Energy Conversion and Management 92 (2015) 149–161

Contents lists available at ScienceDirect

Energy Conversion and Management

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

A hybrid of ant colony optimization and artificial bee colony algorithm

for probabilistic optimal placement and sizing of distributed energy
resources
M. Kefayat a,⇑, A. Lashkar Ara a, S.A. Nabavi Niaki b
a
Department of Electrical Engineering, Dezful Branch, Islamic Azad University, P.O. Box 313, Dezful, Iran
b
Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON M5 S 3G4, Canada

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

Article history: In this paper, a hybrid configuration of ant colony optimization (ACO) with artificial bee colony (ABC)
Received 30 July 2014 algorithm called hybrid ACO–ABC algorithm is presented for optimal location and sizing of distributed
Accepted 14 December 2014 energy resources (DERs) (i.e., gas turbine, fuel cell, and wind energy) on distribution systems. The pro-
posed algorithm is a combined strategy based on the discrete (location optimization) and continuous
(size optimization) structures to achieve advantages of the global and local search ability of ABC and
Keywords: ACO algorithms, respectively. Also, in the proposed algorithm, a multi-objective ABC is used to produce
Ant colony optimization (ACO)
a set of non-dominated solutions which store in the external archive. The objectives consist of minimiz-
Artificial bee colony (ABC)
Multi-objective optimization
ing power losses, total emissions produced by substation and resources, total electrical energy cost, and
Optimal placement improving the voltage stability. In order to investigate the impact of the uncertainty in the output of the
Point estimate method (PEM) wind energy and load demands, a probabilistic load flow is necessary. In this study, an efficient point
Renewable energy estimate method (PEM) is employed to solve the optimization problem in a stochastic environment.
Hybrid ACO–ABC The proposed algorithm is tested on the IEEE 33- and 69-bus distribution systems. The results demon-
strate the potential and effectiveness of the proposed algorithm in comparison with those of other
evolutionary optimization methods.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction technical and economic advantages for operations as well as utili-

In recent years, distribution energy resources (DERs) became
attractive due to its technical, economic and environmental bene-
fits. DER is a small electrical power source located near the cus-
tomer’s site in a distribution network. There are different DER
types such as fuel cell (FC), gas turbine (GT), micro-turbine and
renewable distributed generation (RDG). The environmental
advantage of DER covers a wide range of benefits such as the
reduction of greenhouses gases, new transmission line develop-
ment, and sound pollution. In addition, RDGs are to play a key role
in dealing with environmental concern due to their ability to pro-
duce green power and reduce the dependency on fossil fuels.
Nowadays, renewable energy sources supply 14% of the total world
energy demand and include solar, wind, biomass, etc. [1,2].
Parallel to the environmental benefits, renewable and non-
renewable distributed generation resources can provide significant
⇑ Corresponding author. Tel.: +98 641 6260051; fax: +98 641 6260890.
E-mail addresses: Milad.kefayat@gmail.com (M. Kefayat), Lashkarara@ieee.org
(A. Lashkar Ara), nabavi.niaki@utoronto.ca (S.A. Nabavi Niaki).
ties. Adding DER to the competitive electricity production impacts
on different technical benefits like reducing losses, improving volt-
age profile and power quality [3]. Since the installation of different
energy sources in a power system may have significant impacts on
electric power generation, it requires careful consideration for the
installation of any new energy sources to be very well planned.
However, different benefits of DER depend on location, technology
and capacity [4].
There are different techniques for solving the optimal location
and size of energy sources connected to the distribution network.
Research for the optimal allocation of DER has been investigated
using the conventional algorithms [5] including the analytical [6]
and numerical [7] methods, and evolutionary methods such as bac-
terial foraging optimization [8], particle swarm optimization with
constriction factor approach (PSO–CFA) [9] and imperialist
competition algorithm (ICA) [10]. Therefore, different evolutionary
algorithms (EA) have been applied to the optimization problem of
multiple DERs. Nevertheless, integration of hybrid energy systems
into the current power systems offers various technical
assessments and it has become an essential issue. The authors of

http://dx.doi.org/10.1016/j.enconman.2014.12.037
0196-8904/Ó 2014 Elsevier Ltd. All rights reserved.
150 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161
Ref. [11] have employed a hybrid modified shuffled frog algorithm
and deferential evolution to determine the optimal location with
respect to power losses, energy cost and emissions. Another study
has developed a hybrid genetic algorithm (GA) and particle swarm
optimization method for multi-objective location and sizing of
DERs [12]. The authors have concluded that the combined method
has a higher probability of finding the optimum solution; however,
it does not consider different strategies by installing multiple DERs.
The authors of [13,14] have presented the artificial bee colony
algorithm (ABC) to optimally solve the allocation problem of
DER. However, different DER technologies have various impacts
on power system operation, control, and stability. In Ref. [15], opti-
mal placement of renewable energy sources including three differ-
ent types of RDG: photovoltaic, wind and fuel cell are investigated
by using the multi-objective modified honey bee mating optimiza-
tion algorithm. In [16], fuel cell power plants (FCPPs) have been
used as energy sources for planning with respect to operation
and economic problems while hybrid energy systems are best sui-
ted to reduce the dependence on fossil fuel. Genetic algorithm has
been used to connect multiple wind turbines (WTs) in distribution
networks [17,18]. These approaches are available to find the loca-
tion or size of only a wind turbine. On the other hand, the integra-
tion of energy sources has different impacts on power system
operation. However, most studies do not consider the wind
uncertainty.
Wind turbines as renewable technologies are expected to play
important role for power systems in electricity generation and
operations. According to the statistics, the installed wind power
is 238,351 MW in 2011; the increase of the installed capacity is
nearly 75% over the period of 2005–2011 [19]. In its role as elec-
tricity generation several countries have achieved rapid expansion
such as China, France, Canada, USA and UK with the growth rate of
around 98%, 88%, 87%, 80% and 79% around the period 2005–2011,
respectively. The world wind energy association (WWEA) has pre-
dicted that the total electrical power generated by wind will reach
1,500,000 MW in 2020 [19].
The use of wind turbines can provide significant technical, envi-
ronmental and economic benefits. However, wind is one of the
major sources suffering from uncertainty. Due to increasing pene-
trates of wind power into the systems operation and the inherent
uncertainty in wind, there is an urgent need for a stochastic study
for power system planning, especially in dealing with modern
power systems [20,21]. Many methods have been proposed to deal
with the uncertainty problem [22]. One of the most popular meth-
ods is the Monte Carlo simulation to accommodate probabilistic
analysis through deterministic analysis [23]. With regards to
Monte Carlo simulation, it should be noted that a great deal of sim-
ulation and computing is required to achieve convergence [24]. An
alternative way is through analytical methods; these use mathe-
matical assumptions to simplify the problems such as indepen-
dency or liner dependency of different random variables [25].
Lastly, the approximate method provides a balance between speed
and accurate by an approximate description. Among the well-
known approximate methods, the point estimated methods (PEMs)
stand out [26].
To the best of our knowledge, no research work in the area con-
sidered the combined ant colony optimization (ACO) and artificial
bee colony algorithms called hybrid ACO–ABC algorithm to solve
the optimal placement and sizing of distributed energy resources
(DERs). The main contribution of this paper is to develop the
hybrid ACO–ABC algorithm to optimally locate DERs considering
the wind and load uncertainties as a stochastic programming
framework. The Weibull probability density and normal distribu-
tion functions are used to model the wind and load uncertainties,
respectively. Also, the probabilistic power flow based on the PEM
is utilized to study the stochastic nature of wind generation and
load demands during optimization processes. In addition, the
hybrid ACO–ABC algorithm takes advantages of the global search
ability of ABC and the local search ability of ACO algorithm. The
ACO algorithm is a probabilistic technique for solving computa-
tional problems and gives the optimal solution through graphs
although it may be trapped in a local optimal solution different
from a global optimal one [27]. For this reason, the ABC algorithm
is used to escape from the trap of local optimal solution and to
quickly find the global optimal solution [28]. Therefore, in this
paper, the location of DER and its size are optimized by ACO and
ABC algorithms to correlate discrete and continuous variables,
respectively. The best location and size of DERs (i.e., gas turbine,
fuel cell and wind turbine) are selected to optimize power losses,
voltage stability index (VSI), total emission and electrical energy
cost as objective functions in single- and multi-objective optimiza-
tion frameworks. The proposed algorithm uses the Pareto optimal
solution and saves the non-dominated optima into an external
archive to solve the multi-objective problem. The effective fuzzy
decision making tool is incorporated in the proposed algorithm
to softly switch between the Pareto optimal solutions. So, the fuzzy
decision making tool offers a better judgment among the Pareto
optimal solutions based on the selected weight values and thereaf-
ter select the best compromise one by the proposed algorithm.
Furthermore, the proposed algorithm is implemented on the IEEE
33- and 69-bus distribution systems in two different scenarios.
The results obtained by the proposed algorithm show that the pro-
posed algorithm is outperformed by the other evolutionary optimi-
zation methods such as PSO–CFA [9], ABC [14] and Modified
Teaching–Learning Based Optimization (MTLBO) [29] from techni-
cal, environmental and economic point of view.
The remainder of this paper is organized as follows: Section 2
explains the optimal placement and sizing problem formulation.
In Section 3, the necessary background and fundamentals of the
ABC and ACO algorithms and the implementation of the proposed
hybrid ACO–ABC algorithm are described into three subsections.
Section 4 presents the subject of the point estimated method,
modeling wind and load uncertainty in power systems. Section 5
contains simulation results followed by conclusions.
2. Problem formulation
The main goal of optimization DER is to determine the best
location and size of new energy sources with minimized specific
objective function. In this section the objective functions and con-
straints are explained.
2.1. Objective functions
In this paper, the following objective functions are considered:
2.1.1. Minimization of power losses
Minimizing the total electrical energy losses of distribution net-
work is an important goal of implementing sources, which can be
formulated as follows [21]:
nbr
X
f 1 ðXÞ ¼ Ri jIi j2 ð1Þ
i¼1
where nbr is the number of branches, Ri and Ii are the resistance and
the current magnitude of i th the branch, respectively.
2.1.2. Voltage stability index
Under power systems planning and operation the voltage
stability is one of the most significant security indices. DER has a
profound impacted on the voltage stability index and it will be
changed by connecting DER. In [30], Chakravorty and Das proposed
 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161 151

