www.dergipark.gov.

tr
ISSN:2148-3736

El-Cezerî Journal of Science and Engineering

Vol: 10, No: 3, 2023 (538-553)
DOI: 10.31202/ecjse.1270783
ECJSE
Research Paper

Optimal Reactive Power Injection in Distribution Networks

to Maximise Cost Savings via AOA
G SRINIVASAN1, M LAVANYA2
1
Dept. of Electrical & Electronics Engineering, Thamirabharani Engineering College, Tirunelveli, Tamil Nadu,
India
2
Dept. of Mechatronics Engineering, Sona College of Technology, Salem, Tamil Nadu, India
[email protected]

Received:25.03.2023 Accepted:24.08.2023
Abstract: Utilities have been forced to raise the overall efficiency towards a better position in radial distribution
systems (DS). The literature has proved that reactive power compensation performs well in minimising the
power loss (PLoss) and enhancing the bus voltage profile within the permissible range in radial DSs. This work
presents Archimedes optimisation algorithm (AOA) to resolve the problem efficiently. The merit of this
technique is that it can offer a global or near-global optimum for capacitor siting and sizing. The main intention
of this study is to obtain maximum annual financial benefit (AFB) using the placement and sizing of capacitors
optimally. This can, however, be achieved by minimising the objective function composed of cost-based Power
loss and capacitor investment cost in radial DSs. The proposed technique has been tested on four renowned DS:
the Indian 10-bus, modified 12-bus, PG&E 69-bus, and 94-bus Portugal DSs. The previously published papers
are compared with the outcomes of AOA in terms of Power loss reduction with/without AFB and prove that
AOA yields better performance.

Keywords: Capacitor siting and sizing, Radial distribution system, power loss minimisation, Archimedes
Optimisation algorithm

1. Introduction

The most critical issues that occur in the entire distribution system (DS) are power loss and poor
voltage profile. In developed countries like Europe and US, the power loss is 10% only. On the other
hand, the average transmission and distribution Power loss is roughly 27% of the total power generated
in India [1]. Such a significant quantity of power loss must be addressed since it reflects on financial
aspects and the overall efficiency of the power DS. Further, it is mandatory to maintain an acceptable
voltage profile for the end users. Therefore, Power loss reduction and bus voltage improvement
methods are essential to achieve financial goals.

It is widely recognized that installing capacitors along the DS reduces a portion of the power loss,
increasing the overall efficacy of the power delivery. The other benefits, such as sub-station power
factor improvement, enhancement in bus voltage profile, network stability improvement, reduction
in total apparent power (AP) demand, and feeder capacity release, can be possible only when the
capacitors are located at optimal locations with appropriate capacity [2]. Hence optimal capacitor
placement problem is a complex, combinatorial, mixed integer, and non-linear optimisation problem.

To perform reactive power compensation using capacitors, many researchers used two different
objectives. Either minimisation of Power loss cost against capacitor purchase cost or maximisation
of net annual financial benefit (AFB). However, some authors still consider Power loss reduction as
the only objective [3,4]. Considering economic-based criteria in the capacitor allocation problem is
How to cite this article
G. Srinivasan, M. Lavanya, “Optimal Reactive Power Injection in Distribution Networks to Maximise Cost Savings Via AOA” El-Cezerî Journal of Science and
Engineering, Vol: 10, No: 1, 2023, pp. 538-553.
ORCID: a0000-0003-4030-0182; b0000-0002-9775-8752
 G Srinivasan, M Lavanya ECJSE 2023(3) 538-553

essential to achieve the best solution [5]. Selecting appropriate nodes and determination of optimal
capacitor sizing are the two main steps to obtain the best result in the capacitor allocation problem.
Though sensitivity factor-based identification of appropriate nodes for reactive power compensation
helps reduce the search space during optimisation, the outcome may not indicate the appropriate
location for reactive power compensation [6]. Many research articles consider capacitor sizing as a
continuous type instead of a discrete one. On the other hand, the sizes available in the market are for
discrete kinds only. Also, it has already been proved that continuous variable methodology might not
yield a better result. Therefore, the capacitor sizes are taken as discrete variables [7].

The order of the remaining work is as follows. The literature and inspiration for this work are
reviewed in Section 2. Segment 3 presents the basic distribution system power flow (DSPF), the
objective function (with equality and inequality constraints), the mathematical model of AOA, and
its capability to solve the reactive power optimisation with the block diagram. Segment 4 reveals the
results obtained by AOA. Segment 5 describes the outcomes of AOA after the compensation, and
finally, segment 6 concludes the results and observations, followed by the references.

