COMPEL - The international journal for computation and mathematics in electrical

and electronic engineering

Design technique for leakage current reduction in surge arrester using gravitational search algorithm and
imperialist competitive algorithm
Nurul Ain Abdul Latiff, Hazlee Illias, Ab Halim Abu Bakar, Syahirah Abd Halim, Sameh Ziad Dabbak,
Article information:
To cite this document:
Nurul Ain Abdul Latiff, Hazlee Illias, Ab Halim Abu Bakar, Syahirah Abd Halim, Sameh Ziad Dabbak, "Design technique
for leakage current reduction in surge arrester using gravitational search algorithm and imperialist competitive algorithm",
COMPEL - The international journal for computation and mathematics in electrical and electronic engineering , https://
doi.org/10.1108/COMPEL-03-2017-0116
Permanent link to this document:
https://doi.org/10.1108/COMPEL-03-2017-0116
Downloaded on: 21 November 2017, At: 06:15 (PT)
References: this document contains references to 0 other documents.
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

To copy this document: permissions@emeraldinsight.com

The fulltext of this document has been downloaded 1 times since 2017*

Access to this document was granted through an Emerald subscription provided by emerald-srm:425905 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service
information about how to choose which publication to write for and submission guidelines are available for all. Please
visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of
more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online
products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication
Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.

*Related content and download information correct at time of download.

 Design technique for leakage current reduction in
surge arrester using gravitational search
algorithm and imperialist competitive algorithm

Abstract
Purpose – Leakage current is one of the factors, which can contribute towards degradation of
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

surge arresters. Thus, the purpose of this work is to study on leakage current within surge
arresters and improvement on their design.
Design/methodology/approach - In this work, a three-dimensional model geometry of 11kV
zinc oxide surge arrester was designed in finite element analysis (FEA) and was applied to
calculate the leakage current under normal operating condition and being verified with
measurement results. The optimisation methods were used to improve the arrester design by
minimising the leakage current across the arrester using Imperialist competitive algorithm (ICA)
and gravitational search algorithm (GSA).
Findings - The arrester design in reducing leakage current were successfully optimized by
varying the glass permittivity, silicone rubber permittivity and the width of the ground terminal
of the surge arrester. It was found that the surge arrester design obtained using ICA has lower
leakage current than GSA and the original design of the surge arrester.
Practical implications - The comparison between measurement and simulation enables factors
that affect the mechanism of leakage current in surge arresters to be identified and provides the
ideal design of arrester.
Originality/value –Surge arrester design was optimised by ICA and GSA, which has never been
applied in past works in designing surge arrester with minimum leakage current.
Keywords Surge arrester, leakage current, condition monitoring, finite element analysis, high
voltage engineering
Paper type Research paper

1
 1.0 Introduction
Overhead transmission lines are exposed to lightning interference. Lightning interference can
be minimised by shielding the overhead lines using counter-poise wires, ground rods and
protective devices at the line terminations (Naidu & Kamaraju, 2006). In heavy lightning regions,
surge arresters are usually used. There are many types of surge arresters, depending on the
materials, sizes and rated voltages. Different types of housing can be designed in many forms and
this allows optimum surge arrester to be provided for different applications. Surge arresters are
planned for years of operation in their design and provide a contribution towards lightning
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

protection.
However, many factors can cause the degradation of surge arresters and reducing the stability
of the arrester which is caused by non uniformity of the voltage distribution (Illias, Mohsin Ali,
Bakar, & Mokhlis, 2012). The non uniform voltage distribution can be solve by adding grading
rings to the surge arrester to reduce the effect of stray capacitance (Jinliang, Shanqiang, Bo, &
Jun, 2006; Sjöstedt, Stenström, Pusch, & Ostrowski, 2010). A possible factor of surge arrester
degradation is moisture ingress due to weakening of the housing sealing, which can lead to
internal discharges (Zeinoddini-Meymand, Vahidi, Naghizadeh, & Moghimi-Haji, 2013).
Conversely, varistors in surge arrester are not affected at low humidity level since it is sealed by
glaze layers. The internal flashover will be initiated if the humidity level increases
(Aghaebrahimi, Shariatinasab, & Ghayedi, 2012; Dong, Jiang, Hu, Shu, & Sun, 2012). Since
ZnO varistor is the main component of surge arresters, many studies on the structure of ZnO have
been conducted. Another renowned factor of surge arrester degradation is the overheating. The
heat increases the temperature of a surge arrester due to the intrinsic resistance of ZnO blocks.
Leakage current is also a factor that can cause the degradation of surge arresters (Luca Amicucci
& Mazzetti, 2004). Hence, monitoring of the leakage current is important and there have been
many studies on surge arrester leakage current monitoring techniques (Banik, Dalai, &
Chatterjee, 2016; M. Khodsuz & Mirzaie, 2015; Masume Khodsuz & Mirzaie, 2016; Masume
Khodsuz, Mirzaie, & Seyyedbarzegar, 2015; Song et al., 2015; Sp et al., 2016).
The interest in algorithms has been increasing among researchers since many years before. The
behaviours of natural phenomena have motivated the researchers to explain complex
computational difficulties (Alotto, dos Santos Coelho, C. Mariani, & da C. Oliveira, 2014;
Baraston, Gerbaud, Reinbold , Boussey , & Wurtz, 2016; Caner, Gerada, Asher, & Özer, 2016;

