energies
Article
A Simplified Model of a Surge Arrester and Its Application in
Residual Voltage Tests
Peerawut Yutthagowith 1, *, Sutee Leejongpermpoon 1 and Nawakun Triruttanapiruk 2






Citation: Yutthagowith, P.;
Leejongpermpoon, S.;
Triruttanapiruk, N. A Simplified
Model of a Surge Arrester and Its
Application in Residual Voltage Tests.
Energies 2021, 14, 3132.
https://doi.org/10.3390/en14113132
Academic Editor: Christos
A. Christodoulou
Received: 10 March 2021
Accepted: 11 May 2021
Published: 27 May 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Energies 2021, 14, 3132. https://doi.org/10.3390/en14113132
1 School of Engineering, King Mongkut’s Institute of Technology Ladkrabang, Bangkok 10520, Thailand;
sutee509@hotmail.com
2 Department of Electrical Engineering, Chubu University, 1200 Matsumoto-cho, Kasugai 487-8501, Japan;
nawakunt@gmail.com
* Correspondence: peerawut.yu@kmitl.ac.th; Tel.: +66-81-926-0141
Abstract: A simplified and accurate model of a surge arrester used in the residual voltage test is
proposed in this paper. With the help of a genetic algorithm, the measured impulse current and
residual voltage waveforms are utilized to determine circuit parameters of the proposed model and
the generation circuit precisely. The technique starts from the circuit parameter determination using
the preliminary experimental data with a lower current peak than that specified by the standard.
From the determined model and with the help of the genetic algorithm, the circuit parameters and
the charging voltage to obtain the specified current peak and the residual voltage can be estimated
accurately. The validity of the proposed technique has been verified by experiments for the estimation
of the appropriate current circuit parameters, the charging voltage, and the residual voltage. In
addition, the application of the proposed model in the residual voltage tests is presented. From
comparison of simulated and experimental results with the determined parameters, the impulse
current and residual voltage waveforms are determined precisely. It is confirmed that the proposed
model and technique are attractive in the appropriate circuit parameter determination and the
residual voltage estimation in the residual voltage tests of surge arresters. The proposed method also
provides a good advantage for reduction of the number of trial and error experiments for obtaining
the current waveform according to the standard requirement. Moreover, the unintentional damages
of the arrester during the process of the waveform adjustment will be reduced significantly.
Keywords: circuit parameter determination; impulse currents; surge arresters; residual voltage test;
simplified arrester model
1. Introduction
A surge arrester is a device widely used to protect electrical equipment from over-
voltages due to lightning and switching operation in electrical power and telecommuni-
cation systems [1]. The arrester exhibiting dynamic and nonlinear characteristics diverts
current to the ground to clamp the over-voltage to be lower than the insulation level of the
protected high voltage equipment. To confirm the performance in over-voltage mitigation,
the international standards [2–5] advise performing residual voltage tests on the arrester.
According to the standard [3] and in the residual voltage tests, four kinds of impulse
current waveforms, i.e., steep impulse current (1/20 µs), high impulse current (4/10 µs),
lightning impulse current (8/20 µs), and switching impulse current (30/80 µs), in different
current peaks are injected to the arrester during tests. The residual voltages across the
arrester associated with the different impulse current waveforms and peaks are examined.
The impulse currents can be generated by a simple resistance, inductance, and capacitance
circuit as illustrated in Figure 1. The current generation circuit is composed of the DC
voltage source (Vs ), the charging current limiting resistor (RL ), the charging capacitor
(C), the spark gap (G), the series inductor (Le ), the series resistor (Re ), the shunt current
measuring resistor (RSH ), and the surge arrester (SA).
https://www.mdpi.com/journal/energies
Energies 2021, 14, x FOR PEER REVIEW 2 of 1

Energies 2021, 14, 3132 capacitor (C), the spark gap (G), the series inductor (Le), the series resistor (R ),13the shun
2 eof
current measuring resistor (RSH), and the surge arrester (SA).

SA

Figure 1. The impulse current generation circuit.