2. Literature Survey

Reactive power compensation has been carried out in [9] using the Chu and Beasley genetic algorithm
(CBGA) as an optimisation approach, with decreasing Power loss and minimising capacitor cost as
the objectives. This work includes bus voltage deviation and penalty factor-based line thermal limit
in the objective function. From [9], it is observed that the methodology adopted is different from the
latest one, and the result obtained is not an optimal global value. The bus voltage after optimisation
still needs to be discussed. Salp-swarm algorithm (SSA) based allocation of capacitors in radial DSs
considering three load levels has been presented in [10]. Using a combination of the flower pollination
algorithm (FPA), the voltage stability index (VSI), and the loss sensitivity factor (LSF), the study of
capacitor placement and sizing has been carried out in [11]. LSF and VSI were utilised to identify the
potential nodes for reactive power optimisation. The main objective of [11] is to find the energy loss
reduction with the reduction in capacitor purchase cost, capacitor installation, and operation cost
under three different load levels. 10-bus, 69-bus, and 118-bus test systems have been utilised to prove
the efficacy of FPA. Hybrid FPA and exhaustive search (ES) approach (FPAES) based reactive power
optimisation in radial DS has been performed in [12]. FPA has been utilised to find the optimal buses,
and ES has done the optimal sizing. The authors strongly agree that due to the limited search space
of the predetermined discrete capacitor sizing, the computing effect to determine the optimal capacity
of the capacitor is considerably diminished. 10-bus, 34-bus, and 85-bus test systems determine the
potency. Optimal capacitor allocation and sizing under three load levels, such as 50%, 100%, and
160% in radial DSs using a water cycle algorithm (WCA) and grey wolf optimizer (GWO), is
performed in [13]. Apart from standard test systems (33-bus, 69-bus, and 85-bus), this work considers
three Indian practical radial DSs such as 28-bus, 47-bus, and 52-bus are also taken to prove the
efficacy of the proposed methods. It is obvious that fixed and switchable-based capacitors with
changes in investment cost need to be considered for load-level-based optimization. But [13]
considered only fixed-type capacitors. FFPA-based capacitor siting and sizing optimisation in four
renowned DSs (IEEE 33-bus, 34-bus, PG&E 69-bus, and Indian 85-bus) have been proposed in [14].
Power loss cost and capacitor investment cost minimisation have been considered as objectives. Data
structure-based load flow analysis has been utilised in [14] to determine the Power loss and bus
voltage profile. To determine the optimal allocation of capacitors in two DSs, a combined
optimisation approach (COA) based on SSA and LSF has been considered [15]. VLSF and reactive
power loss sensitivity factor QLSF-based ranking of load buses have been adopted in this paper. The
COA based on a three-stage procedure has been implemented to reduce the search space and achieve
computation time reduction. Power loss minimisation and energy loss minimisation with a reduction
in capacitor investment cost have been taken as the objective function. The algorithms were validated
using 69-bus, 85-bus, and 30-bus (part of the Unified Egyptian Network) DSs. Slime mould
539
 ECJSE2023 (3) 538-553 Optimal Reactive Power Injection …

optimisation algorithm (SMOA) based optimisation of capacitor siting and sizing in two radial DSs
(69-bus and 85-bus) has been reported in [16]. Apart from SMOA, two more metaheuristic
algorithms, such as bonobo optimisation algorithm (BOA) and tunicate swarm algorithm (TSA), have
been utilised for comparison purposes. This work takes the capacitor values as integral multiples of
100 KVAR & 150 KVAR, respectively. The optimisation has been done for both the DSs with and
without voltage constraints. Optimisation of reactive power compensation using capacitors in two
radial DSs via teaching learning-based optimisation algorithm (TLBO) and modified teaching
learning-based optimisation algorithm (MTLBO) to use the energy loss reduction cost, minimise the
capacitor investment expenses and voltage stability enhancement has been carried out in [17].
Besides, it is understood from [17] that there is no guarantee that either TLBO or MTLBO will escape
from sub-optimal solutions. Reactive power optimisation at three optimal nodes using sequential
power loss index (PLI) based method and PSO have been carried out in [18]. 34-bus, Indian 85-bus,
and real Portugal 94-bus systems are evaluated. The optimal buses for reactive power optimisation
are decided using a formula based on the total reactive power demand (QD) and the average range of
capacitors.

Combined overall Power loss cost reduction and capacitor investment cost reduction as objective,
reactive power optimisation using multi-verse optimizer (MVO) has been investigated in [19].
Besides, this work employs the partial and modified use of conventional LSFs. To reduce the search
space, along with modified LSFs, MATLAB ‘is member’ and ‘any’ commands have been utilised.
10-bus, 33-bus, and 69-bus test systems are used to demonstrate the helpfulness of MVO. From [20],
it is understood that the traditional LSF does not yield a realistic representation of the actual change
in the Power loss. Reduction in Power loss, energy loss cost, and reduction in capacitor investment
cost as objective, optimal allocation and sizing of capacitors in radial DSs using four algorithms,
namely stochastic fractal search algorithm (SFSA), modified stochastic fractal search algorithm
(MSFSA) 1 and 2 and the proposed MSFSA have been done in [20]. IEEE 33, PG&E 69, and Indian
85 bus test systems are taken to prove the efficacy of the proposed method. Reactive power injections
at two, three, and four optimal nodes have been carried out to distinguish the power loss reduction. It
is to be noted that the actual bus voltage profile improvement after optimisation has not been
mentioned. Reduction in Power loss cost and costs related to the capacitor such as investment,
installation, and operation as objective, reactive power compensation (four optimal locations) using
mathematical remora optimisation algorithm (ROA) has been performed in [21]. The Power loss of
the network lines index PLNLI has been considered to identify the most critical nodes for reactive
power compensation. Allocation and sizing of Type-I DGs after reactive power compensation at two
optimal locations have also been evaluated in [21]. IEEE 33 and PG&E 69 bus test system has been
taken to prove the efficacy of the proposed method. Enhanced modified particle swarm optimisation
(EMPSO) as an optimisation tool, evaluation of capacitor allocation, and sizing in radial DSs
considering three objective functions have been performed in [22]. Capacitor allocation has been done
in two phases. Identification of potential nodes for capacitor placement using LSF has been taken as
the first phase, and in the second phase, the optimal locations have been chosen by NMPSO. In both
phases, optimal sizing has been done by EMPSO. The first objective function dealt with Power loss
reduction against capacitor integration. At the same time, the second objective function dealt with the
energy loss reduction-based integration of capacitors. The difference between the second and third
objective functions is the time duration. Second and third objective functions considered single and
three load levels, respectively. EMPSO has been evaluated using standard 15 bus, IEEE 33 bus, and
PG&E 69 bus test systems. It is to be noted that the selection of nodes for capacitor integration is
typically three only. On the other hand, [22] considered five nodes for reactive power compensation,
which is abnormal. Considering more nodes for payment and high reactive power penetration (more
than 100%) will reverse the objective of the work, and there is a possibility of profit decrease and
saturation in bus voltage enhancement as well.

Power loss minimisation and economic saving as objectives, optimal allocation and sizing of

540
 G Srinivasan, M Lavanya ECJSE 2023(3) 538-553

capacitors using a reformulation of the same mixed integer non-linear programming MINLP model
through a mixed-integer second-order cone programming (MISOCP) has been performed in [23]. On
the other hand, the major drawback is that it has already been proved that the mathematical-based
optimisation methodology yields poor performance compared to meta-heuristics optimisation
methods. Minimising energy loss cost and capacitor investment cost as objective, placement, and
capacity determination of capacitors at three / five optimal nodes using a multi-verse optimizer
(MVO) as an optimizing tool has been proposed in [24]. This paper discusses the importance of
reactive power compensation at the source and load sides. 34 and Indian 85 bus test system has been
taken for evaluation.

