International Journal of Applied Power Engineering (IJAPE)

Vol. 12, No. 1, March 2023, pp. 71~82

ISSN: 2252-8792, DOI: 10.11591/ijape.v12.i1.pp71-82  71

Optimal capacitor allocation for minimizing cost of energy loss

in active distribution network with different load levels

Mya Myintzu1, Kyaw Myo Lin1,2, Khine Zin Oo1

1
Power System Research Unit, Department of Electrical Power Engineering, Mandalay Technological University, Mandalay, Myanmar
2
Department of Electrical Power Engineering, Pyay Technological University, Pyay, Myanmar

Article Info ABSTRACT

Article history: This paper proposes an effective technique to allocate a shunt capacitor
using a meta-heuristic optimization-approach for minimizing the cost of
Received Oct 14, 2022 energy losses. Particle swarm optimization (PSO) algorithm incorporated
Revised Dec 3, 2022 with MATPOWER is trained for minimizing the targeted function of the
Accepted Dec 15, 2022 optical capacitor placement problem. The 115-node active distribution
network supplied by MESC in the Mandalay distribution area is applied as a
scenario. The results including active power loss and system energy loss,
Keywords: size and location of the optimized capacitor, voltage profiles, and the cost of
energy loss are attained and analyzed. The simulations results revealed that
Cost of energy loss the optimal allocation of the capacitor provides a significant reduction in the
Load levels cost of energy losses as well as improvement in voltage profile with
OCP compensation of reactive power.
PSO algorithm
Voltage profile improvement This is an open access article under the CC BY-SA license.

Corresponding Author:
Kyaw Myo Lin
Power System Research Unit, Mandalay Technological University
Patheingyi, 05072, Mandalay, Myanmar
Email: [email protected]

1. INTRODUCTION
The shunt capacitor installation in a radial power distribution system has become significant
research in power system planning. For advancing power flow control and enhancement of power factor,
managing voltage profile, reducing active power loss and energy loss in the network, the shunt capacitor
injection on the radial feeder is presently applied. In that case, it is necessary to consider economic criteria in
the installation of the capacitor to get the best design procedure.
Dealing with the optimal capacitor placement (OCP) problem and sizing in the radial distribution
network, many algorithms and techniques are addressed [1]–[4]. In literature, different researchers have
introduced different methods considering different fitness functions including minimization of power losses,
reduction in installation cost, and improvement in voltage profile, lessen the burden on existing lines as well
as maximizing system stability, and others.
In earlier stages, the researchers proposed the analytical procedures [5], [6] because of the
computing burdens and lack of resources. After that, numerical programming [7], heuristics techniques [8],
and artificial intelligent methods [9] have been progressed in OCP research. Because of making a high-
quality solution within short calculation-time and stable convergence characteristics, particle swarm
optimization (PSO) approaches [10] have been employed on capacitor installation research in distribution
networks. The [7]–[10], the major focus is on the minimum power losses of the system and the improvement
of the bus voltage regulation.
From the literature review, it is observed that the application of the PSO algorithm targeting with the
cost of energy loss, evaluating with daily load curves, has not been discovered in previous work. Besides, the

Journal homepage: http://ijape.iaescore.com

72  ISSN: 2252-8792

active practical distribution network is rarely tested for OCP. It motivates that the practical case study should
be tested for the optimal allocation for different load levels.
Therefore, the present research mainly focuses on a case study for OCP for practical distribution
network considering the daily load curve segmented by four different load levels. In this paper, the capacitor
sizes are firstly evaluated by using the analytical method, and these values are applied as initiators in the
optimization process. The OCP is implemented by using the PSO algorithm for minimization of cost of
energy loss of a practical 115-bus, 4-feeder radial distribution network of MESC service area, Mandalay,
Myanmar.
The paper is organized as follows. Section 2 designates the methodology and problem formulation
of OCP. The PSO algorithm for OCP is discussed in section 3. Section 4 grants the practical case study
network information while the simulated results of the scenario with discussions are demonstrated in section 5.
Finally, the main conclusion is drawn in section 6.

2. PROBLEM FORMULATION
In this paper, the OCP is implemented with the daily load curve supporting multi-segments to
minimize the cost of energy losses. The mathematical formulations for OCP are discussed in the following
sections.

2.1. Objective function

The main objective function, f, of OCP, in this research, is to minimize the cost of energy losses
summing up the cost of energy losses at each load considering varying load conditions.

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 = 𝐾𝑒 ∑𝐷 ∑𝑛𝑙
𝑗=1 (𝑃𝑙𝑜𝑠𝑠𝑗 . 𝑇𝑗 ) (1)

Where Ke is the cost per unit of energy losses ($/MWh) of the system, D is the number of days in which the
savings are evaluated, Ploss,j is the real loss of power in the distribution network concerning the load level, j,
nl is the number of load levels and Tj is the time interval for respective load intervals.

