Received: March 6, 2023. Revised: April 3, 2023.

378

Optimal Integration of Multiple D-SVCs for Voltage Stability Enhancement in

Radial Electrical Distribution System Using Adaptive Firefly Algorithm

P. Muthukumar1 M. V. Ramesh1 Ponnam Venkata Kishore Babu2* P. Rohinikumar2

S. V. Satyanarayana2

1
Prasad V. Potluri Siddhartha Institute of Technology, Vijayawada-520007, Andhra Pradesh, India
2
RVR & JC College of Engineering, Guntur - 522019, Andhra Pradesh, India
* Corresponding author’s Email: [email protected]

Abstract: Most electrical distribution systems (EDS) are radially structured so that primary and secondary
protection control devices can work together well. Because they have a high R/X ratio in their design, they also have
a bad voltage profile, high distribution losses, and less stability margins. As a result, integrating reactive power
compensation devices like the distribution-static VAr compensator (D-SVC) can solve these issues. Yet, the situation
can get worse if D-SVCs are placed and rated incorrectly in radial distribution networks (RDNs). Determining the
optimum locations and their ratings simultaneously requires an updated version of the Firefly algorithm (FA) with
adaptive parameters, which is introduced in this paper as the adaptive firefly algorithm (AFA). The multi-objective
function that has been presented relates to improving loadability, voltage stability, and reducing active power loss.
On the IEEE 69-bus, simulations are run for three different VAr compensation levels. In comparison to the base case,
the losses are reduced by 34.04% and 33.14% with 50% and 75% VAr compensation, respectively. But for the
optimal VAr compensation of 73.14 percent by AFA, the losses are reduced by 35.29 percent, which is higher than
both under and over compensation cases. Similarly, the loadability margin is increased to 3.099 p.u. with optimal
VAr compensation, but it is observed as only 2.833 p.u. and 2.939 p.u. with 50% and 75% VAr compensation,
respectively. On the other hand, the findings produced with APF demonstrate its efficiency for resolving complex
optimization issues and outperform those obtained with previous research. Also, the proposed D-SVCs allocation has
improved RDN's overall performance, demonstrating how well it adapts to real-time applications.
Keywords: Radial distribution networks, Adaptive parameter, Distribution-static VAr compensator, Firefly
algorithm, Multi-objective optimization.

Thus, many researchers have been focused on

1. Introduction reactive power compensation in power systems after
experiencing blackouts. Flexible AC transmission
Industrialization and automation have resulted in
system (FACTS) devices are highly suggested for
a high demand for electricity in almost all power
reactive power flow control in transmission lines [3].
systems around the world. In comparison to the
On the other hand, capacitor banks (CBs), on-load
growth rate of active power generation sources in
tap-changers (OLTC), booster transformers, voltage
any power system, the growth rate of reactive power
regulators, network reconfiguration (NR), etc. have
compensation sources is very low. And thus, most of
been playing a key role in reactive power
the electrical distribution networks (EDNs) draw
compensation at the distribution side [4].
more reactive power from the main grid, resulting in
The technical (actual power loss reduction,
inadequate voltage magnitudes and voltage
feeder voltage profile improvement, and overall
instability/blackouts [1]. In addition, the radial
voltage stability margin enhancement) and financial
structure and high r/x ratio branches of EDNs cause
(operating cost reduction) benefits of integrating
the network performance to worsen significantly [2].
FACTS in distribution networks are numerous. Yet,
International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30
Received: March 6, 2023. Revised: April 3, 2023. 379