The authors generally consider power loss reduction and capacitor purchase costs as the main
parameters [9-18]. However, [11,15,17,18,21,24] considered installation and O&M cost in addition
to the above parameters. Many authors utilised sensitivity-based identification (LSF, VSF, VLSF,
QLSF) of weak buses for capacitor installation [11,15,17,22]. However, from [18,25,26], it is
understood that SI may not always indicate the appropriate location for reactive power compensation
and also leading to the underutilisation of the optimising tool since it is used only for capacitor sizing.

From the above previously published papers, it is understood that the main objective is to increase
the overall gain by reducing the total Power loss with capacitor investment cost. LSF-based weak
node identification for reactive power injection will not yield a good result. Optimizing optimal nodes
and appropriate capacitor sizing using the meta-heuristic algorithm may help get a better result.

In this study, sensitivity factor-based detection of feeble buses for reactive power injection has been
avoided; instead, the algorithm has to search for both optimal nodes and sizing of capacitors, and also
the capacitor sizes are taken in discrete steps (multiplication of 150 KVAr). Archimedes optimisation
algorithm (AOA) has been engaged to solve the objective function due to its several advantages,
which have been discussed in section 3. A single objective function comprising capacitor purchase
cost with cost-based Power loss reduction has been evaluated with the condition that all the network
constraints from Equations 2 to 5 should get satisfied. Indian 10-bus, modified 12-bus, IEEE 69-bus,
and 94-bus Portugal DSs are used to verify the proposed method.

The purpose and contribution of this work are to yield a better solution for reactive power
compensation using capacitors. A modified 12-bus test system with increased load demand (5 times)
has been considered, which is new for reactive power compensation.

3. Problem Formulation

3.1. Distributed System Power Flow (DSPF)

To evaluate the performance of DSs under seasonal periods, a power flow (PF) study is an essential
tool that needs to be performed frequently under steady-state operating conditions. Due to the DS’s
low X/R ratio and radial character, the renowned matrix-based PF methods used for the transmission
network were ineffective. In this paper, the PF method developed in [8] has been utilised in order to
solve the DS efficiently.

3.2. Objective Function

The main target of this problem is to find the best solution for Power loss and AFB by reactive power
optimisation using capacitors in the DSs with the condition that network constraints must be satisfied.
The objective function has been divided into two parts. From Equation (1), it is apparent that the
denominator discusses the total cost saving pertaining to solving Power loss reduction due to reactive
power compensation. The numerator reveals the capacitor investment cost. The various cost

541
 ECJSE2023 (3) 538-553 Optimal Reactive Power Injection …

parameters discussed above are as follows:

TCN
( KC   QC (l ) )
Minimise Cost = l (1)
( K PLoss  (TP IC
Loss  TP Loss ))
AO

Subject to Equality Constraints

TNC
QMS   QD   QC (l )  TQ Loss
A0
0 (2)
l
Inequality Constraints
min  Q max
QC (l ) C (l )  QC (l )
(3)
TNC
 QC (l )  ( QD  TQ Loss)
A0
l (4)
V min
(i )  Vi  Vi max
(5)
where

TNB Pm2 Qm2

TP Loss   PLOSS ( m, m  1) PLOSS ( m,m1)  2
 R( m,m1)
m 0
Vm
and

K PLoss is the cost per KW of PLoss ($/KW);

Kc is the capacitor investment cost ($/KVAr);
TPLoss represents the total active PLoss s(KW);
TQLoss represents the total reactive QLoss (KVAr);
IC is the initial condition of the DS;
AO indicates after optimisation;
TNC is the total number of nodes that require reactive power compensation
Qci is the reactive power support given at bus ‘i’ (KVAr);
QMS, QD represents the primary source of reactive power supply and reactive power demand (KVAr);
TNB indicates the total number of buses in the DS.
V(i) is the bus voltage at ith bus.

Practical Capacitors are available in standard capacities, which are the multiple integer values of the
smallest size denoted as 𝑄𝐶0 . The per kVAr cost of the capacitor changes across its sizes, which are
available commercially. It is understood that large-capacity capacitors have lower prices. The
available capacitor sizes are typically taken as

Qcmax  A  QC0 (6)

Thus, for each capacitor installation node, the sizes are ‘A’ times that capacitor size (i.e.) {Q C0,
AQC0} where ’A’ is an integer multiplier

3.2. Solution Methodology (AOA)

Hashim et al. propose a population-based metaheuristic optimisation algorithm termed AOA inspired
by the law of physics named Archimedes’ principle [27]. To find globally optimal solutions, AOA
retains inhabitants of solutions and inspects a vast area. Hence this work considers AOA as an
optimisation technique (OT) to solve the capacitor allocation problem and anticipates that AOA
maintains a noble balance between exploration and exploitation. Like other population-based
optimisation techniques, AOA begins the search procedure with initial solution vectors (SV) with

542
 G Srinivasan, M Lavanya ECJSE 2023(3) 538-553

random volumes, densities, and accelerations. Also, each object is set with its arbitrary location in
the fluid. During the evaluation process, AOA updates the thickness and volume of everything in
every iteration. The acceleration is updated based on the condition of its collision with any other
adjacent object. The updated new SVs (density, volume, acceleration) replace the existing positions.
The mathematical model of AOA is discussed below.

Process 1: Initialize the SVs randomly using Equation (7)

obd  BLmin
d + [rand  ( BL d
max -
BL min
d )]; d=1,2,3......N (7)

where obd is the dth object in an SV of N objects. BLmin and BLmax are the search agent's minimum
and maximum values, respectively. Rand is an M-dimensional vector that randomly generates
numbers between 0 and 1.