2.2. Constraints
The power balanced equations for each bus are considered as equality constraints while the technical
limitations are considered as inequality constraints [11]. The constraints that need to be restricted are
itemized below:
- Operational constraints on bus bar voltage

𝑚𝑖𝑛|𝑉𝑖 |𝑚𝑎𝑥
𝑖
𝑉𝑖 (2)

where, Vi is the voltage magnitude of bus i, and in radial power systems, Vmin= 0.9 and Vmax=1.1 [1], [12], [13].
- Constraints on reactive power compensation in network

𝑄𝐶𝑖 ≤ 0.75 ∑𝑛𝑖=1 𝑄𝐿𝑖 (3)

where, QCi is the allowable capacitor size at bus i and QLi is the reactive power demand on each bus.
- Limitations on line flow (LF):

𝐿𝐹𝑘 < 𝐿𝐹𝑘(𝑚𝑎𝑥) (4)

where, LFk is the power flow in the kth line and LFk(max) is the maximum power flow allowed.
- Constraint on capacitor size

𝑄𝐶𝐶,𝑚𝑎𝑥 𝐶,𝑚𝑖𝑛 (5)

2.3. Fitness function

Since the OCP is a non-linear optimization problem, it is required to consider a fitness function for
both the security and economy of the system. The fitness function, F can be formulated as the penalty
function added to the target function. In this research, it can be stated as (6):

Int J Appl Power Eng, Vol. 12, No. 1, March 2023: 71-82
Int J Appl Power Eng ISSN: 2252-8792  73

∑2 𝑙𝑖𝑚
∑2 ∑𝑖=1 𝜆𝑄𝐶𝑖 (𝑄⥂𝐶𝑖 −𝑄𝐶𝑖 ())
𝐹 = 𝑓 + ∑𝑖=1 𝜆𝑣𝑖 (𝑉𝑖 − 𝑉𝑖𝑙𝑖𝑚 ()) (6)

where, λvi and λQci are the penalty factors for voltage limit and reactive power compensation limit,
respectively. In (6), the allowable maximum and minimum limits of voltage magnitude and amount of
reactive power injected by the capacitor are
𝑉
𝑚𝑖𝑛𝑖 𝑚𝑖𝑛
𝑙𝑖𝑚{ 𝑉 𝑖
𝑖 𝑖
𝑚𝑎𝑥𝑖 𝑚𝑎𝑥
𝑖
𝑉𝑖 (7)
𝑄
𝑚𝑖𝑛𝐶𝑖 𝑚𝑖𝑛
𝑙𝑖𝑚{ 𝑄 𝐶𝑖
𝐶𝑖 𝐶𝑖 𝑚𝑎𝑥
𝑚𝑎𝑥𝐶𝑖
𝐶𝑖
𝑄𝐶𝑖 (8)

3. OCP IMPLEMENTATION USING PSO

3.1. Analytical procedure for initial capacitor size
The capacitor size is one of the control variables to be considered in OCP. Before the optimization
process, the initial capacitor sizing for each bus except the station bus is evaluated using the analytical
approach. According to [6], [14], based on the load flow results, the branch current between the bus ‘i’ and ‘k’
is evaluated by:
𝑃𝑖𝑘 −𝑗𝑄𝑖𝑘
𝐼𝑖𝑘 = (9)
𝑉𝑖

The total power loss (TPL) due to active (a) and reactive (r) components of current in the distribution lines is

𝑇𝑃𝐿 = ∑𝑛𝑖𝑘=1(|𝐼𝑖𝑘
𝑎 2 𝑟 2
| + |𝐼𝑖𝑘 | ) 𝑅𝑖𝑘 (10)

For reactive power compensation, the reactive current IC is drawn by the injected capacitor. For a
radial network, it changes just only on the reactive component of the branch current set α, and it does not
affect the current of other branches. Therefore, the new reactive current can be designed by [14].
𝑛𝑒𝑤 𝑟
𝐼𝑖𝑘 = 𝐼𝑖𝑘 + 𝐷𝑖𝑘 𝐼𝐶 (11)

Where Dik is considered as 1 if branch (i, k)  α, 0 for otherwise. After capacitor placement, the compensated
power loss is maximized and it can be expressed in the mathematical formula as (12).