these advantages are only accessible when they are uncertainties [12]. Unified power quality controllers
perfectly networked. In this context, studies on (UPQCs), Distribution- static synchronous
power system planning have given a lot of attention compensators (D-STATCOMs), distribution-static
to the issue of optimal FACTS device integration in VAr compensators (D-SVCs), and distributed
EDN. thyristor-controlled series compensators (D-TCSCs)
In [5], loss sensitivity factors (LSFs) are are the best examples of such D-FACTS devices. In
proposed for determining pre-defined candidate order to accommodate high DG penetration, the
locations for CBs integration, and hybrid artificial need for voltage regulation in uncertain EDNs is
bee colony-particle swarm optimization (ABC-PSO) optimised by using OLTCs and static VAr
with fuzzy logic is introduced for deducing the compensators (SVCs) using robust optimization
optimal locations and ratings. The multi-objective (RO) [13]. In [14], a literature survey on optimal
function is formulated for real power loss and allocation of D-FACTS such as distribution-static
annual loss reduction. In [6], the dragonfly synchronous compensators (DSTATCOMs), unified
algorithm (DFA) and fuzzy expert system are power quality controllers (UPQCs), and CBs is
employed for identifying the optimal sites and presented. In addition, the grasshopper optimization
ratings of CBs in RDNs for real power loss algorithm (GOA) is adapted for the allocation of
reduction. In [7], minimization of active energy loss DSTATCOMs in the 69-bus EDN towards loss
and voltage deviation is aimed by optimally reduction, voltage profile improvement, and voltage
controlling the OLTCs along with distribution stability enhancement. In [15], the impact of D-
generation (DGs) using moth search optimization STATCOM on RDNs with different kinds of load
(MSO). In addition to DGs, CBs, and OLTCs, the models is analyzed, and the optimal location and
NR approach is also highly explored for managing sizes are determined using improved bald eagle
network power flows and improving overall search (IBES) by targeting a multi-objective
performance [8]. In [9], the modified culture function of loss, voltage profile, and voltage stability.
algorithm (MCA) is employed for reducing the In [16], the improved flower pollination algorithm
active power loss in EDNs by using an optimal NR (IFPA) and the voltage stability index (VSI) are
approach. In [10], NR and DGs are proposed for hybridised for solving the CBs and DSTATCOMs in
improving the consistency of EDN in terms of loss RDNs and mitigating the negative impact of electric
reduction and voltage stability enhancement under vehicle (EV) loads considering techno-economic
multiple loading conditions. The optimization benefits. In [17], optimal ratings and locations for
problem is solved using the enhanced marine UPQC along with NR are solved using the improved
predator algorithm (EMPA). In [11], NR with soft whale optimization algorithm (IWOA) for reducing
open points (SOPs) is proposed using artificial the active power loss and cost of UPQC and
rabbit optimization (ARO) for improving the switching operations. In [18], basic open-source
resilience of multi-lateral EDNs under renewable mixed-integer nonlinear programming (BONMIN)
energy (RE)-based DGs and electric vehicle (EV) is proposed for optimal integration of D-SVC and
uncertainty. D-TCSC along with DGs for improving the
CBs can be either fixed or switched efficiency of EDN. In [19], the adaptive differential
configuration and not able to provide appropriate search algorithm (ADSA) is utilised for solving the
and dynamic VAr support. Thus, CBs can lead to SVC location and sizes, along with active power
either under or over compensation, results for either DGs for ensuring minimum distribution losses. In
low voltage or high voltages in the EDNs. On the [20], a gradient-based optimizer with a crossover
other side, NR method needs remote control operator (GBOC) is introduced for solving D-SVCs
switches (RCSs) in each branch and need to in RDN for techno-economic benefits. Further,
cooperatively to control these switches along with different approaches to reactive power compensation
tie-lines. Unfortunately, most of the EDNs are not in EDNs via conventional approaches and D-FACTS
fully automated with RCSs for dynamic NR and this can be reviewed in [21].
approach for performance improvement of EDNs is From the above reviewed works, different
limited. Similarly, OLTC transformers designed with methodologies, like linear programming (LP), non-
fixed tap-settings and are not fine tuneable for linear programming (NLP), and dynamic
dynamic VAr control. In comparison to these programming (DP), were employed for solving the
methods using CBs, NR, and OLTC, now a days, the optimal allocation of D-FACTS in EDN. Meta-
adaptation of FACTS devices at the distribution side, heuristics were highly used due to their multiple
namely D-FACTS devices, is getting high attention advantages. The complexity involved in solving
due to their dynamic and fast response to the real-time engineering problems with multiple
International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30
Received: March 6, 2023. Revised: April 3, 2023. 380

objectives formulated with various simultaneous

linear and non-linear constraints, equal and unequal
constraints, and continuous and discrete variables
can be easily overcome by meta-heuristics. They are
free from derivatives and easy to adapt with
minimum control variables [22]. However, meta-
heuristics suffer from the generation of a random
population in the optimization process, which can
lead to poor exploration and/or exploitation
characteristics and further result in a local minima
trap. On the other hand, the no-free-lunch (NFL)
theorem states that many algorithms may not suit
solving all kinds of optimization problems [23].
Thus, the researchers are still motivated to introduce
new algorithms and also make improvements to the (a) (b)
existing algorithms. Firefly algorithm (FA) is one Figure. 1 Schematic diagram of D-SVC, (a) combination
of TCR and TSC, (b) fixed CB and TCR
such simple and efficient algorithm in recent times
inspired by the flashing patterns and behaviour of
investigated. Hence, adding adaptive parameters to
fireflies [24]. But it has to be strengthened and made
FA's exploration and exploitation stages can enhance
simpler because it sometimes gets stuck in local
its convergence characteristics. In light of this, this
optima and loses its ability to optimise [25]. The
work presents an intriguing subject for studies on
performance of basic FA is dependent on mainly two
power system management and performance
controlling factors i.e., light variation and attraction.
enhancement.
In literature [26], various improvements have been
The mathematical modelling of D-SVC is
suggested for basic FA by modifications to these
covered in section 2. The suggested multi-objective
parameters. On the other side, some researchers are
problem is described in section 3 along with its
also experimented with hybridizing with other
equal and unequal restrictions. The solution
algorithms. In order to overcome these issues,
methodology utilising AFA and its mathematical
adaptive parameters are introduced for improving
relationships in the optimization process is presented
the search ability of the basic FA in the adaptive
in section 4. The simulation results on the IEEE 69-
firefly algorithm (AFA) [27].
bus using the suggested AFA and a comparison of
In light of the above-identified research problem
effeteness to literary works are covered in section 5.
and in comparison to the literature, the following are
Lastly, section 6 projects the thorough study results
the major contributions of this paper:
and major advancements made by this paper.
1) A multi-objective approach for optimal
allocation of D-SVC is proposed.
2. Modelling of D-SVC
2) At the first stage, the preferable locations for One of the shunt type FACTS devices is SVC
installing D-SVC are determined using the which is designed basically with thyristor switched
voltage stability index (VSI). capacitor (TSC) and thyristor controlled reactor
3) In the second stage, AFA is used to get the best (TCR) for working in either reactive power source
reactive power output from D-SVC. (under capacitive mode) or reactive power sink
4) Adaptive parameters are introduced for the basic (under inductive mode). The schematic diagram of
Firefly algorithm (FA) for developing the SVC is given in Fig. 1.
proposed adaptive Firefly algorithm (AFA) for By controlling the firing angles of thyristors, the
improving its search capabilities in the total susceptance of either TCR or TSC can be
optimization process. regulated. The relation between firing angle and
5) The optimal VAr values of D-SVC are evaluated susceptance of the TCR is given by,
under different loading conditions on the IEEE
69-bus radial EDN. 1 2𝛼
𝐵𝐿(𝛼) = 𝜔𝐿 (1 − 𝜋
) and 𝐵𝐶 = 𝜔𝐶 (1)
The D-SVC is not optimally integrated to figure
out how much the RDN can be loaded. The The total susceptance of the D-SVC is equal to
effectiveness of FA in resolving the D-SVC the summation of both 𝐵𝐿(𝛼) and 𝐵𝐶 , given by,
allocation issue in EDN has not yet been
International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30
Received: March 6, 2023. Revised: April 3, 2023. 381

