Tap changer optimisation using embedded differential evolutionary programming technique for loss control in power system

Bulletin of Electrical Engineering and Informatics
Vol. 9, No. 6, December 2020, pp. 2253~2260
ISSN: 2302-9285, DOI: 10.11591/eei.v9i6.2505  2253
Journal homepage: http://beei.org
Tap changer optimisation using embedded differential
evolutionary programming technique for loss control
in power system
Ahmad Faris1
, Ismail Musirin2
, Shahrizal Jelani3
, Saiful Amri Ismail4
,
Mohd Helmi Mansor5
, A. V. Senthil Kumar6
1,2,4
Faculty of Electrical Engineering, Universiti Teknologi MARA, Malaysia
3
Faculty of Engineering, Technology and Built Environment, UCSI University, Malaysia
5
Department of Electrical and Electronic Engineering, College of Engineering, Universiti Tenaga Nasional, Malaysia
6
Hindusthan College of Arts and Science, India
Article Info ABSTRACT
Article history:
Received Jan 18, 2019
Revised Mar 27, 2020
Accepted Apr 14, 2020
Over-compensation and under-compensation phenomena are two undesirable
results in power system compensation. This will be not a good option
in power system planning and operation. The non-optimal values of
the compensating parameters subjected to a power system have contributed
to these phenomena. Thus, a reliable optimization technique is mandatory to
alleviate this issue. This paper presents a stochastic optimization technique
used to fix the power loss control in a high demand power system due to
the load increase, which causes the voltage decay problems leading to current
increase and system loss increment. A new optimization technique termed as
embedded differential evolutionary programming (EDEP) is proposed, which
integrates the traditional differential evolution (DE) and evolutionary
programming (EP). Consequently, EDEP was for solving optimizations
problem in power system through the tap changer optimizations scheme.
Results obtained from this study are significantly superior compared to
the traditional EP with implementation on the IEEE 30-bus reliability test
system (RTS) for the loss minimization scheme.
Keywords:
Differential evolution
Embedded optimisation
Evolutionary programming
Loss control
Tap changer
This is an open access article under the CC BY-SA license.
Corresponding Author:
Shahrizal Jelani,
Faculty of Engineering,
Technology and Built Environment,
UCSI University,
Kuala Lumpur, Malaysia.
Email: shahrizal@ucsiuniversity.edu.my
1. INTRODUCTION
The growing demand in power system network due to increasing load has caused voltage decay,
leading to current increase and system loss. To curb the voltage problems, several power compensation
schemes can be implemented. This requires the use of optimisation processes; among the important
optimisation techniques are evolutionary programming (EP), genetic algorithm (GA) and differential
evolution (DE). In the year 2008, M. Varadarajan et al. [1] reported that the method used to determine
control variables needs to be varied to minimise system loss using the DE method. Several main parameters
can be controlled via the parameter settings of mutation, crossover and population size. The result of system
loss showed that DE provided the best solution compared to sequential quadratic programming [2]. DE is also used in
order to perform parameter estimation in chaotic systems as a way to counter the issues. It also demonstrated
that DE was more effective than partical swarm optimisation (PSO) and GA techniques [3]. Apart from that,

 ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 9, No. 6, December 2020 : 2253 – 2260
2254
in 2012, the study was carried out in order to minimise power losses by selecting the optimal location
and parameter settings of thyristor-controlled series capacitor (TCSC).
The technique had adopted DE and GA methods, which showed that DE was more effective.
Some improvements were made on DE by Z. Huang et al. [4] to establish suitable parameter adjustment
method by adjusting the control parameters following the evolution stage. The results showed that DE was
able to increase the ability of exploration and faster convergence speed. When a higher population is applied
in DE, the effectiveness on DE can be observed through the fast convergence performance [5].
Many parameters can be controlled via DE such as population size, differential weight, crossover
and generation number [6]. An important study related to optimisation technique is the Optimal Power Flow
(OPF). OPF is a problem that occurs in the identification of most acceptable operating levels, such as to
minimise the losses and cost for electric power to make sure it satisfies the consumer demand, which flows
throughout a transmission network. Loss minimisation in OPF is due to current heating [7-9].
In order to perform OPF, N. Sinha et al. [10, 11] found that the technique was suitable for secure
and economic operation in power system. C. Ameur et al. [12] reported that any set of optimisation problems
in electric power systems is known as OPF, which is one of the most practically important research subfields
of constrained nonlinear optimisation. To solve the OPF problems, methods such as DE can be used for
different objectives that reflect fuel cost minimisation, voltage profile improvement, and voltage stability
enhancement, all of which show the effectiveness and robustness of the method used [13]. EP is useful to
solving nonlinear programming problems [14]. Over the past years, OPF problems can be assessed by
various techniques, such as non-linear programming, quadratic programming, mixed-integer programming
and interior-point method. This types of method are categories as traditional methods. Traditional methods
have some disadvantages, as they cannot be used in case of prohibited operating regions and multiple fuels [15].
In addition, they have a high sensitivity to initial solution, consequently, it may be trapped into local results,
thus affecting the findings of the research. The difficulties in implementing OPF based on traditional methods
can now be overcomed by modern stochastic algorithms, such as evolutionary programming (EP), tabu
search (TS), improved evolutionary programming (IEP) and differential evolution (DE).
Genetic algorithm (GA) is an optimisation method for solving both constrained and unconstrained
problems that are based on the natural selection of the population [16, 17]. It is commonly used to generate
the best solutions for optimisation purposes by relying on the algorithm.. Tabu Search is commonly used as
a mathematical optimisation tool to act as a derivative-free optimisation technique in solving OPF problems,
thus significantly reducing computational burden. The main advantages of TS algorithm is its robustness to
its own parameter settings. In addition, TS is characterised by its ability to avoid entrapment in local optimal
solution and prevent cycling by using a flexible memory of search history [18]. Furthermore, PSO is a swarm
technique developed by R. C. Eberhart et al. [19] in 1995. This technique motivated from the simulation of
social behaviour. PSO, like the other optimisation techniques, is able to update the population of individuals
by applying some type of operators according to the fitness information so that the population can be
expected to receive better solution areas [20]. EP was also used in this study. EP is an algorithm that is works
on mutation to breed the offspring [21]. In recent years, EP has been applied with success to many numerical
and other optimisation problems. Nevertheless, there are weaknesses to using EP, which is slow convergence
to a good near optimum [22]. EP can be compared to Genetic Algorithm (GA) in a way that EP controls
the parameter and is based on mutation and selection parameters [23]. EP has two major steps in its
algorithm, which are 1) mutate the solutions in the current population; and 2) select the next generation from
the mutated [24]. On the other hand, J-H. Kim and H. Myung have proposed two EP methods for handling of
nonlinear constrained optimisation problems [25].
This paper presents a stochastic optimization technique used to fix the power loss control in a high
demand power system due to the load increase. A new technique termed as EDEP is developed to solve
optimization involving loss and voltage control. Implementation on a reliability test system produced
promising results; highlighting its superiority over the traditional EP.
2. RESEARCH METHOD
In this section, the processes involved in this study are presented as:
2.1. Problem formulation
Power loss in power system is the big problem for the system. They are happened due to disturbance
on voltage and current as both of them are related. This can be seen through the formula:
∑ ( ) (1)