2
 Golak & Kordos, 2016). Many researchers have studied on design optimisation of surge arresters
to improve their performance. The optimisation used to increase the effectiveness of surge
arresters are by residual voltage value, system reliability, surge arrester location, insulation
coordination, lightning performance, etc. In one of the past works, a 500kV polymeric-housed
arrester design was analysed in (Jinliang, Shuiming, Rong, Jun, & Chenggang, 2006). The
analysis includes the ZnO varistor selection, structure design, polymer housing design,
mechanical structure design and parameter determination. Two ring-shaped discharge electrodes
were included to the gap at two ends of an insulator to enhance the lightning withstand level of
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

transmission lines.
Finite element analysis (FEA), also called the finite element method (FEM), is a method for
numerical solution of field problems (Napieralska-Juszczak, Napieralski, Komeza, & Zeroukhi,
2016). FEM was used to determine the stray and block capacitance of an arrester (Valsalal, Usa,
& Udayakumar, 2011). FEM was also combined with intelligent algorithms to determine the
optimum dimension of a surge arrester grading ring to minimise the value of electric field
magnitude in and outside of the surge arrester (Aghaebrahimi et al., 2012). Therefore, the lifetime
of surge arresters can be increased. Both algorithms, FEM and intelligent algorithms were
verified with particle swarm optimisation (PSO) and differential evolution techniques. In 2013,
PSO method was used to prove the accuracy of surge arresters in minimizing the residual peak
voltage for lightning surge, switching surge and steep front surge (Zeinoddini-Meymand et al.,
2013).
Since leakage current is one of the factors, which can contribute towards degradation of surge
arresters, numerous works on reducing leakage current in surge arrester designs have been
conducted since the past. However, works on improvement on surge arrester design using
optimisation methods are less likely to be found in literature. Therefore, improvement on surge
arrester design by minimising the leakage current across the arrester using optimisation methods
is proposed in this work. The main advantage of using optimisation methods over trial and error
method is the most suitable solutions can be obtained within shorter time.
A three-dimensional model geometry of 11kV Zinc Oxide surge arrester was designed in finite
element analysis (FEA) and was applied to calculate the leakage current under normal operating
condition. To verify the model that has been designed, the results from the simulation were
compared with the measurement results of leakage current on an actual surge arrester. Once the

3
 FEA model has been developed, imperialist competitive algorithm (ICA) and gravitational search
algorithm (GSA) were applied on the model to find the arrester design which can minimise the
leakage current during normal condition. Since ICA and GSA have shown more promising
results compared to other optimisation methods in terms of nonlinear functions and continuous-
optimisation problems, which are similar to optimising surge arrester designs, they were used in
this study. From the comparison results obtained using ICA and GSA, the most suitable method
for surge arrester design optimisation was identified. The results from this work can also assist in
the improvement of surge arrester design in minimising the leakage current.
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

2.0 Modelling and measurement

The surge arrester used in this study is 11kV zinc oxide (ZnO) surge arrester with silicone
rubber housing. The surge arrester was modelled according to the dimension the actual surge
arrester. Table 1 lists the parameters and dimensions of the ZnO surge arrester used in this work.

2.1 11 kV surge arrester model

An 11kV ZnO surge arrester was modelled in three-dimensional (3D) geometry as shown in
Fig. 1 and Fig. 2, using COMSOL Multiphysics software. Two modules were used in the model;
they are ‘electric current’ and ‘heat transfer in solids.’ The ‘electric current’ module was used to
generate the current density and electric field in the surge arrester model while the ‘heat transfer
in solids’ module was used to solve temperature distribution in the model.
The arrester consists of two ZnOs, a glass between the insulator and ZnO and aluminium caps
at both ends. The insulator type of the surge arrester is silicone-housed and the whole model
geometry was surrounded by a layer of air. Each component of the arrester was assigned with
material properties as shown in Table 2. The V-I characteristic of the surge arrester was obtained
by measuring the resistive current across the surge arrester during normal operating condition
from 0kV to 11kV at 50Hz AC sinusoidal voltage. The boundary conditions were set with
relevant interface settings according to Table 3. The aluminium cap of the arrester was assigned
as a terminal source, the aluminium base of the arrester was set as the ground boundary, which is
zero electric potential and the interface condition of the interior boundaries was set to continuity.
After the boundary conditions were assigned, the geometry was meshed for computation as
shown in Fig. 3. The surge arrester was applied with fine mesh elements to increase the accuracy

4
 of the calculation. The surge arrester model was solved by the FEA software using a partial
differential equations (PDE). The leakage current was calculated through surface integration of
the current density, which was obtained on the ground surface. Eq. (1) is used to calculate the
change in current density. Eq. (2) and (3) are used to determine the current density and electric
field that are related to change in the electric potential.

.  =  (1)

 =  +
  +    (2)


where J is the current density, Qj is the current source, Je is the externally generated current
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

density, which equals to zero in the surge arrester model, Ɛr is the relative permittivity of the
material, σ is the electrical conductivity, E is the electric field and V is the electric potential. The
problem to be solved in the model is governed by Maxwell’s equation as follows,

∇×+ 
=0 (3)

Applying quasi-static assumptions, Eq. (3) becomes

∇× =0 (4)
Since the electric field E is conservative and irrotational, E can be written as
 = −
 (5)
Eq. (5) is used to calculate E in eq. (2). Eq. (6) is the heat transfer by conduction equation.