𝐵𝑆𝑉𝐶 = 𝐵𝐿(𝛼) + 𝐵𝐶 (2) SVC integration in the network, respectively; 𝑤1 ,

𝑤2 and 𝑤3 are the weighting factors for the
The reactive power support by D-SVC under objective functions 𝑓1, 𝑓2 and 𝑓3, respectively. .
reactive power source/ sink scenarios is given by,
3.2 Constraints
2
𝑄𝑆𝑉𝐶 = ∓𝐵𝑆𝑉𝐶 𝑉𝑔𝑟𝑖𝑑 (3)
The bus voltage magnitudes, VAr and location
limits for D-SVC and VSI limit are considered.
where 𝐵𝐿(𝛼) , 𝐵𝐶 and 𝐵𝑆𝑉𝐶 are the susceptances
of TCR, CB and D-SVC, respectively, 𝜔 is the |𝑉|𝑚𝑖𝑛 ≤ |𝑉|𝑛 ≤ |𝑉|𝑚𝑎𝑥 , ∀𝑛 = 2: 𝑛𝑏𝑢𝑠 (8)
angular frequency, 𝐿 and 𝐶 are the inductance and
capacitances, respectively; 𝛼 is the firing angle of ∑𝑛𝑑𝑠𝑣𝑐 𝑛𝑏𝑢𝑠
𝑘=1 𝑄𝐷𝑆𝑉𝐶(𝑘) ≤= ∑𝑖=1 𝑄𝑑(𝑖) (9)
thyristor, 𝑄𝑆𝑉𝐶 is the reactive power compensation
by D-SVC; 𝑉𝑔𝑟𝑖𝑑 is the voltage magnitude of grid- 2 ≤ 𝐿𝐷𝑆𝑉𝐶 ≤ 𝑛𝑏𝑢𝑠 (10)
bus or D-SVC incident bus in the network.
In Eq. (3), positive sign indicates the reactive 𝑉𝑆𝐼 > 0, ∀𝑛 = 2: 𝑛𝑏𝑢𝑠 (11)
power consumption state by D-SVC for reducing the
grid-bus voltage magnitude and whereas negative where |𝑉|𝑛 , |𝑉|𝑚𝑖𝑛 and |𝑉|𝑚𝑎𝑥 are the voltage
sign indicates the reactive power support by D-SVC magnitude of bus-n, and its minimum and maximum
for increasing the voltage magnitude of grid-bus. limits, respectively; 𝑄𝐷𝑆𝑉𝐶(𝑘) and 𝑄𝑑(𝑖) are the VAr
Thus by changing the 𝛼, the overall susceptance of capacity of D-SVC at bus-k and VAr load at bus-i,
the D-SVC can be changed and consequently, the respectively; 𝐿𝐷𝑆𝑉𝐶 is the location of D-SVC in the
reactive power output from D-SVC can be adjusted network, 𝑛𝑏𝑢𝑠 and 𝑛𝑑𝑠𝑣𝑐 are the number of buses
dynamically as per the operating conditions. and number of D-SVCs in the network, respectively.
3. Problem formulation 4. Solution methodology
This section introduces the proposed multi- The mathematic modelling of firefly algorithm
objective optimization problem along with its concept and the proposed modifications with
various equal and unequal constraints. dynamic parameter are explained in this section.
3.1 Multi-objective function 4.1 Evaluation of objective functions
Reduction of power losses(𝑓1 ), improvement of The real power distribution losses are evaluated
voltage stability index (𝑓2 ) , and enhancement of by using the Newton Raphson load flow method
loadability (𝑓3 ) are aimed in optimizing the location (NRLF) [28]. The VSI is evaluated by using the
and sizes of D-SVC in this paper. Mathematically, methodology defined in [29]. The maximum
𝑃𝑙𝑜𝑠𝑠(𝐷𝑆𝑉𝐶)
loadability is evaluated by using the repeated power
𝑓1 = 𝑃𝑙𝑜𝑠𝑠(𝑏𝑎𝑠𝑒)
(4) flow (RPF) [29]. Mathematically,

𝑉𝑆𝐼(𝑏𝑎𝑠𝑒) 𝐿 = {∑𝑛𝑏𝑢𝑠
𝑖=1 (𝑃𝑑(𝑖) + 𝑗𝑄𝑑(𝑖) )}(1 + 𝜆𝑚𝑎𝑥 ) (8)
𝑓2 = (5)
𝑉𝑆𝐼(𝐷𝑆𝑉𝐶)
where 𝑃𝑑(𝑖) is the real power demand at bus-i,
𝜆𝑚𝑎𝑥(𝑏𝑎𝑠𝑒) 𝜆𝑚𝑎𝑥 is the maximum load increment factor at
𝑓3 = 𝜆 (6)
𝑚𝑎𝑥(𝐷𝑆𝑉𝐶) which the NR load flow method fails to converge. In
other words, the loading condition at which Jacobian
𝑂𝐹 = 𝑚𝑖𝑛(𝑤1 𝑓1 + 𝑤2 𝑓2 + 𝑤3 𝑓3 ) (7) matrix becomes singular.

