Professional Documents
Culture Documents
A Hybrid of Ant Colony Optimization and Artificial Bee Colony Algorithm PDF
A Hybrid of Ant Colony Optimization and Artificial Bee Colony Algorithm PDF
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.
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 load demands during optimization processes. In addition, the
and deferential evolution to determine the optimal location with hybrid ACO–ABC algorithm takes advantages of the global search
respect to power losses, energy cost and emissions. Another study ability of ABC and the local search ability of ACO algorithm. The
has developed a hybrid genetic algorithm (GA) and particle swarm ACO algorithm is a probabilistic technique for solving computa-
optimization method for multi-objective location and sizing of tional problems and gives the optimal solution through graphs
DERs [12]. The authors have concluded that the combined method although it may be trapped in a local optimal solution different
has a higher probability of finding the optimum solution; however, from a global optimal one [27]. For this reason, the ABC algorithm
it does not consider different strategies by installing multiple DERs. is used to escape from the trap of local optimal solution and to
The authors of [13,14] have presented the artificial bee colony quickly find the global optimal solution [28]. Therefore, in this
algorithm (ABC) to optimally solve the allocation problem of paper, the location of DER and its size are optimized by ACO and
DER. However, different DER technologies have various impacts ABC algorithms to correlate discrete and continuous variables,
on power system operation, control, and stability. In Ref. [15], opti- respectively. The best location and size of DERs (i.e., gas turbine,
mal placement of renewable energy sources including three differ- fuel cell and wind turbine) are selected to optimize power losses,
ent types of RDG: photovoltaic, wind and fuel cell are investigated voltage stability index (VSI), total emission and electrical energy
by using the multi-objective modified honey bee mating optimiza- cost as objective functions in single- and multi-objective optimiza-
tion algorithm. In [16], fuel cell power plants (FCPPs) have been tion frameworks. The proposed algorithm uses the Pareto optimal
used as energy sources for planning with respect to operation solution and saves the non-dominated optima into an external
and economic problems while hybrid energy systems are best sui- archive to solve the multi-objective problem. The effective fuzzy
ted to reduce the dependence on fossil fuel. Genetic algorithm has decision making tool is incorporated in the proposed algorithm
been used to connect multiple wind turbines (WTs) in distribution to softly switch between the Pareto optimal solutions. So, the fuzzy
networks [17,18]. These approaches are available to find the loca- decision making tool offers a better judgment among the Pareto
tion or size of only a wind turbine. On the other hand, the integra- optimal solutions based on the selected weight values and thereaf-
tion of energy sources has different impacts on power system ter select the best compromise one by the proposed algorithm.
operation. However, most studies do not consider the wind Furthermore, the proposed algorithm is implemented on the IEEE
uncertainty. 33- and 69-bus distribution systems in two different scenarios.
Wind turbines as renewable technologies are expected to play The results obtained by the proposed algorithm show that the pro-
important role for power systems in electricity generation and posed algorithm is outperformed by the other evolutionary optimi-
operations. According to the statistics, the installed wind power zation methods such as PSO–CFA [9], ABC [14] and Modified
is 238,351 MW in 2011; the increase of the installed capacity is Teaching–Learning Based Optimization (MTLBO) [29] from techni-
nearly 75% over the period of 2005–2011 [19]. In its role as elec- cal, environmental and economic point of view.
tricity generation several countries have achieved rapid expansion The remainder of this paper is organized as follows: Section 2
such as China, France, Canada, USA and UK with the growth rate of explains the optimal placement and sizing problem formulation.
around 98%, 88%, 87%, 80% and 79% around the period 2005–2011, In Section 3, the necessary background and fundamentals of the
respectively. The world wind energy association (WWEA) has pre- ABC and ACO algorithms and the implementation of the proposed
dicted that the total electrical power generated by wind will reach hybrid ACO–ABC algorithm are described into three subsections.