Bulletin of Electr Eng & Inf ISSN: 2302-9285 
Tap changer optimization using embedded differential evolutionary programming... (Ahmad Faris)
2255
where:
= Number of transmission line
= Conductance of line, vi and vj=voltage magnitude
= Voltage angle difference between busses and
2.2. Differential evolution
DE has three main operators, which are mutation, crossover and selection. These operators are
similar to the EP’s operators. The efficiency of DE is based on mutation and crossover.
a. Mutation
Three different vector xr1, xr2 and xr3 were randomly selected to generate mutant vector using
the following formula:
( ) (2)
b. Crossover
Crossover is a process when a mutant vector and a parent vector are combined to create a trial vector by
the following formula:
{ (3)
c. Selection
In selection, competition between the parent vectors will occur. The selection process is when
the value of the trial vector is less than or equal to the parent vector, the trial vector will enter to the next
generation. Otherwise, the parent vector will survive and go to the next generation [16]. This operator can
be described by the following formula:
{
( ) ( )
(4)
2.3. Evolutionary programming
For EP, there are two major steps which are mutation and selection of fitness. The explanations of
those steps are below:
a. Initialisation
It is a process to generate random numbers, , where k is the number of variables and α is the
number of individuals.
b. Fitness Evaluation
In this phase, fitness values are calculated using generated individuals.
c. Mutation
In mutation, offspring is created using Gaussian Mutation technique by applying this formula:
( ( )( ) (5)
d. Combination
Combination is when the parent and offspring are combined in series (by rows). The number of rows will
be doubled:
[ ] [ ]
[ ] [ ]
(6)
e. Selection
This process proceeds to selecting the survivors from the combination of parent and offspring.
2.4. Proposed embedded differential evolutionary programming (EDEP) technique
The embedded differential evolutionary programming is combination technique that was applied in
this research. The mutation formula from DE was implemented in the evolutionary programming steps to

 ISSN: 2302-9285
2256
produce the second fitness. The major program is EP while DE is the supporting program inserted into EP.
The flowchart of the proposed EDEP technique is shown in Figure 1. Mutation and crossover are steps from
differential evolution which were implemented in the EP algorithm. The steps are discussed in details in
this section:
a. Initialisation
Generation of random number of location (L1, L2,…i) and random number of tap changer
(tap1, tap2….i) are based on the EP’s initialisation approach as mentioned previously. 20 individuals of
each control variable are created in this process.
b. Fitness 1
Parents were created based on the 20 initial individuals using selection tap and location tap from
line data.
Figure 1. Proposed EDEP
c. Mutation
In this section, DE was implemented by using (2) to produce the offspring. The random number,
r for tap changer is selected by using a random selection r=randperm(20,i). F is a constant with a
value set at 1. To form the offspring for a new tap changer and new location, the alpha needs to be less
than the crossover, which is 0.5 or the number of individuals is same to the random number that generated
in the initialisation. If these conditions are not fulfilled, the previous parent value can be taken to form the
tap changer and establish its location. Next, be careful for the value of location of tap changer, if the value
of tap changer is less than zero, use ‘abs’ to form an absolute value. Then, when the location value
equals to zero, take 1. Lastly, if the new location is more than maximum branch number,
Ln_new>maximum branch, take the data value Ln_new=maximum branch number, because the
location must be less than or equal to the data value.
d. Fitness 2
Offspring created during mutation was then used to form Fitness 2.
e. Combination
Offspring (from mutation) and parents are combined in order to perform the selection process.
f. Selection
For this section, the top 20 values for tap changer and location were selected in ascending order.
INITIALISATION
FITNESS 1
MUTATION
FITNESS 2
SELECTION
CONVERGE?
No
START
END
COMBINATION

