2021 IEEE International Conference in Power Engineering Application (ICPEA), 8-9 March 2021

VSI improvement using SVC with aid of a modified

lightning search algorithm
SN Syed Nasir R Ayop JJ Jamian
School of Electrical Engineering School of Electrical Engineering School of Electrical Engineering
Universiti Teknologi Malaysia Universiti Teknologi Malaysia Universiti Teknologi Malaysia
[email protected] [email protected] [email protected]

Abstract—This paper proposes an improved metaheuristic Algorithm (CSA) [13], Bat Algorithm (BA) [14] and
optimization technique to solve the problem of optimal location Dragonfly Algorithm (DA) [15].
and sizing of static var compensator (SVC) in a complex power
system. The main objective of this research is to achieve Researcher at [9] used the GA in their research to ascertain
minimum Voltage Stability Index (VSI) after installing the three the optimal placement and size of an SVC device. The
units of SVC optimally by using Modified Lightning Search authors utilised multi-objective functions, including power
Algorithm (MLSA) technique. MLSA technique is used to find system losses, voltage deviation and the cost of finding the
the best location and sizing for the multiple SVC units. MLSA optimal placement. The results show that the use of a
2021 IEEE International Conference in Power Engineering Application (ICPEA) | 978-1-7281-8546-0/21/$31.00 ©2021 IEEE | DOI: 10.1109/ICPEA51500.2021.9417857

technique basically enhances Lightning Search Algorithm metaheuristic can help provide the adequate placement and
(LSA) which was inspired by the lightning strike phenomenon.
In mathematical analysis, it is proved that MLSA give superior
size for an SVC. Using the GA to improve voltage stability is
solution in 10 nonlinear mathematical benchmark functions reinforced with research [16]. The SVC was able to give
compared to another algorithm including LSA. Next, the better cost-effective than other FACTS devices. Research
effectiveness of MLSA technique is demonstrated on IEEE 30- also shows that the minimum active power losses and
bus test system together with the three SVC units. From the operating costs can be solved by using GSA and PSO in
analysis, MLSA gives better result especially in average, median placing SVC and thyristor-controlled series capacitor [6-7].
and minimum VSI. Thus, it can be concluded that MLSA is a Furthermore, from [8], shows that the injection of suitable
superior method in solving numerical optimization problem reactive power at optimal bus able to stabilize voltage profile
especially for optimal location and sizing of SVC. The of the network. Other research conducted by Nguyen et al.
optimization technique using MLSA proposed in this paper is
useful for power system designers who require consistent and
proposed CSA as an optimisation technique for finding the
accurate location and sizing of SVC in real power system best location and size for the SVC [13]. CSA proved to be
application. better at finding these answers than the Teaching Learning-
Based Optimisation (TLBO). Shiba et al. also used BA to
Keywords—Modified Lightning Search Algorithm, Voltage intensify power stability by corporate proportional integral
Stability Index, Optimization Technique, SVC derivative and SVC in a single machine infinite bus [14]. On
a small, extensive network, the effectiveness of the proposed
I. INTRODUCTION
controllers for voltage and angular speed has been proven.
Electric Power Research Institute have introduced the Song J et.al had to use DA to optimise the cost and size of the
concept of Flexible AC Transmission System (FACTS) SVC with overload factor and voltage deviation [15]. Based
device around 1980s. After that, FACTS devices have on these studies, the techniques used to determine location
become more commonplace in power systems due to its and size of SVC has been introduced repeated times using the
property of having higher controllability in the existing power latest technique aiming to provide better results.
system. Among FACTS devices, the SVC provides more
precise control and is able to be adapted instantaneously to The structure of the paper is as follows: Section II will
new conditions. Other than that, SVC is widespread due to its explain in detail about the optimisation techniques discussed
advantageous properties such as low cost and ease in control. in this paper which are the LSA and the MLSA. Next,
Generally, voltage stability in power system is achieved by describe the SVC modelling in the power system, the
installing SVC in several locations which normally is utilised objective function formulation, the constraints and IEEE 30
when the system needs more reactive power to stabilize the bus system data. Section III focus on MLSA validation uses
system [1-2]. The main problem to install SVC in complex 10 benchmark functions while Sections IV deals with
power system is to determine the best location and size based validation on MLSA for VSI problem in the power system.
on the system requirement such as VSI improvement and Lastly, the conclusion part presented in Section V.
cost. Based on the literature, there are many methods and II. PROBLEM MODELLING AND FORMULATION
approaches that have been used to determine the optimal
location and size of SVC in the power system such as PSO A. Lightning Search Algorithm (LSA)
technique [3-7], Artificial Bee Colony [8] Genetic Algorithm A lightning strike is a natural phenomenon that commonly
(GA) [9-10], Gravitational Search Algorithm [GSA] [6-7], happens around the world, especially during thunderstorms,
Harmony Search Algorithm (HSA) [11-12], Cuckoo Search which is produced from electrostatic discharge effect. The
LSA, which was introduced by Shareef, is a new