1,500,000 MW in 2020 [19]. Section 4 presents the subject of the point estimated method,
The use of wind turbines can provide significant technical, envi- modeling wind and load uncertainty in power systems. Section 5
ronmental and economic benefits. However, wind is one of the contains simulation results followed by conclusions.
major sources suffering from uncertainty. Due to increasing pene-
trates of wind power into the systems operation and the inherent 2. Problem formulation
uncertainty in wind, there is an urgent need for a stochastic study
for power system planning, especially in dealing with modern The main goal of optimization DER is to determine the best
power systems [20,21]. Many methods have been proposed to deal location and size of new energy sources with minimized specific
with the uncertainty problem [22]. One of the most popular meth- objective function. In this section the objective functions and con-
ods is the Monte Carlo simulation to accommodate probabilistic straints are explained.
analysis through deterministic analysis [23]. With regards to
Monte Carlo simulation, it should be noted that a great deal of sim- 2.1. Objective functions
ulation and computing is required to achieve convergence [24]. An
alternative way is through analytical methods; these use mathe- In this paper, the following objective functions are considered:
matical assumptions to simplify the problems such as indepen-
dency or liner dependency of different random variables [25]. 2.1.1. Minimization of power losses
Lastly, the approximate method provides a balance between speed Minimizing the total electrical energy losses of distribution net-
and accurate by an approximate description. Among the well- work is an important goal of implementing sources, which can be
known approximate methods, the point estimated methods (PEMs) formulated as follows [21]:
stand out [26].
nbr
X
To the best of our knowledge, no research work in the area con-
f 1 ðXÞ ¼ Ri jIi j2 ð1Þ
sidered the combined ant colony optimization (ACO) and artificial
i¼1
bee colony algorithms called hybrid ACO–ABC algorithm to solve
the optimal placement and sizing of distributed energy resources where nbr is the number of branches, Ri and Ii are the resistance and
(DERs). The main contribution of this paper is to develop the the current magnitude of i th the branch, respectively.
hybrid ACO–ABC algorithm to optimally locate DERs considering
the wind and load uncertainties as a stochastic programming 2.1.2. Voltage stability index
framework. The Weibull probability density and normal distribu- Under power systems planning and operation the voltage
tion functions are used to model the wind and load uncertainties, stability is one of the most significant security indices. DER has a
respectively. Also, the probabilistic power flow based on the PEM profound impacted on the voltage stability index and it will be
is utilized to study the stochastic nature of wind generation and changed by connecting DER. In [30], Chakravorty and Das proposed
M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161 151
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
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.
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) 149–161 153
store the non-dominated solutions in different place from the bees. concentration and heuristic information. The pheromone update
In the employed bee phase of the MOABC, the position is updated can be formed as follows:
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 sij sij þ Dskij ð27Þ
with a non-dominate solution to go into the archive. Probability for (
each food source will be updated sequentially according to the Eq. k
Q
f ðwk Þ
lij 2 wk
(22). Where fit is the value of the solution in the position i associ-
Ds ¼ ij ð28Þ
0 otherwise
ated with the information provided by the external archive, it can
be defined as: The evaporation update is given by:
fiti ¼ e bni
ð25Þ sij ð1 qÞsij ð29Þ
Here, b is a positive coefficient, and ni is related to the number of where q is the constant factor reduction of all pheromones, f(wk) is
non-dominated food sources in the area of objective space. After the cost of the solution performed by ant k, and Q is a constant. The
all employed bees finish their search processes; they insert all above optimization process is terminated after a certain amount of
new non-dominated solutions into archive. Onlooker bees will iteration.
select an archive member proportional to the quality of food source
with the roulette wheel selection, and pick the leader and update its 3.3. The hybrid ACO–ABC algorithm
position using Eq. (21). When all the different types of bees (i.e.
employed, onlooker, and scouts) are evaluated, the archive is filled Optimal sitting (discrete optimization) and sizing (continuous
with the non-dominated bees. The content of the external archive optimization) of energy sources are non-linear optimization. As
has a bounded size. If the archive is full, the archive gets updated literature review, different standard algorithms such as PSO,
at the end of each iteration. In this work, the updating external ABC and ICA were developed for solving continuous optimization
archive is used similar to the adaptive grid of MOPSO [33]. The problems [35]. The discrete structure of the placement is a main
archive is divided into regions. The non-dominated solutions obstacle to get use of the continuous optimization methods.