a stability index for identifying the node, which is most sensitive to X

N GT X
N FC X
N WT

voltage collapse. It is related to the voltage magnitude of each f 4 ðXÞ ¼ C GT i þ C FC i þ C WT i þ C sub ð9Þ
i¼1 i¼1 i¼1
branch and power of receiving-end, as shown in Fig. 1. The
objective function for voltage stability index can be described as
C ¼ aþb ð10Þ
follows [31]:
 
1 a ¼ Capital cost  Capacity  g ð11Þ
f 2 ðXÞ ¼ min ð2Þ
SIðm2Þ
b ¼ ðFuel cost  O & M costÞ  P ð12Þ
SIðm2Þ ¼ jVðm1Þj4  4½Pðm2ÞX jj  Q ðm2ÞRjj 2  4½Pðm2ÞRjj
þ Qðm2ÞX jj jVðm1Þj2 ð3Þ C sub ¼ Psub  Prsub ð13Þ

where a and b denote fixed cost and variable cost of different

where SI(m2) is the stability index for node m2ð2; 3; . . . ; nbus Þ, V(m1)
resources, respectively; g specifies the complex coefficients of the
is voltage of node m1, P(m2) and Q(m2) are total real and reactive
efficiency and operating region, O & M cost is the operation and
power load fed through node m2, Xjj is reactance of branch jj, and
maintenance cost, Psub is the active power production of substation,
Rjj is resistance of branch jj.
Prsub is cost of power generated at substation, and C is the cost elec-
trical energy generation by each source.
2.1.3. Minimization of emission
Minimization of emissions produced by different sources of
energy and grid is the another objective. The most important pol- 2.2. Constraints
lutants are considered, i.e., carbon dioxide (CO2), nitrogen oxides
(NOx) and sulfur dioxide (SO2). The values of the grid and the In the optimal placement and sizing of distributed sources, the
DER parameters are shown in Table 1. The mathematical formula- systems need to meet certain security constraints and should be
tion of third objective function can be expressed as follows [21]: satisfied. They can be mathematically expressed in the following
subsections.
X
NGT X
NFC X
NWT
f 3 ðXÞ ¼ EGT i þ EFC i þ EWT i þ EGrid ð4Þ
i¼1 i¼1 i¼1 2.2.1. Equality constraints
The equality constraints are based on the principle of equilib-
 
rium. The power flow equations corresponding to both real and
EGT i ¼ COGT GT GT
2 þ NOx þ SO2  PGT i ð5Þ
reactive power balance equations are mathematically defined as
  follows [10]:
EFC i ¼ COFC FC FC
2 þ NOx þ SO2  PFC i ð6Þ
nbus
X
  Pi ¼ V i V j Y ij cosðhij  di þ dj Þ ð14Þ
EWT i ¼ COWind
2 þ NOWind
x þ SOWind
2  PWT i ð7Þ j¼1

  nbus
X
EGrid ¼ COGrid
2 þ NOGrid
x þ SOGrid
2  PGrid ð8Þ Qi ¼ Vi V j Y ij sinðhij  di þ dj Þ ð15Þ
j¼1
where E and P are design emissions produced and active power gen-
eration by the ith energy sources including gas turbine, fuel cell, where Pi and Qi are the real and reactive power at bus i, Vi and di are
wind turbine and grid. Also, NGT, NFC and NWT are the numbers of amplitude and angle of the voltage at ith bus, Yij and hij are ampli-
the GT, FC and WT units, respectively. tude and angle of the admittance between bus i and j, respectively.
Also, the injection real and reactive power considering energy
2.1.4. Minimization of cost sources at the buses of any distribution systems can be expressed
Reducing the total cost of electrical energy produced by substa- as follows:
tion and resources is an important goal, which is given by [11,32]: Pi ¼ Pgi  P di ð16Þ

Q i ¼ Q g i  Q di ð17Þ

V(m1)∠δ(m1) V(m2)∠δ(m2) where P gi and Q gi are active and reactive power output of the ith
+ energy sources, Pdi and Q di are the load active and reactive power
m1 m2 at bus-i, respectively.