𝑇𝑃𝐿𝑐𝑜𝑚
𝑟 = ∑𝑛𝑖𝑘=1(𝐼𝑖𝑘
𝑛𝑒𝑤 2
) 𝑅𝑖𝑘 = ∑𝑛𝑖𝑘=1(𝐼𝑖𝑘
𝑟
+ 𝐷𝑖𝑘 𝐼𝐶 )2 𝑅𝑖𝑘 (12)

The total power loss saving (TLS) can be evaluated as

𝑇𝐿𝑆 = 𝑇𝑃𝐿𝑟 − 𝑇𝑃𝐿𝑐𝑜𝑚

𝑟 = ∑𝑛𝑖𝑘=1(|𝐼𝑖𝑘
𝑟 2
| )𝑅𝑖𝑘 − ∑𝑛𝑖𝑘=1(𝐼𝑖𝑘
𝑟
+ 𝐷𝑖𝑘 𝐼𝐶 )2 𝑅𝑖𝑘 (13)

From the TLS equation, the capacitor current, IC that provides the maximum loss saving can be get
by differentiating TLS for IC and equal to zero. Then,
1
𝐼𝐶 = ∑𝑛 (− ∑𝑛𝑖𝑘∈𝛼 𝐼𝑖𝑘
𝑟
𝑅𝑖𝑘 ) (14)
𝑖𝑘∈𝛼 𝑅𝑖𝑘

assuming that there is no significant change in the voltage after capacitor placement, due to change in the
active component of load current at respective buses, the initial capacitor size at candidate bus can be stated
as (15).

𝑄𝐶 = 𝑉𝑚 𝐼𝐶 (15)

Based on [13], [14], the quantity of reactive power injection should be within their feasible limits.
And also, the total reactive power injection must be less than or equal to the total reactive load demand [4],
[7] and that must be used as inequality constraints as in (3).

Optimal capacitor allocation for minimizing cost of energy loss in active … (Mya Myintzu)
74  ISSN: 2252-8792

3.2. Basic concept of PSO

Since 1995, Kennedy and Eberhart proposed the PSO algorithm [15] based on the computational
simulations of the movements of animals like birds and fishes [10]. E. O. Wilson explained the linkage of
these simulations [16] for optimization problems. In PSO techniques, particles change their positions and
velocities by flying around a multi-dimensional search set which is linked with the simplest solution (fitness).
In each time step, the general best value (pbest for private best) and its location obtained (gbest for global
best) are tracked by rushing the velocity of each particle within the population, and also the acceleration is
weighted by random standings.
The brief practice to implement the global sort of PSO is prearranged by the subsequent phases [10]:
- Initialization: Initialize a population of particles with random positions ad velocities.
- Evaluation: Estimate the fitness value of every particle.
- Comparison 1: Compare each particle’s fitness with the particle’s pbest. If this value is better than pbest,
then state the pbest equaling to the current one following the current position.
- Comparison 2: Compare the fitness with the overall previous finest. If the recent value is improved than
gbest, then reorganize gbest to the current directory and value.
- Updating: Update the rate and position of each particle by using the following calculations.

𝑣𝑖𝑗𝑘+1 = 𝑤 𝑘 𝑣𝑖𝑗
𝑘
+ 𝑐1 𝑟1 (𝑝𝑏𝑒𝑠𝑡𝑖𝑗 − 𝑥𝑖𝑗𝑘 ) + 𝑐2 𝑟2 (𝑔𝑏𝑒𝑠𝑡𝑖𝑗 − 𝑥𝑖𝑗𝑘 ) (16)

𝑥𝑖𝑗𝑘+1 = 𝑥𝑖𝑗𝑘 + 𝛥𝑡. 𝑣𝑖𝑗

𝑘+1
(17)

where, c1, c2 are cognitive and social component factors, w is the inertia weight parameter, and r1, r2 are
two random numbers produced in the range {0, 1}, respectively. In (17), the position is restructured in
line with the previous situation and velocity, as ∆t=1.
The weighting function, w of a particle is given by
𝑤𝑚𝑎𝑥 −𝑤𝑚𝑖𝑛
𝑤 𝑘 = 𝑤𝑚𝑎𝑥 − ( ) (18)
𝑖𝑡𝑒𝑟𝑚𝑎𝑥

where, wmax and wmin are the initial and final inertia weights, and itermax is the maximum number of
iteration and iter is the current number of iterations, respectively.
- Stopping: Go back to Step 2 until the stopping standard is met.

