordinated scheduling
of CHP units in MG is solved with regard to WT and PV units using
MPSO algorithm. Maximizing the total prots for the participation
of an MG in the electrical market is the objective function considered in this study. The uncertainties of electrical market price, solar
radiation, and the speed of wind are considered using a scenariobased method which enhances the operational reliability. In the

M. Bornapour et al. / Energy 117 (2016) 176e189

MPSO, the algorithm performance is signicantly improved using

the proposed mutation technique. Also, the proposed method
signicantly increases the total prot of participation in the electricity market. The simulation results demonstrate that the proposed algorithm has a signicant advantage over other methods.
References
[1] Sreedharan P, Farbes J, Cutter E, Woo CK, Wang J. Microgrid and renewable
generation integration: University of California, San Diego. Appl Energy
2016;169:709e20.
[2] Ferruzzi G, Cervone G, Monache LD, Graditi G, Jacobone F. Optimal bidding in a
Day-Ahead energy market for Micro Grid under uncertainty in renewable
energy production. Energy 2016;106:194e202.
[3] Wang C, Liu Y, Li X, Guo L, Qiao L, Lu H. Energy management system for standalone diesel-wind-biomass microgrid with energy storage system. Energy
2016;97:90e104.

 B.
[4] Marzband M, Sumper A, Ruiz-Alvarez
A, Domnguez-Garca JS, Tomoiaga
Experimental evaluation of a real time energy management system for standalone microgrids in day-ahead markets. Appl Energy 2013;106:365e76.
[5] Holjevac N, Capuder T, Kuzle I. Adaptive control for evaluation of exibility
benets in microgrid systems. Energy 2015;92:487e504.
[6] Niknam T, Bornapour M, Ostadi A, Gheisari A. Optimal planning of molten
carbonate fuel cell power plants at distribution networks considering combined heat, power and hydrogen production. J Power Sources 2013;239:
513e26.
[7] Mohammadia M, Hosseinian SH, Gharehpetian GB. GA-based optimal sizing of
microgrid and DG units under pool and hybrid electricity markets. Int J Electr
Power Energy Syst 2012;35:83e92.
[8] Mohammadi S, Mozafari B, Solymani S, Niknam T. Stochastic scenario-based
model and investigating size of energy storages for PEM-fuel cell unit
commitment of micro-grid considering protable strategies. IET Gener
Transm Distrib 2014;8:1228e34.
[9] Maity I, Rao S. Simulation and pricing mechanism analysis of a solar-powered
electrical microgrid. IEEE Syst J 2010;4:275e84.
[10] Tina G, Brunetto C. Short-term scheduling of a wind generation and hydrogen
storage in the electricity market. Eur Trans Electr Power 2010;20:559e74.
[11] Mashhour M, Golkar MA, Moghaddas-Tafreshi SM. Extending market activities for a distribution comp and reserve markets e Part I: problem formulation. Energy Convers Manage 2011;52:477e86.
[12] Mashhour M, Golkar MA, Moghaddas-Tafreshi SM. Extending market activities for a distribution company in hourly ahead energy and reserve marketsdPart II: numerical results. Energy Convers Manage 2011;52:569e80.
[13] Parastegari M, Hooshmand RA, Khodabakhshian A, Zare AH. Joint operation of
wind farm, photovoltaic, pump-storage and energy storage devices in energy
and reserve markets. Int J Electr Power Energy Syst 2015;64:275e84.
[14] Parastegari M, Hooshmand RA, Khodabakhshian A, Forghani Z. Joint operation
of wind farms and pump-storage units in the electricity markets: modeling,
simulation and evaluation. Simul Model Pract Theory 2013;37:56e69.
[15] Reddy SS, Bijwe PR, Abhyankar AR. Multi-objective market clearing of electrical energy, spinning reserves and emission for wind-thermal power system.
Int J Electr Power Energy Syst 2013;53:782e94.

189

[16] Derakhshandeh SY, Masoum AS, Deilami S, Masoum MAS, Hamedani

Golshan ME. Coordination of generation scheduling with PEVs charging in
industrial micro grids. IEEE Trans Power Syst 2013;28:3451e61.
[17] Nguyen DT, Le LB. Risk-Constrained prot maximization for microgrid
aggregators with demand response. IEEE Trans Smart Grid 2015;6:135e46.
[18] Nguyen DT, Le LB. Optimal bidding strategy for microgrids considering
renewable energy and building thermal dynamics. IEEE Trans Smart Grid
2014;5:1608e20.
[19] Kahlert S, Spliethoff H. Investigation of different operation strategies to provide balance energy with an industrial CHP plant using dynamic simulation.
ASME J Eng Gas Turbines Power 2016. http://dx.doi.org/10.1115/1.4034184.
Accepted manuscript.
[20] Camporeale SM, Fortunato B, Torresi M, Turi F, Pantaleo AM, Pellerano A. Part
Load performance and operating strategies of a natural gasdbiomass dual
fueled microturbine for combined heat and power generation. ASME J Eng Gas
Turbines Power 2015;137:121401.
[21] Colella1 WG, Pilli SP. Energy system and thermoeconomic analysis of combined heat and power high temperature proton exchange membrane fuel cell
systems for light commercial buildings. ASME J Fuel Cell Sci Technol 2015;12:
031008.
[22] Niknam T, Bornapour M, Gheisari A, Bahmani-Firouzi B. Impact of heat, power
and hydrogen generation on optimal placement and operation of fuel cell
power plants. Int J Hydrogen Energy 2013;38:1111e27.
[23] Gunes MB. Investigation of a fuel cell based total energy system for residential
applications. Master of Science Thesis, Dept. Mechanical Engineering, Virginia
Polytechnic Institute and State University; 2001.
[24] Slootweg JG, De Haan SWH, Polinder H, Kling WL. General model for representing variable speed wind turbines in power system dynamics simulations.
IEEE Trans Power Syst 2003;18:144e51.
[25] Singh PP, Singh S. Realistic generation cost of solar photovoltaic electricity.
Renew Energy 2010;35:563e9.
[26] Marwali MKC, Ma H, Shahidehpour SM, AbdulRahman KH. Short-term generation scheduling in photo-voltaic-utility grid with battery storage. IEEE
Trans Power Syst 1998;13:1057e62.
[27] Soroudi A, Caire1 R, Hadjsaid N, Ehsan M. Probabilistic dynamic multiobjective model for renewable and non-renewable distributed generation
planning. IET Gener Transm Distrib 2011;5:1173e82.
[28] Wu L, Shahidehpour M, Li T. Stochastic security-constrained unit commitment. IEEE Trans Power Syst 2007;22:800e11.
[29] Bornapour M, Hooshmand R. An efcient scenario-based stochastic programming for optimal planning of combined heat, power, and hydrogen
production of molten carbonate fuel cell power plants. Energy 2015;83:
734e48.
[30] Kennedy J. The particle swarm: social adaptation of knowledge. IEEE Int Conf
Evol Comput 1997. http://dx.doi.org/10.1109/ICEC.1997.592326.
[31] Silva M, Morais H, Vale Z. An integrated approach for distributed energy
resource short-term scheduling in smart grids considering realistic power
system simulation. Energy Convers Manage 2012;64:273e88.
[32] Baran ME, Wu FF. Network reconguration in distribution systems for loss
reduction and load balancing. IEEE Trans Power Deliv 1989;4:1401e7.
[33] Zare M, Niknam T. A new multi-objective for environmental and economic
management of Volt/Var Control considering renewable energy resources.
Energy 2013;55:236e52.