西南交通大学学报

第 54 卷 第 3 期 Vol.54 No.3
2019 六 3 月
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY June 2019

ISSN - 0258-2724 DOI：10.35741/issn.0258-2724.54.3.21

Research article

OPTIMAL CAPACITOR PLACEMENT IN RADIAL DISTRIBUTION

SYSTEM USING MOTH FLAME OPTIMIZATION ALGORITHM
Ihsan Salman 1, Khalid Mohammed 2, Khalid Shaker3, Hothefa Shaker 4
1
Computer Department, College of Basic Education, University of Diyala, Iraq, [email protected]
2
Computer Department, College of Science, University of Diyala, Diyala, Iraq,
[email protected]
3
College of Computer Science and Information Technology, University of Anbar, Ramadi, Iraq,
[email protected]
4
Modern College of Business & Science, Al-Khuwair Oman, [email protected]

Abstract
The environment friendly designing and running of electricity distribution structures are becoming vaster
with the conjunction of renewable electricity options into power distribution systems. This study presents
Moth Flame Optimization (MFO) approach to minimize total losses kW and energy cost and maximize
net savings with solving the voltage deviation problem through optimal site and size of capacitor banks in
electrical radial distribution networks. Two various load profiles are utilized to take a look at the modified
algorithm on IEEE 123-, and 13-bus check networks. The OpenDSS software is utilized to resolve the
power flowing via MATLAB programming interface. The results obtained are compared with IEEE
normal state and other two heuristic approaches in same study, namely, imperialist competitive algorithm
and particle swarm optimization. The proposed algorithm is more effective to minimise the total power
losses and total cost and maximise net savings and to enhance the voltage profile for various distribution
systems.

Keywords: Distribution Systems; Smart Grids; Voltage Optimization; Total Power Losses.

摘要 : 隨著可再生電力選擇與配電系統的結合，配電結構的環境友好設計和運行變得越來越大。本研究提
出了蛾火焰優化（MFO）方法，通過電氣徑向分佈網絡中電容器組的最佳位置和尺寸解決電壓偏差問題，最
大限度地降低總損耗 kW 和能量成本，並最大化淨節省。利用兩種不同的負載配置文件來查看 IEEE 123 和
13 總線檢查網絡上的修改算法。 OpenDSS 軟件用於通過 MATLAB 編程接口解決功率流。將得到的結果與同
一研究中的 IEEE 正態和其他兩種啟發式方法進行了比較，即帝國主義競爭算法和粒子群優化算法。所提出
的算法更有效地最小化總功率損耗和總成本並最大化淨節省並且增強各種配電系統的電壓分佈。

关键词: 配電系統; 智能電網; 電壓優化; 總功率損耗。

I. INTRODUCTION regulation and integrating of new energy sources

Last few years, electricity systems have been into the planning and working mentalities.
growing, exhibiting these modifications as cancel Through the view of operational point, control
 Salman et al. / Journal of Southwest Jiaotong University/ Vol.54.No.3 June 2019