978-1-7281-8546-0/21/$31.00
Authorized ©2021State
licensed use limited to: San Francisco 56on June 18,2021 at 08:51:16 UTC from IEEE Xplore. Restrictions apply.
IEEEUniv. Downloaded
metaheuristic optimization method based on the movement MLSA updating process. There are 4 improvements made to
of lightning represented as a step leader propagation from LSA in order to produce MLSA. First, the original LSA uses
cloud to ground [16]. The LSA process starts with the the channel time for updating the equation. However, in
generation of the random population before the fitness MLSA, this feature is omitted since it is not suitable when
calculation. The worst step leader will be eliminated every dealing with the equations that have many local minima.
10th iteration before the direction of the step leader is Second, the single direction update used in LSA has been
updated. Next, step leader movement is updated based on modified to become bidirectional. Therefore, in the new
direction, shape parameter and scale parameter as per (1) - updating approach the direction can be either in forward or in
(3). Lastly, the forking phenomena takes place as per (4) for backward direction based on the fitness values. The new
1% of the better step leaders. In general, the process to apply updating approach is stated in Table I. Third, the scale
LSA is as shown in Table I. The basic formula for the parameter (σ ) used in LSA is based on exponential
Li

movement of the step leaders in LSA are: distribution, which causes the step leader’s movement to be
active only at 20% of the early iterations. In order to get more
µ i = xi − xbest (1) possible solutions, Laplacian distribution equation
| |
P s i _ new = P s i ± exp rand(µi ) (2) 1⁄3 is used, so that the movement of the step
leader can be enhanced from 30% up to 70% of the iterations.
P L i _ new = P L i ± normrand ( µ i , σ i ) (3) Equation (5) replaces the former (3) in order to achieve this.
Pi = a + b − Pi (4) Lastly, the learning factors are introduced in MLSA, which
µi - Shape parameter at i th iteration makes the movement of the step leader more active in the
solution searching process. The equation that involves the
σi - Scale parameter at i th iteration (2.05 − 2e −5t ) learning factor is as per (6). Learning factor ( R ) can be vary
from 1.0 to 2.5. The algorithm comparison detail between
xi - Position at i th iteration
LSA and MLSA is shown in Table I.
xbest - Best Individual during minimum VSI
s
New position in space projectile at i th iteration
P L i _ new = P L i ± normrand ( µ i , σ Li ) (5)
P i _ new -
µ i = R ( xi − xbest ) (6)
P s i - Position in space projectile at i th iteration
P L i _ new - New position in lead projectile at i th iteration C. Modelling of SVC
SVC is designed with reactors and capacitors controlled by
P L i - Position in lead projectile at i th iteration thyristor valves. With the reactor and capacitor elements, the
Pi - Position during Forking at i th iteration SVC is able to manipulate reactive or capacitive power, in
order to achieve a specific aim [17]. For example, the SVC
Pi - Position at i th iteration
can be used to provide an improved power systems security,
a - Upper boundary so that more power transfer can be accommodated without
b - Lower boundary expanding the transmission line [18]. On the other hand, the
SVC can also be used to optimise reactive power control,
B. Modified Lightning Search Algorithm (MLSA)
which minimises loss and improves voltage profile [19].
MLSA, which is developed based on the improvements However, the value of reactive power can be varied by
made to the existing LSA, is proposed in this study. The main adjusting the firing angle of the thyristor elements [20]. Thus,
objective of the modification is to get a faster algorithm with to get the suitable reactive power based on network condition
same or better optimal solution by adjusting the updating
equation, so that the performance of MLSA can be improved.
Fig. 1 shows the illustration comparison between LSA and

