Genetic Algorithm based Method for Capacitor

Placement using New Sensitivity based Approach

Neeraj Kanwar, Student Member, IEEE, Pawan Saini, Nikhil Gupta, Member, IEEE, Anil Swarnkar, Member, IEEE
and K. R. Niazi, Senior Member, IEEE
Department of Electrical Engineering,
Malaviya National Institute of Technology
Jaipur, India
nk12.mnit@gmail.com, pawan25saini@gmail.com, nikhil2007_mnit@yahoo.com, mnit.anil@gmail.com,
krn152001@yahoo.co.in

Abstract—This paper proposes a new methodology for QD Nominal reactive power demand (kVAr)
optimal allocation of shunt capacitors in distribution systems. Qci Reactive power injection at bus i (kVAr)
The proposed method combines various objectives and Qci,n Maximum reactive power injection at ith node
constraints into a comprehensive constraint multi-objective
of nth area (kVAr)
function (CCMF). The function has been optimized using a
Genetic Algorithm (GA) based method. Attempts have also been
Qct,n Total reactive power injection at nth area (kVAr)
made to improve the computational efficiency of GA. The search Q0 Size of capacitor bank (kVAr)
space of GA is reduced by introducing a new reactive power flow ΔQ Taping size (kVAr)
sensitivity approach that determines the set of candidate nodes ΔSCR Annual capacity release (kVA)
suitable for capacitor placement and also by employing a new Tj Time duration of jth load level (h)
constrained chromosome structure. The proposed method is VminS Minimum permissible node voltage (p.u.)
tested on 69-bus test distribution system and the application VmaxS Maximum permissible node voltage (p.u.)
results are promising.
Vji Voltage at ith node for the jth load level (p.u.)
Keywords—capacitor placement; distribution systems; Vpf Node voltage deviation penalty function
sensitivity analysis; genetic algorithm; reactive compensation ΔVjimax Maximum node voltage deviation at ith node for
the jth load level (p.u.)
I. NOMENCLATURE II. INTRODUCTION
ΔEj Annual power loss reduction at jth load level Shunt capacitors placement is a classical problem usually
(kW) employed in distribution systems to improve its voltage
F(x) Comprehensive constraint multi-objective profile. The present trend towards competitive business
function ($) environment enforces electric utilities to enhance their annual
gn(x) Total reactive power demand for nth area (kVAr) profits while ensuring reliable and better quality of electric
ImaxS Maximum permissible feeder current (p.u.) supply. Benefits such as reduction in annual energy loss,
Iji Current in ith branch for the jth load level (p.u.) improvement in node voltage profiles, system capacity release,
Ipf Feeder current deviation penalty function etc. depend greatly on how optimally shunt capacitors are
ΔIjimax Maximum feeder current deviation at ith branch installed. In the context of modern distribution the aim of the
for the jth load level (p.u.) capacitor placement problem is to maximize the profit on the
investment of capacitor placement subject to operational &
Ke Energy cost of losses ($/kWh)
power quality constraints.
Kp Peak energy cost of losses ($/kW)
Ks Cost of system capacity release ($/kVA) Initial efforts for capacitor placement used analytical
Kci Annual cost per unit of the reactive power methods under some unrealistic or impractical assumptions [1-
injection at bus i ($/kVAr/year) 4]. Other method employed complex mathematical
programming techniques [5-8]. Later on, to overcome the
k1,k2,k3 Weighting factors
complexity of the programming methods some heuristic
NL Number of load level approaches have been developed [9-13]. In recent years
Nn,a ath node of nth area population based meta-heuristic techniques such as Genetic
Nn,b bth node of nth area Algorithm (GA), Particle Swarm Optimization (PSO),
nc Total number of shunt capacitors to be installed Teaching Learning Based Optimization (TLBO), Plant Growth
PF Penalty Function Simulation Algorithm (PGSA), Artificial Bee Colony (ABC)
ΔP Annual peak power loss reduction (kW) approach and Micro-Genetic Algorithms (MGA) [14-21] have
been applied to solve capacitor placement problems.