Figure 1. The impulse current generation circuit.
The most essential problem in residual voltage tests is that it is quite laborious for test
The to
engineers most essential
generate problem
an accurate in residual
waveform and a voltage
current peaktestsaccording
is that ittoisthe
quite laborious fo
standard
test engineers
requirement, to generate
because an accurate
the uncontrolled waveform
parasitic impedanceand a current
in the currentgeneration
peak according
circuit to the
and the nonlinear
standard characteristic
requirement, because of the arrester affect the
uncontrolled currentimpedance
parasitic waveform significantly.
in the current gen
Therefore, an accurate
eration circuit and the arrester model
nonlinear is necessary to
characteristic ofuse
theinarrester
controlling thethe
affect current wave-
current waveform
form. There are a lot of researchers developing the accurate model
significantly. Therefore, an accurate arrester model is necessary to use in controlling the used for designing
insulation coordination and for obtaining successful high-voltage testing [6–14]. Further-
current waveform. There are a lot of researchers developing the accurate model used fo
more, the effect of thermal stress under the arrester operation and experimental tests can
designing insulation coordination and for obtaining successful high-voltage testing [6–
be included in the model successfully [15]. However, the generation circuit and parasitic
14]. Furthermore,
impedance are neglected theineffect of thermal
the process stress underIt leads
of model construction. the to
arrester
the modelsoperation
not and
experimental tests can be included in the model successfully
being appropriate in the residual voltage test in real practice. Therefore, trial and error [15]. However, the
generationforcircuit
approaches selecting andthe parasitic impedance
circuit parameters and the arecharging
neglected in are
voltage thestillprocess
necessary of mode
to employ in theItreal
construction. experiment.
leads Several not
to the models experiments for waveforminadjustment
being appropriate the residual are carried
voltage test in
out
realtopractice.
obtain the requiredtrial
Therefore, waveform
and error andapproaches
it sometimesfor unintentionally
selecting the causes
circuit damage
parameters and
on the arrester during the test. It will be advantageous for
the charging voltage are still necessary to employ in the real experiment. Severatest engineers if an effective
technique for selecting the circuit parameters and the charging voltage is developed.
experiments for waveform adjustment are carried out to obtain the required waveform
This paper aims to propose a simplified and accurate model of a surge arrester and an
and it sometimes unintentionally causes damage on the arrester during the test. It will be
effective technique for selecting the circuit parameters and the charging voltage to obtain
advantageous
the specified impulsefor test
currentengineers
waveformifinan effective
residual voltagetechnique for selecting
tests. Moreover, the residualthe circui
parameters
voltage and the
associated thecharging voltagewaveform
impulse current is developed. is estimated precisely. With the help
This paper
of a genetic aimsthe
algorithm, to unknown
propose aparasitic
simplified and accurate
impedance model of and
was determined, a surge arrester and
the other
an effective technique for selecting the circuit parameters and the chargingthe
appropriate circuit parameters and the charging voltage can be selected. The validity of voltage to
proposed
obtain the technique
specified wasimpulse
verified current
by the experiments
waveformfor indetermination
residual voltage of thetests.
appropriate
Moreover, the
current
residual circuit parameters,
voltage associated thethe
charging
impulse voltage,
currentandwaveform
the residualisvoltage associated
estimated withWith the
precisely.
the specified current waveform. From comparison of the experimental and simulated
help of a genetic algorithm, the unknown parasitic impedance was determined, and the
results, it was confirmed that the method is fairly highly accurate and very effective in
other appropriate circuit parameters and the charging voltage can be selected. The validity
the residual voltage tests of surge arresters. From this achievement, the proposed method
ofalso
is thevery
proposed
useful for technique was verified
the significant reductionby the experiments
of unintentional for determination
damages to the arrester of the
appropriate
during current
the process circuit parameters,
of waveform adjustment. the charging voltage, and the residual voltage
associated with the specified current waveform. From comparison of the experimenta
2.
andConventional Surge Arrester
simulated results, Model
it was confirmed that the method is fairly highly accurate and very
There in
effective are the
many proposedvoltage
residual models of surgeofarresters
tests surge but most of them
arresters. Fromarethis