means which include the coordinated technique of operation of musical group (harmony search) [12],
tap changeable transformers, like capacitors, are and the direction finding method of moths in
wanted due to the fact masses fee are not fixed via nature, which is called transverse direction (moth
time and the production of new energy capacity flame optimization [MFO]) [13] were developed.
sources are sporadic. Optimization of Voltage As shown in the literature summary, the
(VO) is a tremendous technological know-how application of MFO has not been discussed so far
that has been rescuing a lot of money in lost to solve for capacitor size and placement in radial
electrical power since the commencing of the new distribution system. Therefore, this motivates to
millenary [1]. Rise request utilized to be utilize MFO to find best place and size of
administer by voltage decrease [2]. The capacitors capacitor in power distribution system to reduce
are using as another way to helping system total power loss and total cost and maximize net
operation for get better power factor, minimize savings and to enhance the voltage profile.
power losses and improve voltage profile [3]. The OpenDSS software is applied to implement power
capacitors are usually utilized to supply imaginary influx and load influx to discovery the parameters
power kVAr reparation in power distribution of power system such as voltage profile, power
systems. The installation of capacitors bank factor, power losses , real and imaginary power
includes definition of size kVAr and capacitors flow in every line, and etc. [14], [15]. Two test
location. Selecting better site of the capacitor will systems IEEE 13-, and 123-node are applied in
minimize the demand of the imaginary power, this work. Results acquired from the suggested
which therefore increase the pure savings and algorithm has been compared with those gained
utilized to keep voltage profile values within from IEEE standard case, PSO and ICA.
allowable limits. For example, in an algorithm MATLAB and an open power distribution system
that rely on genetic algorithm, Fuzzy Logic and simulation gadget OpenDSS [14], [15] are utilized
Dynamic Programming techniques are utilized for in the emulations.
best capacitor allocation in distribution network. The remnant of the paper is orderlies as
Flower pollination optimization algorithm and follows. Section II suggested the mathematical
power loss index was used for best sizing and problem formulation. Section III briefly dissect
locations of capacitors in power radial distribution MFO, ICA, and PSO and modified algorithms for
networks in [4]. Gravitational search algorithm solving the optimal capacitor placement problem.
was used for optimal capacitor placement in [5], Section IV presents the experiments and the
whereas a teaching–learning-based optimization simulation results. Finally, Section V presents the
was used for the same aim in [6]. Capacitors can conclusions, followed by the References used in
as well be utilized to reduce the effects of this paper.
harmonics in distribution networks; the harmony
search algorithm was utilized for this aim in [7]. 1. MATHEMATICAL PROBLEM FORMULATION
Capacitor size and site problem have been
resolved through other heuristics, for example ant A. Objective function
colony algorithm [8], clonal selection algorithm The capacitors costs and the power losses cost
[8], and particle swarm optimization [9]. represent the total cost in power system. The aim
Making the better potential results with the of best capacitor placement and size of
ready resources are constantly an objective in distribution network is to improvement a specific
scientific and problems of engineering. Generally aim function such minimize losses and cost and
two approaches are utilizing to solve problems of enhance the voltage values and maximize annual
optimization. The first method is depend on savings. The aim function in this work can be
arithmetic analysis, whereas the second method is written as follows:
depend on numerical computations. Numeral
Nb
optimization procedures can be classified into
derivative and non-derivative. The non-derivative Minimize x = Pcost . P i loss
process are applied when the derivatives of the i=1
performed sample are difficult to discover or not Nc
exist a numerical task connected to the sample. + Capj cost . Capj
These techniques are generally enthused through j=1
nature. The greatest public sample is GA, which Nb
reverses the development operation in the natural
+ |Vk − 1|
world [10]. Afterward, approaches motivated
through the attitude of fish and birds (particle k=1
swarm optimization [PSO]) [11] improvised
 西南交通大学学报
第 54 卷 第 3 期 Vol.54 No.3
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
2019 六 3 月 June. 2019

Minimize x = 2 ≤ li ≤ Lmax (4)

Nb Where, li indicate the site of the capacitor in bus i,
Pcost . Pi loss + and Lmax indicate the maximum bus location.
i=1
Nc j j
j=1 Cap cost . Cap +
Nb 2. HEURISTIC ALGORITHMS
k=1
|Vk − 1| (1)
Where, x denotes the aim of better capacitor A. Moth Flame Optimization (MFO)
placement and size, Nb indicate the number of Public method: MFO is a meta-heuristics
buses, Pcost represents cost of real power losses algorithm sophisticated by means of Mirjalili
[168 $/kW], Pi loss represents real power losses on in 2015; the optimization of this method
bus i [kW], Nc represents number of capacitors. reflects transversal direction, which is the
Capj cost is the cost of capacitor j [$/kVAr], Capj style of move of moths in the natural world at
is the size of capacitor j [kVAr], and Vk is the night [13]. Nearly 160,000 various groups of
voltage magnitude of bus k.The capacitors size insects, include moths, are present in nature.
and cost as shown in Table 1. Moths have two life phases: maggot and
mature phase. These insects are a lot like to
Table 1: Possible choice of capacitor size in kVAr and cost in the family of butterflies only possess a
$/kVAr specific function when shifting at night-time
Q kVAr $/kVAR Q kVAr $/kVAR Q kVAr $/kVAR [16]. Moths go lengthy spaces and rectum
50 0.5 1350 0.207 2850 0.183
100 0.5 1500 0.201 3000 0.180
lines through way of preserve a fixed
150 0.5 1650 0.193 3150 0.195 perspective together with the moonlight. This
300 0.35 1800 0.187 3300 0.174 method is so nice for touring these insects
450 0.253 1950 0.211 3450 0.188 specifically while the lighting supply is far
600 0.220 2100 0.176 3600 0.170 from them. When the lighting source is near,
750 0.276 2250 0.197 3750 0.183 the moths flies round it with a spiral
900 0.183 2400 0.170 3900 0.182
1050 0.228 2550 0.189 4050 0.179
direction, and collect with it in last. The
1200 0.170 2700 0.187 moths appear the nominee resolutions and
moth’s site in examine space appear
B. Constraints problem’s directions. So, the moths have the
The best capacitor siting of distribution ability to fly in one dimensions area or more
network problem has the following constraints: with replace the site vectors. Figure 1
 The voltage volume on the bus is the initial illustrates a MFO algorithm flowchart.
constraint in optimal capacitor placement and This optimization algorithm is executed as
sizing problem. It is necessary preserved within follows:
the specific limits : 1. Set the fixed parameters: size of dimensions,
size of moths, higher number of repetitions,
0.95 ≤ Vi ≤ 1.05 (2) and maximum and minimum boundaries.
Where, Vi is the voltage size on bus i. 2. Randomly be the initial population
 The installed capacitors size at each bus must be depending on the population size, size of
less than the total effective reactive power kVAr dimensions, maximum and minimum
in the test system as follows: bounds as follow:

Nb Mo11 ⋯ Mo1d
i
Nc
𝑖=1 Cap < Qj (3) Mo21 … Mo2d
j=1 Mo = (5)
⋮ ⋱ ⋮
Where, Nc indicate to the number of
Mon1 ⋯ Mond
capacitors, Capi show the size of capacitor i, Nb
j
indicate the number of buses, and Q is the where n denotes size of moths, and d represent
reactive power kVAr in bus j. size of dimensions.
 The optimal capacitor location between bus 3. Compute and order the identical cost values
number two and number of buses in simulation for each an individual's.
system. The bus number one is represent
substation.
 Salman et al. / Journal of Southwest Jiaotong University/ Vol.54.No.3 June 2019

OMo1 OF1
OMo2 OF2
OMo = (6) OF = (8)
⋮ ⋮
OMon OFn
Where n indicate to number of moths Where n indicate to size of moths, and d
4. Initialize inhabitance of flames and flames indicate to size of dimensions.
efficiency that is evenness sort population of 5. Reiterate until the stopping state is met.
moths, and sort moth’s cost values, 6. Compute the space of the ith moth for the jth
respectively. flame through neutralization:
Di = |Fj-Moi| (9)
7. Modernize moths site utilizing spiral task as
follow:
F11 ⋯ F1d Moi = Di . ebt . cos(2ℼt)+Fj (10)
F … F2d Where, Di space, b is a fixed number (b=1),
F = 21 (7)
⋮ ⋱ ⋮ and t is a random number in [-1, 1].
Fn1 ⋯ Fnd 8. Modernize flames site that is the same better
former and actual moth’s site, identical as
with flames efficiency.

Begin

Initialize parameters values

Initialize initial populations utilize specified limits

Compute the aim function to all moths in population

No Yes
Repetition = 1

Modernize flames site that is equal better previous and Sort the population of moths and save as an
current site of moths initial flames population

Sort moth’s efficiency and save as an initial

Update flames efficiency that is equal better previous
flames efficiency
and current efficiency values of moths

Update moth’s site

No
Achieve the stop
Modernize flames size state?

Yes
End

Figure 1: Flowchart of MFO algorithm

 西南交通大学学报
第 54 卷 第 3 期 Vol.54 No.3
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
2019 六 3 月 June. 2019

Begin

Initialize parameters values

At random generate the site and size of the capacitors by fixed lower and upper limits as in
Eq. (5)

Implement a load and power flow utilize the identified load profile

Compute the moth’s efficiency as in Eq. (6) using Eq. (1)

Choice better moth’s site and efficiency as a flames site and efficiency utilize Eqs. (11) and (12).

Modernize moths site (capacitors site and size) using Eqs. (9) and (10)

No Achieve the stop state?

Yes

End

Figure 2. Flowchart of modified MFO algorithm.

F = Sort (Moi-1, Moi) (11)

OF = Sort (OMoi-1, OMoi) (12) Step 3: Utilize the particular load profile
Where, i indicate to actual repetition. and solution candidates to implement a
load and power flux , then compute the
 MFO for resolve optimal capacitor moth efficiency of the trial system use
site problem: Figure 2 shows the Equations (1) as in Equation (6).
modify MFO algorithm flowchart, and Step 4: Choice the better moth site as
the proceedings as follows: flame site and the better moth cost value
Step 1: Initialize set fixed variables such as the flame cost value using Equations
as: size of moths N, size of (11) and (12), as shown in Equations (7)
dimensions D, maximum repetitions and (8), respectively.
Maxit, lower bound value lb, and upper Step 5: Compute the space among
bound value ub. Set voltage size, and flames and moths, then compute
capacitor sites and size limits. different moth site by Equations (9) and
Step 2: At random generate the site and (10).
size of the capacitors, then form the  Step 6: Reiterate procedures 3–5 till
initial moth site as in Equation (5). the ending case is seen.
 Salman et al. / Journal of Southwest Jiaotong University/ Vol.54.No.3 June 2019

brought with the aid of Esmaeil Atashpaz-

Gargari and Lucas in 2007 [17]. This algorithm

Begin

Setting the empires

Transfer the colonies to their pertinent imperialist

N
A colony in empire has been cost less of
imperialist?
o

Y
es
Exchange site of colony and imperialist

Calculate the full all empires cost

Choice the feeblest colony from the frailest empire then give it to the empire which
owns the best probability to possess it

N
There an empire without colonies? o

Y
es
Remove this empire

N
Achieve the stop state?
o

Y
es
End

Figure 3: Flowchart of ICA [17]

is a new socio-politically motivated global
search strategy
B. Imperialist Competitive Algorithm (ICA)
General approach: ICA is a new meta-heuristic
algorithm suggested for continuous optimization that has been added for dealing with various
issues. This algorithm stimulated by means of optimization tasks. Figure.3 presents a flowchart
the imperialistic competition process. It is of ICA [17].
 西南交通大学学报
第 54 卷 第 3 期 Vol.54 No.3
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
2019 六 3 月 June. 2019