978-1-4799-5141-3/14/$31.00 ©2014 IEEE
 GA is one of the widely accepted population-based meta- where,
heuristic techniques and is based on the philosophy of |VminS/maxS – Vji | ; Vmin < Vji < VminS
“Survival of the fittest” [22]. However, the disadvantage of ΔVji = 0 ; VmaxS ≥ Vji ≥VminS (4)
GA is poor convergence and high processing time associated. a very large number ; else
That is due to its evolutionary concept, based on random
processes that make the algorithm quite slow [21], and
therefore needs remedial measures. • Feeder current deviation penalty function (Ipf)

In this work, the search space of GA is reduced by The feeder current deviation penalty function is formulated
proposing new reactive power flow sensitivity based method as the maximum of feeder current deviations of each branch of
and also by introducing a constrained chromosome structure. the system from the specified maximum feeder current ImaxS,
The proposed sensitivity method determines the set of while considering all load levels, i.e.
candidate nodes for reactive compensation. In addition, the
constrained chromosome structure provides pace to genetic Ipf = 1/(1+k3 (Max (ΔIijmax))) (5)
evolutions. The capacitor placement has numerous benefits but where,
involves capital investments, therefore a comprehensive |Iji – ImaxS| ; Imax < Iji ≤ ImaxS
constraint multi-objective function has been proposed which ΔIji = 0 ; Iji ≤ Imax (6)
considers loss reduction, voltage profile improvement, feeder
a very large number ; else
overloading, kVA enhancement etc., in such a way that the
annual savings is maximized. The proposed method is applied
on a standard 69-bus test distribution system and the results B. Operating Constraints
are promising when compared with other existing methods.
• Power flow balance constraint
III. PROBLEM FORMULATION The power flow constraints are expressed in a vector form
as follows
In this paper, the optimal capacitor problem is formulated
to maximize annual savings in such a fashion that results in
H(x,u)=0 (7)
minimization of annual energy losses, peak power losses,
enhancement of substation capacity release and improvement
of node voltage and feeder current profiles by proposing where, x and u are vectors of state (dependent) variables and
following CCMF control (independent) variables, respectively.
• Capacitor capacity and control setting constraint
⎛ NL ⎞ nc
Max F ( x) = k1 ⎜ Ke ∑Tj ΔE j + K p ΔP + KS ΔSCR ⎟ × PF − ∑ KciQci (1) Capacitor banks and their control settings are
⎝ j =1 ⎠ i =1 commercially available in discrete sizes. Therefore, the shunt
capacitors to be dealt are integral multiples of the available
Where, the first term within the bracket represents annual capacitor bank/tapping size, i.e.
cost of energy loss reduction, second terms represents, the
annual cost of peak power loss reduction and the third term
represents the annual cost of the substation capacity release. Qci ≤ LQo; L= 0, 1, 2, …, nc (8)
The negative term in the expression represents annual charges
on capacitor placement. The term penalty function (PF) is Qci = k∆Q; k = 0, 1, 2, …, Qci /∆Q (9)
explained as under.
It is intuitively assumed that the total reactive power
A. Penalty Function injection is not to exceed the nominal reactive power demand
of the distribution system.
The penalty function PF in (1) is incorporated to check
node voltage and feeder current constraints and is defined by
∑Qci = QD (10)
the geometric mean of node voltage deviation and feeder
current deviation penalty functions as below
The CCMF defined by (1) is optimized using a proposed
GA based method. The above problem is solved in two steps.
PF = √(Vpf × Ipf) (2)
In the first step optimal installed capacity of shunt capacitors
is determined on the basis of annual loading. In second step
• Node voltage deviation penalty function (Vpf) optimal control settings of installed capacitor, for each load
The node voltage deviation penalty function is formulated level, is determined.
as the maximum of node voltage deviations of each system
nodes from the specified minimum node voltage VminS, at all IV. PROPOSED GENETIC ALGORITHM
load levels, i.e.
The Genetic Algorithm is a derivative free population
based stochastic optimization method inspired by the concept
Vpf = 1/(1+k2 (Max (ΔVijmax))) (3) of natural selection and evolutionary process. The
development of GA is largely credited to the work of Holland
[23] and Goldberg [24]. Since then GA have evolved and
become a promising tool to solve complex combinatorial
optimization problems. However, the optimal capacitor
placement problem offers enormous search space to GA and
that makes it computationally demanding. Therefore, suitable
measures are required to make GA computationally efficient.
In the present work this goal is achieved by squeezing search
space by proposing a reactive power flow sensitivity approach
and also by employing constrained bound chromosome
structure.