3.3. Strategy implementation of PSO algorithm incorporated with MATPOWER

This paper provides an optimal algorithm for OCP in the active distribution network. The primary
computational work of the proposed approach is mainly based on power loss calculation, which is completed
using MATPOWER [17]. It is a simulation tool for researchers easy to use and modify. It also supports for
optimal process and state variable initializations [18]–[20]. Therefore, MATPOWER is incorporated in PSO
algorithm calculations to support power loss calculation [21] and bus voltage profile evaluation, calling the
simulation function “runpf.” The OCP is determined based on the load demand curve of the system. In
practice, consumer loads of networks vary with time over a wide range. Therefore, daily load curves are
considered to define the operational program of shunt capacitor placement. In this work, a typical load curve
with four different load levels is taken into account to estimate the average load demand profile of the test
system. Therefore, four different system information data files are scripted initially in MATPOWER
casefiles. In each case file, the range of voltage magnitude is set from 0.9 p.u to 1.1 p.u while the shunt var
limit is 0.001 MVAr to 1.05 MVAr, respectively.
The comprehensive flowchart of the proposed PSO technique for OCP considering four-segment
load curves to evaluate the energy loss is depicted in Figure 1. In this figure, a step-by-step procedure to
optimize the capacitor allocation is demonstrated. These steps are intended for one candidate bus selection
with the target of cost of energy loss minimization. To evaluate the power loss and voltage profile through the
Newton load flow solver, MATPOWER toolbox is incorporated during the optimization process. In each step
of optimization (particles of PSO), the load flow solver is called upon as a function and it is evaluated the
energy loss in each iteration of the optimization algorithm. In the program setting, the parameters of PSO
should be selected to seek out the global solution as well as to increase the computational complexity. The
initialization of PSO parameters that require to be tuned for assessing the objective function is tabulated as
basic parameters in Table 1.
As explained in section 3.1, the value of randomly generated numbers is within the range of {0, 1}
to randomly weigh the social and cognitive behaviors meanwhile the stopping standard is touched to 100
times. To keep increasing until approaching infinity in the optimization process, the penalty multipliers for
bus voltage, λVi, and shunt compensation, λQc are assigned big numbers with 25,500 and 10,000, respectively.

Int J Appl Power Eng, Vol. 12, No. 1, March 2023: 71-82
Int J Appl Power Eng ISSN: 2252-8792  75

Figure 1. Flowchart of the proposed method for solving OCP

Table 1. Basic parameters of PSO

Sr. No. Parameters Value
1 Population size 40
2 Maximum inertia weight 0.9
3 Minimum inertia weight 0.4
4 Acceleration constants [2.05, 2.05]
5 Maximum number of iterations 100
6 Random number [0, 1]

4. CASE STUDY NETWORK

To validate the performance of the proposed methodology, the active distribution network of the
Mandalay distribution area in Myanmar is employed. The scenario network [22] is described in Figure 2.

Optimal capacitor allocation for minimizing cost of energy loss in active … (Mya Myintzu)
76  ISSN: 2252-8792

There are four feeders in the active network which distribute to the consumers. The system consists of 115
buses, 4 radial lines, 114 branches, and 102 load points, and also the rated capacity is 20 MVA.

Figure 2. 115-bus system of Mandalay distribution area

To examine energy loss, the load variation in the system for a given period are taken under
consideration. Based on the monthly load reports of September 2018 through August 2020 [23], the typical
daily load curve is considered, eventually, taken at substation bus. The load profile is segmented into four
intervals (L1-L4) based on system average mode condition during the loads are assumed to be constant
during the time interval. The prominent load data during the whole day (24-hour) are summarized hourly and
plotted on a graph. The typical load duration curve of the case study network is shown in Figure 3.

Figure 3. Typical load curve of the test system

Int J Appl Power Eng, Vol. 12, No. 1, March 2023: 71-82
Int J Appl Power Eng ISSN: 2252-8792  77

For research purposes, the maximum demand of the system is considered as the base case, reference
load level. The total time for study is related to one year of system operation and the energy price is taken to
be considered as 0.065 $/MWh [24], [25]. The load demand, load-level factor, and the corresponding period
of every load level are recorded in Table 2. The initial power flow solutions of base case and different load
level cases are carried out and recorded. These parameters are applied to evaluate the initial capacitor sizing
and to compare with the compensated results.

Table 2. Load levels 114-bus test system analysis

Load level Level 1 Level 2 Level 3 Level 4
Total load (MW) 8.978 12.111 13.044 10.606
Load factor (p.u) 0.741 1.0 1.077 0.875
Time (h) 8 10 3 3

5. RESULTS AND DISCUSSION

In this section, the consequences of optimal size and site allocation using the proposed technique
will be presented. Firstly, the analytical results for capacitor initial sizing will be explained, and followed by
optimization results. The performance investigation of OCP also will be established. Finally, the OCP results
will be compared and verified with the loss sensitivity factor (LSF) approach [1], [13] for capacitor
allocation.

5.1. Analytical results for initial capacitor size

In this work, the initial load flow with the Newton load flow solver of MATPOWER is done to
estimate the initial capacitor size at an allied bus. The procedure described in section 3.1 was employed.
According to the results, the maximum rating of capacitor size is found at bus 2 with 0.4246 MVAr while the
minimum is at bus 34 with 0.3441 MVAr, individually. These capacitor sizes obtained from the analytical
approach were prepared as start-up sizings during the optimization process. The respective capacitor sizings
are injected on each candidate bus and the different case files are implemented in M-file for four different
load levels.