  =
. 
! +  (6)

where ρ is the density, Cp is the specific heat capacity, T is the temperature and Q is the heat
energy.

2.2 Leakage current measurement setup

Fig. 4 shows the measurement setup of leakage current in a surge arrester. The measurement
setup consists of an AC sinusoidal variable voltage source, a high voltage step-up transformer Tx
with a rating of 0.22/100kV, a 2.4MΩ protective resistor Rs, a 1nF measuring capacitor C1, which
is used to monitor the applied voltage, a 20kΩ shunt resistor Rsh, a two-channel digital storage
oscilloscope to capture the applied voltage and leakage current that flows through the surge
arrester. The arrester used in this experiment is 11kV, 227mm height and silicone rubber-housed
with four sheds.

5
 Measurements of surge arrester leakage current were conducted under different applied
voltages and temperatures. For the first measurement, the applied voltage magnitude of 50 Hz AC
sinusoidal was varied at 6, 6.5, 7, 7.5, 8, 8.5, 9, 9.5 and 10 kV while maintaining its temperature
at 26oC. For the second measurement, the temperature of the surge arrester was varied at 26, 33,
36, 40, 45, 50, 55, 60 and 65oC while maintaining its applied voltage at 50 Hz, 10 kV AC
sinusoidal. The maximum temperature was selected at 65oC because the maximum operating
temperature of the arrester is 65oC. First, the sample was tested at room temperature, 26oC at 50
Hz, 10 kV. Then, the arrester was placed inside an oven for 4 hours to ensure the inner and outer
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

temperatures of the arrester are the same. The arrester was reconnected to the system and the
voltage was reapplied. These steps were repeated until all temperatures were tested.
The resistive current was extracted from the leakage current waveform recorded by the
oscilloscope using an improved time-delay addition method (ITDAM) (Masume Khodsuz &
Mirzaie, 2016).

3.0 Optimisation procedure

This section describes in details the optimisation methods used in this work, which are
imperialist competitive algorithm (ICA) and gravitational search algorithm (GSA). There are six
parameters adjusted on the surge arrester. The parameters are the dimension of the aluminum
block, aluminum base, curve of the housing shed, thickness of the silicone rubber, permittivity of
the silicone rubber and glass, as shown in Fig. 5. These parameters are optimised in order to
achieve the lowest leakage current that flows through the arrester. These parameters were
selected because they have strong influence on the leakage current in the surge arrester. The size
of the design can be changed while for the material permittivity, there is variation in the
permittivity value of the glass and silicone rubber available in the market. The objective function
of the problem is defined as the minimum leakage current that flows through the arrester. The
optimized leakage current through the surge arrester must be less than the leakage current
through the original surge arrester design, which is 408µA.

3.1 Imperialist competitive algorithm (ICA)

ICA is a method recently used to deal with different optimization problems. This algorithm was
announced since 1989 by Baran (Baran & Wu, 1989). It is inspired based on the imperialist
competition. The empires comprise of imperialists and colonies. The imperialists are chosen from

6
 the best countries in the empires and the remaining forms the colonies of those imperialists. The
steps involved in ICA in reducing the leakage current in the surge arrester are as follows:
Step 1: Initial empires generation
The first step is to form an empire, which is determined by the parameters of the surge arrester to
be optimized. The number of parameters of the surge arrester is equal to the number of cells in
each country’s array. The value of every cell is 0 or 1, depending on the width or length of the
arresters. The array for n parameters to be optimized c is formed, which is written as
Country = [c1, c2 c3, ….., cn] (7)
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

The parameters with the best objective function are chosen as the imperialists and the rest of the
parameters are colonies.
Step 2: Assimilation and revolution
Assimilation is when the imperialists try to move forward to their colonies and make them a part
of themselves. This step is applied to the parameters of the arrester in the colonies’ array. The
colony moves towards the imperialist by
x ~ U (0, β x d ) (8)
where d is the distance between the imperialist and colony and β is a constant larger than one.
Step 3: Position exchange of imperialist and colony
If a colony reaches a position with more powerful than an imperialist, the colony and the
imperialist exchange their positions. In this step, the colony and imperialist will exchange their
position if it reaches a lower leakage current value.
Step 4: Empire total cost calculation
The cost of imperialist country affects the total cost (TC) of an empire where in other words, the
value of the leakage current will be computed for all parameters using
TCn=Ɛmean[cost(empire coloniesn)]+Cost(imperialistn) (9)
Step 5: Competition of imperialistic
The power of weaker empires decreases while the power of powerful empires becomes greater. In
the weakest empire, the weakest colony, which has the highest leakage current, is sent to other
empires. The normalized total cost (NTC) of each empire is determined by
NTCn = TC.n =max[TC.i] (10)
Step 6: Convergence check

7
 Steps 2 to 5 are repeated until all of the empires collapse except the most powerful empire, which
has the lowest value of the leakage current.