A. Proposed Reactive Power Flow Sensitivity Approach
Shunt capacitors should be placed at very few selected
locations in distribution networks. Many researchers [14, 17,
20, 25] used perturbations to determine these optimal locations
by placing a small unit capacity and measuring the change in
power losses. These perturbations provides ranking of nodes
and high ranking nodes are selected for capacitor placement.
However, the sensitivity based schemes commonly used to
select attractive buses in capacitor placement, can lead to poor
quality solutions [26]. The capacitors primarily regulate
reactive power flow in distribution feeders. Therefore, a new
reactive power flow sensitivity based approach is proposed to
identify prominent areas for capacitor placement. The nodes in
the identified prominent areas are considered possible
candidate for capacitor placement.
B. Proposed Constrained Chromosome Structure
When GA is initialized, the initial population may appear
with very poor average fitness and thus may results in poor
convergence. Therefore, the chromosome structure should be
bounded by certain intuitive constraints. The individuals
contain the information about the siting and sizing of the
candidate shunt capacitors.
repeated. The term gn(x) is taken proportional to total reactive
power loading of respective area. Initially, such p numbers of
genetic chromosome are created to constitute the initial
population. These constraints are also used while repairing
infeasible individuals.
The conventional “Roulette-wheel” selection, single point
crossover and mutation are used in the proposed real-coded
GA. The crossover rate and mutation rate are kept at 0.9 and
0.03, respectively. These rates are obtained after well trade off
through numerous simulations. The elite child is preserved,
and is replaced, if a better child evolves. When all individuals
acquire the same fitness or the maximum generations
exhausted, the genetic evolutions terminates.
In the proposed method, the distribution network is
reconfigured at each load level after placing optimal capacities
to extract its benefits.
V. SIMULATION RESULTS AND DISCUSSIONS
The proposed method is tested on one of the most popular
69-bus radial distribution system [14] having 68 sectionalizing
lines, 5 tie-lines and 8 feeders whose single line diagram for
the base case is shown in Fig. 2. The real and reactive power
demand at nominal load level is 3801.5 kW and 2694.6 kVAr,
respectively. The feeder current constraints are set at 300 A,
except for the upstream branches 1-9, whose limits are set at
600A.

N1,1 ... Nn,1 Qc,1 … Qc,1 N1,2 … Nn,2 Qc.2 … Qc,2

Candidate nodes Capacitor sizing Candidate nodes Capacitor sizing
for Area 1 for Area 1 for Area 2 for Area 2

Fig. 1. Proposed constrained chromosome structure

In this work, the site is taken randomly from the list of
candidate nodes prepared through sensitivity analysis
approach and the size represents randomly generated number
of candidate capacitor banks between its minimum and
maximum bounds. In the proposed constrained chromosome
structure, the chromosome has been divided area-wise as
shown in Fig. 1. The genetic information of the chromosome
is selected in accordance to the constraints defined by (10)-
(13).
Qci,n ≤ gn(x)
Qct,n ≤ gn(x)
Nn,a ≠ Nn,b
These constraints are pertaining to maximum reactive
power injection at a candidate node and area of the distribution
network. The constraint (13) ensures that no candidate node is
Fig. 2. Single line diagram of 69-bus radial distribution system
The sensitivity test is applied on this system and the result
obtained is presented Fig. 3. The figure shows a 3-dimensional
view of the change in reactive power flow in the network
when small capacity perturbations employed. It can be
observed from the figure that two areas of the network are
prominent, i.e., Area-1 containing nodes 16-27 and Area-2
with nodes 53-65. On the basis of this information, gn (x) has
been calculated proportionately, as shown in Table I.
Following three options for capacitor placement have been
(11)
investigated:
(12) Case 1: three optimal locations
Case 2: four optimal locations
(13) Case 3: five optimal locations
The rated substation voltage is considered as 1.0 p.u. The
system constraints and GA parameters considered for
simulations are shown in Table II. Three different load levels
and corresponding load durations considered are shown in
Table III along with the power loss and minimum node
voltage for the bare system. The cost of annual energy loss for
the bare system is US $135905.
compared with existing established methods in Table VI. The
table compares power losses and minimum node voltage at all
load levels. It can be observed from the table that the optimal
solution obtained using proposed GA provides least power
losses and better minimum node voltages at all load levels
than other established methods. Therefore, proposed method is
capable to generate promising results.