5.2. Numerical results for minimizing the cost of energy loss

The operation of the proposed strategy is done using intel®coreTM i7-4770 CPU, 3.40 GHz with
4 GB RAM computer. In the optimization process, the capacitor sizes are taken into account as continuous
variables starting with initial values of analytical results. The total 114 candidate casefiles (except substation
bus) with four different load segments are supported in the optimization process. The convergence curve of
OCP for one candidate solution (@case94) is given in Figure 4. In the respective profile summary, the
optimized capacitor size at the affiliated bus for the cost of energy loss minimization, respective voltage
profiles are principally recorded and compared.

4
x 10 Minimization of Cost of Energy Losses
2.291

2.29 X: 1
Cost of Energy Losses($)

Y: 2.291e+04
2.289

2.288

2.287 X: 3
Y: 2.286e+04
2.286 X: 29 X: 53 X: 75
Y: 2.285e+04 Y: 2.285e+04 Y: 2.285e+04
2.285

2.284
0 10 20 30 40 50 60 70 80 90 100
Iteration

Figure 4. Minimization of cost of energy loss with capacitor

One of the recorded profile summaries is demonstrated in Figure 5. It can be seen from the figure
that the “runpf” function is called upon 16160 times to compute the power flow solution getting the active

Optimal capacitor allocation for minimizing cost of energy loss in active … (Mya Myintzu)
78  ISSN: 2252-8792

power loss during the optimization process. It is shown that the PSO is well incorporated with MATPOWER
and all of the four different load segments are well evaluated to induce the optimal solution for capacitor
placement.

Figure 5. Profile summary of CPU time for PSO (@case94)

The optimization results associated with the cost of energy losses at each bus are investigated and
compared to induce the optimal one. The optimal shunt capacitor sizes compared with initial sizes are
demonstrated as a bar-chart in Figure 6. It can be seen from the figure that the optimal sizes are greater in
values than that of initial cases and all are satisfied with predefined constraints. According to the results, the
optimal sizes are between 0.573 MVAr @ bus 2 and 0.4764 MVAr @ bus 112. Accordingly, the obtained
results of the cost of energy losses are compared and analyzed. The result summary of the cost minimization
for each bus is demonstrated in Figure 7. In comparison results, the lowest value of the cost of energy losses
is observed in bus 94 with the optimal size of 0.564 MVAr. Therefore, bus 94 is identified as the optimal
location for the capacitor in the practical network.

Capacitor Sizing (Initial & Optimal)

0.7
Initial Size Optimal Size
0.6

0.5
Q (MVAr)

0.4

0.3

0.2

0.1

0
10 20 30 40 50 60 70 80 90 100 110
Bus Number

Figure 6. Comparison of initial and optimal capacitor sizes at allied buses

Int J Appl Power Eng, Vol. 12, No. 1, March 2023: 71-82
Int J Appl Power Eng ISSN: 2252-8792  79
4
x 10
2.35

2.3

Cost of Energy Loss

2.25 X = 94
Y = 2.28e+04
2.2

2.15

2.1

2.05

2
0 20 40 60 80 100
Bus Number

Figure 7. Comparison of cost of energy loss

5.3. Performance analysis on OCP

Based on the optimal results, the effect of loss reduction to capacitor placement, the respective
voltage profiles, and the cost of energy loss impact is studied. The active power loss reduction for the four-
segment load curve is summarized in Table 3. It can be found that the power loss reduction on different load
levels is decreased obviously. Comparing with the base case and the initial size injection, the optimal
capacitor placement with optimized size gets superior upshots in loss reduction than the two other cases in all
load levels.

Table 3. Effect of loss reduction to capacitor placement

Total active power losses (MW) % Reduction in power losses
Load Level
Base case Initial size Optimal size Base case Initial size Optimal size
L1 0.841 0.812 0.75 - 3.448 10.82
L2 1.496 1.4584 1.270 - 2.513 15.11
L3 3.108 3.058 2.757 - 1.608 11.29
L4 1.144 1.122 1.02 - 1.923 10.83

In this study, the impressions of cost minimization can also be accomplished by the reduction in
power loss as well as bus voltage profile improvement. The uncompensated voltage profile is shown in
Figure 8 while the improvement of voltage profile after capacitor installation for all four load cases is
illustrated in Figure 9. Comparing the voltage features of the system with and without the capacitor, it can be
said that, the optimal allocation of capacitor regulates the voltage profile within allowable voltage regulation
limits (as in Figure 8, the bus voltages are out of limitation with 0.844 p.u @ bus 2 in level 1, however, as in
Figure 9, all profiles are within the allowable limits). Predominantly, the proposed method regulates the
minimum bus voltage of each feeder achieving improvement above 0.911 p.u after capacitor placement (it
will be described in Table 4).