4- The colonies are shift to the site of the

The method is conducted as follows: imperialist through X units to start the process
1- Generating initial Empires. For the Dn- of optimization. The DN country of
dimensional optimization problem a 1 x Dn population is generated, and Nimp represents
dimensional vector regarded as a solution the most powerful population, whose
that is named country as shown: members have been selected as imperialists
countryi = [P1, P2, P3, … , PDN] (13) (the controller coefficients sets with same cost
Where, DN denotes dimensions of function). The remaining Ncol countries are
problem. Unplanned for Pi values must the colonies (the controller coefficients sets
be within the upper and lower domain with a high cost function), each of which is a
for every variable. part of one of the abovementioned empires.
2- The cost of every country is calculated the colonies move toward the imperialists
through the function f as shown: with the attraction policy, along Mx units and
f (countryi) = f (P1, P2, P3,…, PDN) (14) are situated in a new position. Mx is a random
3- Divided the countries into imperialists, which variable with regular distribution and can be
represents the most powerful countries, and expressed as follows:
colonies, which represents the others left. To Mx ~ U (0, β. ds) (16)
distribute the colonies between imperialists Where, β is a digit bigger than 1 (β = 2)
proportionally, the following normalized cost and ds is the space between colony and

Begin

Initialize variables value

At random generate the country as in Eq. (17)

Implement a load and power flow utilize the identified load profile

Compute the aim function for all the inhabitants as in Eq. (1)

Choice the colonized countries and imperialist as in Eqs. (14) and (15)

Modernize capacitors site and size as in Eq. (16)

No
Achieve the stop
state?

Yes

End

Figure 4. Modified ICA

of each imperialist is described as shown: imperialist.

Cn = cn - max (ci) (15)
Where, cn represent cost of nth imperialist,
and Cn represents normalized value.
 Salman et al. / Journal of Southwest Jiaotong University/ Vol.54.No.3 June 2019

ICA for resolve optimal capacitor site Step 5: Modernize the site and size of the
problem: Figure 4 shows a flowchart of the capacitors for wholly empires as in Equation
modify ICA algorithm, and the proceedings as (16).
follows: Step 6: Reiterate procedures 3–5 till the
ending case is seen.
Step 1: Initialize set fixed variables such
as: inhabitance size Npop, maximum repetition C. Particle Swarm Optimization (PSO)
number Maxit, number of imperialist General approach: PSO is initially developed
countries Nimp, and number of colony by Eberhart and Kennedy in 1995, and was once
countries Ncol. Set voltage size, and capacitor stimulated by way of the public conduct of birds
sites and size limits. clustering and fishes education [18]. The birds in
Step 2: Create randomly the site and size a herd are described as particles in the PSO
of capacitors, then form the first country as method. These particles can be flying over a
follows: examine space. A particle’s site in the examine
countryi = [l1, . . . , lm, Cap1, . . . , Capm] (17) space look as one answer for the issue. New
Where, m is the amounts of capacitors various answer is created when a particle shift to
bank. a new site in the examine space. Each answer
Step 3: Utilize the particular load profile can be examine by the usage of an aim function
and solution candidates to implement a load that affords a value of the advantage of the
and power flux of the exam system as in answer. The path and speed of every element
Equations (1). can be transferring alongside every dimension of
Step 4: Define the colonized countries and the examine space; so, will be shifting with each
imperialist depend on cost values as in generation of motion. PSO is typically viewed to
Equations (14) and (15). be an evolutionary computation (EC) model.

Begin

At random generate velocities and individual sites

utilize specific upper and lower limits

Compute the aim purpose for all particle in the inhabitants

Locate the global and local better site

Modernize particle velocity and site

No
Achieve the
stop state?

Yes

End

Figure 5. PSO algorithm

 西南交通大学学报
第 54 卷 第 3 期 Vol.54 No.3
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
2019 六 3 月 June. 2019

Other EC samples consist of evolutionary By utilize the above note, the method is execute
strategies (ES), evolutionary programming (EP), as follows:
genetic programming (GP), and Genetic 1. Initialize a set of fixed parameters
Algorithms (GA) [19]. Every particle (i) such as: size of the swarm, variable
preserve the following data [20]: of the problem, high quantity of
 Xi = The actual site of the particle. iterations, maximum and minimum
 Vi = The actual speed of the particle. bounds and etc.
 Yi = The personal better site of the particle 2. At random initialization of particle
(pbest); the better site visited until now by locations depending on the fixed
the particle. parameters.
 Ŷ = The global better site of the swarm 3. Randomly initialization of particle
(gbest); the better site visited until now by speeds depending on the fixed
the entire swarm. parameters.
Figure 5 shows a flowchart of the PSO 4. Reiterate till the ending case is seen.
algorithm. 5. Estimate the cost values utilize the
aim function.
6. Locate gbest and pbest value.
7. Locate the change element speed
vector as follows:

Begin

Set variables value

At random generate velocities and individual sites utilize specific

upper and lower limits

Implement a load and power flow utilize the identified load profile

Compute the aim function for all particle in the inhabitants as in Eq. (1)

Choice global and local best [(gb) , (lb)]

Modernize particle velocity, capacitors site and size utilize Eqs. (18)- (20)

No
Achieve the
stop state?