not applicable
achievement, the
in
proposed method is also very useful for the significant reduction not
residual voltage tests since the parasitic impedance and generation circuit are taken
of unintentiona
into account in the process of model construction. For the sake of clarity, an experiment of
damages to the arrester during the process of waveform adjustment.
the residual voltage test on a surge arrester with the rated voltage of 9 kV and the nominal
discharge current of 10 kA was considered as a test case. As shown in Figure 1, the setting
2. Conventional Surge Arrester Model
circuit parameters in the experiment were composed of the charging capacitance of 2 µF,
There
the series are many
resistance of 0.1 Ω, themodels
(Re ) proposed of surge (L
series inductance arresters butand
e ) of 25 µH, most the of them are no
charging
voltage (Vch )in
applicable of residual
69 kV. Thevoltage
IEEE model
tests[6] as illustrated
since in Figure
the parasitic 2 was utilized
impedance and in a circuit circui
generation
simulator to calculate
are not taken the impulse
into account in thecurrent
processand residualconstruction.
of model voltage waveforms.
For theThe sakecircuit
of clarity, an
parameters of the model were composed of two series
experiment of the residual voltage test on a surge arrester with resistors (R and R
0 the rated
1 ), two series of 9 kV
voltage
inductors (L0 and L1 ), two nonlinear resistors (A0 and A1 ), and a parallel capacitor (C0 ).
and the nominal discharge current of 10 kA was considered as a test case. As shown in
With the help of a genetic algorithm as described in Section 4, those circuit parameters
Figure 1, the setting circuit parameters in the experiment were composed of the charging
of the model were determined by matching the residual voltage from simulation with
capacitance
that from the of 2 µF, the As
experiment. series resistance
shown in Figures (R3e) and
of 0.1 Ω, the
4, the series current
calculated inductance (Le) of 25 µH
waveform
and the charging voltage (Vch) of 69 kV. The IEEE model [6] as illustrated in Figure 2 wa
utilized in a circuit simulator to calculate the impulse current and residual voltage
 capacitor (C0). With the help of a genetic algorithm as described in Section
waveforms.
waveforms.
parameters Theof
The circuit
circuit parameters
theparameters
model were of the
of the model
model were composed
were
determined composed of two
of
by matching two series
seriesthe resistors
resistors (R00
(R
residua
and RR11),), two
two series
series inductors
inductors (L(L00 and
and LL11),), two
two nonlinear
nonlinear resistors
resistors (A(A00 and
and A A11),), and
and aa parallel
and
simulation with that from the experiment. As shown in Figures 3parallel
and 4
Energies 2021, 14, 3132
capacitor (C
capacitor (C00).). With
With the the help
help of
of aa genetic
genetic algorithm
algorithm as as described
described in in Section
Section 4, 4, those
those
3 of 13
circuit
circuit
current waveform
parameters
parameters of the
of the model
modelagreed
werewell
were with the
determined
determined by experimental
by matching the
matching result. The
the residual
residual voltage
voltage calculate
from
from
deviatedwith
simulation
simulation fromthat
with thefrom
that experimental
from the experiment.
the experiment. oneAs byshown
As only 2%.
shown However,
in Figures
in Figures and the
33 and 4, the
4, calculated
the calculated r
calculated
deviated
current
current from agreed
waveform
waveform the experimental
agreed well with
well with the one. It was result.
the experimental
experimental foundThe
result. that
The the calculated
calculated
calculated peak voltage
peak tim
voltage
agreed
deviated
deviatedwell withthe
from
from theexperimental
the experimental one
experimental result.
one by
by The
only
only calculated
2%.
2%. peak voltage
However,
However, the
the deviatedresidual
calculated
calculated from thevoltage
residual voltage
the residual
experimental one by wasonlylonger that the
2%. However, thecalculated
experimental one, deviated
residual voltage and thefrom spurious
the os
deviated from
deviated from the the experimental
experimental one. one. It
It was
was found
found that
that the
the calculated
calculated time
time to to the
the peak
peak ofof
calculatedone.
experimental residual voltage
It was found occurred
that the calculated intime
thetolatethe time
peak of ortheinresidual
the non-conduc
was
the residual
the residual was was longer
longer that
that the
the experimental
experimental one, one, and
and the the spurious
spurious oscillation
oscillation of of the
the
longer that the
arrester.residual experimental one, and the spurious oscillation of the calculated residual
calculated
calculated residual voltage
voltage occurred
occurred in
in the
the late
late time
time or
or
voltage occurred in the late time or in the non-conductive state of the arrester. in
in the
the non-conductive
non-conductive state
state of
of the
the
arrester.
arrester.
L0 L1
L00
L L11
L