Table 4. Performance analysis on optimal capacitor placement

Description Without capacitor With capacitor Voltage profile of each feeder
Feeder Without capacitor With capacitor
Min@bus Max@bus Min@bus Max@bus
Total P demand (MW) 12.11 12.11 F5 0.876@ 34 0.995 @ 2 0.913@ 34 0.997@ 2
Total Q demand (MVAr) 4.25 4.25 F6 0.878 @ 65 0.989@ 35 0.911 @65 0.992@ 35
Optimal location @ bus - 94 F7 0.879@ 96 0.986@66 0.920 @ 96 0.990@66
Optimal capacity (MVAr) - 0.5640 F8 0.878@115 0.989@97 0.927@115 0.992@ 97
Energy cost ($) 27480 22768.85 System voltage 0.876 @ 34 1.000 @ 1 0.911@65 1.000@1
Net saving ($) - 4711.15
% saving - 17.14

Table 4 summarizes the technical benefits and energy savings per year. The table describes how the
proposed method has achieved optimal results with remarkable savings. The results of the proposed method
yield a significant reduction in power loss and the cost of energy losses. The result accuracy value of the

Optimal capacitor allocation for minimizing cost of energy loss in active … (Mya Myintzu)
80  ISSN: 2252-8792

proposed approach states about 17.14% of energy cost saving with optimal capacitor injection. Therefore, the
current work proves well effectiveness on OCP.

Voltage Profile of Test System with Different Load Levels (Base Case)
1.02

1 Level1 Level 2 Level 3 Level 4

0.98
Voltage Magnitude (p.u)

0.96

0.94

0.92

0.9

0.88

0.86

0.84
0 20 40 60 80 100
Bus Number

Figure 8. Voltage profile for different load levels (before capacitor placement)

Bus Voltage Profile (After Capacitor Placement) of Test System with Different Load Levels

1 Level1 Level2 Level3 Level4

0.98
Voltage Magnitude (p.u)

0.96

0.94

0.92

0.9
0 20 40 60 80 100
Bus Number

Figure 9. Voltage profile improvement for different load levels (after capacitor placement)

5.4. Comparative results with PSO and LSF

The LSF [1], [13] is used to find the candidate bus in the practical test network based on the initial
load flow results. The norm of (Vi/0.095) [13] is accepted to pick the candidate buses having normalized
voltage magnitudes (<1.01). After that, these candidates are tested with the initial capacitor sizing of section 5.1.
The candidate buses obtained by LSF are not enough to be satisfied with the optimal location of the
capacitor. And, it is necessary to contemplate at which bus should be installed by comparing the total power
losses. According to the tested results, among the nominees, the optimal site is at bus 34 with 0.3908 MVAr
size, which gives the least power loss of 1.475 MW comparing the base case of 1.496 MW. The best
allocation of bus 34 with the optimal size is tested with four different load levels to compare with the
proposed PSO technique. The comparative results of the PSO algorithm-based OCP and LSF are tabulated in
Table 5.

Int J Appl Power Eng, Vol. 12, No. 1, March 2023: 71-82
Int J Appl Power Eng ISSN: 2252-8792  81

According to the results, even though there are different locations supported by different
approaches, the results obtained by PSO is more operative than that of LSF. It can be noted that meta-
heuristic optimization-based OCP considering more than one load level (segment) in daily load curve yields a
reasonable solution and the practical network has resulted in more loss reduction as well as better voltage
profile improvement.

Table 5. Effect of los reduction to capacitor placement

Compensated
Items Base case
LSF [1], [13] PSO
Optimal location - 34 94
Optimal capacity (MVAr) - 0.3908 0.564
Cost of energy loss ($) 27480 24615.76 22768.85
Net saving ($) - 2864.24 4711.25
% Saving - 10.42 17.14

6. CONCLUSION
In this study, the optimization method for optimal positioning and rating of the shunt capacitor in the
practical distribution network was presented. Incorporating with MATPOWER toolbox, the PSO algorithm is
employed to adjust the controlled variables to minimize the cost of energy losses through optimal capacitor
allocation. The proposed technique successfully achieved in finding optimal size and site of the capacitor
seeking the most effective candidate bus to install in which the technical and commercial benefits are
attained. It is found that the proposed technique identified the optimal allocation of the capacitor that
provides a noteworthy power loss reduction and voltage profile improvement of the system.