2.2.2. Inequality constraints

Sending End Receiving End
2.2.2.1. Voltage limits. The following range for voltage must be kept
at each bus:
Fig. 1. Single line model of a typical two-bus system.

V min 6 V 6 V max ð18Þ

Table 1
Emission related for resources. 2.2.2.2. Operating limits of sources. This includes the upper active
Emission type Emission factors (lb/MW h) power generation limits of energy sources and ensures that all
Grid Gas turbine Micro turbine Wind FC
DERs can operate just within their power limits as follows:

NOx 5.06 1.15 0.44 0 0.03

SO2 11.6 0.008 0.008 0 0.006
Pgi 6 P limt
gi ð19Þ
CO2 2031 1494 1596 0 1078
where Plimt
gi is the maximum operating limit of energy sources.
152 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161

3. Proposed hybrid algorithm

Initial food source

The proposed optimization approach in this paper is based on a

hybrid of ACO and ABC. This section presents a brief description of
Evaluate the population
ABC, ACO and hybridization process as the hybrid ACO–ABC
algorithm.
Determine food source position
3.1. Artificial bee colony algorithm

Employed bee phase

Produce a new position
The artificial bee colony algorithm is one of the recently meta-
heuristic search technique defined by Karaboga in 2005 [13]. The
ABC is derived from honey bee foraging behavior. Colonies of Evaluate a function values
honey bees have a highly structured social organization. In the
bee colony, the honey bees are divided into three different groups:
employed bees, onlooker bees and scout bees. Each employed bee
NO All employed
exploits one specific food source, and then returns to the hive to
distributed?
share information about that food source with the other bees
through specialized dances. The related dance is called a waggle Yes
dance and it is proportional to the quality of the food source. In
other words, all information are available on the dance floor. The Determine food source position
other kinds of bees (onlookers) are waiting on the dance area in
the hive. They watch the dances of the employed bees in order to

Onlooker bee phase

make decisions about which food source to choose. The indirect Select a position base on roulette wheel
communication between bees via dance allows for good food
sources to attract more onlooker bees. If the number of visiting
bees does not improve the food source, then the source is aban- Compare and update best position
doned and the employed bee becomes a scout. Every bee colony
has scout bees to avoid local optima whenever a source is exploited
fully. They do not have any guidance while looking for food. Scout NO All onlookers
bees carry out a random search to explore the environment in distributed?
order to find new food source locations (solution). Yes
In the used ABC algorithm, the employed bee is searching
around the source. Each employed bee moves from one old loca-

Scout bee phase

tion xij to new candidate location vij by using the following Abandoned
solution? NO
expression:

v ij ¼ xij þ /ij ðxij  xkj Þ ð20Þ Yes

Generate the new solution

where k 2 f1; 2; . . . ; SNg and j 2 f1; 2; . . . ; Dg are randomly chosen
while SN is the number of food sources and D denotes the problem
dimension. Based on Eq. (20) k has to be different from i, also /ij is a
NO Is maximum
uniform random number in the range of [1, 1]. If the new location
generation?
value vij is better than xij the position is updated and replaced with
the older value. The onlooker bee selects a food source according to
Yes
the probability value by using a roulette wheel selection method
and then, this new position is calculated by using the following Output solution to ACO
equation:
Fig. 2. Flowchart of the ABC algorithm.
v ij ¼ xij þ wi /ij ðxij  xkj Þ ð21Þ

where wi is the weight coefficient of employed bee information. The 3.1.1. Multi-objective artificial bee colony algorithm
food source probability is calculated as follows: Many optimization problems involve different interacting and
conflicting objectives function. Multi-objective optimization
fit entails finding the best optimal solution from all number of objec-
pi ¼ PSN i ð22Þ tives simultaneously. Recently, the multi-objective artificial bee
j¼1 fit j
colony algorithm (MOABC) method has been proposed to find a
set of non-dominated solutions that are known as the Pareto opti-
where fit denoted the fitness values of the food sources of employed
mality. A solution X1 is said to dominate another solution X2 if the
bees and it is defined as follows:
following condition are satisfied [15].
(
1
f ðxi Þ P 0 (
fit i ¼ 1þf ðxi Þ
ð23Þ 8i 2 f1; 2; . . . ; mg f i ðX 1 Þ 6 f i ðX 2 Þ
1 þ jf ðxi Þj f ðxi Þ < 0 ð24Þ
9j 2 f1; 2; . . . ; mg f j ðX 1 Þ < f j ðX 2 Þ