3.2Gravitational search algorithm (GSA)

GSA is an optimisation technique motivated by the Newton law of motion and gravity. E.
Rashedi first introduced this algorithm in 2009 (Esmat Rashedi, Hossein Nezamabadi-pour *, &
Saryazdi, 2009). In GSA, the performance of the objects, called as agents, are measured by their
masses. The solution of the problem is the position of the mass and the objective function is used
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

to specify its gravitational and inertial masses. The steps involved in GSA in optimising the surge
arrester design in this work are as follows:
Step 1: Initializing agents
The first step is agent initialization, which determines the upper and lower limit of the
independent variables. The parameters of surge arrester that will be optimised are called as
agents. The positions of N parameters to be optimised are initialized, where xid represents the
positions of ith agent in dth dimension and n is the space dimension, where
Xi = (xi1,..,xid,…,xin), where i is 1 to N (11)
Step 2: Evolution of fitness and best fitness computation
The evolution of the fitness is performed to minimize or maximize the dimensions of the
parameters by determining the worst and best value of the leakage current for each combination
of the parameters.
Step 3: Computation of constant of gravitational (G)
G is determined by
G(t) = G0(-αt/T) (12)
where G0 = 100 and α = 10 control the search accuracy, t is iteration and T is the total iteration.
Step 4: Calculation of masses of the agents
Gravitational and inertia masses for every agent are determined at each iteration. Mii is the inertia
mass of ith agent and Mpi and Mai are the passive and active gravitational masses, where
Mai = Mpi = Mii = Mi, where i is 1 to N (13)
In this step, every iteration will decide the position and dimension of the parameters.
Step 5: Calculation of agents’ accelerations
At t-th iteration, the acceleration of ith agent aid is computed by

8
 aid (t) = Fid (t) / Mii (t) (14)
where Fid(t) is the total force on ith agent.
Step 6: Position and velocity of agents
In this step, the next position of each parameter is updated. The position and velocity of the
agents at t+1 iteration are calculated by
xid (t+1) = xid (t) + vid (t+1) (15)
vid (t+1) = randi vid (t) + aid (t) (16)
d
where t is the current iteration, t+1 is the next iteration, vi is the i agent velocity, xi is the ith
th d

agent position, randi is a random number between 0 and 1 and aid is the ith agent acceleration.
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

Step 7: Convergence check

Steps 2 to 6 are repeated until the final iteration is reached. The lowest leakage current value at
the final iteration is taken as the global fitness while the parameters’ position with the lowest
leakage current is taken as the global solution of the problem.

4.0 Measurement and simulation results

Table 4 shows the simulation and measurement results of the r.m.s. (roots mean square) value
of the leakage current in the surge arrester under different applied voltage amplitudes. It is
noticed that the leakage current increases with the applied voltage from both measurement and
simulation results. The average error between the measurement results and simulation using FEA
model is considerably small, which is 2.74%. Therefore, the proposed model of the surge arrester
in this work under various applied voltage magnitudes can be considered reasonable.
Table 5 shows the measurement and simulation results of the r.m.s. leakage current in the surge
arrester under different temperatures at 10kV applied voltage. From this table, it is noticed that
the leakage current increases with the temperature of the arrester (Cheng, Mei, Wang, Guan, &
Zhang, 2016; Lee, Lee, & Lee, 2010; S. M. Seyyedbarzegar & Mirzaie, 2016; Seyyed Meysam
Seyyedbarzegar & Mirzaie, 2015; Sp et al., 2016). The average error between the measurement
and simulation results from the FEA model is 4.78%, which is considerably small. Therefore, the
proposed model of the surge arrester in this work under different material temperature can also be
considered reasonable.
Fig. 6 shows the leakage current signals obtained from the measurement and FEA model at 8
kV rms AC sinusoidal voltage at 26oC and at 10 kV rms AC sinusoidal voltage at 40oC

9
 respectively. From Fig. 6a, the rms leakage current magnitude obtained from the measurement
and FEA model is 338µA and 333µA respectively. From Fig. 6b, the rms leakage current
magnitude is 438µA from the measurement and 419µA from the FEA model. Comparison of the
leakage current signals between the measurement and FEA model shows that both leakage
current signals are within reasonable agreement with each other. Thus, this shows that the
proposed FEA model of ZnO surge arrester is practical since the leakage current signals and
magnitudes agree well with the measurement results.

5.0 Optimisation results

Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

In this work, designing a surge arrester with reduced leakage current was proposed using
optimisation methods. The results obtained from the FEA model of the surge arrester were
interfaced with MATLAB programming code to evaluate the objective function. The results from
imperialist competitive algorithm (ICA) and gravitational search algorithm (GSA) were
compared and analysed. Optimisation techniques were used to obtain a surge arrester design
which yields the lowest leakage current through the arrester. The fitness function of the problem
is defined as the minimum leakage current across the surge arrester.
Two case studies of the optimisation were performed by using four variables. For Case 1, the
parameters to be optimized are the permittivity of the material and the width of the ground
terminal of the arrester. For Case 2, the width of the ground terminal of the surge arrester, the
thickness of the silicone rubber and curve of housing sheds were optimized. The descriptions of
the case studies are summarised in Table 6.