REFERENCES
[1] A. Elsheikh, Y. Helmy, Y. Abouelseoud, and A. Elsherif, “Optimal capacitor placement and sizing in radial electric powe r
systems,” Alexandria Engineering Journal, vol. 53, no. 4, pp. 809–816, 2014, doi: 10.1016/j.aej.2014.09.012.
[2] A. R. Salehinia, M. R. Haghifam, M. Shahabi, and F. Mahdloo, “Energy loss reduction in distribution systems using GA-based
optimal allocation of fixed and switched capacitors,” 2010 IEEE International Energy Conference and Exhibition, EnergyCon
2010, pp. 835–840, 2010, doi: 10.1109/ENERGYCON.2010.5771798.
[3] L. W. De Oliveira, S. Carneiro, E. J. De Oliveira, J. L. R. Pereira, I. C. Silva, and J. S. Costa, “Optimal reconfiguration and
capacitor allocation in radial distribution systems for energy losses minimization,” International Journal of Electrical Power and
Energy Systems, vol. 32, no. 8, pp. 840–848, 2010, doi: 10.1016/j.ijepes.2010.01.030.
[4] S. Gopiya Naik, D. K. Khatod, and M. P. Sharma, “Optimal allocation of combined DG and capacitor for real power loss
minimization in distribution networks,” International Journal of Electrical Power and Energy Systems, vol. 53, pp. 967–973,
2013, doi: 10.1016/j.ijepes.2013.06.008.
[5] M. M. A. Salama, E. A. A. Mansour, A. Y. Chikhani, and R. Hackam, “Control of Reactive Power in Distribution Systems with
an End-Load and Varying Load Condition,” IEEE Transactions on Power Apparatus and Systems, vol. PAS-104, no. 4, pp. 941–
947, Jul. 1985, doi: 10.1109/TPAS.1985.319095.
[6] M. H. Haque, “Capacitor placement in radial distribution systems for loss reduction,” IEE Proceedings: Generation,
Transmission and Distribution, vol. 146, no. 5, pp. 501–506, 1999, doi: 10.1049/ip-gtd:19990495.
[7] H. M. Khodr, F. G. Olsina, P. M. D. O. De Jesus, and J. M. Yusta, “Maximum savings approach for location and sizing of
capacitors in distribution systems,” Electric Power Systems Research, vol. 78, no. 7, pp. 1192–1203, 2008, doi:
10.1016/j.epsr.2007.10.002.
[8] M. Ramalinga Raju, K. V. S. Ramachandra Murthy, and K. Ravindra, “Direct search algorithm for capacitive compensation in
radial distribution systems,” International Journal of Electrical Power and Energy Systems, vol. 42, no. 1, pp. 24–30, 2012, doi:
10.1016/j.ijepes.2012.03.006.
[9] G. Boone and H. D. Chiang, “Optimal capacitor placement in distribution systems by genetic algorithm,” International Journal of
Electrical Power and Energy Systems, vol. 15, no. 3, pp. 155–161, 1993, doi: 10.1016/0142-0615(93)90030-Q.
[10] C. S. Lee, H. V. H. Ayala, and L. D. S. Coelho, “Capacitor placement of distribution systems using particle swarm optimization
approaches,” International Journal of Electrical Power and Energy Systems, vol. 64, pp. 839–851, 2015, doi:
10.1016/j.ijepes.2014.07.069.
[11] A. K. Singh and S. K. Parida, “Novel sensitivity factors for DG placement based on loss reduction and voltage improvement,”
International Journal of Electrical Power and Energy Systems, vol. 74, pp. 453–456, 2016, doi: 10.1016/j.ijepes.2015.04.010.
[12] J. Sardi, N. Mithulananthan, and D. Q. Hung, “A loss sensitivity factor method for locating ES in a distribution system with PV
units,” Asia-Pacific Power and Energy Engineering Conference, APPEEC, vol. 2016-January, 2016, doi:
10.1109/APPEEC.2015.7380873.
[13] A. Y. Abdelaziz, E. S. Ali, and S. M. Abd Elazim, “Flower Pollination Algorithm and Loss Sensitivity Factors for optimal sizing
and placement of capacitors in radial distribution systems,” International Journal of Electrical Power and Energy Systems, vol.
78, pp. 207–214, 2016, doi: 10.1016/j.ijepes.2015.11.059.
[14] M. Ghiasi and J. Olamaei, “Optimal capacitor placement to minimizing cost and power loss in Tehran metro power distribution
system using ETAP (A case study),” Complexity, vol. 21, pp. 483–493, 2016, doi: 10.1002/cplx.21828.
[15] J. Kennedy and R. Eberhart, “Particle swarm optimization,” in Proceedings of ICNN’95 - International Conference on Neural
Networks, vol. 4, pp. 1942–1948. doi: 10.1109/ICNN.1995.488968.
[16] B. Brandstaätter and U. Baumgartner, “Particle swarm optimization - Mass-spring system analogon,” IEEE Transactions on
Magnetics, vol. 38, no. 2, pp. 997–1000, 2002, doi: 10.1109/20.996256.