Yes

End

Figure 6: Modified PSO algorithm

Vi (t+1) = W.Vi(t)+c1.r1(t).(Yi(t)-
Xi(t))+c2.r2(t).(Ŷ(t)-Xi(t)) (18)
 Salman et al. / Journal of Southwest Jiaotong University/ Vol.54.No.3 June 2019

high speed value Vmax. Set the

where, t represents actual repetition, voltage size, possible capacitor sites,
r1(t) and r2(t) denote regular random and size limits.
numbers among 0 and 1, and W  Step 2: Create randomly initial
represents the inertia weight, which particle speed, define the site and size
is calculated as follows: of the capacitors, then form particle
W = (Maxit-t)/Maxit (19) site as follows:
8. Locate the change element site vector particlei = [l1, . . . , lm, Cap1, . . . ,
as follows: Capm] (21)
Xi (t+1) = Xi (t) +Vi (t+1) (20) Where, m denote the numbers of
capacitors bank.
 PSO for resolve optimal capacitor  Step 3: Utilize the particular load
site problem: Figure 6 shows a profile and solution candidates to
flowchart of the modify PSO algorithm, implement a load and power flux of
and the proceedings as follows: the exam system as in Equations (1).
 Step 1: Initialize set fixed variables  Step 4: Select local and global best,
such as: group size Psize, element of then locate change particle speed and
the problem Npar, high number of sites as in Equations (18)–(20).
repetitions Maxit, cognitive parameter  Step 5: Reiterate procedures 3 and 4
c1, public parameter c2, high bound till the ending case is seen.
value ub, lesser bound value lb, and

Figure 7. IEEE 13-node map [21]

 西南交通大学学报
第 54 卷 第 3 期 Vol.54 No.3
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
2019 六 3 月 June. 2019

system. The node maps of the trial system are be

seen in Figures 7, and 8, respectively. The sum
3. SIMULATION RESULTS of population and iteration are taken 30 and
The suggested optimization sample for 1000, respectively for all simulation in this study.
placement of capacitor and size kVAr has been
experimented on IEEE 13- and 123-node trial

Figure 9: Simulation I voltage values on IEEE 13-node trial system, MFO optimal capacitors
placement with standard case and without capacitor.
 Salman et al. / Journal of Southwest Jiaotong University/ Vol.54.No.3 June 2019

A. IEEE 13-node trial system respectively, and in simulation II is 3463 kW

The all information around this test system and 2102 kVAr respectively. The MFO optimal
such as line and bus data have been explain in outcomes are matched with the standard IEEE
[21]. The gross active power load kW and state, without capacitors installation case and
reactive power load kVAr of this trial system in other methods which is shown in Table 2 for
simulation I is 2364 kW and 2102 kVAr simulation I and Table 3 for simulation II.

Table 3: Simulation II, performance of MFO over without control, standard case, PSO method, ICA method on 13 bus
RDN.
Title Witho Standa PSO ICA MFO
ut rd case
contro
l
Total losses (kW) 130.3 110.8 107.7 107.9 107.5
% Loss reduction - 15% 17.3% 17.2% 17.5
%
Candidate buses - 675 675 675 675
611 633 684 652
634 634 634
652 633 645
Optimal capacitor size (kVAr) - 600 50 50 50
100 50 600 50
1500 1050 1500
50 50 50
Total kVAr - 700 1650 1750 1650
Annual cost for kW loss ($168/kW) 21890 18614 18093 18127 18060
.4 .4 .6 .2
Annual capacitor cost ($/kVAr) - 182 376.5 421.4 376.5
Total annual cost ($) 21890 18796 18470 18548 18436
.4 .4 .1 .6 .5
Net savings ($) 0 3094 3419. 3341. 3453.
9 8 9
% Savings 0.0% 14.1% 15.6% 15.3% 15.8
%
Minimum voltage (p.u.) 0.892 0.973 0.967 0.966 0.968
Maximum voltage (p.u.) 0.999 1.047 1.05 1.048 1.048
Mean voltage (p.u.) 0.946 1.004 1.010 1.009 1.01
Standard deviation voltage (p.u.) 0.035 0.021 0.024 0.024 0.023
 西南交通大学学报
第 54 卷 第 3 期 Vol.54 No.3
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
2019 六 3 月 June. 2019

Figure 10: Simulation II voltage values on IEEE 13-node trial system, MFO optimal capacitors
placement with standard case and without capacitor

Figure 11: MFO iteration versus total power losses (kW) on 13-node trial system in simulation I

Figure 12: MFO iteration versus total power losses (kW) on 13-node trial system in simulation II.