where 𝑃𝑙𝑜𝑠𝑠(𝑏𝑎𝑠𝑒) and 𝑃𝑙𝑜𝑠𝑠(𝐷𝑆𝑉𝐶) are the real 4.2 Strategy for D-SVC locations
power losses with and without D-SVC in the
network, respectively; 𝑉𝑆𝐼(𝑏𝑎𝑠𝑒) and 𝑉𝑆𝐼(𝐷𝑆𝑉𝐶) are The VSI should be more than 0 and less than 1,
the voltage stability index (VSI) of the network according to [28]. The buses that are getting close to
before and after D-SVC integration in the network, 0 can be thought of as having more potential for
respectively; 𝜆𝑚𝑎𝑥(𝑏𝑎𝑠𝑒) and 𝜆𝑚𝑎𝑥(𝐷𝑆𝑉𝐶) are the voltage collapse. Hence, enhancing the voltage
profile and consequently the VSI at those locations
maximum loadability margin before and after D-
can lead to an improvement in overall stability of
International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30
Received: March 6, 2023. Revised: April 3, 2023. 382

the network. All of the locations are ranked in 4.4 Adaptive firefly algorithm
descending order based on their respective VSI
values after being determined. As a pre-defined An adaptive parameter strategy is used in the
search space for D-SVC integration, the top ten AFA to control the step factor and attractiveness. In
locations are taken into consideration. Next, by the basic FA, the parameters 𝛿 and 𝛾 are constant
employing the suggested optimization approach, the and thus subjected to local optima trap. In AFA, they
optimal locations are deduced from them together are dynamically tuned for improving search
with the ratings. characteristics.
𝑡
4.3 Basic firefly algorithm (−𝑘 )
𝛾(𝑡 + 1) = 𝛾(𝑡) × 𝑒 𝑡𝑚𝑎𝑥 (12)
In nature, fireflies' flashing inspired the firefly 𝑡
algorithm (FA). Fireflies release quick, rhythmic (−𝑚 )
𝛿(𝑡 + 1) = 𝛿(𝑡) − 𝜗𝑒 𝑡𝑚𝑎𝑥 (13)
bioluminescent flashes. Flashing lights attract
companions, prey, and predators. Hence, the where 𝜗= 0.9, 𝑘{= 1,2,3, . . } and 𝑚{= 1,2,3, . . }
intensity of the light affects other fireflies' approach. are used to defined the rate of decrease and increase
The FA's main objective function is brightness, the dynamic movements, respectively; 0.5 and 1 are
based on the fireflies’ 𝑓𝑖 and 𝑓𝑗 . The particle's lesser used for 𝛾(0) and 𝛿(0), respectively; 𝑡 and 𝑡𝑚𝑎𝑥 are
brightness helps find and brighten the brightest the number of present and maximum iteration,
particles. Distance between particles reduces respectively.
brightness. Fireflies are ordered by brightness. The
particle will locate their ideal partner until the 5. Results and discussion
number of generations is limited.
The light intensity (𝐿𝑖 ) of the firefly associated The proposed AFA is implemented for solving
with the solution is proportional to the target value the D-SVC locations and ratings in IEEE 69-bus
of the fitness function and is defined by: feeder. The load data and branch data are taken from
[29]. It has real and reactive power loading levels as
2
𝛾𝑖(𝑑) = 𝛾𝑖(0) 𝑒 −𝜎𝑑 , 𝑑 ≥ 1 (9) 3802.1 kW and 2694.7 kVAr, respectively.

5.1 Simulations with different algorithms

where 𝑑 is the distance between two fireflies,
𝐿𝑖(0) and 𝛾𝑖(𝑑) are the firefly’s initial ( 𝑑 =0) and Base case: For the standard test system data [30],
distance based light intensity, respectively; 𝜎 is the NRLF is performed for determining the performance
light absorption factor. of EDN. It is noted that the total real and reactive
The distance (𝑑) is modelled using Euclidian power losses of 𝑃𝑙𝑜𝑠𝑠(𝑏𝑎𝑠𝑒) = 225 kW and 102.2
distance formula, as follows: kVAr, respectively. The minimum voltage
magnitude is registered at bus-65 as 0.9092 p.u. The
2 least VSI is determined as 𝑉𝑆𝐼(𝑏𝑎𝑠𝑒) = 0.55 at bus-60.
𝑑(𝑖,𝑗) = ‖𝑓𝑖 − 𝑓𝑗 ‖ = √∑𝑛𝑠
𝑘=1(𝑓𝑖,𝑘 − 𝑓𝑗,𝑘 ) (10) By implementing repeated power flow (RPF) [31],
the maximum loadability of the network is
where 𝑓𝑖,𝑘 and 𝑓𝑗,𝑘 are the kth member of ith and determined as 𝜆(𝑏𝑎𝑠𝑒) = 2.211 p.u. These results
jth firefly, respectively; 𝑛𝑠 is the number of search treated as base case for comparison.
variables of the problem. Ideal case: Before optimization the locations
The movement between ith and jth fireflies is and ratings of D-SVCs, the global optima values or
modelled by, ideal case values are determined by considering total
reactive power loading as zero, which is aimed by
2 1 VAr compensation with D-SVCs in the network.
𝑓𝑖 = 𝑓𝑖 + 𝛾𝑖(0) 𝑒 −𝜎𝑑 (𝑓𝑖 − 𝑓𝑗 ) + 𝛿 (𝑟 − 2) (11)
The total real and reactive power losses reduced to
𝑃𝑙𝑜𝑠𝑠(𝑏𝑎𝑠𝑒) = 143.523 kW and 65.298 kVAr,
Here, the first and second parts in Eq. (11) are
respectively. The minimum voltage magnitude is
used to define attraction and third is for defining
registered at bus-65 as 0.9317 p.u. The least VSI is
random step movement by a parameter 𝛿 and 𝑟 is a
determined as 𝑉𝑆𝐼(𝑏𝑎𝑠𝑒) = 0.7012 at bus-60. By
randomly generated uniform number between 0 and
1. implementing RPF, the maximum loadability of the
network is determined as 𝜆(𝑏𝑎𝑠𝑒) = 3.064 p.u. These
results treated as ideal case for comparison.
International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30
Received: March 6, 2023. Revised: April 3, 2023. 383