Equation (8) indicates the acceleration initialization of the dth object. Estimate the object with the best
fitness value.

acd  BL min
d + [rand  ( BL max
d - BL max
d )] (8)

Process 2: The volume and density for each object ‘d’ for the iteration IT+1 is updated using Equation
(9). Assign xbt, debt, vobt and acbt

IT 1  IT 1
ded + [rand  (ded - ded )] & vod  vod + [rand  ( vo d - vo d )]
IT bt IT IT bt IT
ded (9)

Where vobt and debt are the volume and density connected with the best object established.

Process 3: During the commencement of the process in AOA, a collision between the objects occurs
and drives the objects toward the equilibrium state after a specified period done by a Transfer
Operator (TO), which changes the search from exploration to exploitation as given in Equation (10).
The value of TO increases gradually towards 1.

 IT  IT max 
TO  exp  
 IT max  (10)

where IT and ITmax represents the current iteration and maximum iterations, respectively. In the same
way, the density decreasing factor ‘g’ also helps AOA achieve global to local search concerning time
using Equation (11).

 IT  IT max   IT 
g IT 1  exp  IT 
 max   IT max 
(11)

where gIT+1 decreases over time, allowing converging in previously recognized promising good value.
Appropriate control of this variable must be confirmed to balance the exploration and exploitation
process well.

Process 4: As already discussed, a collision between the object occurs if the value of TO is less than
or equal to 0.5. Select an MR and the acceleration of the object is updated for iteration IT+ 1 using
Equation (12):

543
 ECJSE2023 (3) 538-553 Optimal Reactive Power Injection …

IT 1 de MR  vo MR  ac MR (12)
ac d = IT 1 IT 1
de d  vo d

Where ded, vod, and acd are the density, volume, and acceleration of object ‘d’. The parameters such
as acMR, deMR, and voMR are the acceleration, density, and volume of MR, respectively. It is significant
to state that TO is less than or equal to 0.5 conforms to the exploration during one-third of iterations.
However, if the TO value is more important than 0.5, no collision between the objects occurs, hence
updating the object’s acceleration for iteration IT+1 using Equation (13).

bt  bt  bt
ac d
IT 1
= de IT vo ac (13)
1 IT 1
de d  vo d

where acbt is the acceleration of the best object.

Process 5: To calculate the percentage of change, normalize the acceleration using Equation (14):
IT 1 - acmin
IT 1 ac
ac d nor = b  d k (14)
acmax - acmin

Where ‘b’ and ‘k’ are the normalization range, set to 0.9 and 0.1, respectively. The LHS of eqn. (14)
regulates the % step that each agent will change. The acceleration value is high when the object ‘d’
is far away from the global optimum, which indicates that the object will be in the exploration phase;
or else in the exploitation phase. Under the typical case, the acceleration factor starts with a larger
value and moves towards a lower value with time.

Process 6: If the object ‘d’ is in the exploration phase, the updating has been done using Equation
(15), and if the object ‘d’ is in the exploitation phase, then updating has been done using Equation
(16)

IT 1
xd = x dIT  P1  rand  ac dITnor
1
 g  ( x rand - x dIT 1)
(15)
IT 1
xd = IT
xbt  F  P 2  rand  ac dITnor
1
 g  (T  x rand - IT 1
xd ) (16)

where ‘T’ increases concerning time, is directly proportional to TO, and is defined as T= P3× TO. ‘F’
is the flag to change the direction of motion. The value of ‘F’ is +1 for ‘P’ is less than or equal to 0.5;
otherwise, -1. The value of ‘P’ is calculated using Equation (XVII)

P = 2 × rand – P4 (17)

The AOA optimisation process begins with producing a random set of candidate solutions
(populations) according to their given value. Assign the P1, P2, P3, and P4 values as 2, 6, 2, and 0.5,
as mentioned in [27]. Amid the direction of reiteration, acceleration, density, and volume of MR,
access the attainable positions of the immediate ideal solution (Equations (7) and (8). Each answer
re-establishes its function from the optimally obtained solution with the help of ‘TO’ and ‘g’
(Equations (9) – (11)). Exploration and exploitation (Equations (12) to (16)) of the parameters are
expanded directly. Candidate solutions look to separate from the near-optimal solutions. Update ‘F’
after calculating ‘P’ (Equation (17)). Inevitably, the AOA calculation is halted by coming to the
fulfillment of the convergence measure. Figure 1 reveals the block diagram of the entire process.

544
 G Srinivasan, M Lavanya ECJSE 2023(3) 538-553

Figure 1. Block diagram – Reactive power compensation optimisation using AOA

4. Results

To demonstrate the usefulness of AOA in Power loss reduction and progress in node voltage with an
increase in AFB, four radial power DSs, such as 10-Bus, modified 12-bus, IEEE 69-bus, and 94-bus
Portuguese DSs, are considered for assessment. Figures 2 to 5 show the bus arrangement for all four
test systems under IC.

Fig. 2. Indian 10-Bus system (IC)

Figure 2. Indian 12-Bus system (IC)

Figure 3. IEEE 69-Bus system (IC)

545
 ECJSE2023 (3) 538-553 Optimal Reactive Power Injection …

Figure 4. Portugal 94-Bus system (IC)

4.1. Test System Details

The first test system is a well-known single-feeder Indian DS operating at 23 KV with considerable
loads in all the nodes. The details about this DS can be seen in [9]. Like the 10-bus system, the 12-
bus test system is also a single-feeder Indian 11 KV system with loads in all the buses. Further, details
of this network can be found in [28]. However, similar to [29], the AP demand on each bus is
multiplied by five. 12.66 KV and 100 MVA are taken as KV (base) and MVA (base) for the renowned
69-bus DS. The details about 69-bus DS can be taken from [11]. For all the test cases, bus number 1
has been considered a substation bus/slack bus whose bus voltage is fixed as 1 p.u. The remaining
buses are considered load buses, and capacitors will be installed in any potential nodes requiring
compensation. The total number of compensation buses decides the loss reduction (PLoss and QLoss).
However, there is a limitation to the installation of capacitors in a DS. Injection of reactive power
more than the required number of nodes and through non-optimal buses will increase the Power loss
economically. Hence, this paper chooses capacitor installation at four optimal nodes for the 10-bus
system and three optimal nodes for the remaining test systems. Algorithm parameter details such as
the total population and iteration number have been taken as 800 and 100, respectively. DSPF-
embedded AOA has been done using MATLAB coding. To find out the net AFB, the Power loss cost
has been taken as $168/kW/year, and the cost related to prevailing capacitor sizes ($/KVAr) has been
taken from [15].