The results of the proposed algorithm in algorithms, without control, and standard case.
Tables 2 and 3 are better than those of other Figures 9 and 10 shows the voltage profile for
 Salman et al. / Journal of Southwest Jiaotong University/ Vol.54.No.3 June 2019

standard case, without capacitors case and MFO matched with the standard IEEE state, without
optimal capacitors placement on IEEE 13-node capacitors installation case and other methods
trial system in simulation I and simulation II which is shown in Table 4 for simulation I and
respectively. The propose algorithm iteration Table 5 for simulation II.
versus total active power losses (kW) of The results of the proposed algorithm in
simulation I and II on 13-node trial system Tables 4 and 5 are better than those of other
shown in Figures 11 and 12 respectively. algorithms, without control, and standard case.
Figures 13 and 14 shows the voltage profile for
B. IEEE 123 bus exam system standard case, without capacitors case and MFO
The information around this test system like optimal capacitors placement on IEEE 123-node
line and bus data have been explain in [22]. The trial system in simulation I and simulation II
gross active power load kW and reactive power respectively. The proposed algorithm iteration
load kVAr of this trial system in simulation I is versus total active power losses (kW) of
2305 kW and 1920 kVAr respectively, and in simulation I, and II on 123-node trial system be
simulation II is 3435 kW and 1920 kVAr seen in Figures 15, and 16 respectively.

Table 4: Simulation I, performance of MFO over without control, standard case, PSO and ICA on 123 bus RDN.
Title Witho Standa PSO ICA MFO
ut rd case
contro
l
Total losses (kW) 63 47.5 44.5 43.9 43.1
% Loss reduction - 24.6% 29.4% 30.3% 31.6%
Candidate buses - 83 76 72 76
88 49 47 70
90 114 114 114
92 15 160 44
Optimal capacitor size (kVAr) - 600 50 50 50
50 50 50 100
50 50 50 50
50 1650 100 150
Total kVAr - 750 1800 250 350
Annual cost for kW loss ($168/kW) 10584 7980 7476 7375.2 7240.8
Annual capacitor cost ($/kVAr) - 207 393.45 125 175
Total annual cost ($) 10584 8187 7869.45 7500.2 7415.8
Net savings ($) 0 2397 2714.55 3083.8 3168.2
% Savings 0.0% 22.7% 25.6% 29.1% 29.9%
Minimum voltage (p.u.) 0.933 0.996 0.997 0.999 0.99
Maximum voltage (p.u.) 0.999 1.048 1.04 1.039 1.045
Mean voltage (p.u.) 0.964 1.024 1.023 1.019 1.019
Standard deviation voltage (p.u.) 0.016 0.015 0.013 0.014 0.014

Figure 13: Simulation I voltage values on IEEE 123-node trial system, MFO optimal capacitors placement with
standard case and without capacitor.

respectively. The MFO best outcomes are

 西南交通大学学报
第 54 卷 第 3 期 Vol.54 No.3
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
2019 六 3 月 June. 2019

Figure 14: Simulation II voltage values on IEEE 123-node trial system, MFO optimal
capacitors placement with standard case and without capacitor.

Figure 15: MFO iteration versus total power losses (kW) on 123 bus system in simulation I

Figure 16: MFO iteration versus total power losses (kW) on 123 bus system in simulation II.

enhance the voltage profile, cost minimization,

and maximize net savings. The proposed MFO
algorithm and open source OpenDSS software
4. CONCLUSION through MATLAB coding interface are applied
A newly meta-heuristics Moth Flame to resolve the power inflow and optimization
optimization algorithm for optimum capacitor issues. To prove the implementation of the
size and site on radial distribution system suggested approach, the test system IEEE 13,
resulted the total power losses (kW) reduction, and 123-bus are executed. Moreover, the
 Salman et al. / Journal of Southwest Jiaotong University/ Vol.54.No.3 June 2019