Fig. 1. Illustration of LSA & MLSA concept in searching optimal solution

57on June 18,2021 at 08:51:16 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: San Francisco State Univ. Downloaded
 TABLE I. COMPARISON BETWEEN LSA & MLSA

Step LSA MLSA

1 Generate random Population for step leader Generate random Population for step leader
2 Calculate VSI and set Best VSI Calculate VSI and set Best VSI
3 Eliminate worst channel every 10 iteration -
4 Update Direction Update Direction
f ( x + value ) = VSI current f ( x + value ) = VSI 1
if VSI current < VSI best f ( x − value ) = VSI 2
Assign direction to positive if VSI 1 < VSI best & VSI 1 ≤ VSI 2
Else Assign direction to positive
Assign direction to negative if VSI 2 < VSI best & VSI 2 < VSI 1
Assign direction to negative
5 Update step leader (SL) and fork Update step leader (SL) and fork
Dist = SL t −1 − SL best Dist = R ( SL t −1 − SL best )
if Dist < 0 if Dist < 0
Calculate SL using equation (2) with positive sign Calculate SL using equation (2) with positive sign
if Dist > 0 if Dist > 0
Calculate SL using equation (2) with negative sign Calculate SL using equation (2) with negative sign
Else Else
Calculate SL using equation (3) with direction as per step 4 Calculate SL using equation (5) with direction as per
(shape parameter using exponential distribution) step 4 (shape parameter using Laplace distribution)

Compute VSI Compute VSI

if VSI < VSI best & rand < 0.01 if VSI < VSI best & rand < 0.01
Fork procedure Fork procedure
6 Repeat Step 2 Repeat Step 2

is the challenges task to the power system operator in

Busmin ≤ Busi ≤ Busmax (8)
ensuring the system performance at the best condition.
Q min ≤ Q i ≤ Q max (9)
D. Objective Function & Constraint
The optimisation approach in this study refers to the The standard IEEE 30 bus test network which consists of
identification of the optimal location and the sizes of three 30 buses and 41 lines is used as test system in this research.
SVC units simultaneously. The objective for SVC placement Based on the standard 30 bus system, there are two capacitor
is to minimise the VSI of the system network. VSI for a given banks installed at bus 10 and 24 with four step-down
network is defined as the summation of voltage magnitude transformers.
(Vi) deviation of each and every bus (n) from unity (1.0).
III. VALIDATION ON MLSA IN MATHEMATICAL
Based on the simulation for the default IEEE 30 bus system, EQUATION
VSI is recorded at 0.9066. Thus, objective function (VSI) that
is used in this study is defined as follows: - In this section, the ability of the proposed MLSA will be
n validated using 10 mathematical standard benchmark
VSI =  |1−U
i =1
i | (7) functions as per Table II. These benchmark functions are
commonly used to validate the ability of new proposed
In this research, there are six parameters that will be varied, method such as in [20-25]. In this validation test, the stopping
which are the individual location and sizing of three SVC criterion is set at 500 iterations with 50 populations. The
units. The constraints for these parameters are SVC bus result obtained by using MLSA will be compared with PSO,
Firefly Algorithm (FA) and LSA. Standard benchmark is a
location and SVC reactive limit. The number of bus in this
complex equation which has many local minima that can
test system is the limited for SVC placement. In this study
make unstable method easily trapped. Generally, separable
IEEE 30 bus system will be used. The values of Bus min and
functions are much easier to be solved rather than non-
Busmax are 1 and 30 respectively. The values of Qmin and separable functions due to independent variable. In this
validation test, stopping criteria is set at 500 iteration with 50
Qmax are -50MVAr and 50MVAr respectively. Thus, the
population. Table III represents the result of 50 independent
constraints are as follow: -