2257
3. RESULTS AND DISCUSSION
In this study, the proposed EDEP optimisation technique was implemented on the IEEE 30-bus RTS
for loss minimisation process.
3.1. Implementation of EDEP on bus 3
Figure 2 illustrates the loss profile results using the proposed EDEP with load variation at bus 3,
Qd3, which was gradually increased from 10 to 40 MVar. In general, the loss increased as the load was
increased from 10 to 40 MVar. The loss profile is lower when EDEP was implemented to the system through
the transformer tap changer optimisation exercise. On the other hand, Table 1 shows the results for optimal
location and sizing of transformer tap setting values of load variations subjected to bus 3.
For instance, at Qd30=10 MVar, the optimal locations for tap settings were at lines 33, 32 and 36, with their
corresponding optimal tap setting values of 0.5951, 1.107, and 0.6736. The same table can be referred
for other load variations.
Figure 2. Result for loss profile with load variations at bus 3
Table 1. Result for optimal location and sizing of transformer taps
10 MVar 20 MVar 30 MVar 35 MVar 40 MVar
L1 33 23 36 36 38
L2 32 41 9 16 36
L3 36 36 33 13 28
Tap1 0.5951 0.9763 0.8300 0.8932 0.9771
Tap2 1.1070 1.0296 1.0594 0.9718 0.7956
Tap3 0.6736 0.7293 0.9040 0.9800 0.9870
3.2. Implementation of EDEP on bus 30
Figure 3 illustrates the result for loss profile using the proposed EDEP when the load at bus 30, Qd30
was gradually increased from 5 MVar to 35 MVar. In general, the loss increased as the load was increased
from 5 MVar to 35 MVar. The loss profile is lower when EDEP was implemented to the system through
the transformer tap changer optimization exercise. On the other hand, Table 2 portrays the result for optimal
location and sizing of transformer tap setting values of load variations subjected to bus 3. For instance,
at Qd30=5 Mvar, the optimal location for tap settings are lines 37, 38, and 2 with their corresponding
optimal tap setting values of 0.8308, 0.8239, and 0.9596. Results for other load variations can be referred to
the same table.
28,38 28,8826 29,5821 29,8223 30,087
21,0371 21,4002 21,7591
23,1617 23,969
0
5
10
15
20
25
30
35
1 0 2 0 3 0 3 5 4 0
POWER
LOSS
BUS VALUE,MVAR
B E FORE AND AFT E R E DE P
before after EPDE

 ISSN: 2302-9285
2258
Figure 3. Result for loss profile with load variation at bus 30
L1 37 36 33 5 13
L2 38 14 3 36 11
L3 2 33 36 15 36
Tap1 0.8308 0.7658 0.8981 0.9460 0.9394
Tap2 0.8239 0.9795 0.9922 0.7603 0.9128
Tap3 0.9596 0.8081 0.8611 0.9629 0.8039
3.3. Implementation of EP
Figure 4 illustrates the result for loss using the EP when the load at bus 30, Qd30 was gradually
increased from 5 MVar to 35 MVar. In general, the loss increased as the load was increased from 5 MVar to
35 MVar. The loss profile is lower when EP was implemented to the system through the transformer tap
changer optimization exercise. On the other hand, Table 3 demonstrates the result for optimal location
and sizing of transformer tap setting values of load variations subjected to bus 3. For instance, at Qd30=5
Mvar, optimal tap setting values are 0.9733, 0.9478, and 0.9910. Results for other load variations can be
referred to the same table.
Figure 4. Result for loss profile with load variation at bus 30
Tap1 0.9733 0.9783 0.9744 0.9533 0.9128
Tap2 0.9478 0.9570 0.9588 0.9695 0.9054
Tap3 0.9910 0.9926 0.9897 0.9423 0.9058
17,444 17,735
19,1
22,675
28,1362
17,0358 17,331
18,229
20,0839
21,0923
15
17
19
21
23
25
27
29
5 1 0 2 0 3 0 3 5
POWER
LOSS
BUS VALUE, MVAR
GRAPH POWE R L OSS WHE N APPL YING E DE P
AT B US 3 0
before after EDEP
17,444 17,735
19,1
22,675
28,1362
17,201 17,4112
18,7254
22,2027
25,852
17
19
21
23
25
27
29
5 1 0 2 0 3 0 3 5
POWER
LOSS
BUS VALUE,MVAR
GRAPH APPL YING E P AT B US 3 0
before with EP