4.2. Indian 10- Bus System

The first test system is a single-feeder 10-bus Indian system whose data can be viewed in [9]. Fig. 2
shows the structure of 10-bus Indian DS, which has 10 nodes and nine branches. 23 KV and 100
MVA have been taken as KV (base) and MVA (base), respectively. The total AP demand is
(12368+j4186) KVA. The real and reactive Power loss and minimum bus voltage under IC are
(783.7784+j 1036.5) KVA and 0.8375 p.u. (@ bus 10) respectively. The total Power loss cost under
IC is $131674.7712.

Table 1 shows that the Power loss has been reduced by 12.75787% compared to IC by optimally
injecting 96.895% of the total reactive power (QD + Q Loss(AO)) at four buses. The minimum node
voltage has enriched from 0.8375 p. u to 0.8877 p.u. which is around 6% compared to IC. The
enhancement in node voltage after compensation is uniformly distributed in all the nodes.
Considering the cost factor, the power loss (∆ PLoss) change is $16798.89. After considering the
capacitor purchase cost, the net AFB is around 12.1% compared to the IC of Power loss of the DS.
Table 1 shows that the AOA minimises Power loss in an enhanced way compared to [9-12]. The AFB
achieved by AOA is $1909.74 greater than [9]. However, the benefit gained between AOA and [10]
is minuscule. The minimum bus voltage profile after compensation is greater than [12]. However, the
minimum bus voltage improvement equals SSA [10] and less than FPA [11]. Though [19] achieves
a better result in Power loss reduction than the proposed method, the number of compensation nodes
is six. Moreover, the reactive power compensation given is 2.4367 times greater than the total reactive
546
 G Srinivasan, M Lavanya ECJSE 2023(3) 538-553

power demand of the system. Figure 6 shows the graph of the bus voltages before and after
compensation. Since it is a single feeder, the bus voltage profile faces a drastic fall in voltage from
buses 1 to 10. However, after capacitor placement, the bus voltage has improved by 6%.

Table 1. Performance of AOA over other methods – 10-Bus DS

FPA FPAES
Parameters E S [12] F P A [11] S S A [10] CBGA [9] MVO [19] AOA
[12] [12]
𝑃𝐿𝑜𝑠𝑠 (AO)
694.4 / 695 / 694.4 / 688.28 / 683.8012 / 691.99 / 675.6971 / 683.785 /
/𝑃𝐿𝑜𝑠𝑠 (BO)
783.7 783.7 783.7 783.77 783.7784 783.79 783.7895 783.7784
(KW)
% 𝑃𝐿𝑜𝑠𝑠
11.3947 11.318 11.39467 12.1834 12.7558 11.7123 13.79 12.75787
reduction
4050 (3)
1200 (5) 1200(5) 1200 (5) 1500 (5) 2400 (5) 2100 (4) 2100 (4) 2400 (5)
Optimal
1100 (6) 1200(6) 1100 (6) 300 (7) 1050 (6) 1950 (5) 2100 (5) 1050 (6)
Capacitor
500 (9) 300 (9) 500 (9) 600 (9) 450 (8) 1950 (6) 1200 (6) 450 (7)
Size
200 (10) 200(10) 200 (10) 1100 (10) 300 (10) 750 (10) 450 (9) 300 (10)
300 (10)
𝑉𝑚𝑖𝑛 (p.u.) 0.86 0.86 0.86 0.9509 0.8877 ------ 0.9 0.8877

𝑃𝐿𝑜𝑠𝑠 cost –
116659.2 116760 116659.2 115631.04 114878.60 116254.32 113517.113 114875.88
AC ($/Year)
Capacitor
Cost ------- ------- ------- ------- 866.25 1399.5 1887 866.25
($/Year)
Net AFB
------- ------- ------- ------- 15929.92 14022.9 16272.5232 15932.64
($/Year)
% AFB ------- ------- ------- ------- 12.098 10.6495 12.3579 12.09999

4.3. Indian Modified 12-Bus System

The second radial test system is a modified 12-bus system with 12 nodes and 11 branches with loads
in all the nodes. The KV (base)and MVA (base)are 11 KV and 100 MVA, respectively. This system
supplies an AP demand of (2175+j2025) KVA. The Apparent Power loss and minimum bus voltage
under IC are(1090.7+j416.8654) KVA and 0.5689 p.u. @ bus no.12 respectively. The Power loss cost
under IC is $183237.6.

Figue 6. Bus voltage profile – 10-Bus DS

The Power loss has reduced from 1090.7 KW to 418.3909 KW, 61.64015% compared to IC. This has
been achieved after the reactive power injection of 98.57% of the total (QD + Q(Loss(AO))) at three
optimal nodes {1050 (4), 600 (7), and 600 (10)}. The minimum bus voltage observed is 0.7525 p.u.

547
 ECJSE2023 (3) 538-553 Optimal Reactive Power Injection …

@ bus number 12, which is 0.1836 p.u. increase compared to IC. The improvement in bus voltage is
found to be around 32.273%. Considering the cost factor, the change in the Power loss cost (∆PLoss)
cost is $112947.93, and the capacitor purchase cost is $503.4. Thus, the total AFB is found to be
61.3654% compared to the IC of the DS. Figure 7 shows the graph of the bus voltages before and
after compensation. Fig.7 shows a drastic voltage drop for buses from 1 to 5 and 7 to 9 compared to
other buses. However, after capacitor placement, the bus voltage has improved by 31.306%.

Figure 7. Bus Voltage profile – Modified 12-Bus DS

4.4. PG&E 69- Bus DS

The next DS is a renowned system with 69 nodes, 68 main switches, and five tie-switches formed as
loop branches, as exposed in Fig. 4.