TABLE IV. OPTIMAL CAPACITOR PLACEMENT USING PROPOSED GA

Case Node Control Settings (kVAr)
Light Nominal Peak
1 18 200 300 300
61 500 1100 1200
64 100 200 600
Total 800 1600 2100
2 18 200 300 300
59 100 100 300
61 400 1000 1200
64 100 200 600
Total 800 1600 2400
3 18 200 300 300
Fig. 3. A 3-dimensional view for change in reactive power flow 59 100 200 300
61 400 900 1200
TABLE I. SYSTEM CONSTRAINTS AND GA PARAMETERS
64 100 200 300
Parameter Value Parameter Value
VminS (p.u.) 0.09 Vmin (p.u.) 0.90 65 - - 300
VmaxS (p.u.) 1.05 ImaxS (p.u.) 1.1 Total 800 1600 2400
ImaxS 1.10 Imax (p.u.) 1.0
Ke (US $/kWh) 0.06 Q0 (kVAr) 300 TABLE V. NETWORK PERFORMANCE AFTER OPTIMAL CAPACITOR
Kci (US $/kVAr) 3.0 ΔQ (kVAr) 100 PLACEMENT
Kp (US $/kW) 42.6 Population size 100
Ks (US $/kVA) 19.8 Maximum generations 50 Load Proposed GA
k1, k2, k3 each 1.0 Crossover rate 0.90 level Particulars Case 1 Case 2 Case 3
VmaxS (p.u.) 1.1 Mutation rate 0.03 Power Loss (kW) 34.32 34.33 34.33
Light
VminS (p.u.) 0.95 Vmin (p.u.) 0.9666 0.9664 0.9664
Power Loss (kW) 146.17 145.99 145.95
Nominal
TABLE II. RESULTS OF PROPOSED SENSITIVITY METHOD Vmin (p.u.) 0.9320 0.9317 0.9315
Power Loss (kW) 417.21 411.72 412.25
Peak
Particulars Area 1 Area 2 Vmin (p.u.) 0.8814 0.8854 0.8863
Bus Locations 16- 27 53- 65
Area reactive power demand (kVAr) 225.1 1226.7 TABLE VI. COMPARISON RESULT FOR PROPOSED GA
System reactive power demand (kVAr) 2694.6
gn(x) (kVAr) 414 2280 Method Load Power loss Minimum Voltage
levels (kW) (p.u.)
DSA [13] Light 35.52 0.9683
TABLE III. LOAD LEVELS, LOAD DURATIONS AND INITIAL CONDITION OF 69- Nominal 147.0 0.9318
BUS SYSTEM Peak 427.3 0.8936
Load Load Vmin Power Loss Energy Loss GA [14] Light 40.48 0.9622
Level Duration(h) (p.u.) (kW) Cost (US$) Nominal 156.52 0.9369
Light 2000 0.9567 51.60 6192 Peak 460.45 0.9001
(0.5) Dedicated GA Light 34.95 0.9668
Nominal 5260 0.9092 224.96 70997 [16] Nominal 146.88 0.9302
(1.0) Peak - -
Peak 1500 0.8445 652.40 58716 TLBO [18] Light 34.43 0.9662
(1.6) Nominal 146.80 0.9321
Peak 417.28 0.8795
Annual Cost of Energy Loss 135905
Proposed GA Light 16.29 0.9810
The application results of proposed GA are summarized in Nominal 66.57 0.9680
Peak 179.11 0.9413
Tables IV and V. Table IV depicts the optimal nodes, total
installed capacities and respective optimal control setting of Fig. 4 and 5 compare the performance of optimal capacitor
shunt capacitors at all load levels. The table shows that the placement and network reconfiguration with initial conditions
total installed capacity is 2100 kVAR for Case 1 while it is of the network under all load scenerios. It is clearly seen that
2400 kVAr for the remaining two cases. From Table V it may for all three cases, the power losses reduced and minimum
be observed that all three cases of capacitor placement provide node voltage improves remarkebly when compared with initial
more or less similar results. The application results are
conditions of the network.