58on June 18,2021 at 08:51:16 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: San Francisco State Univ. Downloaded
 TABLE II. STANDARD BENCHMARK FUNCTION
No Name Expression n Search Space Function Minimum
n
 xi
2
1 Sphere f1 ( x ) = 30 [-100,100] n
0
i =1
n n
2 Schwefel 2.22 f2 (x) =  xi + ∏ xi 30 [-10,10]n 0
i =1 i =1
n
3 Dixon-Price f 3 ( x ) = ( x 1 − 1) 2 +  i ( 2 x i2 − x i −1 ) 2 30 [-10,10]n 0
i= 2

4 Schwefel 2.21 f 4 ( x ) = max i { xi ,1 ≤ i ≤ n} 30 [-100,100]n 0

n n
5 Shubert f 5 ( x ) = [  i cos( i + ( i + 1 ) x 1 ) ] × [  i cos( i + ( i + 1 ) x 2 ) ] 2 [-10,10] n
-186.731
i =1 i =1
n
6 Step f6 (x) =  ( x i + 0 . 5 ) 2 30 [-100,100]n 0
i =1
n
7 Quartic f 7 ( x ) =  ix i4 + rand ( 0 ,1 ) 30 [-1.28,1.28]n 0
i
6-Hump Camel 1
8 f 8 ( x ) = 4 x 12 − 2 . 1 x 14 + x 16 + x 1 x 2 − 4 x 22 + 4 x 24 2 [-5,5]n −1.0316285
Back 3
n
9 Schwefel 2.26 f9 ( x) =  x i sin( xi ) 30 [-500,500]n −12569.50
i =1
n
10 Rastrigin [
f 10 ( x ) =  x i2 − 10 cos( 2π x i ) + 10 ] 30 [-5.12,5.12]n 0
i =1

run at standard benchmark function considering the best, 300 iterations. However, the MLSA still gave the best or the
worst, average, and standard deviation. same optimal compared to other methods.
From the analysis, the MLSA gives better result for best, In conclusion, MLSA always provides superior results
worst, average and standard deviation for four benchmark compared to LSA, PSO and FA for all the tested benchmark
functions , , and as per Fig. 2 until Fig. 5. This functions used in this study. The consistency of MLSA in this
shows that the improvement made on MLSA gives positive validation has shown the capability of this algorithm in
impact in finding better optimal solution. Next, for functions solving complex optimization problem. Based on the result,
and , MLSA shows better convergence compared to FA MLSA has the ability to explore and exploit in finding better
and same convergence for PSO and LSA. Also, for function optimal fitness. Therefore, it can be concluded that the MLSA
, MLSA gives better results compared to PSO and FA
although it gives the same result with LSA. MLSA gives best
result for , and compared to other algorithms although
LSA provide better average and standard deviation.
However, the best or known as global solution is the main
criteria in optimization analysis.
In terms of convergence curve performance for , , ,
, , , and , LSA and MLSA were able to perform
for the first 200 iterations compared to FA and PSO. This
shows the capability of LSA and MLSA in ability to explore
in early iteration. Also, and also gave better
performance, although the optimal solution was found after
Fig. 3. Convergence Curve for