Optimal capacitor allocation for minimizing cost of energy loss in active … (Mya Myintzu)
82  ISSN: 2252-8792

[17] R. D. Zimmerman, C. E. Murillo-Sánchez, and R. J. Thomas, “MATPOWER: Steady-state operations, planning, and analysis
tools for power systems research and education,” IEEE Transactions on Power Systems, vol. 26, no. 1, pp. 12–19, 2011, doi:
10.1109/TPWRS.2010.2051168.
[18] K. Buayai, K. Chinnabutr, P. Intarawong, and K. Kerdchuen, “Applied MATPOWER for power system optimization research,”
Energy Procedia, vol. 56, no. C, pp. 505–509, 2014, doi: 10.1016/j.egypro.2014.07.185.
[19] C. E. Murillo-Sánchez, R. D. Zimmerman, C. Lindsay Anderson, and R. J. Thomas, “Secure planning and operations of systems
with stochastic sources, energy storage, and active demand,” IEEE Transactions on Smart Grid, vol. 4, no. 4, pp. 2220–2229,
2013, doi: 10.1109/TSG.2013.2281001.
[20] A. J. Lamadrid, D. Munoz-Alvarez, C. E. Murillo-Sanchez, R. D. Zimmerman, H. Shin, and R. J. Thomas, “Using the matpower
optimal scheduling tool to test power system operation methodologies under uncertainty,” IEEE Transactions on Sustainable
Energy, vol. 10, no. 3, pp. 1280–1289, 2019, doi: 10.1109/TSTE.2018.2865454.
[21] K. Z. O. Kyaw Myo Lin, Pyone Lai Swe, “Optimal Distributed Generator Sizing and Placement by Analytical Method and PSO
Algorithm Considering Optimal Reactive Power Dispatch,” International Journal of Electronics and Communication
Engineering, 13(1), 13 - 20., vol. 13, no. 1, pp. 13–20, 2019, doi: doi.org/10.5281/zenodo.2580016.
[22] M. Myintzu and K. M. Lin, “Particle swarm optimization based optimal allocation of capacitor in active distribution network
minimizing cost of energy losses,” Journal of Science, Engineering and Technology, vol. 5, no. 3, pp. 107–111, 2018.
[23] MESC, “Pyigyitagon Township Distribution Network Data” Mandalay Area Network Data Collection Notes, Myanmar, 2018.
[24] D. Dapice, “Electricity Supply, Demand and Prices in Myanmar—How to Close the Gap?” ASH Center for Democratic
Governmance and Innovation, Harvard Kennedy School. [Online]. Available: https://ash.harvard.edu/publications/electricity-
supply-demand-and-prices-myanmar-–-how-close-gap
[25] Energy Demand and Supply of the Republic of the Union of Myanmar 2010-2017, “Oil and Gas Planning Department, Ministry
of Electricity and Energy, Republic of the Union of Myanmar” Economic Research Institute for ASEAN and East Asia,
Indonesia, 2020. [Online]. Available: https://www.eria.org/uploads/media/Research-Project-Report/Energy-Demand-and-Supply-
of-the-Republic-of-the-Union-of-Myanmar-2010-2017.pdf

BIOGRAPHIES OF AUTHORS

Mya Myintzu received her M.E degree in Electrical Power Engineering from
Mandalay Technological University (MTU), Mandalay, Myanmar, in 2019. She joined the
power system research unit of Department of Electrical Power Engineering, MTU as a
research assistance under the supervison of Dr. Kyaw Myo Lin. She is also a desing engineer
of Power System Development Sector. Her research interests include distribution system
planning, optimization techniques and its application on active networks. She can ben
contacted at email: [email protected].

Kyaw Myo Lin received his Ph. D degree in Electrical Engineering from
Mandalay Technological University (MTU), Mandalay, Myanmar, in 2014. After that, he set
up the Power System Research Unit at MTU and has done the research area of power system
stability and optimization as well as reliability and energy management of the distribution
system. Now, he is the professor and head of the Department of Electrical Power Engineering
at Pyay Technological University, Pyay, Myanmar. He can be contacted at email:
[email protected].

Khine Zin Oo received her M.E degree in Electrical Power Engineering from
Mandalay Technological University (MTU), Mandalay, Myanmar, in 2017. She is an
assistance lecturer at Department of Electrical Power Engineering, Technological University
of Meiktila, Mandalay Region. She is also a research assistance at Power System Research
Unit of MTU. Her research interests include the optimization techniques, distributed
generation and planning, and energy management. She can be contacted at email:
[email protected].

Int J Appl Power Eng, Vol. 12, No. 1, March 2023: 71-82