Fig. 4 Comparison of power loss before and after capacitor placement and
reconfiguration
Fig. 5 Comparison of minimum node voltage before and after capacitor
placement and reconfiguration
Fig. 6. Node voltage profile for base case, after capacitor placement and after
capacitor placement & reconfiguration at peak load level
Fig. 6 and 7 shows the impact of cpapcitor placement
alone and capacitor placement & reconfiguration on system
node voltage and feeder current profiles under peak load
conditions. It can be concluded from these figures that the
performance of distribution networks can be partially
Fig. 7. Feeder current profile for base case, after capacitor placement and
after capacitor placement & reconfiguration at peak load level
TABLE VII. COMPARISON RESULT OF COST BENEFIT APPRAISAL
Particulars (US $) Case 1 Case 2 Case 3
Annual Investment on shunt capacitors 6300 7200 7200
Annual cost of energy loss reduction 48106 48656 48621
Annual cost of peak power loss reduction 10019 10253 10230
Annual cost of capacity release 24077 26350 26350
Annual Savings 75902 78059 78001
Benefit/Cost Ratio 12.05 10.84 10.83
improved by optimal capacitor placement and network
reconfiguration, even during peak load hours. Finally, the
impact of optimal capacitor placement on the benefits and cost
involved for all three cases is presented in Table VII. The table
reveals that all three cases generates comparable results using
the proposed method and the selection adhere upon choice of
the operator.
V. CONCLUSIONS
This paper presents a GA based method for optimal
capacitor allocation in radial distribution systems. The search
space of GA is reduced by introducing a new reactive power
flow sensitivity-based approach and also by suggesting a new
constrained bound chromosome structure. The proposed
sensitivity based approach identifies those areas of distribution
network which are suitable for capacitor placement on the
basis of reactive power flow in the distribution network and
thereby encompass required candidate nodes. The proposed
chromosome structure divides itself according to candidate
areas and furnishes genetic information accordingly subjected
to operational and network constraints. Further, a
comprehensive constraint multi-objective function (CCMF) is
formulated to combine various objectives related to energy
and power loss savings, substation capacity release and
investments on capacitors along with system operational
constraints pertaining to allowable node voltage deviations
and thermal limits of feeders for solving the optimal capacitor
allocation problem. The proposed CCMF maximizes net
annual savings by reducing annual energy and peak power
losses and enhancing substation capacity release while
simultaneously improving system node voltage and feeder
current profiles with optimal investments on shunt
compensation. The proposed method first determines the
optimal siting and sizing of installed capacities of shunt
capacitors and then the optimal control settings are determined
to avoid over-voltages and also to improve the operating
efficiency of distribution system under variable load scenarios.
The proposed method is tested on 69-bus test distribution
system and the application results are promising when
compared with other established techniques available in
literature. The effectiveness of the proposed strategy is
ensured by the lower amount of annual investments on
capacitors, higher savings in annual energy losses, lower peak
power losses, better system node voltage and feeder current
profiles and also results in a better benefit to cost ratio for the
reactive compensation project.
REFERENCES
[1] N. M. Neagle, and D. R. Samson, “Loss reduction from capacitors
installed on primary feeders,” AIEE Trans., vol. 75, pp. 950–959,
October 1956.
[2] R. F. Cook, “Optimizing the application of shunt capacitors for reactive
voltampere control and loss reduction,” AIEE Trans., vol. 80, pp. 430–
444, August 1961.
[3] N. E. Chang, “Generalized equations on loss reduction with shunt
capacitors,” IEEE Trans. Power App. Syst., vol. PAS-91 (5), pp. 2189
2195, September 1972.
[4] J. J. Grainger, and S. H. Lee, “Optimum size and location of shunt
capacitors for reduction of losses on distribution feeders,” IEEE Trans.
Power App. Syst., vol. PAS-100, pp. 1105–1118, March 1981.
[5] M. E. Baran, and F. F. Wu, “Optimal capacitor placement on radial
distribution systems,” IEEE Trans. Power Del., vol. 4 (1), pp. 725–734,
January 1989.
[6] S. Ertem, “Optimal shunt capacitor sizing for reduced line loading,
voltage improvement and loss reduction of distribution feeders,” Proc.
21st Annu. North Amer. Power Symp., pp. 262–269, October 1989.
[7] R. Baldick, and F. F. Wu, “Efficient integer optimization algorithms for
optimal coordination of capacitors and regulators,” IEEE Trans. Power
Syst., vol. 5 (3), pp. 805–812, August 1990.
[8] A. Augugliaro, L. Dusonchet, and S. Mangione, “Optimal capacitive
compensation on radial distribution systems using nonlinear
programming,” Elect. Power Syst. Res., vol 19 (2), pp. 129–135, July
1990.