5.1 Optimised value of leakage current

The lower and upper limits of each parameter used in optimisation were fixed based on the
limits relative to the dimension and material properties of the surge arrester as shown in Table 7.
The starting values in the optimization methods are chosen based on the actual surge arrester
dimensions and must fulfil the IEC60815 standard, where the base of the surge arrester should
have a minimum cross section of 35mm2 and the arrester housing should have a suitable creepage
distance. The initial value of the glass permittivity and silicone rubber permittivity are 4.2 and
11.7 respectively. The width of the grounded aluminium block and base are 86mm and 70mm.

10
 The thickness of the silicone rubber is 3mm and the curve of the housing sheds is 0.4mm. The
number of iterations used is 150.
Fig. 7 and Fig. 8 show the parameters obtained using GSA and ICA optimisation for two case
studies. For Case 1, it shows that the permittivity value of glass for GSA is higher than ICA and
its initial value. The difference between GSA and ICA is around 3.0003. The permittivity of
silicone rubber for both GSA and ICA are higher than initial value. And the width of the ground
terminal for GSA is lower than ICA for both block and base. For Case 2, the thickness of the
silicon rubber of 3mm increase up to 4.924mm and 4.998mm for GSA and ICA. The curve of
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

housing sheds for both algorithm also increase double from the initial value. The value of
aluminium block for ICA is higher than GSA and the value of aluminium base for GSA is lower
than the initial value and ICA.
The minimized leakage current using GSA and ICA for all case studies is shown in Fig. 9. For
Case 1, which is varying the permittivity of the materials and arrester dimension, it yields the
leakage current value of 383.316µA from GSA and 382.927µA from ICA which is lower than the
leakage current from the original arrester design. This shows a reduction of 6.05% and 6.15% in
the leakage current obtained by GSA and ICA respectively compared to the original arrester
design. In Case 2, which is varying only dimensions of the arrester, the leakage current obtained
by GSA is higher than ICA, which is 375.34 µA for GSA and 374.073 µA for ICA. This shows a
reduction of 8.00% and 8.32% in the leakage current obtained by GSA and ICA respectively
compared to the original arrester design. From both cases, it shows that the value of the leakage
current obtained using ICA is slightly lower than GSA. The convergence curves of ICA and GSA
for Case 1 and Case 2 are shown in Fig. 10a and 10b. From both cases, it shows that GSA
converges slightly faster than ICA.
Table 8 shows the calculated value of the active power for the original design and designs
optimised by ICA and GSA. The active power for both cases was calculated based on the
obtained resistive leakage current. The results show that the active power is lower in the
optimised arrester design than the original design due to lower resistive leakage current.

5.2 Model of optimised design under different applied conditions

From the results shown in Fig. 9, ICA in Case 2 shows the best result in reducing the leakage
current. Hence a new FEA model was developed according to the parameters obtained using ICA

11
 from Case 2. The thickness of the silicone rubber was set to 4.998mm and the curve of housing
shed was set to 0.8mm. The ground aluminium block and base were fixed to 74.56mm and
86.716mm respectively. Since surge arresters are mainly affected by the applied voltage and
temperature, the new 11kV surge arrester model was simulated under various applied voltage
amplitudes and operating temperatures to compare the results between the original and optimised
designs of 11kV surge arrester.
The voltage was applied between 6kV and 10kV. Table 9 and Fig. 11 show the simulation
results of the original and optimised designs of the surge arrester under different applied voltage
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

amplitudes is between 2.48% and 9.80%. Negative values show that the leakage current is
reduced in the new design of the surge arrester compared to the original surge arrester design
from 6kV to 10kV.
Table 10 and Fig. 12 show the simulation results of the original and optimised 11kV surge
arrester designs under different temperatures at 10kV applied voltage. It can be observed that the
optimised design of the arrester has lower leakage current compared to the original surge arrester
design under different temperatures is between 2.32% and 3.52%. Therefore, this shows that the
surge arrester design has been successfully improved using gravitational search algorithm (GSA)
and imperialist competitive algorithm (ICA).

6.0 Conclusions
In this work, the effect of different applied voltages and temperatures on the leakage current
within an 11kV ZnO surge arrester has been successfully studied. It was observed that the
leakage current is significantly influenced by the applied voltage and temperature. The leakage
current results were also successfully reproduced through modelling of a three-dimensional
model of 11kV surge arrester in a finite element analysis (FEA) method. The FEA solves for the
current density and temperature distributions in the model.
A surge arrester design has been successfully improved using two optimisation methods, which
are gravitational search algorithm (GSA) and imperialist competitive algorithm (ICA). The
parameters being optimized are the glass permittivity, silicone rubber permittivity and the width
of the ground terminal of the surge arrester. It was found that the surge arrester design obtained
using ICA has lower leakage current than GSA and the original design of the surge arrester under
different applied voltages and temperatures. Therefore, ICA can be considered a preferred

12
 optimisation method over GSA in designing surge arresters with lower leakage current.
Consequently, this can lead to the improvement of surge arrester designs.
From the obtained results, it was shown that the leakage current could be reduced by only a few
percent. However, a small change could improve the life of the surge arrester slightly but
experiments will be required to assess the actual improvement. Other ways of reducing leakage
current in the future are by improving the condition monitoring methods and reduce pollution on
surge arresters.
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