Table 1. Results of different algorithms the maximum loadability of the network is

Method
D-SVC in ± kVAr Ploss
VSI determined as 𝜆(𝐷𝑆𝑉𝐶) = 3.099 p.u. These results
(bus #) (kW) treated as optimal VAr comparison level of 73.14%.
604 (66), 699 (47), In addition to APFA, basic FA, cuckoo search
CSA 147.553 0.7178
1250 (61)
algorithm (CSA) [32], flower pollination algorithm
172 (52), 1255 (61),
FPA 145.985 0.7170 (FPA) [33] and teaching learning based optimization
340 (19)
510 (50), 1252 (61), (TLBO) [34] also used solve the proposed objective
TLBO 145.778 0.7155 function. With respect to base case values, the best
368 (19)
303 (23), 1169 (61), results obtained by each algorithm over 25
FA 145.721 0.7166 independent runs are given in Table 1. The
469 (53)
281 (21), 1184 (61), convergence characteristics of these algorithms are
AFA 145.605 0.7175
506 (53) given in Fig. 2.
In order to compare the effectiveness of VAr
compensation on the performance of EDN, the AFA
0.800 is used for determining the D-SVCs’ locations and
TLBO ratings. The corresponding results for different VAr
0.780
Objective Function

FPA
compensation levels are given in Table 2. From this
0.760 CSA
analysis, it is evident that the under compensation
FA
0.740 (less than optimal) or over compensation (more than
AFA
0.720 optimal) of VAr can result for adverse effects on the
0.700
network performance. The improved voltage
profiles for different VAr compensation levels are
0.680
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49
compared and given in Fig. 3.
Iteration
5.2 Comparison with literature
Figure. 2 Convergence characteristics
5.2.1. Simulations with D-SVCs
1.020
In this section, the efficiency of AFA is
Bus voltage magnitude (p.u.)

1.000 compared with literature for two different case

0.980 studies. In case 1, the VAr compensation limit is
0.960
defined as 50% of total reactive power demand of
the network and whereas in case 2, it is taken as
0.940 Base Optimal VAr 75%, respectively. From [20], the results of
0.920 50 % VAr 75 % VAr gradient-based optimizer (GBO), GBO with
0.900 crossover operator (GBOC), dwarf mongoose
1 4 7 1013161922252831343740434649525558616467 optimization algorithm (DMOA), salp swarm
Bus number algorithm (SSA), differential evolution (DE),
bernstein-levy search DE (BSDE) and honey badger
Figure. 3 Comparison of voltage profile
algorithm (HBA) are compared with the proposed
AFA.
Optimal Case: For improving this operating
Under Compensation (50%): In this case, three
state, three D-SVCs are proposed to integrate
D-SVCs are optimally integrated for 50% reactive
optimally. The search dimension is equal to 6 (i.e., 3
power compensation. This is less than optimal VAr
for locations and 3 for ratings). In the proposed
compensation and treated as under compensation.
multi-objective function, the weighting factors 𝑤1 ,
The best sizes of D-SVCs by AFA in ± kVAr (bus #)
𝑤2 and 𝑤3 are taken as 0.6, 0.2 and 0.2, respectively.
are as follows: 163 (63), 973 (61) and 212 (18). The
The best results obtained by AFA are as follows:
total real and reactive power losses reduced to
The sizes (locations) in ± kVAr are 281 (21),
𝑃𝑙𝑜𝑠𝑠(𝑏𝑎𝑠𝑒) = 148.342 kW and 69.197 kVAr,
1184 (61), and 506 (53), respectively. The total real
respectively.
and reactive power losses reduced to 𝑃𝑙𝑜𝑠𝑠(𝑏𝑎𝑠𝑒) =
The minimum voltage magnitude is registered at
145.605 kW and 67.844 kVAr, respectively. The
bus-65 as 0.9287 p.u. The least VSI is determined as
minimum voltage magnitude is registered at bus-65
𝑉𝑆𝐼(𝑏𝑎𝑠𝑒) = 0.7029 at bus-60. By implementing RPF,
as 0.9315 p.u. The least VSI is determined as
the maximum loadability of the network is
𝑉𝑆𝐼(𝑏𝑎𝑠𝑒) = 0.7175 at bus-63. By implementing RPF,
determined as 𝜆(𝐷𝑆𝑉𝐶) = 2.833 p.u. The comparison
International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30
Received: March 6, 2023. Revised: April 3, 2023. 384

Table 2. Network performance for different VAr compensation levels

VAr Locations Ratings Ploss Qloss
Vmin (p.u.) VSI
Comp (%) (bus #) (± kVAr) (kW) (kVAr
Base - - 225 102.2 0.9092 (65) 0.55 (60)
Ideal 143.523 65.298 0.9317 (65) 0.7012 (60)
10 27, 26, 64 13, 13, 1995 192.034 88.448 0.9424 (61) 0.6266 (60)
20 59, 61, 64 13, 981, 1026 175.018 79.694 0.9434 (61) 0.7367 (60)
30 22, 61, 64 9, 1398, 614 170.679 77.598 0.9424 (65) 0.7739 (64)
40 62, 61, 22 697, 1128, 196 158.363 72.832 0.9387 (65) 0.7408 (63)
50 64, 61, 69 292, 1448, 281 155.952 71.774 0.9385 (65) 0.7612 (64)
60 61, 61, 18 579, 902, 540 149.921 69.403 0.9347 (65) 0.7279 (63)
70 61, 26, 11 1336, 138, 548 146.063 67.996 0.9331 (65) 0.7227 (63)
73.14 21, 61, 53 281, 1184, 506 145.605 67.844 09315 (65) 0.7175 (61)
80 53, 17, 61 643, 255, 1123 145.892 67.967 0.9310 (65) 0.7160 (63)
90 61, 15, 29 927, 319, 776 151.868 71.159 0.9257 (65) 0.6902 (60)
100 2, 62, 11 511, 859, 651 152.907 70.805 0.9260 (65) 0.5920 (60)