R0 R1
R00
R R11
A00
A A
C0
C CR
0 A11
A A1
00

Figure
Figure
Figure
Figure 2.IEEE
2.2. IEEE
IEEE model
model
2. IEEE
model of
of of
model aa surge
surge
a surge arrester.
aarrester.
ofarrester.
surge arrester.

Figure3.3.
Figure
Figure 3.Comparison
Comparison
Comparison of the
of of
thethe injected
injected
injected impulse
impulse
impulse current
current
current waveforms.
waveforms.
waveforms.

Figure 3. Comparison of the injected impulse current waveforms.

Experiment
Experiment
Simulation
Simulation
Voltage (kV)

Experiment
Simulation
Voltage (kV)

Time (µs)
Time (µs)

Figure4.4.
Figure
Figure 4.Comparison
Comparison
Comparison of the
of of
thethe residual
residual
residual voltage
voltage
voltage waveforms
waveforms
waveforms across
across
across the arrester.
the arrester.
the arrester.

3. Simplified and Accurate Model of a Surge Arrester

3. Simplified
3. Simplified and
and Accurate
Accurate Model
Model of
of aa Surge
Surge Arrester
Arrester
In the results presented in Section 2, the conventional arrester model provided inaccu-
In the
In the results
results presented
presented in in Section
Section 2, 2, the
the conventional
conventional arrester
arrester model
model provided
provided
rate results when they were compared to the experimental ones. In this section, a simplified
inaccurate
inaccurate results
results when
when they
they
and accurate model of a surge arrester were
were compared
compared
Time to
to the
the experimental
experimental ones.
ones. In
In
(µs) in Figure 5 is proposed to overcome the
as shown this
this section,
section, aa
simplified
simplified and
and accurate
accurate model
model of
of aa surge
surge arrester
arrester as
as shown
shown in
in
problems of the conventional model. The circuit parameters of the model were composed Figure
Figure 55 is
is proposed
proposed to
to
overcome
overcome
ofFigure the
4.the
the internal problems
problems
Comparison
resistor of
of
(R0 ) in the
the conventional
theconventional
ofparallel
residual
with the voltagemodel.
model.
internal The circuit
The circuit
waveforms
inductor parameters
parameters
across
(L0 ), the of the
of the model
the arrester.
nonlinear resistor model
were
were composed
composed of
of the
the internal
internal resistor
resistor (R
(R )
) in
in parallel
parallel with
with the
the
(A0 ), and the damping resistor (R2 ) in series with the internal capacitor (C2 ).
0
0 internal
internal inductor
inductor (L
(L ), the
00), the
3. Simplified and Accurate Model of a Surge Arrester
In the results presented in Section 2, the conventional arrester m
inaccurate results when they were compared to the experimental ones. I
simplified and accurate model of a surge arrester as shown in Figure 5
overcome the problems of the conventional model. The circuit paramete
 Energies 2021, 14, x FOR PEER REVIEW 4 of 13
Energies 2021, 14, x FOR PEER REVIEW
Energies 2021, 14, 3132 4 of 13

nonlinear resistor (A0), and the damping resistor (R2) in series with the internal capacitor
(C2).
L0

R0 R2
A0
C2

Figure 5. Simplified arrester model.