References
Aghaebrahimi, M. R., Shariatinasab, R., & Ghayedi, M. (2012, 25-28 March 2012). Optimal design of grading ring
of surge arresters due to electric field distribution. Paper presented at the 2012 16th IEEE Mediterranean
Electrotechnical Conference.
Alotto, Piergiorgio, dos Santos Coelho, Leandro, C. Mariani, Viviana, & da C. Oliveira, Camila. (2014). A modified
lambda algorithm for optimization in electromagnetics. COMPEL - The international journal for
computation and mathematics in electrical and electronic engineering, 33(3), 759-767. doi:
doi:10.1108/COMPEL-10-2012-0224
Banik, A., Dalai, S., & Chatterjee, B. (2016). Autocorrelation aided rough set based contamination level prediction
of high voltage insulator at different environmental condition. IEEE Transactions on Dielectrics and
Electrical Insulation, 23(5), 2883-2891. doi: 10.1109/TDEI.2016.7736849
Baran, Mesut E, & Wu, Felix F. (1989). Network reconfiguration in distribution systems for loss reduction and load
balancing. Power Delivery, IEEE Transactions on, 4(2), 1401-1407.
Baraston, Arnaud, Gerbaud, Laurent, Reinbold , Vincent, Boussey , Thomas, & Wurtz, Frédéric. (2016).
Multiphysical approach including equivalent circuit models for the sizing by optimization. COMPEL - The
international journal for computation and mathematics in electrical and electronic engineering, 35(3), 871-
884. doi: doi:10.1108/COMPEL-11-2015-0418
Caner, Murat, Gerada, Chris, Asher, Greg, & Özer, Tolga. (2016). Design optimization of Halbach array permanent
magnet motor to achieve sensorless performance using genetic algorithm. COMPEL - The international
journal for computation and mathematics in electrical and electronic engineering, 35(5), 1741-1759. doi:
doi:10.1108/COMPEL-06-2015-0218
Cheng, L., Mei, H., Wang, L., Guan, Z., & Zhang, F. (2016). Research on aging evaluation and remaining lifespan
prediction of composite insulators in high temperature and humidity regions. IEEE Transactions on
Dielectrics and Electrical Insulation, 23(5), 2850-2857. doi: 10.1109/TDEI.2016.7736845
Dong, B., Jiang, X., Hu, J., Shu, L., & Sun, C. (2012). Effects of artificial polluting methods on AC flashover
voltage of composite insulators. IEEE Transactions on Dielectrics and Electrical Insulation, 19(2), 714-
722. doi: 10.1109/TDEI.2012.6180267
Esmat Rashedi, Hossein Nezamabadi-pour *, & Saryazdi, Saeid. (2009). GSA: A Gravitational Search Algorithm.
Information Sciences(179), 2232–2248. doi: 10.1016/j.ins.2009.03.004
Golak, Slawomir, & Kordos, Mirosław. (2016). Shaping inductor geometry for casting functionally graded
composites. COMPEL - The international journal for computation and mathematics in electrical and
electronic engineering, 35(1), 16-26. doi: doi:10.1108/COMPEL-10-2014-0288
Illias, H., Mohsin Ali, Tunio, Bakar, A. H. A., & Mokhlis, H. (2012, 2-5 Dec. 2012). Distribution of electric field in
capacitor and surge arrester bushings. Paper presented at the 2012 IEEE International Conference on
Power and Energy (PECon).
Jinliang, He, Shanqiang, Gu, Bo, Zhang, & Jun, Hu. (2006, 0-0 0). Analysis and Improvement of Potential
Distribution of 1000. Paper presented at the 2006 12th Biennial IEEE Conference on Electromagnetic Field
Computation.
Jinliang, He, Shuiming, Chen, Rong, Zeng, Jun, Hu, & Chenggang, Deng. (2006). Development of polymeric surge
ZnO arresters for 500-kV compact transmission line. Power Delivery, IEEE Transactions on, 21(1), 113-
120. doi: 10.1109/TPWRD.2005.860234

13
 Khodsuz, M., & Mirzaie, M. (2015). Evaluation of ultraviolet ageing, pollution and varistor degradation effects on
harmonic contents of surge arrester leakage current. IET Science, Measurement & Technology, 9(8), 979-
986. doi: 10.1049/iet-smt.2014.0372
Khodsuz, Masume, & Mirzaie, Mohammad. (2016). An improved time-delay addition method for MOSA resistive
leakage current extraction under applied harmonic voltage. Measurement, 77, 327-334. doi:
http://dx.doi.org/10.1016/j.measurement.2015.09.027
Khodsuz, Masume, Mirzaie, Mohammad, & Seyyedbarzegar, Seyyedmeysam. (2015). Metal oxide surge arrester
condition monitoring based on analysis of leakage current components. International Journal of Electrical
Power & Energy Systems, 66, 188-193. doi: http://dx.doi.org/10.1016/j.ijepes.2014.10.052
Lee, Su-Bong, Lee, Seung-Ju, & Lee, Bok-Hee. (2010). Analysis of thermal and electrical properties of ZnO arrester
block. Current Applied Physics, 10(1), 176-180. doi: http://dx.doi.org/10.1016/j.cap.2009.04.019
Luca Amicucci, Giovanni, & Mazzetti, Carlo. (2004). Probabilistic method for reliability assessment of metal oxide
varistors under lightning stress. COMPEL - The international journal for computation and mathematics in
electrical and electronic engineering, 23(1), 263-276. doi: doi:10.1108/03321640410509237
Naidu, M, & Kamaraju, S. (2006). High Voltage Engineering (Third ed.). Singapore: Mc Graw Hill.
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