[9] M. Kaplan, “Optimization of number, location, size, control type and
control setting of shunt capacitors on radial distribution feeders,” IEEE
Trans. Power App. Syst., vol. PAS-103 (9), pp. 2659–2665, September
1984.
[10] Y. Baghzouz, and S. Ertem, “Shunt capacitor sizing for radial
distribution feeders with distorted substation voltages,” IEEE Trans.
Power Del., vol. 5 (2), pp. 650–657, April 1990.
[11] T. S. Abdel-Salam, A. Y. Chikhani, and R. Hackam, “A new technique
for loss reduction using compensating capacitors applied to distribution
systems with varying load condition,” IEEE Trans. Power Del., vol. 9
(2), pp. 819–827, April 1994.
[12] M. Chis, M. M. A. Salama, and S. Jayaram, “Capacitor placement in
distribution systems using heuristic search strategies,” Proc. Inst. Elect.
Eng., vol. 144 (2), pp. 225–230, May 1997.
[13] M. Ramalinga Raju, K.V.S. Ramachandra Murthy and K. Ravindra,
“Direct search algorithm for capacitive compensation in radial
distribution systems,” Electrical Power and Energy Systems, vol. 42, pp.
24–30, November 2012.
[14] D. Das, “Optimal placement of capacitors in radial distribution system
using a Fuzzy-GA method,” Electrical Power & Energy Systems, vol. 30
(6-7), pp. 361–367, July-September 2008.
[15] G. Levitin, A. Kalyuzhny, A. Shenkman and M. Chertkov “Optimal
Capacitor Allocation in Distribution Systems Using a Genetic Algorithm
and a Fast Energy Loss Computation Technique,” IEEE Trans Power
Syst., vol. 15 ( 2), pp. 623–628, April 2000.
[16] M.A.N. Guimaraes, C.A. Castro and R. Romero, “Distribution systems
operation optimization through reconfiguration and capacitor allocation
by a dedicated genetic algorithm,” IET Gener. Transm. Distrib. , vol. 4,
(11), pp. 1213–1222, April 2010.
[17] S.P. Singh and A.R. Rao, “Optimal allocation of capacitors in
distribution systems using particle swarm optimization, “Electr. Power
and Energy Syst., vol. 43 (1), pp. 1267–1275, December 2012.
[18] Sneha Sultana and Provas Kumar Roy, “Optimal capacitor placement in
radial distribution systems using teaching learning based optimization,”
Electr. Power and Energy Syst., vol. 54, pp. 387–398, January 2014.
[19] R. Srinivasas Rao, S.V.L. Narasimham and M. Ramalingaraju,
“Optimal capacitor placement in a radial distribution system using Plant
Growth Simulation Algorithm,” Electr. Power and Energy Syst., vol. 33,
pp. 1133–1139, June 2011.
[20] Attia A. El-Fergany and Almoataz Y. Abdelaziz, “Capacitor placement
for net saving maximization and system stability enhancement in
distribution networks using artificial bee colony-based approach,” Electr.
Power and Energy Syst., vol. 54, pp. 235–243, January 2014.
[21] Benemar Alencar de Souza, Helton do Nascimento Alves and Helvio
Alves Ferreira, “Microgenetic algorithms and fuzzy logic applied to the
optimal placement of capacitor banks in distribution networks,” IEEE
Trans. Power Syst., vol. 19 (2), May 2004.
[22] Lingfeng Wang and Chanan Singh, “Stochastic economic emission load
dispatch through a modified particle swarm optimization algorithm,”
Electr. Power Syst. Res., vol. 78, pp. 1466–1476, August 2008.
[23] J. H. Holland, Adaptation in natural and artificial systems, University of
Michigan Press, Michigan, 1975.
[24] D. E. Goldberg, Genetic algorithms in search, optimization and machine
learning, Addison-Wesley; 1989.
[25] Damanjeet Kaur and Jaydev Sharma, “Multiperiod shunt capacitor
allocation in radial distribution systems,” Electr. Power and Energy
Syst., vol. 52, pp. 247–253, November 2013.
[26] Ramon A. Gallego, Alcir José Monticelli and Rubén Romero, “Optimal
Capacitor Placement in Radial Distribution Networks,” IEEE Trans.
Power Syst., vol. 16 (4), November 2001.