Simplified
Figure 5.5.
Figure Simplifiedarrester model. model.
arrester
As mentioned above, the generation circuit, the parasitic impedance, i.e., series in-
ductance As (mentioned
Lp) and seriesabove, the generation
resistance (Rp), and the circuit, the parasitic
nonlinear impedance,
characteristic i.e.,arrester
of a surge series induc-
tance As
(L
must be takenp ) mentioned
and series above,
resistance
into account in the real (R the
), and generation
the nonlinear circuit, the
characteristic parasitic
of a surge
p impulse current generation. The simplified equiva- impedance,
arrester must i .e
be taken into account in the real impulse current generation. The simplified equivalent
ductance
lent (Lp) and
circuit including theseries
parasitic impedance (and
resistance Rp)the
, and the nonlinear
proposed characteristic
arrester model is illus- of a sur
circuit
trated in including
Figure 6. In thereal
parasitic
practice,impedance andtothe
it is possible proposed
measure the arrester
chargingmodel is illustrated
capacitance, the in
must
Figure be 6. In taken into account
real practice, in the
it is possible real impulse
to measure current
the charging generation
capacitance, . The simplif
the charging
charging voltage (Vch), and the arrester capacitance precisely, but it is quite hard to meas-
lent
voltage
ure othercircuit
circuit including
(Vch ), and the arrester
parameters thecapacitance
. However, parasitic impedance
precisely,
the characteristic
but it isand
quite
of the nonlinearthe
hard proposed
to measurearrester
resistor (A0) can
other mod
circuit parameters. However, the characteristic of the nonlinear resistor (A0 ) can be ap-
betrated in Figure
approximated from 6 the. In real practice,
experimental results it
of is possible
voltage to measure
and current
proximated from the experimental results of voltage and current across the arrester. As
across thethe charging
arrester . capac
As shown in Figure 7, in this paper, a piecewise linear curve drawn through the center of
charging
shown in Figure voltage7, in (this ), anda the
Vchpaper, arrester
piecewise linearcapacitance
curve drawn precisely,
through the centerbut itofisthequite ha
the data of voltage and current of the arrester was employed to represent the characteristic
data of voltage and current of the arrester was
. However, employed to represent the characteristic of
ofure other circuit
the nonlinear resistorparameters
. A genetic algorithm the characteristic
was employed to search the unknownof thecircuit
nonlinear resist
the nonlinear resistor. A genetic algorithm was employed to search the unknown circuit
parameters
be of the generation
approximated fromcircuit
the and of the arresterresults
experimental model to ofmatch
voltageresidual voltage
parameters of the generation circuit and of the arrester model to match and residualcurrent
voltage across t
(vr) waveforms from the experiment.
As shown in from
(vr ) waveforms Figure 7, in this paper, a piecewise linear curve drawn through th
the experiment.
the data of voltage and current of the arrester was employed to represent the ch
of the nonlinear resistor. A genetic algorithm was employed to search the unkn
parameters of the generation circuit and of SAthe arrester model to match resid
(vr) waveforms from the experiment.

G L R
+
i(t)
vC C
Lp and Rp
SA
LA

_
Energies 2021, 14, x FOR PEER REVIEW 5 of 13
Figure 6. Equivalent circuit of the impulse current generation circuit including the parasitic
Arrester modelim-
Figure 6. Equivalent circuit of the impulse current generation circuit including the parasitic
pedance and the arrester model.
impedance and the arrester model.
Le = L+Lp Re = R+Rp L0
G
+ +
R0 R2
vC C i(t) vr A0
C2
_ _

Figure 6. Equivalent circuit of the impulse current generation circuit including the para
pedance and the arrester model.

Figure
Figure 7. Approximated
7. Approximated characteristicofofthe
characteristic thenonlinear
nonlinearresistor
resistor in
in the
the proposed
proposedmodel.
model.

4. Genetic Algorithm
A genetic algorithm (GA) is inspired by the process of natural selection. The genetic
algorithm is a kind of stochastic optimization. In this paper, the genetic algorithm [16,17]
was utilized to search the appropriate circuit parameters, i.e., the equivalent inductance
Energies 2021, 14, 3132 5 of 13

4. Genetic Algorithm
A genetic algorithm (GA) is inspired by the process of natural selection. The genetic
algorithm is a kind of stochastic optimization. In this paper, the genetic algorithm [16,17]
was utilized to search the appropriate circuit parameters, i.e., the equivalent inductance
(Le ), the equivalent resistance (Re ), the internal arrester inductance (L0 ), the internal arrester
resistance (R0 ), the damping resistance (R2 ), and the charging voltage (Vch ). The genetic
algorithm searches the input parameters (the circuit parameter in the vector form; X) to
minimize objective function (Ob(X)), as given in Equation (1).
n  2
Ob( X ) = ∑ vr(exp) (i ) − vr(sim) ( X, i ) (1)
i =1

where vr (exp) and vr (sim) are the residual voltages from experiment and simulation using
the arrester model. i is the ith residual voltage point and n is a number of points of the
measured and simulated residual voltages.
Energies 2021, 14, x FOR PEER REVIEW 6 of 13
The flow chart based on the genetic algorithm for searching the most proper circuit
parameters is presented in Figure 8.

Start

Step 1:
Import the recorded waveform data

Step 2:
1) Initial Lumped parameters
Le, Re, Lo,Ro, R2, and Vch
2) Determine upper bound and lower
bound (only positive values)

Choose the initial population

Evaluation

Selection Step 3:

Crossover

Mutation

Step 4:
Calculate the residual voltage of the
arrester model

Step 5:
1) Calculate objective function
2) Minimization of objective
function

i = i+1

i=n

Save value

Stop

Figure
Figure 8. Flowchart
8. Flowchart ofof thepresented
the presentedgenetic algorithm.
genetic algorithm.