Napieralska-Juszczak, Ewa, Napieralski, Piotr, Komeza, Krzysztof, & Zeroukhi, Youcef. (2016). The influence of
localized mechanical stresses on the non-homogeneous distribution of electrical conductivity. COMPEL -
The international journal for computation and mathematics in electrical and electronic engineering, 35(6),
1983-1995. doi: doi:10.1108/COMPEL-02-2016-0060
Seyyedbarzegar, S. M., & Mirzaie, M. (2016). Thermal balance diagram modelling of surge arrester for thermal
stability analysis considering ZnO varistor degradation effect. IET Generation, Transmission &
Distribution, 10(7), 1570-1581. doi: 10.1049/iet-gtd.2015.0728
Seyyedbarzegar, Seyyed Meysam, & Mirzaie, Mohammad. (2015). Heat transfer analysis of metal oxide surge
arrester under power frequency applied voltage. Energy, 93, Part 1, 141-153. doi:
http://dx.doi.org/10.1016/j.energy.2015.09.031
Sjöstedt, H., Stenström, L., Pusch, D., & Ostrowski, J. (2010, 11-14 Oct. 2010). Voltage grading design of UHV
surge arresters using 3D transient capacitive-resistive field simulations. Paper presented at the 2010
International Conference on High Voltage Engineering and Application.
Song, W., Shen, W. W., Zhang, G. J., Song, B. P., Lang, Y., Su, G. Q., . . . Deng, J. B. (2015). Aging
characterization of high temperature vulcanized silicone rubber housing material used for outdoor
insulation. IEEE Transactions on Dielectrics and Electrical Insulation, 22(2), 961-969. doi:
10.1109/TDEI.2015.7076797
Y. Späck-Leigsnering, E. Gjonaj, H.DeGersem, T.Weiland, M.Gießel, and V. Hinrichsen (2016). Electroquasistatic-
Thermal Modeling and Simulation of Station Class Surge Arresters. IEEE Transactions on Magnetics,
52(3), 1-4. doi: 10.1109/TMAG.2015.2490547
Valsalal, P., Usa, S., & Udayakumar, K. (2011). Modelling of metal oxide arrester for very fast transients. Science,
Measurement & Technology, IET, 5(4), 140-146. doi: 10.1049/iet-smt.2010.0129
Zeinoddini-Meymand, H., Vahidi, B., Naghizadeh, R. A., & Moghimi-Haji, M. (2013). Optimal Surge Arrester
Parameter Estimation Using a PSO-Based Multiobjective Approach. IEEE Transactions on Power Delivery,
28(3), 1758-1769. doi: 10.1109/TPWRD.2013.2257880

14
 Aluminium cap
(Applied voltage)
Aluminium block
(Applied voltage)
Silicone rubber
housing
Zinc Oxide (ZnO)
block

Glass
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

Aluminium
block (Ground)
Aluminium
base (Ground)

Figure 1. Surge arrester model geometry

Surge arrester

Air

Figure 2. Three-dimensional model geometry of the surge arrester

Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

Figure 3. Mesh elements of the model

Rs

Surge
arrester
C1

AC Oscilloscope
Rsh
Tx

Figure 4. Setup for leakage current measurement

Silicon rubber
permittivity

Housing sheds curve

Glass permittivity

Silicone thickness
Ground terminal (block)
Ground terminal (base)

Figure 5. Optimised parameters of 11kV surge arrester

 15 Applied voltage 1000
FEA model 800
10 Measurement
600

Leakage current (uA)

Applied voltage (kV)

400
5
200

0 0

-200
-5
-400
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

-600
-10
-800

-15 -1000
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Time (s)

(a)

-15 Applied voltage 1000

FEA model
Measurement
-10

Leakage current (uA)

500
Applied voltage (kV)

-5

0 0

5
-500
10

15 -1000
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Time (s)

(b)

Figure 6. Leakage current signals from measurement and FEA model at (a) 50 Hz, 8 kV

r.m.s., 26oC and (b) 50 Hz, 10kV r.m.s., 40oC

 100
86 90
90 75.975 85.708
80 76
70
70
60
Value

50
40
30 12.319
20 7.1873 11.7 12.303
10 4.2 4.187
0
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

Permittivity of Permittivity of Width of Width of

glass silicone rubber aluminium block aluminium base
(ground) (ground)
Initial GSA ICA

Figure 7. Parameters obtained using GSA and ICA for Case 1

100
86 86.716
90 74.56 83.279
80 74.424
70
70
60
Value

50
40
30
20 4.998 0.8
10 4.924 0.7996
3 0.4
0
Thickness of Curve of housing Width of Width of
silicone rubber sheds aluminium block aluminium base
(ground) (ground)
Initial GSA ICA

Figure 8. Parameters obtained using GSA and ICA for Case 2

 420

r.m.s. leakage current (µA)