Table 2. Performance of AOA over other methods– IEEE 69-Bus DS

GWO WCA FPA SMO MSFS

Parameters C O A [15] AOA
[13] [13] [14] A [16] [20]
146.74 / 146.73 / 145.86 / 146.269 / 145.78 / 145.129 145.775 /
𝑃𝐿𝑜𝑠𝑠 (IC) / 𝑃𝐿𝑜𝑠𝑠 (BO)(KW)
225 225 225 224.96 225 7 / 225 225
34.7866 35.1733 35.2088
% 𝑃𝐿𝑜𝑠𝑠 reduction 34.78222 34.98 35.4979 35.211
6 3 8
406.55
450 (11) 150 (12) (11)
300 (16) 300 (16) 300 (17) 450 (12)
Capacitor Size (KVAR) / 150 (22) 300 (18) 246.59
900 (60) 450 (59) 150 (57) 150 (21)
Nodes 1350 1200 (19)
450 (61) 900 (60) 1200 (61) 1200 (61)
(61) (61) 1236.45
(61)
0.9312
Vmin (p.u) 0.9322 (65) 0.933 0.93131 ----- 0.93 0.9314
(65)
24650.6 24504.4 24491.0 24381.7
𝑃𝐿𝑜𝑠𝑠 cost (AC) ($) 24652.32 24573.192 24489.36
4 8 4 9
Cost of Capacitor ($/(KVAR-
383.55 383.55 468.3 384 384 ----- 392.85
year))
12765.8 12827.2 12924.9
Net AFB ($) 12764.13 12836.088 ----- 12917.79
1 2 6
33.9344
%AFB 33.76754 33.772 33.96394 34.193 ----- 34.1782
4

This test system KV (base) and MVA (base) of this test system are 12.66 KV and 100 MVA,
respectively. This DS supplies a total AP demand of (4027+j 2797.115) KVA. The total AP and
minimum bus voltage profile under IC is (225+j 102.115) KVA and 0.90918 p.u. respectively. The
Power loss cost under IC is $37800.

548
 G Srinivasan, M Lavanya ECJSE 2023(3) 538-553

From Table 2, it is evident that the Power loss has reduced by 35.181% after optimal reactive power
support of 64.352% of the total (QD + Q(Loss(AO))) at three optimal nodes. The bus voltage has enhanced
from 0.90918 to 0.9314 p.u. The change in the Power loss cost is $13312.2, and the net AFB after
considering the capacitor cost is 34.178%. Table 2 shows that the Power loss reduction achieved by
AOA is better than [13-16]. The minimum and maximum AFB differences between AOA and other
methods [13-15], as discussed in Table 2, are $81.702 and $153.66. However, the difference in Power
loss reduction and net AFB achieved by AOA and [16] is minuscule. This is because of the increased
capacitor cost. The Power loss reduction difference between [20] and the proposed method is around
0.5% only. This may be due to increased reactive power compensation. Fig. 8 reveals the bus voltage
profile of the IEEE 69-bus system. The enhancement in node voltage between AOA and other
methods [13-16] is minuscule. Yet AOA’s minimum bus voltage profile after compensation is less
than GWO [13] and FPA [14]. From Fig. 8, it is evident that the bus voltage has improved well in the
maximum number of load buses.

4.5. Portugal 94-Bus System

The final test system taken for evaluation is a real 94-bus Portugal DS with 94 nodes, 93 branches,
and 22 laterals. The KV (base)and MVA (base) are 15 KV and 100 MVA, respectively. This real test
system's line and load data can be viewed in [11]. The total AP demand is (4797+j2323.9) KVA. The
total AP Power loss and minimum bus voltage under IC are (361.67636+j 503.7688) KVA and
0.85413561 p.u. @ bus no. 33 respectively.

Table 3. Performance of AOA over other methods – real Portugal 94-Bus DS

MTLBO
Parameters GA [17] PSO [17] TLBO [17] PSO [18] AOA
[17]
𝑃𝐿𝑜𝑠𝑠 (AO) (KW) / 𝑃𝐿𝑜𝑠𝑠 (IC) 279.1 / 301.5 / 278.98 / 269.91/ 271.777 / 268.386 /
(KW) 362.858 362.858 362.858 362.858 362.858 362.8578
% 𝑃𝐿𝑜𝑠𝑠 reduction 23 16.91 23.1 25.63 25.10 26.035
450 (65) 650 (58) 800 (59) 850 (58)
750 (20) 750 (10)
Capacitor Size (kVAr) / 450 (73) 450 (73) 450 (72) 400 (72)
250 (25) 750 (20)
Nodes 600 (84) 450 (84) 500 (83) 500 (84)
800 (58) 900 (58)
250 (87) 300 (90) 300 (90) 250 (89)
𝑉𝑚𝑖𝑛 (p.u) 0.9094 0.9124 0.9039 0.9065 0.8485 0.9065
45658.53
𝑃𝐿𝑜𝑠𝑠 cost (AC) ($) 46888.8 50652 46868.64 45344.88 45088.848
6
10308.14 15615.26 15301.60
∆𝑃𝐿𝑜𝑠𝑠 Cost ($) 14071.344 14091.504 15871.296
4 4 8
Cost of Capacitor ($/(KVAR-
----- ----- ----- ----- ----- 578.7
year))
-----
% AFB ----- ----- ----- ----- 25.0862

The total Power loss cost under IC is $60761.62848. From Table 3, it is observable that the Power
loss has reduced from 362.8578 to 268.386 KW, which is 26.035% compared to the IC after reactive
power injection of 97.71615% of the total (QD + QLoss (AO)), at three optimal nodes. The difference in
bus voltage enhancement is found to be 0.0523644 p.u. After reactive power compensation, the Power
loss cost (∆PLoss) change is $15871.296. Thus, the net AFB, $15292.596, is more than 25%. By
comparing the P(Loss(AO)) with other methods [17,18], AOA achieves better performance. Figure 9
shows the graph of the bus voltages before and after optimisation. Fig. 9 shows that the enhancement
in bus voltage is uniformly distributed in all the buses, and the overall voltage enhancement is
identified as 6%. The minimum node voltage achieved by the proposed method is less compared to
549
 ECJSE2023 (3) 538-553 Optimal Reactive Power Injection …

the genetic algorithm (GA) [17] and particle swarm optimisation (PSO) [17] and equals MTLBO
[17].