Here f(xi) stands for the amount of the objective function values to Pareto dominance is used to compare the solutions. If X1
be optimized. The whole process of the algorithm may also be dominates X2, the related food source is stored in the archive.
explained through the flow chart given in Fig. 2. Transforming the ABC to MOABC requires an external archive to
 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161
store the non-dominated solutions in different place from the bees.
In the employed bee phase of the MOABC, the position is updated
using Eq. (20) but k is chosen from the archive randomly and
stored in its position to xkj. The employed bee comes into the hive
with a non-dominate solution to go into the archive. Probability for
each food source will be updated sequentially according to the Eq.
(22). Where fit is the value of the solution in the position i associ-
ated with the information provided by the external archive, it can
be defined as:
fiti ¼ e bni
ð25Þ
Here, b is a positive coefficient, and ni is related to the number of
non-dominated food sources in the area of objective space. After
all employed bees finish their search processes; they insert all
new non-dominated solutions into archive. Onlooker bees will
select an archive member proportional to the quality of food source
with the roulette wheel selection, and pick the leader and update its
position using Eq. (21). When all the different types of bees (i.e.
employed, onlooker, and scouts) are evaluated, the archive is filled
with the non-dominated bees. The content of the external archive
has a bounded size. If the archive is full, the archive gets updated
at the end of each iteration. In this work, the updating external
archive is used similar to the adaptive grid of MOPSO [33]. The
archive is divided into regions. The non-dominated solutions
located in more populated areas are randomly selected and
automatically discarded. The implemented search of the probability
proportional selection strategy spares regions of the Pareto fronts;
therefore, the diversity has been increased among the stored solu-
tions in the archive and explores the search space to a greater
extend. Thus convergence would be quick and the onlooker bees
explore better to the true region.
3.2. Ant colony optimization
Ant colony optimization is one of the population based meta-
heuristic techniques based on the foraging behavior of real ants.
They forage for food and establish the shortest paths from their
nest to the food source. Ant colony optimization is a class of algo-
rithms whose construct solutions are based on the problem data,
and has been presented for application to discrete optimization
problems. In reality, ants search for food sources in a random
way. As soon as an ant finds a food source, they carry some food
back to their colony. When they move along the paths, they lay a
chemical substance called pheromone as they travel. In turn,
shorter paths will have the higher rate of pheromone trails. Each
ant makes decisions by using pheromone trails as a communica-
tion mechanism. Strength of pheromone trail deposited on the
ground depends on the quality of the solution (food source) found.
Pheromone trails accumulate with multiple ants in shorter paths,
which causes a higher density than longer paths, thus increasing
its attractiveness. All pheromone trails are eventually reduced by
an evaporation rate. On the other side, a process of evaporation
presents the exploration and prevents stalling in a local minimum.
However, the pheromone values are updated at the end of each
iteration [32,34].
8 a
< P ðsij Þ ðgijaÞ
b
ðsim Þ ðgim Þb
j 2 N ki
Pkij ¼ m2N k
i
ð26Þ
: as a decision maker can choose the best compromise solution
0 otherwise
where Pkij is the probability which ant k chooses to move from node i
to node j. This decision depends on the pheromone level and heuris-
tic information. While N ki is the set of feasible neighborhoods that
have not yet been visited by ant k, gij is heuristic function, sij is
the amount of pheromone on edge i and j, and a and b are the
parameters which determine the relative importance of pheromone
153
concentration and heuristic information. The pheromone update
can be formed as follows:
sij sij þ Dskij ð27Þ
(
k
Q
f ðwk Þ
lij 2 wk
Ds ¼ ij ð28Þ
0 otherwise
The evaporation update is given by:
sij ð1  qÞsij ð29Þ
where q is the constant factor reduction of all pheromones, f(wk) is
the cost of the solution performed by ant k, and Q is a constant. The
above optimization process is terminated after a certain amount of
iteration.
3.3. The hybrid ACO–ABC algorithm
Optimal sitting (discrete optimization) and sizing (continuous
optimization) of energy sources are non-linear optimization. As
literature review, different standard algorithms such as PSO,
ABC and ICA were developed for solving continuous optimization
problems [35]. The discrete structure of the placement is a main
obstacle to get use of the continuous optimization methods.
Among the previously mentioned optimization methods, ACO
has attracted a lot of attention in the fields of discrete problems
due to its population-based search capability as well as simplicity
and robustness. ACO used heuristic technique to produce a good
initial solution and determine a good search direction depending
on the experience. It is worthy of note that this strategy often
helps ACO find a good solution and it causes ACO to be trapped
in local minima.
To overcome these disadvantages and integrates the merits of
both ACO and ABC, the hybrid ACO–ABC algorithm is introduced.
Hybridizing ant colony optimization with artificial bee colony algo-
rithm takes advantage of ACO and ABC by dividing optimization
problem into two parts. In each iteration, ant builds candidate solu-
tion to locate the DERs; thereafter, the size of DERs is optimized by
ABC. The proposed algorithm has several characteristic features.
Firstly, the triple search capability of the ABC algorithm is used
for escaping from the local minimum solution. This causes ABC gen-
erally affect the search process since it is able to quickly find the
correct global optimum. Secondly, the performance of algorithm
is improved by dividing the optimization problem into continuous
and discrete categories stages; so, it decreases the size of the search
space. Finally, it is able to overcome the drawback of classical ant
colony algorithm which is not suitable for continuous optimiza-
tions. The schematic diagram of the proposed algorithm is shown
in Fig. 3.
3.3.1. Multi-objective ACO–ABC algorithm
When solving single objective optimization problems by using
hybrid ACO–ABC, the best solution can be readily found. But, in
multi-objective optimization problems, have set of Pareto optimal
solution. The non-dominated solution considers several objectives
that increase the diversity of the final solution. One of the impor-
tant advantages of the Pareto method is that the system operator
according to the requirements of the power system operations. It
is necessary to select one optimal size of resource as the best com-
promise solution. In this paper, a fuzzy optimization approach is
employed to calculate each objective function for all the non-dom-
inated solutions in the archive. The main significant issue in the
fuzzy method is solving the optimization problem with the nature
of human judgment [36].
154 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161

X
n
Start ki ¼ 1; ki 2 ½0; 1 ð32Þ
i¼1

Here, F1, F2, F3 and F4 are membership function of power loss, VSI,
Initialization emission and cost objective function, respectively. Also, ki is the rel-
ative weight of the ith objective function and n is the number of
objective functions. The weight values can be specified by the expe-
Select possible paths rience of the expert power system operator based on the impor-
tance of objectives.

Tour construction 4. Uncertainty modeling

4.1. Point estimated method

Using ABC to calculate
optimal source size The development of the new renewable energy, especially wind
turbines, has its own features and impact on power flow. It is nec-
essary for power systems operators to recognize the significant
ABC determine best size problems which are influenced by the uncertainty in the networks.
Therefore, they use statistic studies to judge the performance of
electrical power systems. Different types of approximation meth-
ods are utilized in the field of engineering systems under uncer-
NO tainty [22]. In mathematics, these methods are able to use
Last ant?
approximate formulas for calculating the statistical moment out-
put network quantities of power systems. This paper uses PEM
Yes
for solving the problem of optimal sitting and sizing of energy
Update pheromones values sources. PEM is a stochastic techniques developed by Rosenblueth
[25]. This method has been made as approach to approximate a
function with little statistical data information and is satisfactorily
Evaporation phase accurate for a wide range of probabilistic problems. When some of
the input variables are uncertain, the PEM only needs the probabil-
ity density function for each input variable to propagate uncer-
tainty in power systems studies. Assume Y = h(X), a function of n
Compare the solution
variable with probability density function, f xk , is [37]
Y ¼ hðXÞ ¼ hðx1 ; x2 ; . . . ; xn Þ: ð33Þ
Store best cost The ith moment of kth variable is defined as:
M j ðxk Þ
kk;j ¼ ð34Þ
NO ðrk Þ j
Convergence?
Z þ1
Mj ðxk Þ ¼ ðxk  lk Þj f xk dxk ð35Þ
Yes 1

Print the results where kk;1 is equal to zero, kk;2 is equal to one, kk;3 and kk;4 are the
coefficient of skewness and the coefficient of kurtosis, respectively.
Also lk and rk are the mean and standard deviation of f xk . The pre-
End sented method chooses two estimate points from every distribu-
tion. The location of concentration can be calculated using the
Fig. 3. The procedure of the hybrid ACO–ABC algorithm.
following equations:

8 xk;i ¼ lk þ nk;i rk i ¼ 1; 2 ð36Þ

min
>
> 1 f i ðXÞ 6 f i
>
< sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f i ðXÞf min min max  2
F i ðXÞ ¼ 1 i
fi 6 f i ðXÞ 6 f i ð30Þ kk;3 kk;3
>
> f max
i f min
i nk;i ¼ þ ð1Þ3i nþ ð37Þ
>
: max 2 2
0 f i ðXÞ P f i
min max The probability concentration at location xk,i can be obtained
where f i and f i are extracted from results of each objective
by:
function independently. As described before, the goal of optimiza-
tion is optimal location and sizing of different sources into the dis- ð1Þi  nk;3i
tribution network in order to minimize different objective Pk;i ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2 ð38Þ
k
functions. The following weighted sum approach can be selected 2n  n þ k;3 2
to find the best compromise solution under the system constraint.
The multi-objective function is maximizing the sum of the functions The above procedure is repeated for different variables. The
as follows: mean and variance are approximated by:
( )
Xn X
n X
2

FðXÞ ¼ max ki F i ðXÞ ð31Þ EðYÞ ffi ðPk;i  hðl1 . . . ; xk;i ; . . . ln ÞÞ ð39Þ

i¼1 k¼1 i¼1
 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161 155

23 24 25

26 27 28 29 30 31 32 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

19 20 21 22

Fig. 4. Single line diagram of the IEEE 33-bus distribution system.