Fig. 4. Convergence Curve for

Fig. 2. Convergence Curve for

59on June 18,2021 at 08:51:16 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: San Francisco State Univ. Downloaded
 TABLE III. BEST, WORST, AVERAGE AND STANDARD DEVIATION TABLE V. THE BEST RESULT FOR LSA AND MLSA
VALUES OBTAINED BY PSO, FA, LSA AND MLSA THROUGH 50 INDEPENDENT LSA MLSA
RUNS Selected Value Selected Value
f Statistic PSO FA LSA MLSA Bus (MVAr) Bus (MVAr)
1 Best 1.76300E− 0.005027795 5.6167E-19 8.94978E-25 SVC 1 9 -49.6103 9 -49.9841
Worst 06 0.026468336 2.3902E-09 3.28407E-24 SVC 2 13 -34.2969 13 -28.2699
Average 0.00028190 0.011961069 8.9292E-11 1.85178E-24 SVC 3 26 3.6120 26 3.6298
SD 9 0.004295896 4.3026E-10 4.90912E-25 VSI 0.3725 0.3724
2.76240E−
05 in finding the best solution. Both algorithms give the same
4.32130E− SVC locations, which are at buses 9, 13 and 26. However,
05 the MLSA able to get a better size for SVC, 49.9841
2 Best 0.00145646 0.172154000 2.9506E-07 4.32010E-12
MVAr, -28.2699 MVar and 3.6298 MVAr respectively,
Worst 7 0.653905407 3.8760E-01 1.61028E-02
and consequently giving the lowest VSI value, 0.3724 as
Average 0.01890286 0.373326823 1.0916E-02 3.68708E-04
SD 8 0.101431051 0.0547611 0.002293620
per Table V. Furthermore, Fig. 6 shows the voltage profile
0.00492327 1 before and after the MLSA implementation. Based on the
4 voltage profile, it clearly shows the improvement of
0.00333475 voltage, especially at bus 14 to 30 after installing three SVC
4 units based on the placement and sizing given by MLSA.
3 Best 0.66666666 0.679300000 0.6666666 0.000155823 In term of algorithm performance, Fig. 7 shows the
Worst 7 11.67240000 7 4.740040000 convergence curve between MLSA and LSA, which shows
Average 6671.66666 2.380902000 4.8251100 1.059489012 that MLSA has better convergence for this analysis with
SD 7 2.475021000 0 1.059868099 0.3724 compared to 0.3725. The MLSA can provide a
191.899100 1.0432011 better solution faster than LSA due to the learning factor
0 2
(R) in the algorithm. To summarise, MLSA is a useful
962.944500 0.8478845
optimisation strategy for the problem that combines
0 6
4 Best 0.31174171 0.055032609 0.3225510 4.13764E-13
continuous and discrete numbers. From the results, the
Worst 3 0.109483015 0 7.24675E-01 proposed algorithm can be used either in mathematical as
Average 0.93597885 0.076694719 5.0311600 8.73730E-02 well as power system problem.
SD 7 0.014606062 0 0.167281355
0.61023703 1.5793686 V. CONCLUSION
0 0
MLSA technique is proposed in this study for obtaining
0.14601838 1.1340763
8 6
a better optimisation solution, either in mathematical
5 Best - -186.731000 -186.73100 -186.731000 function or power system problem. With the modification
Worst 186.668100 -173.256000 -186.73100 -186.731000 of the existing LSA technique and the introduction of the
Average - -186.148000 -186.73100 -186.731000 learning factor, MLSA is able to improve the solution that
SD 14.1925000 2.05206900 0.0000000 0.000000000 is obtainable by LSA. In mathematical analysis, MLSA’s
- ability has been proven using 10 well utilised nonlinear
99.5396000 mathematical benchmark functions with various
49.0091000 characteristics. The MLSA also provides a more suitable
6 Best 0.00000000 0.00440000 0.0000000 0.000000000 SVC output size after producing better best, average and
Worst 1.00000000 0.01800000 0 4.000000000 median compared to LSA. As a consequence, the MLSA
able to provide significant improvement in finding the global gives better VSI value compared to LSA. Hence, it can be
optimization result. concluded that MLSA is a superior optimisation method in
IV. VALIDATION ON MLSA IN IMPROVING VSI
In order to validate the effectiveness of MLSA
performance, the algorithm is used to determine optimal
placement and sizing of three units SVC devices
simultaneously, and the results will be compared with LSA.
According to the result in Table IV, after 50 independent
trials, the MLSA, again, give slightly better result than LSA
TABLE IV. RESULT FOR LSA AND MLSA
50 independent trials
VSI Value
LSA MLSA
Minimum 0.372500 0.372400
Maximum 0.417600 0.409300
Median 0.398000 0.374450
Average 0.395540 0.385122 Fig. 5. Convergence Curve for
Standard Deviation 0.011015 0.013735