Table 3. Comparison of AFA with various methods Table 5. Comparison of AFA in solving optimal VAr
reported in [20] for of 50% VAr compensation compensation using CBs
D-SVC in ±kVAr Ploss CB in kVAr Ploss
Method VSI Method VSI
(bus #) (kW) (bus #) (kW)
BSDE 31 (61), 259 (62) 196.124 0.5656 126 (11), 93 (13),
FIS-ABC- 145 (17), 117
SSA 226 (61), 259 (64) 180.630 0.5944 214.28* -
POS [5] (21), 94 (28), 104
393 (62), 236 (39), 96 (45)
DMOA 158.448 0.5889
(63), 309 (64) 1230 (61), 190
HBA 321 (61), 618 (62) 158.221 0.6232 FES-DA [6] (64), 100 (59), 150.43* -
557 (62), 382 100 (65), 360 (21)
DE 156.658 0.5897 526 (12), 1035
(63), 82 (69)
AFA (61), 43 (69), 208 146.27 0.7177
GBO 189 (21), 876 (61) 154.847 0.6813
(64)
204 (21), 589 * Indicates, results are revised as per the load flow
GBOC [20] 152.690 0.5912
(62), 355 (64)
308 (22), 818
AFA 149.405 0.7085
(61), 221 (20)
is given in Table 3 and the results of AFA are
observed as superior to all other algorithms by
having least objective function value.
Table 4. Comparison of AFA with various methods
reported in [20] for of 75% VAr compensation
Over Compensation (75%): In this case, the
D-SVC in ±kVAr Ploss reactive power compensation target is set as 75% by
Method VSI integrating three D-SVCs are optimally. This is
(bus #) (kW)
560 (22), 233 more than optimal VAr compensation and treated as
BSDE 196.124 0.5656 over compensation. The best sizes of D-SVCs by
(62), 413(63)
SSA 226 (61), 259 (64) 180.630 0.5944 AFA in ± kVAr (bus #) are as follows: 308 (22), 818
393 (62), 236 (61) and 221 (20). The total real and reactive power
DMOA 158.448 0.5889 losses reduced to 𝑃𝑙𝑜𝑠𝑠(𝑏𝑎𝑠𝑒) = 148.405 kW and
(63), 309 (64)
HBA 321 (61), 618 (62) 158.221 0.6232 69.24 kVAr, respectively.
557 (62), 382 The minimum voltage magnitude is registered at
DE 156.658 0.5897 bus-65 as 0.9282 p.u. The least VSI is determined as
(63), 82 (69)
GBO 189 (21), 876 (61) 154.847 0.6813 𝑉𝑆𝐼(𝑏𝑎𝑠𝑒) = 0.7074 at bus-63. By implementing RPF,
204 (21), 589 the maximum loadability of the network is
GBOC [20] 152.690 0.5912 determined as 𝜆(𝐷𝑆𝑉𝐶) = 2.939 p.u. The comparison
(62), 355 (64)
681 (63), 546 is given in Table 4 and the results of AFA are
AFA 150.442 0.6020
(41), 793 (57) observed as superior to all other algorithms by
having least objective function value.

International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30
Received: March 6, 2023. Revised: April 3, 2023. 385