408 408
410
400
383.316
390 382.927
375.34
380 374.073
370
360
350
Case 1 Case 2
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

Initial GSA ICA

Figure 9. r.m.s. value of leakage current obtained using GSA and ICA

410
ICA
405 GSA
Leakage current (uA)

400

395

390

385

380
0 50 100 150
Iteration
(a)
 410
ICA
GSA
Leakage current (uA)

400

390

380
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

370
0 50 100 150
Iteration
(b)
Figure 10. Convergence curve of GSA and ICA for (a) Case 1 and (b) Case 2

450

400
Leakage current (uA)

350

300

250
Original design
Optimised design
200
6 6.5 7 7.5 8 8.5 9 9.5 10
Applied voltage (kV)

Figure 11. Leakage current under different applied voltages from the FEA model
 440
Leakage current (uA)

430

420

410

400 Original design

Optimised design
390
25 30 35 40 45 50 55 60 65
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

o
Temperature ( C)

Figure 12. Leakage current under different temperatures from the FEA model
 Table 1. Dimensions and parameters of the surge arrester
Characteristics Data
Housing material Silicon rubber
Height 227 mm
Nominal discharge current 8/20 µs 10 kA peak
Number of sheds 4
Maximum continuous operating voltage 10 kVrms
Rated voltage 12.5 kVrms
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

Table 2. Properties assigned on each material

Material Relative Electrical Conductivity, σ
Permittivity, εr (Sm-1)
From the measured V-I characteristics
Zinc Oxide 2250
of ZnO surge arrester
Silicone 11.7 1x10-12
Glass 4.2 1x10-14
Aluminium 1 3.77x107
Air 1 0

Table 3. Boundary settings

Boundary condition Expression
Applied voltage V=Vapp*sin(2*pi*50*t)
Ground V=0
All interior boundaries n.(J1 – J2) = 0
 Table 4. Leakage current from measurement and simulation under different applied voltages
r.m.s. value of leakage current (µA) Error of
Applied
FEA model Measurement total current
voltage (kV)
Icapacitive Iresistive Icapacitive Iresistive (%)
6.0 225 16 232 18 3.60
6.5 243 19 259 21 6.43
7.0 277 20 278 23 1.33
7.5 297 22 298 24 0.93
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

8.0 310 23 312 26 1.48

8.5 323 24 324 29 1.70
9.0 349 29 349 33 3.66
9.5 378 30 375 35 3.41
10.0 383 31 386 37 2.13

Table 5. Leakage current from measurement and simulation under different material
temperature
r.m.s. value of leakage current (µA)
Temperature of the Error of total
FEA model Measurement
test sample (oC) current (%)
Iconductive Iresistive Iconductive Iresistive
26 380 34 386 37 2.12
33 381 35 395 38 3.92
36 381 36 398 39 4.57
40 382 37 398 40 4.34
45 384 38 398 42 4.09
50 387 39 399 44 3.84
55 386 41 400 46 4.26
60 386 43 408 47 5.71
65 386 45 429 51 10.21
 Table 6. Description of the case studies
Case Parameter: Description
Permittivity value of glass
Permittivity value of silicone rubber
Case 1
Width of aluminium block (ground)
Width of aluminium base (ground)
Thickness of silicone rubber
Curve of housing sheds
Case 2
Width of aluminium block (ground)
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

Width of aluminium base (ground)

Table 7. Lower and upper limits of each parameter used in optimisation

Lower Upper
Parameter Initial
limit limit
Permittivity of glass 3.7 8 4.2
Permittivity of silicone rubber 11.0 12.9 11.7
Thickness of silicone rubber 1 6 3
Curve of housing sheds 0.05 0.8 0.4
Width of aluminum block (ground) 68 76 70
Width of aluminum base (ground) 80 90 86

Table 8. Calculation of active power due to leakage current

Original Optimised using GSA Optimised using ICA
Case
Iresistive (uA) P (W) Iresistive (uA) P (W) Iresistive (uA) P (W)
Case 1 30 0.285 24 0.228 23 0.219
Case 2 30 0.285 18 0.171 16 0.152
 Table 9. Comparison between the original design and optimised design using ICA under
different applied voltages
r.m.s. value of leakage
Difference between
Applied current (µA) Difference
original and
voltage (kV) Original Optimised (%)
optimised design (µA)
design design
6.0 241 235 -6 2.48
6.5 262 250 -12 4.58
7.0 297 277 -20 6.73
Downloaded by UNIVERSITY OF ADELAIDE At 06:15 21 November 2017 (PT)

7.5 319 301 -18 5.64

8.0 333 313 -20 6.01
8.5 347 327 -20 5.76
9.0 378 349 -19 7.67
9.5 408 368 -40 9.80
10.0 414 375 -39 9.42

Table 10. Comparison between the original design and optimised design using ICA under
different temperatures
r.m.s. value of leakage
Difference between
Temperature current (µA) Difference
original and
(oC) Original Optimised (%)
optimised design (µA)
design design
26 414 402 -12 2.89
33 416 404 -12 2.88
36 417 405 -12 2.87
40 419 409 -10 2.38
45 422 410 -12 2.84
50 426 411 -15 3.52
55 427 415 -12 2.81
60 429 419 -10 2.33
65 431 421 -10 2.32