60on June 18,2021 at 08:51:16 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: San Francisco State Univ. Downloaded
 Network Reconfiguration,” Indonesian Journal of Electrical
Engineering and Computer Science, vol.5, pp. 488-495, 2017.
[9] O. Arouna, M. I. Adolphe, A. O. Robert, A. Z. Kenneth, V. Antoine,
B. Ramanou, T. Herman, D. Celestin, “Technico-economic
optimization of Distributed Generation (DG) and Static Var
Compensator (SVC) positioning in a real radial distribution network
using the NSGA-II genetic algorithm,” in Proc. 2019 IEEE PES/IAS
PowerAfrica, 2019.
[10] O. M. A. Gabl, E. D. Amgad, E. M. Magdy, “Optimal Location and
Size of FACTS Devices for Distributed Voltage Control,” in Proc.
2018 Twentieth International Middle East Power Systems Conference
Fig. 6. Voltage Profiles before and after implement MLSA (MEPCON), 2018.
[11] S. K. Bali, M. Suryakalavathi, V. N. K. Gundavarapu, “Harmony
search algorithm and combined index-based optimal reallocation of
generators in a deregulated power system,” Neural Computing and
Applications vol. 31 n.6, pp. 1949-1957, 2019.
[12] H. Suyono, N. H. Rini, P. W. Eka, “Power system optimization of static
VAR compensator using novel global harmony search method,”
International Journal of Electrical and Electronic Engineering &
Telecommunications (IJEETC), 2019.
[13] A. L. Ahmad, and S. Reza, “Optimal placement and sizing of multi-
type FACTS devices in power systems using metaheuristic
optimisation techniques: An updated review,” Ain Shams Engineering
Journal, 2019.
[14] B. Baadji, B. Hamid, and B. Azzeddine, “Comprehensive learning bat
algorithm for optimal coordinated tuning of power system stabilizers
Fig. 7. Convergence curve for MLSA & LSA and static VAR compensator in power systems,” Engineering
Optimization, pp. 1-19, 2019.
solving nonlinear benchmark function and optimal location
[15] D. Diptanu, A. Bhattacharya, R. N. Ray, “Dragonfly Algorithm for
and sizing of SVC solving probabilistic Economic Load Dispatch problems,” Neural
Computing and Applications, pp. 1-17, 2019.
ACKNOWLEDGMENT
[16] N. G. Hingorani, L. Gyugyi, Understanding FACTS: concepts and
The scholars would like to prompt their gratitude to the technology of flexible AC transmission system, Wiley, 1999.
Universiti Teknologi Malaysia (UTM) for supporting this [17] Y. C. Chang, “Multi-Objective Optimal SVC Installation for Power
research under UTMER Grant (17J76). System Loading Margin Improvement,” IEEE Transactions on Power
Systems, vol. 27 n.2, pp 984-92, 2012.
REFERENCES [18] P. Balachennaiah, P. H. Reddy, U. N. K. Raju, “A novel algorithm for
voltage stability augmentation through optimal placement and sizing of
[1] J. Machowski, Z. Lubosny, J. W. Bialek, J. R. Bumby, Power system SVC,” in Proc. 2015 IEEE International Conference on Signal
dynamics: stability and control, John Wiley & Sons, 2020. Processing, Informatics, Communication and Energy Systems,
[2] B. Banerjee, D. Jayaweera, S. Islam, “Modelling and simulation of SPICES 2015.
power systems,” in Smart Power Systems and Renewable Energy [19] H. Ambriz-Perez, E. Acha, C. R. Fuerte-Esquivel, “Advanced SVC