X
n X
2
2
where p is constant, rPL is standard deviation and lPL is the mean
VarðYÞ ffi ðPk;i  h ðl1 . . . ; xk;i ; . . . ln ÞÞ  fEðYÞg2 ð40Þ value.
k¼1 i¼1

4.2. Probabilistic modeling of wind 5. Simulation results

For many years, wind has been utilized as the cleanest type of In this section, the optimal location of a gas turbine, fuel cell,
renewable energy source. Using wind power can reduce emissions and a wind turbine and their settings to optimize the objective
and diversify the energy mix. The characteristics of the wind functions are obtained and discussed. The objective functions,
depend on various factors like geography and topography. which are considered here, are the total energy losses, voltage sta-
Collected wind speed data is one of the most important steps for bility index, total emissions produced by substation and resources,
determining the optimal location for wind turbines. A very good and total electrical energy cost. The proposed hybrid ACO–ABC
expression often recommended for modeling the behavior of wind algorithm is implemented to solve problems for the standard IEEE
speed is the Weibull probability density function which can be for- 33- and 69-bus distribution systems. The mentioned optimization
mulated as follows [37]: problem is solved in two scenarios; in the first scenario it is
  k1   k  assumed that the resources units include GT and WT. In second
k v v
f ðv Þ ¼ exp  ð41Þ scenario, it is considering GT, WT and FC in order to investigate
c c c the effect of FC on the grid. In the case of ACO and ABC implemen-
where c and k are called the scale and shape factor, and v represents tation, the population size for each algorithm is 20 and the itera-
the wind speed. The output power of wind power generation is cor- tion number parameters are set as: 40 and 35, respectively. Their
responding to a Weibull distribution with the shape parameter of 2. performances are compared to PSO–CFA, ABC and MTLBO. In
The accurate structure to compute relation between output power multi-objective optimization problem, the size of archive is chosen
and wind speed is given by: to equal 100 for the two case studies. The simulation studies are
8 carried out in Matlab software. Due to the high R/X ratios in distri-
> 0 0 6 v 6 v ci bution systems, the forward–backward sweep process based on
>
>
< av 3 þ b v 6 v 6 v
ci ra Kirchhoff’s laws is used. The maximum generation limits of GT
Pðv Þ ¼ ð42Þ
> Prated
>
> v ra 6 v 6 v co and FC are 2 MW and 0.2 MW, respectively.
:
0 v co 6 v
5.1. IEEE 33-bus distribution system
Prated
a¼ ð43Þ
v  v 3ci
3
ra The test system is 33-bus radial distribution with the total load
of 3.715 MW, 2.3 MVAr, 33 bus and 32 branches. The real and reac-
P rated v 3ci
b¼ ð44Þ tive power losses are 0.2027 MW and it is shown in Fig. 4 [9].
v 3ra  v 3ci In order to verify the effectiveness and feasibility of the pro-
where Prated is the rate power of the wind turbine, P(v) is the power posed hybrid approach in solving optimization problem over the
obtained from wind energy, and vci, vra and vco denote the cut in other well-known algorithms, hybrid ACO–ABC algorithm is used
speed, rated speed and cut off speed of wind turbine, respectively. to determine optimum size and location for the seven simulated
The wind turbine parameters are taken from [38]. cases. In fact, increasing candidate sources for connecting raised
the problem complexity and local minimum. The results are com-
4.3. Load model pared with PSO–CFA and ABC method. Cleary, as shown in Table 2,
the proposed technique achieves better loss reduction. It proves
Load in distribution networks can be affected by different vari- that the proposed method has better performance in obtaining
ables such as weather, temperature, etc. For taking these facts into high-quality solutions with respect to the other algorithms.
consideration, the power of the load buses are modeled as normal Hybrid GT and WT is considered as the first scenario in the opti-
distributions. The probability density function of the load power mal sizing and location procedure. For the actual operation, the
(PL) is expressed as follows [21]: output power of a wind turbine is modeled as a Weibull distribu-
tion due to the intermittent and uncertainty nature of the wind
ðP L lP Þ2
 L power. For this reason, the best scenario for the WT is operated
1 2r2
FðPL Þ ¼ pffiffiffiffiffiffiffi e PL
ð45Þ in combination with other types of energy sources, which have a
r P L 2p lower uncertainty in their output power.
156 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161

Table 2
Comparison of the simulation results for IEEE 33-bus distribution system.

Simulation case Method Optimal location (bus number) DER optimal output (MW) Total capacity added (MW) Power losses (MW)
1 DER unit ABC [14] 6 2.5775 2.5775 0.1050
PSO–CFA [9] 6 2.5752 2.5752 0.1039
ACO–ABC 6 2.5753 2.5753 0.1039
2 DER units ABC [14] 6 1.9707 2.5464 0.0899
15 0.5757
PSO–CFA [9] 14 0.7876 2.0363 0.0862
29 1.2487
ACO–ABC 13 0.8464 2.0052 0.0859
30 1.1588
3 DER units ABC [14] 6 1.7569 3.1152 0.0792
15 0.5757
25 0.7826
PSO–CFA [9] 10 1.0491 2.7326 0.0760
25 0.8786
33 0.8049
ACO–ABC 14 0.7547 2.9260 0.0714
24 1.0999
30 1.0714
4 DER units ABC [14] 6 1.0765 3.0884 0.0665
15 0.5757
25 0.7824
32 0.6538
PSO–CFA [9] 13 0.7484 3.0213 0.0689
24 1.0759
28 0.5994
31 0.5976
ACO–ABC 7 0.9114 3.2000 0.0659
14 0.5858
24 0.9877
31 0.7151
5 DER units PSO–CFA [9] 4 0.5172 3.2906 0.0662
15 0.5839
25 0.7268
26 0.8241
32 0.6386
ACO–ABC 6 0.9041 3.1121 0.0651
11 0.3614
16 0.3554
25 0.8003
31 0.6909
6 DER units PSO–CFA [9] 5 0.8620 3.5477 0.0667
8 0.4617
15 0.5661
24 0.8018
28 0.1367
30 0.7194
ACO–ABC 6 0.8423 3.5003 0.0640
11 0.3585
16 0.3710
20 0.2820
24 0.9638
31 0.6827
7 DER units PSO–CFA [9] 6 0.8923 3.4355 0.0651
14 0.6055
19 0.1445
22 0.2084
25 0.7882
28 0.3201
33 0.4765
ACO–ABC 6 0.8465 3.4210 0.0635
15 0.6022
22 0.2052
24 0.6631
25 0.3524
30 0.2857
32 0.4659

The results of the single objective for obtaining optimal place- power output of WT. The hybrid ACO–ABC determines the optimal
ment and sizing of GT and WT are given in Table 3. There are location of the WT and the power obtained from wind energy is
two identical wind turbines in the energy sources. In order to calculated by adding probabilistic power flow. As expected, for sin-
reflect variability of wind power, PEM is employed to estimate gle objective cost optimization, the size of a gas turbine is equal to
 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161 157

Table 3
Results obtained by hybrid ACO–ABC algorithm for GT and WT in the IEEE 33-bus distribution system.

Power losses (MW) VSI Emission (lb/h) Cost ($/h)

Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW)
GT 13 0.7627 12 2.0000 12 0.9457 – 0
25 0.8514 24 2.0000 24 1.5665 – 0
30 1.0015 29 2.0000 30 1.1742 – 0
WT 7 – 16 – 11 – 16 –
27 – 22 – 17 – 32 –
Function values 0.0712 1.0143 5512.8093 302.6087

Table 4
Results obtained by hybrid ACO–ABC algorithm for GT, FC and WT in the IEEE 33-bus distribution system.