outcomes of proposed MFO are compared [8] D. Kaur and J. Sharma, "Multiperiod shunt
together with standard state, without capacitor capacitor allocation in radial distribution
placement, PSO, and ICA as in Tables 2-5. The systems," International Journal of
comparison affirm the effectiveness and surpass Electrical Power & Energy Systems, vol.
of the proposed MFO algorithm over the other 52, pp. 247-253, 2013.
techniques. http://dx.doi.org/10.1016/j.ijepes.2013.03.0
26.
REFERENCES [9] K. Prakash and M. Sydulu, "Particle swarm
[1] L. M. Rios and N. V. Sahinidis, optimization based capacitor placement on
"Derivative-free optimization: a review of radial distribution systems," in Power
algorithms and comparison of software Engineering Society General Meeting,
implementations," Journal of Global 2007. IEEE, 2007, pp. 1-5.
Optimization, vol. 56, pp. 1247-1293, 2013. [10] D. E. Goldberg and J. H. Holland, "Genetic
[2] R. Uluski and M. Mcgranaghan, "Load algorithms and machine learning," Machine
models for voltage optimization," in learning, vol. 3, pp. 95-99, 1988.
Proceedings of the International [11] R. Eberhart and J. Kennedy, "A new
Conference on Electricity Distribution, optimizer using particle swarm theory," in
2011, p. 1312. Micro Machine and Human Science, 1995.
[3] A. Elsheikh, Y. Helmy, Y. Abouelseoud, MHS'95., Proceedings of the Sixth
and A. Elsherif, "Optimal capacitor International Symposium on, 1995.
placement and sizing in radial electric [12] Z. W. Geem, J. H. Kim, and G. V.
power systems," Alexandria Engineering Loganathan, "A new heuristic optimization
Journal, vol. 53, pp. 809-816, 2014. algorithm: harmony search," simulation,
http://dx.doi.org/10.1016/j.aej.2014.09.012. vol. 76, pp. 60-68, 2001.
[4] A. Abdelaziz, E. Ali, and S. A. Elazim, [13] S. Mirjalili, "Moth-flame optimization
"Optimal sizing and locations of capacitors algorithm: A novel nature-inspired heuristic
in radial distribution systems via flower paradigm," Knowledge-Based Systems, vol.
pollination optimization algorithm and 89, pp. 228-249, 2015.
power loss index," Engineering Science and http://dx.doi.org/10.1016/j.knosys.2015.07.
Technology, an International Journal, vol. 006.
19, pp. 610-618, 2016. [14] R. Dugan, "Open distribution simulations
http://dx.doi.org/10.1016/ system workshop: Using open dss for smart
j.jestch.2015.09.002. distribution simulations," in EPRI PQ
[5] Y. M. Shuaib, M. S. Kalavathi, and C. C. Smart Distribution 2010 Conference and
A. Rajan, "Optimal capacitor placement in Exhibition, Québec, 2010, pp. 14-17.
radial distribution system using [15] P. Anbarasan and T. Jayabarathi, "Optimal
gravitational search algorithm," Reactive Power Dispatch Using Moth-
International Journal of Electrical Power Flame Optimization Algorithm,"
& Energy Systems, vol. 64, 2015. International Journal of Applied
http://dx.doi.org/10.1016/j.ijepes.2014.07.0 Engineering Research, vol. 12, pp. 3690-
41 3701, 2017.
[6] S. Sultana and P. K. Roy, "Optimal [16] Benjamin Durakovic, Muris Torlak,
capacitor placement in radial distribution "Simulation and experimental validation of
systems using teaching learning based phase change material and water used as
optimization,"International Journal of heat storage medium in window
Electrical Power & Energy Systems, vol. applications", J. of Mater. and Environ.
54, pp. 387-398, 2014. Sci., Vol. 8, No. 5, (2017), ISSN: 2028-
http://dx.doi.org/10.1016/j.ijepes.2013.07.0 2508, pp 1837-1846..
11. [17] Kennedy J, Eberhart R , “Particle swarm
[7] K. Muthukumar and S. Jayalalitha, optimization”, Proceedings of the 1995
"Harmony search approach for optimal IEEE International Conference on Neural
capacitor placement and sizing in Networks, pp. 1942-1948, IEEE Press.
unbalanced distribution systems with [18] Ogudo, K.A.; Muwawa Jean Nestor, D.;
harmonics consideration," in Advances in Ibrahim Khalaf, O.; Daei Kasmaei, H. A
Engineering, Science and Management Device Performance and Data Analytics
(ICAESM), 2012 International Conference Concept for Smartphones’ IoT Services and
on, 2012, pp. 393-398.
 西南交通大学学报
第 54 卷 第 3 期 Vol.54 No.3
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
2019 六 3 月 June. 2019

Machine-Type Communication in Cellular [20] W. H. Kersting, "Radial distribution test