System Integration. Springer, Cham, pp. 15-28, 2016 models for Newton-Raphson load flow and Newton optimal power
[3] Y. D. Valle, J. C. Hernandez, G. K. Venayagamoorthy, R.G. Harley, flow studies,” IEEE Transactions on Power Systems, vol. 15, pp 129-
“Optimal STATCOM sizing and placement using particle swarn 136, 2000.
optimization,” in Proc. 2006 IEEE/PES Transmission & Distribution [20] H. Shareef, A. A. Ibrahim, A. H. Mutlag, “Lightning search algorithm,”
Conference and Exposition: Latin America. IEEE, 2006. Applied Soft Computing vol. 36, pp 315-333, 2015.
[4] R. Eberhart, and K. James, “A new optimizer using particle swarm [21] H. R. Tizhoosh, “Opposition-based learning: A new scheme for
theory,” in Proc. Sixth International Symposium on Micro Machine machine intelligence.” in Proc. International Conference on
and Human Science. IEEE, 1995. Computational Intelligence for Modelling, Control and Automation,
[5] R .J. Singh and K. G. Vikash, "Comparison of GSA and PSO-Based CIMCA 2005 and International Conference on Intelligent Agents, Web
Optimization Techniques for the Optimal Placement of Series and Technologies and Internet; 2005, pp. 695-701.
Shunt FACTS Devices in a Power System." Artificial Intelligence and [22] A. Naserbegi, M. Aghaie, A. Minuchehr, G. Alahyarizadeh, “A novel
Evolutionary Computations in Engineering Systems, pp. 375-385. exergy optimization of Bushehr nuclear power plant by gravitational
Springer, Singapore, 2017. search algorithm (GSA),” Energy, vol. 148, pp. 373-385, 2018.
[6] M. C. Pandya, and G. J. Jitendra, “Coordinated control of SVC and [23] R. Rao, “Review of applications of TLBO algorithm and a tutorial for
TCSC for voltage profile improvement employing particle swarm beginners to solve the unconstrained and constrained optimization
optimization,” in Proc. 2017 International Conference On Smart problems,” Decision science letters, vol. 5 no. 1, pp. 1-30. 2016
Technologies For Smart Nation (SmartTechCon), IEEE, 2017.
[24] S. N. Syed Nasir, J. J. Jamian, and M. W. Mustafa, “Minimization of
[7] E. R. Sanseverino, Q. T. Tran, M. L. Di Silvestre, B. V. Doan, N. Q. harmonic distortion impact due to large‐scale fast charging station
Nguyen, “Voltage Profile Improvement for Soc Son's Low-Voltage using Modified Lightning Search Algorithm and Pareto‐Fuzzy
Grid with High Penetration of PV Systems by Optimizing the Location synergistic approach,” IEEJ Transactions on Electrical and Electronic
of SVC Devices,” in Proc. 2018 IEEE International Conference on Engineering, vol.13 no.6, pp. 815-822, 2018.
Environment and Electrical Engineering and 2018 IEEE Industrial and [25] S. N. Syed Nasir, J. J. Jamian, and M. W. Mustafa, “Minimizing
Commercial Power Systems Europe (EEEIC/I&CPS Europe), IEEE, harmonic distortion impact cause by CS using meta heuristic
2018. technique,” Telkomnika, vol. 17 n.4, August 2019.
[8] M. N. B. Muhtazaruddin, N. A. Bani, S. A. M. Aris, H. M. Kaidi, A.
Y. A. Fatah, J. J. Jamian, F. M. Sukki, S. H. A. Bakar, “Distribution
Power Loss Minimization via Distributed Generation, Capacitor and

61on June 18,2021 at 08:51:16 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: San Francisco State Univ. Downloaded