Power losses (MW) VSI Emission (lb/h) Cost ($/h)

Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW)
GT 15 0.6165 10 2.0000 3 1.9145 – 0
24 1.0654 24 2.0000 14 0.5351 – 0
31 0.7128 30 2.0000 31 0.8313 – 0
FC 8 0.2000 4 0.2000 9 0.2000 17 0.2000
29 0.2000 22 0.2000 25 0.2000 32 0.2000
WT 9 – 21 – 7 – 15 –
27 – 32 – 24 – 29 –
Function values 0.0683 1.0023 5337.2974 312.3476

Table 5
Simulation results of the three-objective optimization for scenario one.

Case I Case II Case III Case IV

Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW)
GT 15 0.7062 13 0.7789 13 0.7132 15 0.8868
24 1.2738 25 0.6589 24 1.1351 25 0.7970
31 1.1847 30 1.2189 30 1.0169 30 2.0000
WT 2 – 11 – 9 – 6 –
18 – 29 – 15 – 9 –
f1 0.0780 0.0758 0.0727 –
f2 1.0957 1.1078 – 1.0313
f3 5646.3845 – 5751.6190 5559.7691
f4 – 375.3435 381.2605 410.5946

Table 6
Simulation results of the three-objective optimization for scenario two.

Case I Case II Case III Case IV

Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW)
GT 11 1.0595 10 0.9808 12 0.8011 13 1.3568
24 1.3522 25 0.8807 25 0.8803 24 0.9138
30 0.9165 33 0.7219 30 0.8498 32 1.0951
FC 22 0.1502 6 0.1239 7 0.1871 6 0.1783
26 0.2000 16 0.2000 28 0.2000 33 0.1901
WT 7 – 4 – 6 – 27 –
9 – 28 – 29 – 29 –
f1 0.0754 0.0723 0.0715 –
f2 1.1081 1.1197 – 1.0340
f3 5361.3816 – 5558.3102 5429.4329
f4 – 384.8677 383.0101 414.0513

zero because energy costs produced by the substation are lower
than those attained by them. Cost coefficient of the substation
bus (Prsub) is 0.044 ($/kW h) [21].
In the configuration of the second scenario, a combination of GT,
WT and FC is applied. The best results obtained by single optimiza-
tion of the objectives are shown in Table 4. Note that total environ-
mental pollution is the summation of the emissions produced by
the sources and substation. The best results obtained by optimizing
total emissions are shown in Tables 3 and 4 and it should be noted
that WT and FC have active roles in emission reduction. In other
words, to minimize the emission objective function, the active
power of all FCs are set to their maximum value. As it is obvious
from the results, FCs active powers for cost objective function is
equal to maximum value because the variable cost of fuel cell units
is lower than substation and GT.
The study system is a distribution network having the substa-
tion located at bus 1 and has some loads. Therefore, the simulation
results show the optimal location of many energy sources close to
the end of feeder. It can be seen from Table 4 that the result
obtained by optimization set all fuel cell in their maximum output.
158 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161

Table 7
Simulation results of multi-objective optimization for different scenario.

Scenarios Scenario one Scenario two

DG type GT WT GT FC WT
Location (bus no.) Size (MW) Location (bus no.) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.)
15 0.8005 5 11 0.5562 9 0.0395 5
25 0.9063 2 24 1.3006 16 0.1519 22
30 1.0972 – 30 1.1515 – – –
f1 0.0726 0.0714
f2 1.1282 1.1435
f3 5784.9381 5549.4019
f4 379.4800 399.2160

Table 8
Simulation results of multi-objective optimization for different scenario considering for other Weibull factors (c = 7 and k = 2.4).

Scenarios Scenario one Scenario two

DG type GT WT GT FC WT
Location (bus no.) Size (MW) Location (bus no.) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.)
16 0.6439 8 14 0.6281 8 0.0331 7
24 1.4012 27 24 1.2936 26 0.1182 9
31 0.9436 – 30 1.0239 – – –
f1 0.0737 0.0705
f2 1.1166 1.1220
f3 5653.6119 5558.4848
f4 384.3842 396.3933

Table 9
Simulation results of multi-objective optimization for different scenario considering load uncertainty.

Scenarios Scenario one Scenario two

DG type GT WT GT FC WT
Location (bus no.) Size (MW) Location (bus no.) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.)
16 0.7195 13 11 0.9965 7 0.1546 17
24 0.9057 26 25 0.8409 22 0.1515 21
30 1.0468 – 31 0.8343 – – –
f1 0.0738 0.0718
f2 1.1296 1.1209
f3 5790.6787 5508.5316
f4 374.3725 386.5685

36 37 38 39 40 41 42 43 44 45 46

47 48 49 50

51 52 68 69

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

66 67

53 54 55 56 57 58 59 60 61 62 63 64 65

28 29 30 31 32 33 34 35

Fig. 5. Single line diagram of the IEEE 69-bus distribution system.

 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161 159

Table 10
Comparison of the simulation results for IEEE 69-bus distribution system.

Simulation case Method Optimal location (bus number) DER optimal output (MW) Total capacity added (MW) Power losses (MW)
1 DER unit MTLBO [29] 61 1.8197 1.8197 0.083323
ACO–ABC 61 1.8726 1.8726 0.083189
2 DER units MTLBO [29] 17 0.5197 2.2517 0.071776
61 1.7320
ACO–ABC 18 0.5309 2.3127 0.071657
61 1.7818
3 DER units MTLBO [29] 11 0.4938 2.5447 0.069539
18 0.3784
61 1.6725
ACO–ABC 11 0.5597 2.6224 0.069429
21 0.3468
61 1.7159

Table 11
Results obtained by hybrid ACO–ABC algorithm for GT and WT in the IEEE 69-bus distribution system.

Power losses (MW) VSI Emission (lb/h) Cost ($/h)

Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW)
GT 22 0.3574 50 2.0000 17 0.4995 – 0
61 1.7100 64 2.0000 49 1.4696 – 0
66 0.4575 68 2.0000 61 1.7978 – 0
WT 28 – 13 – 55 – 62 –
54 – 59 – 64 – 64 –
Function values 0.0698 1.0080 5638.6417 307.6084

Table 12
Results obtained by hybrid ACO–ABC algorithm for GT, FC and WT in the IEEE 69-bus distribution system.

Power losses (MW) VSI Emission (lb/h) Cost ($/h)

Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW)
GT 11 0.4877 27 2.0000 4 1.2729 – 0
19 0.2387 49 2.0000 14 0.4416 – 0
61 1.6948 62 2.0000 61 1.6471 – 0
FC 24 0.1342 44 0.2000 17 0.2000 62 0.2000
50 0.2000 60 0.2000 65 0.2000 64 0.2000
WT 46 – 13 – 15 – 13 –
56 – 45 – 63 – 65 –
Function values 0.0690 1.0041 5474.5920 317.1060

Table 13
Simulation results of the three-objective optimization for scenario one.