Table 6. Comparison CBs and D-SVCs Author contributions

Ploss
Method Total kVAr (%) VSI Conceptualization, methodology, writing-
(Kw)
- - 225 0.55 original draft preparation: P Muthukumar; Software,
validation, review and editing: M V Ramesh;
CBs 1812 (67.24) 146.27 0.7172
Formal analysis, investigation, visualization:
D-SVCs 1971 (73.14) 145.605 0.7175 Ponnam Venkata Kishore Babu; Resources, data
curation: S V Satyanarayana; Supervision, project
administration: P Rohinikumar.
5.2.1. Comparison of CBs and D-SVCs
Notation list
In this case study, VAr compensation by means
of CBs and D-SVCs is compared. In [5, 6], the 𝐵𝐿(𝛼) Susceptance of TCR
impact of CBs on the performance of EDNs is 𝐵𝐶 Susceptance of CB
analyzed. Thus, the effectiveness of AFA in solving 𝐵𝑆𝑉𝐶 Susceptance of SVC
the CBs allocation problem is first analyzed, and 𝜔 Angular frequency
then the performance of EDN with CBs and D- 𝐿 Inductance
SVCs is compared. The simulation results obtained 𝐶 Capacitance
with AFA are given in Table 5. 𝛼 Thyristor firing angle
Table 6 provides a comparison of CBs and D- 𝑸𝑺𝑽𝑪 Reactive power by SVC
SVCs. CBs only reach the ideal VAr compensation 𝑽𝒈𝒓𝒊𝒅 Grid bus voltage magnitude
of 67.24%, but it should be closer to 73.14%. D- 𝑃𝑙𝑜𝑠𝑠(𝑏𝑎𝑠𝑒) Base case real power loss
SVCs produced a result of 145.605 kW, while the 𝑷𝒍𝒐𝒔𝒔(𝑫𝑺𝑽𝑪) Real power loss with D-SVC
losses decreased to 146.27 kW from the base case of 𝑽𝑺𝑰(𝒃𝒂𝒔𝒆) Voltage stability index at base case
225 kW. Moreover, the stability index is greater than 𝑽𝑺𝑰(𝑫𝑺𝑽𝑪) Voltage stability index at base case
the base case of 0.55 and equivalent to 0.7175 with 𝝀𝒎𝒂𝒙(𝒃𝒂𝒔𝒆) Loadability factor at base case
D-SVCs. Nonetheless, CBs' stability index is 0.7172, 𝝀𝒎𝒂𝒙(𝑫𝑺𝑽𝑪) Loadability factor with D-SVC
which is once again lower than D-SVCs. In this |𝑉|𝑛 Voltage magnitude of bus-n
fashion, D-SVCs are demonstrated to perform better |𝑉|𝑚𝑖𝑛 Minimum voltage magnitude
than CBs. |𝑉|𝑚𝑎𝑥 Maximum voltage magnitude
𝑄𝐷𝑆𝑉𝐶(𝑘) Reactive power by D-SVC at bus-k
6. Conclusion 𝑷𝒅(𝒊) Real power load at bus-i
In this paper, a novel meta-heuristic approach is 𝑄𝑑(𝑖) Reactive power load at bus-i
presented for solving the optimal locations and 𝐿𝐷𝑆𝑉𝐶 Location of D-SVC
ratings of distribution-static VAr compensators (D- 𝑛𝑏𝑢𝑠 Number of buses
SVCs) for reducing active power loss, voltage 𝑛𝑑𝑠𝑣𝑐 Number of D-SVCs
stability index, and loadability margin in radial 𝝀𝒎𝒂𝒙 Maximum loadability factor
electrical distribution networks (EDNs). An adaptive 𝐿𝑖 Light intensity
firefly algorithm (AFA) with improved search 𝑑 Distance between two fireflies
capabilities is proposed for solving the multi- 𝐿𝑖(0) Firefly’s intial light intensity
objective function with different equal and unequal 𝛾𝑖(𝑑) Firefly’s distance based light intensity
constraints. The computational efficiency of AFA is 𝜎 Light absorption factor
compared with that of other algorithms and 𝑡 Present iteration number
literature. From the comparative study, it is observed 𝑡𝑚𝑎𝑥 Number of maximum iterations
that the AFA is performing well by resulting in a
global optimum. Simulations are performed on the References
IEEE 69-bus EDN for different scenarios. For [1] H. H. Alhelou, M. E. H. Golshan, T. C. Njenda,
different VAr compensation levels, the network and P Siano, “A survey on power system
performance is evaluated and compared with the blackout and cascading events: Research
base case, ideal case, and optimal case. The results motivations and challenges”, Energies, Vol. 12,
emphasise the need for optimal VAr compensation No. 4, 682, 2019.
for ensuring the network's performance. [2] H. R. Bouchekara, Y. Latreche, K. Naidu, H.
Mokhlis, W. M. Dahalan, and M. S. Javaid,
Conflicts of interest “Comprehensive review of radial distribution
The authors declare no conflict of interest. test systems for power system distribution
International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30
Received: March 6, 2023. Revised: April 3, 2023. 386

education and research”, Resource-Efficient Renewable Energy and Environment, Vol. 10,
Technologies, Vol. 3, pp. 1-12, 2019. No. 1, pp. 43-58, 2023.
[3] Z. Čonka, M. Kolcun, M. K. Jr, J. Dudiak, M. [13] S. Wang, S. Chen, L. Ge, and L. Wu,
Mikita, and M. Vojtek, “Improvement of power “Distributed generation hosting capacity
system stability using FACTS device”, Power evaluation for distribution systems considering
and Electrical Engineering, Vol. 33, pp: 12-15, the robust optimal operation of OLTC and
2016. SVC”, IEEE Transactions on Sustainable
[4] D. Stanelyte and V. Radziukynas, “Review of Energy, Vol. 7, No. 3, pp. 1111-1123, 2016.
voltage and reactive power control algorithms [14] M. Ebeed, S. Kamel, S. H. A. Aleem, and A. Y.
in electrical distribution networks”, Energies, Abdelaziz, “Optimal allocation of
Vol. 13, No. 1, 58, 2019. compensators”, Electric Distribution Network
[5] S. Sharma and S. Ghosh, “FIS and hybrid Planning. Power Systems, pp. 321-253, 2018.
ABC-PSO based optimal capacitor placement [15] K. R. Rani, P. S. Rani, N. Chaitanya, and V.
and sizing for radial distribution networks”, Janamala, “Improved bald eagle search for
Journal of Ambient Intelligence and Humanized optimal allocation of D-STATCOM in modern
Computing, Vol. 11, pp. 901-916, 2020. electrical distribution networks with emerging
[6] E. A. A. Ammar, G. A. Ghazi, and W. Ko, loads”, Int. J. Intell. Eng. Syst, Vol. 15, No. 2,
“Optimal capacitor placement in radial pp. 554-563, 2021, doi:
distribution systems using a fuzzy-dragonfly 10.22266/ijies2022.0430.49.
method”, Int. J. Smart Grid Clean Energy, Vol. [16] R. Puppala and K. Chandrasekhar, “Optimal
8, No. 2, pp. 116-124, 2019. Allocation of Capacitor Banks and
[7] P. Singh, S. K. Bishnoi, and N. K. Meena, DSTATCOMs in Radial Distribution System
“Moth search optimization for optimal DERs Considering Electric Vehicle Load Growth”,
integration in conjunction to OLTC tap Int. J. Intell. Eng. Syst, Vol. 15, No. 6, pp. 45-
operations in distribution systems”, IEEE 53, 2022, doi: 10.22266/ijies2022.1231.05.
Systems Journal, Vol. 14, No. 1 pp. 880-888, [17] Y. Priyanka and R. Raghu, “Performance
2019. Analysis of Distribution System with Optimal
[8] O. Badran, S. Mekhilef, H. Mokhlis, and W. Allocation of Unified Power Quality
Dahalan, “Optimal reconfiguration of Conditioner Considering Distribution Network
distribution system connected with distributed Reconfiguration”, Int. J. Intell. Eng. Syst, Vol.
generations: A review of different 16, No. 1, pp. 364-374, 2023, doi:
methodologies”, Renewable and Sustainable 10.22266/ijies2023.0228.32
Energy Reviews, Vol. 73, pp. 854-867, 2017. [18] E. M. Ahmed, S. Rakočević, M. Ćalasan, Z. M.
[9] H. K. Verma and P. Singh, “Optimal Ali, H. M. Hasanien, R. A. Turky, and S. H.
reconfiguration of distribution network using Aleem, “BONMIN solver-based coordination
modified culture algorithm”, Journal of The of distributed FACTS compensators and
Institution of Engineers (India): Series B, Vol. distributed generation units in modern
99, pp. 613-622, 2018. distribution networks”, Ain Shams Engineering
[10] A. M. Shaheen, R. A. E. Sehiemy, S. Kamel, E. Journal, Vol. 13, No. 4, p. 101664, 2022.
E. Elattar, and A. M. Elsayed, “Improving [19] B. Mahdad and K. Srairi, “Adaptive differential
distribution networks’ consistency by optimal search algorithm for optimal location of
distribution system reconfiguration and distributed generation in the presence of SVC
distributed generations”, IEEE Access, Vol. 9, for power loss reduction in distribution system”,
pp. 67186-67200, 2021. Engineering Science and Technology, An
[11] V. Janamala, K. R. Rani, P. S. Rani, A. N. International Journal, Vol. 19, No. 3, pp. 1266-
Venkateswarlu, and S. R. Inkollu, “Optimal 1282, 2016.
Switching Operations of Soft Open Points in [20] G. Moustafa, M. Elshahed, A. R. Ginidi, A. M.
Active Distribution Network for Handling Shaheen, and H. S. Mansour, “A Gradient-
Variable Penetration of Photovoltaic and Based Optimizer with a Crossover Operator for
Electric Vehicles Using Artificial Rabbits Distribution Static VAR Compensator (D-SVC)
Optimization”, Process Integr. Optim. Sustain, Sizing and Placement in Electrical Systems”,
pp. 1-9, 2022. Mathematics, Vol. 11, No. 5, p. 1077, 2023.
[12] G. Shahgholian, “A Brief Overview of [21] A. A. Téllez, G. López, I. Isaac, and J. W.
Microgrid Performance Improvements Using González, “Optimal reactive power
Distributed FACTS Devices”, Journal of compensation in electrical distribution systems
International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30
Received: March 6, 2023. Revised: April 3, 2023. 387