2259
3.4. Comparing EDEP and EP
Figure 5 illustrates the comparison results for loss using the proposed EDEP, EP and without
optimisation where the load at bus 30, Qd30 had gradually increased from 5 MVar to 35 MVar. In general,
the loss increased as the load was increased from 5 MVar to 35 MVar. The loss profile is lower when EP
and EDEP were implemented to the system through the transformer tap changer optimization exercise.
The lowest loss profile that can be seen is when EDEP was implemented in this research.
Figure 5. Result for loss profile with load variations at bus 30
4. CONCLUSION
A new technique was developed and termed as embedded differential evolutionary programming
(EDEP) for solving optimisation problem in power system through tap changer optimisation scheme. EDEP
was found to be more efficient than EP, and the mutation method from differential evolution (DE) was used
to replace the Gaussian mutation method in the original EP. The comparison result between EDEP and EP
proved that EDEP is the best method for power loss minimisation the loss.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the UCSI University for the financial support given for this
project. This research is supported by the UCSI University under the Centre of Excellence for Research,
Value Innovation, and Entrepreneurship (CERVIE).
REFERENCES
[1] M. Varadarajan and K. S. Swarup, “Network loss minimization with voltage security using differential evolution,”
Electr. Power Syst. Res., vol. 78, no. 5, pp. 815-823, 2008.
[2] B. Peng, B. Liu, F. Y. Zhang, and L. Wang, “Differential evolution algorithm-based parameter estimation for
chaotic systems,” Chaos, Solitons and Fractals, vol. 39, no. 5, pp. 2110-2118, 2009.
[3] G. I. Rashed, Y. Sun, and H. I. Shaheen, “Optimal location and parameter setting of TCSC for loss minimization
based on differential evolution and genetic algorithm,” Phys. Procedia, vol. 33, pp. 1864-1878, 2012.
[4] Z. Huang and Y. Chen, “An improved differential evolution algorithm based on adaptive parameter,” J. Control
Sci. Eng., vol. 2013, pp. 1-5, 2013.
[5] B. Kazimipour, X. Li, and A. K. Qin, “Effects of population initialization on differential evolution for large scale
optimization,” Proc. 2014 IEEE Congr. Evol. Comput. CEC 2014, pp. 2404-2411, 2014.
[6] J. L. Carpentier, “Optimal power flows: uses, methods and developments,” IFAC Proceeding, vol. 18, no. 7,
pp. 11-21, 2017.
[7] K. R. C. Mamandur and R. D. Chenoweth, “Optimal control of reactive power flow for improvements in
voltage profiles and for real power loss minimization,” IEEE Trans. Power Appar. Syst., vol. PAS-100, no. 7,
pp. 3185-3194, 1981.
[8] J. Ma, L. Yuan, Z. Zhao, and F. He, “Transmission loss optimization-based optimal power flow strategy by
hierarchical control for DC microgrids,” IEEE Trans. Power Electron., vol. 32, no. 3, pp. 1952-1963, 2017.
[9] A. A. Abou El Ela, M. A. Abido, and S. R. Spea, “Optimal power flow using differential evolution algorithm,”
Electr. Power Syst. Res., vol. 80, no. 7, pp. 878-885, 2010.
5 10 20 30 35
before 17,444 17,735 19,1 22,675 28,1362
after EDEP 17,0358 17,331 18,229 20,0839 21,0923
with EP 17,201 17,4112 18,7254 22,2027 25,852
17
19
21
23
25
27
29
GRAPH FOR COMPARING POWE R L OSS B E FORE
OPT IMIZAT ION, WIT H E DE P AND WIT H E P
before after EDEP with EP