Figure 8. Bus voltage profile – IEEE 69-Bus DS

Figure 9. Bus voltage profile – Portugal 94-Bus DS

5. Discussions

As we have already seen, even without using SI, the proposed method yields better results than any
other methods taken for comparison [9-18]. If the algorithm has optimal location and sizing, the AFB
will be better than [9-18].

(i) Considering the Indian 10 Bus system, 100% reactive power support will yield 12.75787% of real
power loss reduction. However, optimal results will not be achieved if the reactive power support is
more [9] or less than adequate [11, 12]. It is noticeable that the difference between the result obtained
by AOA and SSA [10] is found to be minuscule. FPA [11] achieves better voltage enhancement than
any other method considering the bus voltage improvement. Understandably, the total reactive power
support given to any DS should be at the search for most of the network's total reactive power demand
(QD). On the other hand, CBGA [9] violates the previous statement by injecting 160% of the total
QD. It is logical from [9] that over-compensation will reverse the objective of the work and will lead
to financial loss.

(ii) The capacitor investment cost yielded by GWO, WCA, COA, and SMOA is almost the same, and
the difference in capacitor size is only 150 KVAr. Though AOA recorded the least Power loss
reduction after compensation compared to other methods, the % AFB is less than SMOA. This is
because of the increased capacitor cost. The difference between AOA and other algorithms mentioned
in Table 2 could be more manageable.

(iii) Regarding 94-bus Portugal practical DS, the change in power loss cost achieved by AOA due to
reactive power compensation at three optimal nodes is more than GA, PSO, TLBO, MTLBO, and
PSO [18], and the bus voltage also seems to be better. By analyzing the overall performance of the

550
 G Srinivasan, M Lavanya ECJSE 2023(3) 538-553

proposed method taking into consideration the above test systems, it is evident that AOA yields better
performance. However, it is noticeable that, though AFB is imperative from an economic point of
view, the performance of AOA in achieving voltage profile enhancement is found to be less. This
study has not considered (i) Capacitor installation, operation, and maintenance costs and (ii) Load
levels because of time and space constraints.

6. Conclusions

In this paper, a powerful physics-inspired intelligence algorithm named AOA has been utilised to
solve the cost-based objective function for the capacitor placement problem: the combination of
Power loss cost with capacitor investment cost to get more AFB. The merits of adopting AOA for
this problem have already been discussed. The proposed method has been successfully applied to an
Indian 10-bus, a new modified 12-bus, an IEEE 69-bus, and a real Portugal 94-bus test system. The
following are the key points that are worth noting.

i. As already discussed, this paper has not considered the sensitivity-based index for identifying
the most critical buses for reactive compensation. AOA must search for optimal location and
sizing to yield the least cost value.

ii. The performance of the Single feeder 10-bus system, IEEE 69-bus system, and 94-bus
Portuguese system have been analyzed and compared with the recent methods presented in
the literature. The simulation results show that the difference in PLoss reduction and AFB
achieved by AOA are better and more significant. Hence AOA has been recommended to be
another strong and efficient method for solving reactive power optimisation. Thus, AOA has
been acknowledged as an effective optimisation tool for solving problems related to DSs.

Authors’ Contributions

Srinivasan G: Conceptualization, Data collection, Running the MATLAB coding, taking the result,
and partial manuscript preparation. Lavanya M: Conceptualization, Analysis, and Interpretation of
Results and Review & Editing of the Manuscript.

Competing Interests

The authors declare that they have no competing interests.

References

[1]. S. G.Ankaliki, A. D. Kulkarni, S. I. Katti “Hierarchical Compensation Technique for Loss

Minimization in DistributionSystems,” IOSR Journal of Engineering (IOSRJEN), vol. 2, no.
2, pp. 356-362, February 2012.
[2]. G. G. Soma, “Optimal Sizing and Placement of Capacitor Banks in Distribution Networks
Using a Genetic Algorithm. Electricity,” MDPI, vol. 2, no. 2, pp. 187-204, May 2021.
[3]. P. S. Babu, B. Pamuleti, M. P. Lalitha, “Optimal Capacitor Placement in Radial Distribution
Systems Using Fuzzy and Advanced Harmony Search Algorithm for Loss Reduction,”
International Journal of Engineering Research &Technology, vol. 2, no. 11, pp. 3256-60,
November 2013.
[4]. M. N.l Muhtazaruddin, J. J. Jamian, D. Nguyen, N. A. Jalalludin and G. Fujita, “Optimal
capacitor placement and sizing via artificial bee colony,” International Journal of Smart Grid
and Clean Energy, vol. 3, no. 2, pp. 200-206, April 2014.