located in more populated areas are randomly selected and Among the previously mentioned optimization methods, ACO
automatically discarded. The implemented search of the probability has attracted a lot of attention in the fields of discrete problems
proportional selection strategy spares regions of the Pareto fronts; due to its population-based search capability as well as simplicity
therefore, the diversity has been increased among the stored solu- and robustness. ACO used heuristic technique to produce a good
tions in the archive and explores the search space to a greater initial solution and determine a good search direction depending
extend. Thus convergence would be quick and the onlooker bees on the experience. It is worthy of note that this strategy often
explore better to the true region. helps ACO find a good solution and it causes ACO to be trapped
in local minima.
3.2. Ant colony optimization To overcome these disadvantages and integrates the merits of
both ACO and ABC, the hybrid ACO–ABC algorithm is introduced.
Ant colony optimization is one of the population based meta- Hybridizing ant colony optimization with artificial bee colony algo-
heuristic techniques based on the foraging behavior of real ants. rithm takes advantage of ACO and ABC by dividing optimization
They forage for food and establish the shortest paths from their problem into two parts. In each iteration, ant builds candidate solu-
nest to the food source. Ant colony optimization is a class of algo- tion to locate the DERs; thereafter, the size of DERs is optimized by
rithms whose construct solutions are based on the problem data, ABC. The proposed algorithm has several characteristic features.
and has been presented for application to discrete optimization Firstly, the triple search capability of the ABC algorithm is used
problems. In reality, ants search for food sources in a random for escaping from the local minimum solution. This causes ABC gen-
way. As soon as an ant finds a food source, they carry some food erally affect the search process since it is able to quickly find the
back to their colony. When they move along the paths, they lay a correct global optimum. Secondly, the performance of algorithm
chemical substance called pheromone as they travel. In turn, is improved by dividing the optimization problem into continuous
shorter paths will have the higher rate of pheromone trails. Each and discrete categories stages; so, it decreases the size of the search
ant makes decisions by using pheromone trails as a communica- space. Finally, it is able to overcome the drawback of classical ant
tion mechanism. Strength of pheromone trail deposited on the colony algorithm which is not suitable for continuous optimiza-
ground depends on the quality of the solution (food source) found. tions. The schematic diagram of the proposed algorithm is shown
Pheromone trails accumulate with multiple ants in shorter paths, in Fig. 3.
which causes a higher density than longer paths, thus increasing
its attractiveness. All pheromone trails are eventually reduced by
3.3.1. Multi-objective ACO–ABC algorithm
an evaporation rate. On the other side, a process of evaporation
When solving single objective optimization problems by using
presents the exploration and prevents stalling in a local minimum.
hybrid ACO–ABC, the best solution can be readily found. But, in
However, the pheromone values are updated at the end of each
multi-objective optimization problems, have set of Pareto optimal
iteration [32,34].
solution. The non-dominated solution considers several objectives
8 a
< P ðsij Þ ðgijaÞ
b
that increase the diversity of the final solution. One of the impor-
ðsim Þ ðgim Þb
j 2 N ki
Pkij ¼ m2N k
i
ð26Þ tant advantages of the Pareto method is that the system operator
: as a decision maker can choose the best compromise solution
0 otherwise
according to the requirements of the power system operations. It
where Pkij is the probability which ant k chooses to move from node i is necessary to select one optimal size of resource as the best com-
to node j. This decision depends on the pheromone level and heuris- promise solution. In this paper, a fuzzy optimization approach is
tic information. While N ki is the set of feasible neighborhoods that employed to calculate each objective function for all the non-dom-
have not yet been visited by ant k, gij is heuristic function, sij is inated solutions in the archive. The main significant issue in the
the amount of pheromone on edge i and j, and a and b are the fuzzy method is solving the optimization problem with the nature
parameters which determine the relative importance of pheromone 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.
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:
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
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
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.
Table 4
Results obtained by hybrid ACO–ABC algorithm for GT, FC and WT in the IEEE 33-bus distribution system.