with distributed resources. Review”, Heliyon, [32] X. S. Yang, and S. Deb, “Engineering
Vol. 4, No. 8, p. e00746, 2018. optimisation by cuckoo search”, International
[22] R. P. Parouha and P. Verma, “State-of-the-art Journal of Mathematical Modelling and
reviews of meta-heuristic algorithms with their Numerical Optimisation, Vol. 1, No. 4, pp. 330-
novel proposal for unconstrained optimization 343, 2010.
and applications”, Archives of Computational [33] X. S. Yang, “Flower pollination algorithm for
Methods in Engineering, pp. 4049-4115, 2021. global optimization”, In: Proc. of
[23] S. P. Adam, S. A. Alexandropoulos, P. M. Unconventional Computation and Natural
Pardalos, and M. N. Vrahatis, “No free lunch Computation: 11th International Conference,
theorem: A review”, Approximation and UCNC 2012, Orléan, France, 2012.
Optimization: Algorithms, Complexity and [34] R. V. Rao, V. J. Savsani, and D. P. Vakharia,
Applications, pp. 57-82, 2019. “Teaching–learning-based optimization: a
[24] X. S. Yang, “Firefly algorithm, stochastic test novel method for constrained mechanical
functions and design optimisation”, design optimization problems”, Computer-
International Journal of Bio-Inspired Aided Design, Vol. 43, No. 3, pp. 303-315,
Computation, Vol. 2, No. 2, pp. 78-84, 2010. 2011.
[25] M. Ghasemi, S. K. Mohammadi, M. Zare, S.
Mirjalili, M. Gil, and R. Hemmati, “A new
firefly algorithm with improved global
exploration and convergence with application
to engineering optimization”, Decision
Analytics Journal, Vol. 5, p. 100125, 2022.
[26] V. Kumar and D. Kumar, “A systematic review
on firefly algorithm: past, present, and future”,
Archives of Computational Methods in
Engineering, Vol. 28, pp. 3269-3291, 2021.
[27] Y. Wang and S. Song, “An adaptive firefly
algorithm for multilevel image thresholding
based on minimum cross-entropy”, The Journal
of Supercomputing, Vol. 78, No. 9, pp. 11580-
11600, 2022.
[28] R. D. Zimmerman, C. E. M. 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, 2010.
[29] V. Janamala and T. K. Pandraju, “Static voltage
stability of reconfigurable radial distribution
system considering voltage dependent load
models”, Mathematical Modelling of
Engineering Problems, Vol. 7, No. 3, pp. 450-
458, 2020.
[30] M. A. Kashem, V. Ganapathy, and G. B.
Jasmon, “A geometrical approach for network
reconfiguration based loss minimization in
distribution systems”, International Journal of
Electrical Power & Energy Systems, Vol. 23,
No. 4, pp. 295-304, 2001.
[31] J. P. Ham, J. H. Kim, B. H. Lee, and J. R. Won,
“Calculation of Active Power Transfer
Capability using Repeated Power Flow
Program”, KIEE International Transactions on
Power Engineering. Vol. 12, No. 1, pp. 15-19,
2002.
International Journal of Intelligent Engineering and Systems, Vol.16, No.3, 2023 DOI: 10.22266/ijies2023.0630.30