Case I Case II Case III Case IV

Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW)
GT 21 0.5389 15 0.2145 6 0.8757 23 0.7052
49 1.1060 23 0.2790 16 0.4704 62 2.0000
61 2.0000 61 2.0000 61 1.7959 67 1.0849
WT 32 – 3 – 2 – 24 –
52 – 53 – 43 – 49 –
f1 0.0728 0.0731 0.0716 –
f2 1.0537 1.0542 – 1.0226
f3 5673.1709 – 5803.4255 5673.1581
f4 – 374.4723 393.6908 417.3784

FC do not introduced uncertainty in planning and have lower emis- Case II: Considering functions f1, f2 and f4 (k1 = 0.6, k2 = 0.2,
sion compared to the subsystems and GT. Moreover, the number of k4 = 0.2).
energy sources affect on the results. Case III: Considering functions f1, f3 and f4 (k1 = 0.6, k3 = 0.2,
In order to show the improvement of each objective function k4 = 0.2).
with respect to the other objective functions the combinations of Case IV: Considering functions f2, f3 and f4 (k2 = 0.6, k3 = 0.2,
the objective functions are classified into the following cases: k4 = 0.2).

Case I: Considering functions f1, f2 and f3 (k1 = 0.6, k2 = 0.2, The results of three objective functions optimization for the sit-
k3 = 0.2). ting and sizing are shown in Tables 5 and 6. As seen from these
160 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161

Table 14
Simulation results of the three-objective optimization for scenario two.

Case I Case II Case III Case IV

Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.) Size (MW)
GT 22 0.4221 10 0.2603 11 0.6232 23 0.7108
49 1.0911 25 0.2047 24 0.2673 49 0.6827
61 1.9338 61 2.0000 61 1.7739 63 2.0000
FC 7 0.0587 8 0.1237 46 0.1921 15 0.1826
51 0.2000 17 0.1997 47 0.1813 59 0.2000
WT 15 – 3 – 5 – 9 –
69 – 50 – 32 – 52 –
f1 0.0721 0.0719 0.0700 –
f2 1.0576 1.0500 – 1.0170
f3 5545.5748 – 5680.0891 5505.6250
f4 – 385.3991 390.8508 417.0460

Table 15
Simulation results of multi-objective optimization for different scenario.

Scenarios Scenario one Scenario two

DG type GT WT GT FC WT
Location (bus no.) Size (MW) Location (bus no.) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.)
10 0.5510 40 18 0.4052 48 0.0860 5
19 0.4111 44 61 1.6720 54 0.2000 51
61 1.8483 – 66 0.5521 – – –
f1 0.0705 0.0699
f2 1.0677 1.0847
f3 5962.3268 5783.9988
f4 383.5500 390.2620

Table 16
Simulation results of multi-objective optimization for different scenario considering load uncertainty.

Scenarios Scenario one Scenario two

DG type GT WT GT FC WT
Location (bus no.) Size (MW) Location (bus no.) Location (bus no.) Size (MW) Location (bus no.) Size (MW) Location (bus no.)
8 0.6298 29 19 0.4536 41 0.1300 2
15 0.6283 39 48 0.8319 67 0.1634 7
61 1.7372 – 61 1.7743 – – –
f1 0.0713 0.0699
f2 1.0778 1.0851
f3 5800.0042 5543.9293
f4 387.7860 402.9898