Table 5
Simulation results of the three-objective optimization for scenario one.
Table 6
Simulation results of the three-objective optimization for scenario two.
zero because energy costs produced by the substation are lower words, to minimize the emission objective function, the active
than those attained by them. Cost coefficient of the substation power of all FCs are set to their maximum value. As it is obvious
bus (Prsub) is 0.044 ($/kW h) [21]. from the results, FCs active powers for cost objective function is
In the configuration of the second scenario, a combination of GT, equal to maximum value because the variable cost of fuel cell units
WT and FC is applied. The best results obtained by single optimiza- is lower than substation and GT.
tion of the objectives are shown in Table 4. Note that total environ- The study system is a distribution network having the substa-
mental pollution is the summation of the emissions produced by tion located at bus 1 and has some loads. Therefore, the simulation
the sources and substation. The best results obtained by optimizing results show the optimal location of many energy sources close to
total emissions are shown in Tables 3 and 4 and it should be noted the end of feeder. It can be seen from Table 4 that the result
that WT and FC have active roles in emission reduction. In other 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.
Table 8
Simulation results of multi-objective optimization for different scenario considering for other Weibull factors (c = 7 and k = 2.4).
Table 9
Simulation results of multi-objective optimization for different scenario considering load uncertainty.
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
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.
Table 12
Results obtained by hybrid ACO–ABC algorithm for GT, FC and WT in the IEEE 69-bus distribution system.
Table 13
Simulation results of the three-objective optimization for scenario one.
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.
Table 15
Simulation results of multi-objective optimization for different scenario.
Table 16
Simulation results of multi-objective optimization for different scenario considering load uncertainty.
tables, the multi-objective algorithm gives a slightly higher objec- in both wind energy sources and load demands. It is assumed that
tive function value than the single-objective methods for all the the loads on buses 9, 17 and 28 have the normal distribution func-
optimization levels. It shows that the gas turbine as a bigger energy tion. In this regard, the proposed algorithm is implemented again
sources can provide higher amount of active power with less emis- and the optimal location and capacity of each DER with consider-
sion than subsystems, but imposes higher electrical energy cost. ing load and wind uncertainty in the case of one and two scenarios
Energy source installation improves the voltage stability of neigh- are given in Table 9.
boring nodes, and it is also clear that the cost increases as the volt-
age stability index increases. To have better comparison between 5.2. IEEE 69-bus distribution system
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 In this part, the proposed algorithm is tested on 69-bus radial
and k4 are: 0.7, 0.1, 0.1 and 0.1, respectively. It can be concluded distribution system and it is demonstrated in Fig. 5 [39]. The origi-
from the results that the second scenario has slightly lower power nal total real power loss is 0.225 MW and most of load is far from
losses than first one. the substation. In the case of minimizing power losses, the results
The characteristics of the wind speed depend on target region. obtained by the proposed algorithm are compared to that of
Based on the Weibull distribution, higher c factor makes a greater MTLBO and tabulated in Table 10. Comparison of performances
amount of higher wind speed will be probable. In this regard, the in this case reveals that the proposed algorithm gives a better
Weibull distribution function with specified scale and shape response. In the first scenario, each case is contained 3 GT and 2
parameter is implemented to show that hybrid systems usually WT. Table 11 lists the results attained by hybrid ACO–ABC algo-
present an acceptable performance. The results of this experiment rithm. In the second scenario, it is assumed the hybrid energy
are shown in Table 8. It is interesting to note that for this particular sources are applied to the optimization problem. The simulation
data set, the values of the Weibull distributions are c = 7 and results for optimizing hybrid GT, FC and WT are explained in
k = 2.4. In order to show the effectiveness of the optimization pro- Table 12. The results of different combinations of the objective
cedure, the proposed methodology is applied to the uncertainties functions are shown in Tables 13 and 14. The detailed results of
M. Kefayat et al. / Energy Conversion and Management 92 (2015) 149–161 161
the best multi-objective solution for both scenarios are given in [12] Moradi MH, Abedini M. A combination of genetic algorithm and particle
swarm optimization for optimal DG location and sizing in distribution
Table 15. Regarding the uncertainty in the load demand, it is
systems. Int J Elect Power Energy Syst 2012;34:66–74.
assumed that buses 26, 27 and 33 have the normal distribution [13] El-Zonkoly A. Optimal placement and schedule of multiple grid connected
load. The optimal placements and capacities for the multi- hybrid energy systems. Int J Elect Power Energy Syst 2014;61:239–47.
objectives are shown in Table 16. [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.