551
 ECJSE2023 (3) 538-553 Optimal Reactive Power Injection …

[5]. S. M. Tabatabaei, B. Vahidi, “Bacterial foraging solution based fuzzy logic decision for
optimal capacitor allocation in radial distribution system,” Electric Power Systems Research,
vol 81, pp. 1045-1050, April 2011.
[6]. A. A. El-Fergany, “Optimal capacitor allocations using evolutionary algorithms,” IET
Generation Transmission & Distribution, vol. 7, no. 6, pp. 593-601, June 2013.
[7]. A. Elsheikh, Y. Helmy, Y. Abouelseoud, A. Elsherif, et al., “Optimal capacitor placement and
sizing in radial electric power systems,” Alexandria Engineering Journal, vol. 53, no. 4, pp.
809-816, December 2014.
[8]. B. Venkatesh, B. Ranjan, “Data structure for radial distribution load flow analysis,” In Proc.
Of the IEE Proceeding in Generation, Transmission and Distribution, vol. 150, no. 1, pp. 101-
106, January 2003.
[9]. F. E. Riaño, J. F. Cruz, O. D. Montoya, H. R. Chamorro, L. Alvarado-Barrios, et al.,
“Reduction of Losses and Operating Costs in Distribution Networks Using a Genetic
Algorithm and Mathematical Optimization,” Electronics, vol. 10, no. 419, pp. 1-25, February
2021.
[10]. C. K. Reddy, G. Srinivasan, B. S. Lokasree, “Optimization of reactive power support in Power
Distribution Network – An economically based study,” In Proceedings of the 7th International
Conference on Electrical Energy Systems, February 2021.
[11]. A. Y. Abdelaziz, E. S. Ali, S. M. Abd Elazim, “Flower Pollination Algorithm for Optimal
Capacitor Placement and Sizing in Distribution Systems,” Electric Power Components and
Systems, vol. 44, no. 5, pp. 544-555, February 2016.
[12]. D. Jose da Silva, E. Antonio Belati, E. Werley Silva dos Angelos. “FPAES: A Hybrid
Approach for the Optimal Placement and Sizing of Reactive Compensation in Distribution
Grids,” Energies, vol. 13, pp. 1-18, December 2020.
[13]. S. K. Sampangi, T. Jayabarathi, “Optimal capacitor allocation in distribution networks for
minimization of power loss and overall cost using water cycle algorithm and grey wolf
optimizer,” International Transactions on Electrical Energy Systems (Wiley), vol. 30, no. 5,
pp. 1-32, January 2020.
[14]. V. Tamilselvan, T. Jayabarathi, T. Raghunathan, Xin-She Yang, et al., “Optimal capacitor
placement in radial distribution systems using flower pollination algorithm,” Alexandria
Engineering Journal, vol. 57, pp. 2775-2786, December 2018.
[15]. A. R. Youssef, S. Kamel, M. Ebeed, J. Yu, et al., “Optimal Capacitor Allocation in Radial
Distribution Networks using a combined optimization approach,” Electric Power Components
and Systems, vol. 46, no. 19-20, pp.2084-102, December 2018.
[16]. L. C. Kien, T. T. Nguyen, T. D. Pham, T. T. Nguyen, “Cost reduction for energy loss and
capacitor investment in radial distribution networks applying novel algorithms,” Neural
Computing and Applications, vol. 22, pp. 1-28, June 2021.
[17]. A. Rahiminejad, A. Foroughi-Nematollahi, B. Vahidi, S. Shahrooyan, et al., “Optimal
Placement of Capacitor Banks Using a New Modified Version of Teaching- Learning- Based
Optimization Algorithm,” AUT Journal of Modeling and Simulation, vol. 50, no. 2, pp. 171-
80, December 2018.
[18]. J. Gholinezhad, R. Noroozian, A. Bagheri, “Optimal Capacitor Allocation in Radial
Distribution Networks for Annual Costs minimization Using Hybrid PSO and Sequential
Power Loss Index Based Method,” Journal of Operation and Automation in Power
Engineering, vol. 5, no. 2, pp. 117-130, December 2017.
[19]. T. P. M. Mtonga, K. K. Kaberere, G. K. Irungu, “Optimal Shunt Capacitors’ Placement and
Sizing in Radial Distribution Systems Using Multiverse Optimizer,” IEEE Canadian Journal
of Electrical and Computer Engineering, vol. 44, no. 1, pp. 10-21, February 2021.
[20]. L. C. Kien, T. T. Nguyen, b. H. Dinh, T. T. Nguyen, “Optimal Reactive Power Generation for
Radial Distribution Systems Using a Highly Effective Proposed Algorithm,” Complexity
(Hindawi), vol. 2021, pp. 1-36, January 2021.

552
 G Srinivasan, M Lavanya ECJSE 2023(3) 538-553

[21]. M. Babanezhad, S. Arabi Nowdeh, A.Y.Abdelaziz, K.M. AboRas, and H. Kotb, “Reactive
power based capacitors allocation in distribution network using mathematical remora
optimization algorithm considering operation cost and loading conditions,” Alexandria
Engineering Journal, vol. 61, no. 12, pp. 10511-10526, December 2022.
[22]. M. J. Tahir, M.B. Rasheed, and M. K. Rahmat, “Optimal Placement of Capacitors in Radial
Distribution Grids via Enhanced Modified Particle Swarm Optimization,” Energies, vol. 15,
pp. 1-27, March 2022.
[23]. O. D. Montoya, W. Gil-González and A. Garcés, “On the Conic Convex Approximation to
Locate and Size Fixed-Step Capacitor Banks in Distribution Networks,” Computation, vol. 10,
no. 32, February 2022.
[24]. T. P. M. Mtonga, K. K. Kaberere, and G. K. Irungu, “A Novel Optimal Shunt Capacitors
Placement and Sizing Technique for Cost Minimization,” TechRxiv. Preprint, pp. 1-8,
December 2021.
[25]. R. A. Gallego, A. J. Monticelli, R. Romero, “Optimal Capacitor Placement in Radial
Distribution Networks,” IEEE Transactions on Power Systems, vol. 16, no. 4, pp. 630-637,
November 2001.
[26]. A. A. El-Fergany, A. Y. Abdelaziz, “Capacitor placement for net saving maximization and
system stability enhancement in distribution networks using artificial bee colony-based
approach,” Electrical Power and Energy Systems, vol.54, pp. 235-243, January 2014.
[27]. F. A. Hashim, K. Hussain, E. H. Houssein, M. S. Mabrouk, A. A. W, “Archimedes
optimization algorithm: a new metaheuristic algorithm for solving optimization problems,”
Applied Intelligence, vol. 51, pp. 1531-1551, September 2020.
[28]. D. Das, H. S. Nagi, D. P. Kothari, “Novel method for solving radial distribution networks,”
IEEE Proc. on Generation, Transmission & Distribution, vol. 141, no. 4, pp. 291-8, July 1994.
[29]. M. M. Aman, G. B. Jasmon, H. Mokhlis, A. H. A. Bakar, “Optimal Placement and sizing of
DG placed on a power stability index and line losses,” Electrical Power and Energy Systems,
vol. 43, pp.1296-1304, December 2012.

553