tables, the multi-objective algorithm gives a slightly higher objec-
tive function value than the single-objective methods for all the
optimization levels. It shows that the gas turbine as a bigger energy
sources can provide higher amount of active power with less emis-
sion than subsystems, but imposes higher electrical energy cost.
Energy source installation improves the voltage stability of neigh-
boring nodes, and it is also clear that the cost increases as the volt-
age stability index increases. To have better comparison between
the scenarios one and two, the results of the four objective func-
tions are shown in Table 7. In this table, the values for k1, k2, k3
and k4 are: 0.7, 0.1, 0.1 and 0.1, respectively. It can be concluded
from the results that the second scenario has slightly lower power
losses than first one.
The characteristics of the wind speed depend on target region.
Based on the Weibull distribution, higher c factor makes a greater
amount of higher wind speed will be probable. In this regard, the
Weibull distribution function with specified scale and shape
parameter is implemented to show that hybrid systems usually
present an acceptable performance. The results of this experiment
are shown in Table 8. It is interesting to note that for this particular
data set, the values of the Weibull distributions are c = 7 and
k = 2.4. In order to show the effectiveness of the optimization pro-
cedure, the proposed methodology is applied to the uncertainties
in both wind energy sources and load demands. It is assumed that
the loads on buses 9, 17 and 28 have the normal distribution func-
tion. In this regard, the proposed algorithm is implemented again
and the optimal location and capacity of each DER with consider-
ing load and wind uncertainty in the case of one and two scenarios
are given in Table 9.
5.2. IEEE 69-bus distribution system
In this part, the proposed algorithm is tested on 69-bus radial
distribution system and it is demonstrated in Fig. 5 [39]. The origi-
nal total real power loss is 0.225 MW and most of load is far from
the substation. In the case of minimizing power losses, the results
obtained by the proposed algorithm are compared to that of
MTLBO and tabulated in Table 10. Comparison of performances
in this case reveals that the proposed algorithm gives a better
response. In the first scenario, each case is contained 3 GT and 2
WT. Table 11 lists the results attained by hybrid ACO–ABC algo-
rithm. In the second scenario, it is assumed the hybrid energy
sources are applied to the optimization problem. The simulation
results for optimizing hybrid GT, FC and WT are explained in
Table 12. The results of different combinations of the objective
functions are shown in Tables 13 and 14. The detailed results of
 M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161
the best multi-objective solution for both scenarios are given in
Table 15. Regarding the uncertainty in the load demand, it is
assumed that buses 26, 27 and 33 have the normal distribution
load. The optimal placements and capacities for the multi-
objectives are shown in Table 16.
6. Conclusion
This paper presented a hybrid algorithm, namely hybrid ACO–
ABC, which takes the advantage of ant colony optimization and
artificial bee colony algorithm. This technique is based on ACO
for finding the location and uses the ABC to calculate the capacity
of energy sources. Also, the hybrid ACO–ABC is validated by com-
parison with results adopted from the literature. The simulation
results show that the hybrid ACO–ABC is more effective in the mul-
tiple DER allocation optimization problems. In addition, the pro-
posed algorithm is extended by multi-objective optimization
equipped with a fuzzy decision making approach to consider the
technical, environmental and economic optimization issues simul-
taneously. In this method, the decision maker could be chosen a
most preferred solution by updates the weight factors. The pro-
posed algorithm is used to determine the location and size of var-
ious sources including GT, FC and WT. The optimal allocations are
calculated for two different scenarios with different power sources
to show the most efficient solution. The objectives functions
consist of total active power losses, voltage stability index, total
emissions and electrical energy costs. Furthermore, one of the
main pivotal features of this paper is investigating the impact of
stochastic wind power generation and load demands in the system
by using PEM which provides an appropriate tool to taking
uncertainty in power flow. The methodology presented in this
paper could be easily implemented in the distribution networks
for planning and operation studies.
References
[1] Manzano-Agugliaro F, Alcayde A, Montoya FG, Zapata-Sierra A, Gil C. Scientific
production of renewable energies worldwide: an overview. Renew Sustain
Energy Rev 2013;18:134–43.
[2] Georgilakis PS, Hatziargyriou ND. Optimal distributed generation placement in
power distribution networks: models, methods, and future research. IEEE
Trans Power Syst 2013;28:3420–8.
[3] Pepermans G, Driesen J, Haeseldonckx D, Belmans R, D’haeseleer W.
Distributed generation: definition, benefits and issues. Energy Policy
2005;33:787–98.
[4] Viral R, Khatod DK. Optimal planning of distributed generation systems in
distribution system: a review. Renew Sustain Energy Rev 2012;16:5146–65.
[5] Tan W-S, Hassan MY, Majid MS, Abdul Rahman H. Optimal distributed
renewable generation planning: a review of different approaches. Renew
Sustain Energy Rev 2013;18:626–45.
[6] Gözel T, Hocaoglu MH. An analytical method for the sizing and siting of
distributed generators in radial systems. Elect Power Syst Res 2009;79:912–8.
[7] Al Abri RS, El-Saadany EF, Atwa YM. Optimal placement and sizing method to
improve the voltage stability margin in a distribution system using distributed
generation. IEEE Trans Power Syst 2013;28:326–34.
[8] Devi S, Geethanjali M. Application of modified bacterial foraging optimization
algorithm for optimal placement and sizing of distributed generation. Expert
Syst Appl 2014;41:2772–81.
[9] Mistry KD, Roy R. Enhancement of loading capacity of distribution system
through distributed generator placement considering techno-economic
benefits with load growth. Int J Elect Power Energy Syst 2014;54:505–15.
[10] Soroudi A, Ehsan M. Imperialist competition algorithm for distributed
generation connections. IET Generation Transm Distrib 2012;6:21–9.
[11] Doagou-Mojarrad H, Gharehpetian GB, Rastegar H, Olamaei J. Optimal
placement and sizing of DG (distributed generation) units in distribution
networks by novel hybrid evolutionary algorithm. Energy 2013;54:129–38.
161
[12] Moradi MH, Abedini M. A combination of genetic algorithm and particle
swarm optimization for optimal DG location and sizing in distribution
systems. Int J Elect Power Energy Syst 2012;34:66–74.
[13] El-Zonkoly A. Optimal placement and schedule of multiple grid connected
hybrid energy systems. Int J Elect Power Energy Syst 2014;61:239–47.
[14] Lalitha MP, Reddy NS, Reddy VV. Optimal DG placement for maximum loss
reduction in radial distribution system using ABC algorithm. Int J Rev Comput
2010;3:44–52.
[15] Niknam T, Taheri SI, Aghaei J, Tabatabaei S, Nayeripour M. A modified honey
bee mating optimization algorithm for multiobjective placement of renewable
energy resources. Appl Energy 2011;88:4817–30.
[16] Niknam T, Bornapour M, Gheisari A. Combined heat, power and hydrogen
production optimal planning of fuel cell power plants in distribution
networks. Energy Convers Manage 2013;66:11–25.
[17] Mokryani G, Siano P. Optimal wind turbines placement within a distribution
market environment. Appl Soft Comput 2013;13:4038–46.
[18] Ugranlı F, Karatepe E. Optimal wind turbine sizing to minimize energy loss. Int
J Elect Power Energy Syst 2013;53:656–63.
[19] Islam MR, Mekhilef S, Saidur R. Progress and recent trends of wind energy
technology. Renew Sustain Energy Rev 2013;21:456–68.
[20] Sheen J-N, Tsai M-T, Wu S-W. A benefits analysis for wind turbine allocation in
a power distribution system. Energy Convers Manage 2013;68:305–12.
[21] Malekpour AR, Niknam T, Pahwa A, Fard AK. Multi-objective stochastic
distribution feeder reconfiguration in systems with wind power generators
and fuel cells using the point estimate method. IEEE Trans Power Syst
2013;28:1483–92.
[22] Morales JM, Perez-Ruiz J. Point estimate schemes to solve the probabilistic
power flow. IEEE Trans Power Syst 2007;22:1594–601.
[23] Vahidinasab V. Optimal distributed energy resources planning in a
competitive electricity market: multiobjective optimization and probabilistic
design. Renew Energy 2014;66:354–63.
[24] Ruiz-Rodriguez FJ, Gomez-Gonzalez M, Jurado F. Optimization of radial
systems with biomass fueled gas engine from a metaheuristic and
probabilistic point of view. Energy Convers Manage 2013;65:343–50.
[25] Malekpour AR, Niknam T. A probabilistic multi-objective daily Volt/Var control
at distribution networks including renewable energy sources. Energy
2011;36:3477–88.
[26] Mohammadi S, Mozafari B, Solimani S, Niknam T. An adaptive modified firefly
optimisation algorithm based on Hong’s point estimate method to optimal
operation management in a microgrid with consideration of uncertainties.
Energy 2013;51:339–48.
[27] Lin Y-K, Yeh C-T, Huang P-S. A hybrid ant-tabu algorithm for solving a
multistate flow network reliability maximization problem. Appl Soft Comput
2013;13:3529–43.
[28] Nasiraghdam H, Jadid S. Optimal hybrid PV/WT/FC sizing and distribution
system reconfiguration using multi-objective artificial bee colony (MOABC)
algorithm. Solar Energy 2012;86:3057–71.
[29] Martín García JA, Gil Mena AJ. Optimal distributed generation location and size
using a modified teaching–learning based optimization algorithm. Int J Elect
Power Energy Syst 2013;50:65–75.
[30] Chakravorty M, Das D. Voltage stability analysis of radial distribution
networks. Int J Elect Power Energy Syst 2001;23:129–35.
[31] Aman MM, Jasmon GB, Bakar AHA, Mokhlis H. A new approach for optimum
DG placement and sizing based on voltage stability maximization and
minimization of power losses. Energy Convers Manage 2013;70:202–10.
[32] Niknam T. A new approach based on ant colony optimization for daily Volt/Var
control in distribution networks considering distributed generators. Energy
Convers Manage 2008;49:3417–24.
[33] Coello CAC, Pulido GT, Lechuga MS. Handling multiple objectives with particle
swarm optimization. IEEE Trans Evol Comput 2004;8:256–79.
[34] Dorigo M, Blum C. Ant colony optimization theory: a survey. Theor Comput Sci
2005;344:243–78.
[35] AlRashidi MR, AlHajri MF. Optimal planning of multiple distributed generation
sources in distribution networks: a new approach. Energy Convers Manage
2011;52:3301–8.
[36] Peng C, Sun H, Guo J, Liu G. Multi-objective optimal strategy for generating and
bidding in the power market. Energy Convers Manage 2012;57:13–22.
[37] Ghofrani M, Arabali A, Etezadi-Amoli M, Fadali MS. A framework for optimal
placement of energy storage units within a power system with high wind
penetration. IEEE Trans Sustain Energy 2013;4:434–42.
[38] Atwa YM, El-Saadany EF. Probabilistic approach for optimal allocation of wind-
based distributed generation in distribution systems. IET Renew Power
Generation 2011;5:79–88.
[39] Aman MM, Jasmon GB, Bakar AHA, Mokhlis H. Optimum network
reconfiguration based on maximization of system loadability using
continuation power flow theorem. Int J Elect Power Energy Syst
2014;54:123–33.