6. Conclusion [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.
This paper presented a hybrid algorithm, namely hybrid ACO– [16] Niknam T, Bornapour M, Gheisari A. Combined heat, power and hydrogen
ABC, which takes the advantage of ant colony optimization and production optimal planning of fuel cell power plants in distribution
networks. Energy Convers Manage 2013;66:11–25.
artificial bee colony algorithm. This technique is based on ACO
[17] Mokryani G, Siano P. Optimal wind turbines placement within a distribution
for finding the location and uses the ABC to calculate the capacity market environment. Appl Soft Comput 2013;13:4038–46.
of energy sources. Also, the hybrid ACO–ABC is validated by com- [18] Ugranlı F, Karatepe E. Optimal wind turbine sizing to minimize energy loss. Int
J Elect Power Energy Syst 2013;53:656–63.
parison with results adopted from the literature. The simulation
[19] Islam MR, Mekhilef S, Saidur R. Progress and recent trends of wind energy
results show that the hybrid ACO–ABC is more effective in the mul- technology. Renew Sustain Energy Rev 2013;21:456–68.
tiple DER allocation optimization problems. In addition, the pro- [20] Sheen J-N, Tsai M-T, Wu S-W. A benefits analysis for wind turbine allocation in
posed algorithm is extended by multi-objective optimization a power distribution system. Energy Convers Manage 2013;68:305–12.
[21] Malekpour AR, Niknam T, Pahwa A, Fard AK. Multi-objective stochastic
equipped with a fuzzy decision making approach to consider the distribution feeder reconfiguration in systems with wind power generators
technical, environmental and economic optimization issues simul- and fuel cells using the point estimate method. IEEE Trans Power Syst
taneously. In this method, the decision maker could be chosen a 2013;28:1483–92.
[22] Morales JM, Perez-Ruiz J. Point estimate schemes to solve the probabilistic
most preferred solution by updates the weight factors. The pro- power flow. IEEE Trans Power Syst 2007;22:1594–601.
posed algorithm is used to determine the location and size of var- [23] Vahidinasab V. Optimal distributed energy resources planning in a
ious sources including GT, FC and WT. The optimal allocations are competitive electricity market: multiobjective optimization and probabilistic
design. Renew Energy 2014;66:354–63.
calculated for two different scenarios with different power sources [24] Ruiz-Rodriguez FJ, Gomez-Gonzalez M, Jurado F. Optimization of radial
to show the most efficient solution. The objectives functions systems with biomass fueled gas engine from a metaheuristic and
consist of total active power losses, voltage stability index, total probabilistic point of view. Energy Convers Manage 2013;65:343–50.
[25] Malekpour AR, Niknam T. A probabilistic multi-objective daily Volt/Var control
emissions and electrical energy costs. Furthermore, one of the
at distribution networks including renewable energy sources. Energy
main pivotal features of this paper is investigating the impact of 2011;36:3477–88.
stochastic wind power generation and load demands in the system [26] Mohammadi S, Mozafari B, Solimani S, Niknam T. An adaptive modified firefly
optimisation algorithm based on Hong’s point estimate method to optimal
by using PEM which provides an appropriate tool to taking
operation management in a microgrid with consideration of uncertainties.
uncertainty in power flow. The methodology presented in this Energy 2013;51:339–48.
paper could be easily implemented in the distribution networks [27] Lin Y-K, Yeh C-T, Huang P-S. A hybrid ant-tabu algorithm for solving a
for planning and operation studies. 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
References system reconfiguration using multi-objective artificial bee colony (MOABC)
algorithm. Solar Energy 2012;86:3057–71.