 ISSN: 2302-9285
2260
[10] N. Sinha, R. Chakrabarti, and P. K. Chattopadhyay, “Evolutionary programming techniques for economic load
dispatch,” IEEE Trans. Evol. Comput., vol. 7, no. 1, pp. 83-94, 2003.
[11] M. F. Shaari, I. Musirin, M. F. M. Nazer, S. Jaelani, F. A. Jamaludin, M. H. Mansor, and A. S. Kumar, “Supervised
evolutionary programming based technique for multi-DG installation in distribution system,” International Journal
of Artificial Intelligent, vol. 9, no. 1, pp. 11-17, 2020.
[12] C. Ameur, S. Faquir, and A. Yahyaouy, “Intelligent optimization and management system for renewable energy
systems using multi-agent,” International Journal of Artificial Intelligent, vol. 8, no. 4, pp. 352-359, 2019.
[13] H. Yuan, F. Li, Y. Wei, and J. Zhu, “Novel linearized power flow and linearized OPF models for active distribution
networks with application in distribution LMP,” IEEE Trans. Smart Grid, vol. 9, no. 1, pp. 438-448, 2018.
[14] L. Thurner, A. Scheidler, F. Schäfer, J-H. Menke, J. Dollichon, F. Meier, S. Meinecke, and M. Braun
“Pandapower-an open-source python tool for convenient modeling, analysis, and optimization of electric power
systems,” IEEE Trans. Power Syst., vol. 33, no. 6, pp. 6510-6521, 2018.
[15] N. S. Razali, N. Roslan, M. H. Mansor, I. Musirin, S. A. Shaaya, and S. Jelani, “Optimal reactive power control
using compensating capacitor based on artificial immune system,” Int. J. Adv. Trends Comput. Sci. Eng., vol. 8,
no. 1.3, pp. 381-386, 2019.
[16] G. I. Rashed, Y. Sun, and H. I. Shaheen, “Optimal location and parameter setting of TCSC for loss minimization
based on differential evolution and genetic algorithm,” Phys. Procedia, vol. 33, pp. 1864-1878, 2012.
[17] D. P. Kothari, “Power system optimization: Second edition,” 2012 2nd National Conference on Computational
Intelligence and Signal Processing (CISP), pp. 18-21, 2011.
[18] R. C. Eberhart and Y. Shi, “Comparison between genetic algorithms and particle swarm optimization,”
International conference on evolutionary programming. Springer, pp. 611-616, 1998.
[19] M. A. Abido, “Optimal power flow using tabu search algorithm,” Electr. Power Components Syst., vol. 30, no. 5,
pp. 469-483, 2002.
[20] B. Niu, H. Huang, L. Tan, and Q. Duan, “Symbiosis-based alternative learning multi-swarm particle swarm
optimization,” IEEE/ACM Trans. Comput. Biol. Bioinforma., vol. 14, no. 1, pp. 4-14, 2017.
[21] H. Han, W. Lu, L. Zhang, and J. Qiao, “Adaptive gradient multi objective particle swarm optimization,” IEEE
Trans. Cybern., vol. 48, no. 11, pp. 3067-3079, 2018.
[22] R. C. Eberhart and Y. Shui, “Particle swarm optimization: developments, applications and resources,” Proceedings
of the 2001 Congress on Evolutionary Computation (IEEE Cat. No.01TH8546), vol. 1, pp. 81-86, 2001.
[23] Y. Shi and R. C. Eberhart, “Empirical study of particle swarm optimization,” Proc. 1999 Congr. Evol. Comput.
CEC 1999, vol. 3, pp. 1945-1950, 1999.
[24] S. A. Shaaya, I. Musirin, S. I. Sulaiman, M. H. Masor, and S. S. Mustafa, “Performance comparison of distributed
generation installation arrangement in transmission system for loss control,” Bulletin of Electrical Engineering and
Informatics, vol. 8, no. 1, pp. 39-45, 2019.
[25] Jong-Hwan Kim and Hyun Myung, “Evolutionary programming techniques for constrained optimization
problems,” IEEE Trans. Evol. Comput., vol. 1, no. 2, pp. 129-140, 1997.

Tap changer optimisation using embedded differential evolutionary programming technique for loss control in power system

Recommended

More Related Content

What's hot (20)

Similar to Tap changer optimisation using embedded differential evolutionary programming technique for loss control in power system (20)

More from journalBEEI (20)

Recently uploaded (20)

Tap changer optimisation using embedded differential evolutionary programming technique for loss control in power system