Networks. Symmetry 2019, 11, 593 feeders," in 2001 IEEE Power Engineering
[19] D. Van der Merwe and A. P. Engelbrecht, Society Winter Meeting. Conference
"Data clustering using particle swarm Proceedings (Cat. No. 01CH37194), 2001,
optimization," in Evolutionary pp. 908-912.
Computation, 2003. CEC'03. The 2003 [21] Radial Distribution Test Feeders. [Online].
Congress on, 2003, pp. 215-220. Available:ttps://ewh.ieee.org/soc/pes/dsaco
m/testfeeders/testfeeders.pdf .
[9] K. Prakash 和 M. Sydulu，“基於粒子群優
参考文:
化的徑向分佈系統電容放置”，電力工程
[1] L. M. Rios 和 N. V. Sahinidis，“無衍生優 學會大會，2007 年.IEEE，2007，pp.1-5。
化：算法評估和軟件實施比較”，全球優 [10] D. E. Goldberg 和 J. H. Holland，“遺傳算
化 雜 誌 ， 第 一 卷 。 56 ， pp.1247-1293, 法和機器學習”，機器學習，第一卷。 3，
2013。 pp.95-99, 1988。
[2] R. Uluski 和 M. Mcgranaghan，“用於電壓 [11] R. Eberhart 和 J. Kennedy，“使用粒子群
優化的負載模型”，載於“國際電力分配 理論的新優化器”，“微機械和人類科學”，
會議論文集”，2011 年，第 7 頁。 1312。 1995 年.MHS'95。，第六屆國際研討會
[3] A. Elsheikh，Y。Helmy，Y。Abouelseoud 論文集，1995 年。
和 A. Elsherif，“徑向電力系統中的最佳 [12] Z. W. Geem ， J 。 H. Kim 和 G. V.
電 容 器 放 置 和 尺 寸 ” ， Alexandria Loganathan ， “ 一 種 新 的 啟 發 式 優 化 算
Engineering Journal，vol。 53，pp.809- 法：和諧搜索”，模擬，第一卷。 76，
816, 2014。 pp.60-68, 2001。
http://dx.doi.org/10.1016/j.aej.2014.09.012。 [13] S. Mirjalili，“蛾火焰優化算法：一種新
[4] A. Abdelaziz，E。Ali 和 S. A. Elazim，“通
穎的自然啟發式啟發式範例”，“知識系
過花授粉優化算法和功率損耗指數在徑
統”，第一卷。 89，pp.228-249, 2015。
向分佈系統中優化尺寸和電容器位置”，
http://dx.doi.org/10.1016/j.knosys.2015.07.006 。
“工程科學與技術”，國際期刊，第一卷。
[14] R. Dugan，“開放分佈模擬系統研討會：
19，pp.610-618, 2016。
使用開放式 dss 進行智能分佈模擬”，在
http://dx.doi.org/10.1016/j.jestch.2015.09.002。
EPRI PQ Smart Distribution 2010 會議和
[5] Y. M. Shuaib，M。S. Kalavathi 和 C. C. A.
展覽中，魁北克，2010 年，第 14-17 頁。
Rajan，“使用重力搜索算法的徑向分佈
[15] P. Anbarasan 和 T. Jayabarathi，“使用蛾
系統中的最佳電容器放置”，國際電力和
火焰優化算法的最佳無功功率調度”，國
能源系統期刊，第一卷。 64，2015。
http://dx.doi.org/10.1016/j.ijepes.2014.07.041 際應用工程研究期刊，第一卷。 12，
[6] S. Sultana 和 P. K. Roy，“使用基於教學學 pp.3690-3701, 2017。
習的優化的徑向分佈系統中的最佳電容 [16] Benjamin Durakovic，Muris Torlak，“用
器放置”，國際電力與能源系統雜誌，第 於窗戶應用中作為儲熱介質的相變材料
一卷。 54，第 387-398 頁，2014 年。 和水的模擬和實驗驗證”，J.of Mater。和
http://dx.doi.org/10.1016/j.ijepes.2013.07.011。 環 境 。 Sci 。 ， Vol 。 8 ， No 。 5 ，
[7] K. Muthukumar 和 S. Jayalalitha，“在諧波 （2017），ISSN：2028-2508，pp 1837-
考慮的非平衡配電系統中優化電容器放 1846 ..
置和尺寸的和諧搜索方法”，“工程，科 [17] Kennedy J，Eberhart R，“Particle swarm
學和管理進展”（ICAESM），2012 年國 optimization” ， Proceedings of the 1995
際會議，2012 年，pp 393-398 IEEE International Conference on Neural
[8] D. Kaur 和 J. Sharma，“徑向分佈系統中的 Networks，pp.1942-1948，IEEE Press。
多周期並聯電容器分配”，“國際電力與 [18] Ogudo，K.A。; Muwawa Jean Nestor，
能源系統雜誌”，第一卷。 52，pp.247- D。; Ibrahim Khalaf，O。; Daei Kasmaei，
253, 2013。 H。智能手機在蜂窩網絡中的物聯網服
http://dx.doi.org/10.1016/j.ijepes.2013.03.026。 務和機器類通信的設備性能和數據分析
概念。 Symmetry 2019, 11, 593
 Salman et al. / Journal of Southwest Jiaotong University/ Vol.54.No.3 June 2019

[19] D. Van der Merwe 和 A. P. Engelbrecht，

“使用粒子群優化的數據聚類”，在
Evolutionary Computation ， 2003
中.CEC'03。 2003 年大會，2003 年，第
215-220 頁。
[20] W. H. Kersting，“徑向分佈測試饋線”，
2001 年 IEEE 電力工程學會冬季會議。
會 議 論 文 集 （ 目 錄 號 01CH37194 ） ，
2001 年，第 908-912 頁。
[21] 徑向分佈測試饋線。 [線上]。可用：
ttps ：
//ewh.ieee.org/soc/pes/dsacom/testfeeders/t
estfeeders.pdf 。