[1] Manzano-Agugliaro F, Alcayde A, Montoya FG, Zapata-Sierra A, Gil C. Scientific [29] Martín García JA, Gil Mena AJ. Optimal distributed generation location and size
production of renewable energies worldwide: an overview. Renew Sustain using a modified teaching–learning based optimization algorithm. Int J Elect
Energy Rev 2013;18:134–43. Power Energy Syst 2013;50:65–75.
[2] Georgilakis PS, Hatziargyriou ND. Optimal distributed generation placement in [30] Chakravorty M, Das D. Voltage stability analysis of radial distribution
power distribution networks: models, methods, and future research. IEEE networks. Int J Elect Power Energy Syst 2001;23:129–35.
Trans Power Syst 2013;28:3420–8. [31] Aman MM, Jasmon GB, Bakar AHA, Mokhlis H. A new approach for optimum
[3] Pepermans G, Driesen J, Haeseldonckx D, Belmans R, D’haeseleer W. DG placement and sizing based on voltage stability maximization and
Distributed generation: definition, benefits and issues. Energy Policy minimization of power losses. Energy Convers Manage 2013;70:202–10.
2005;33:787–98. [32] Niknam T. A new approach based on ant colony optimization for daily Volt/Var
[4] Viral R, Khatod DK. Optimal planning of distributed generation systems in control in distribution networks considering distributed generators. Energy
distribution system: a review. Renew Sustain Energy Rev 2012;16:5146–65. Convers Manage 2008;49:3417–24.
[5] Tan W-S, Hassan MY, Majid MS, Abdul Rahman H. Optimal distributed [33] Coello CAC, Pulido GT, Lechuga MS. Handling multiple objectives with particle
renewable generation planning: a review of different approaches. Renew swarm optimization. IEEE Trans Evol Comput 2004;8:256–79.
Sustain Energy Rev 2013;18:626–45. [34] Dorigo M, Blum C. Ant colony optimization theory: a survey. Theor Comput Sci
[6] Gözel T, Hocaoglu MH. An analytical method for the sizing and siting of 2005;344:243–78.
distributed generators in radial systems. Elect Power Syst Res 2009;79:912–8. [35] AlRashidi MR, AlHajri MF. Optimal planning of multiple distributed generation
[7] Al Abri RS, El-Saadany EF, Atwa YM. Optimal placement and sizing method to sources in distribution networks: a new approach. Energy Convers Manage
improve the voltage stability margin in a distribution system using distributed 2011;52:3301–8.
generation. IEEE Trans Power Syst 2013;28:326–34. [36] Peng C, Sun H, Guo J, Liu G. Multi-objective optimal strategy for generating and
[8] Devi S, Geethanjali M. Application of modified bacterial foraging optimization bidding in the power market. Energy Convers Manage 2012;57:13–22.
algorithm for optimal placement and sizing of distributed generation. Expert [37] Ghofrani M, Arabali A, Etezadi-Amoli M, Fadali MS. A framework for optimal
Syst Appl 2014;41:2772–81. placement of energy storage units within a power system with high wind
[9] Mistry KD, Roy R. Enhancement of loading capacity of distribution system penetration. IEEE Trans Sustain Energy 2013;4:434–42.
through distributed generator placement considering techno-economic [38] Atwa YM, El-Saadany EF. Probabilistic approach for optimal allocation of wind-
benefits with load growth. Int J Elect Power Energy Syst 2014;54:505–15. based distributed generation in distribution systems. IET Renew Power
[10] Soroudi A, Ehsan M. Imperialist competition algorithm for distributed Generation 2011;5:79–88.
generation connections. IET Generation Transm Distrib 2012;6:21–9. [39] Aman MM, Jasmon GB, Bakar AHA, Mokhlis H. Optimum network
[11] Doagou-Mojarrad H, Gharehpetian GB, Rastegar H, Olamaei J. Optimal reconfiguration based on maximization of system loadability using
placement and sizing of DG (distributed generation) units in distribution continuation power flow theorem. Int J Elect Power Energy Syst
networks by novel hybrid evolutionary algorithm. Energy 2013;54:129–38. 2014;54:123–33.