5. Verification of the Proposed Model

To express the validity of the proposed model, two experiments were carried out on
two arresters. The first and the second arresters had the nominal discharge currents of 10
kA and 2.5 kA, respectively. In the first experiment, the lightning impulse current of 10
 Energies 2021, 14, 3132 6 of 13

Energies
Energies 2021,
2021, 14,
14, xx FOR
FOR PEER
PEER REVIEW
REVIEW 77 of
of 13
13

5. Verification of the Proposed Model

To express the validity of the proposed model, two experiments were carried out on
agreements
two arresters.of
agreements the
ofThe generated
thefirst and thecurrents
generated and
and the
second arresters
currents residual
the had voltages
the nominal
residual voltages were
were found
discharge in
in both
currents
found of cases
both cases
As
As shown in Figures 3 and 4, the calculated current waveform agreed quite well with the
10 shown
kA and in
2.5 Figures
kA, 3 and
respectively.4, the
In calculated
the first current
experiment, waveform
the lightning agreed
impulse quite well
current ofwith the
experimental
10 kA was generated
experimental result. The
The time
through
result. to
to peak
the first
time of
arrester.
peak the
of The calculated
the nonlinear
calculated residual
residual voltage
characteristic was
was almost
of the arrester
voltage as
almost the
the
showntime
same in Figure
of the7 was used in the one.
experimental model construction.
The calculated The genetic
peak algorithm
current deviatedwasfrom
utilized
the exper-
same time of the experimental one. The calculated peak current deviated from the exper-
to searchone
imental the by
appropriate
5%. The unknown
high circuit of
oscillation parameters.
the Asvoltage
residual shown in Figures around
occurred 9 and 10,the peak
imental
the calculated current and residual voltage waveforms agreed well with the experimentalthe peak
one by 5%. The high oscillation of the residual voltage occurred around
time,
time, which is
results.which is quite
The time quite different
different
to peak
from
from the model
theresidual
of the calculated one.. was almost the same time of the
one
model voltage
It
It was
was confirmed
experimental confirmed that
that the
one. The calculated proposed
thepeak current model
proposed model is
is fairly
and voltage fairly highly
highly
deviated fromaccurate
accurate for
for the
the residual
the experimental residua
voltage
ones by tests
only
voltage tests. .
0.25% and 0.2, respectively.

Experiment
Experiment
Simulation
Simulation
(kA)
Current(kA)
Current

Time (µs)
Time (µs)

waveforms..
Figure 9. Comparison
Figure of thethe
simulated andand
experimental lightning impulse current current
waveforms.
Figure 9.
9. Comparison
Comparison of
of the simulated
simulated and experimental
experimental lightning
lightning impulse
impulse current waveforms

Experiment
Experiment
Simulation
Simulation
(kV)
Voltage(kV)
Voltage

Time (µs)
Time (µs)
Figure 10. Comparison
Figure of thethe
residual voltage waveforms across the arrester in a case of lightning
Figure 10.
10. Comparison
Comparison of of the residual
residual voltage
voltage waveforms
waveforms across
across the
the arrester
arrester in
in aa case
case of
of lightning
lightning
impulse
impulse current
currentinjected.
injected .
impulse current injected.
In the second experiment, the steep impulse current of 2.5 kA was generated through
the second one. The comparisons of the generated currents and the residual voltages
Experiment
from the experiment and from the proposed model are shown in Figures 11 and 12. Good
Experiment
Simulation
Simulation voltages were found in both cases.
agreements of the generated currents and the residual
As shown in Figures 3 and 4, the calculated current waveform agreed quite well with
the experimental result. The time to peak of the calculated residual voltage was almost
(kA)
Current(kA)

the same time of the experimental one. The calculated peak current deviated from the
Current

experimental one by 5%. The high oscillation of the residual voltage occurred around the
peak time, which is quite different from the model one.

Time (µs)
Time (µs)

Figure
Figure 11.
11. Comparison
Comparison of
of the
the simulated
simulated and
and experimental
experimental steep
steep impulse
impulse current waveforms..
current waveforms
 Time (µs)

Energies 2021, 14, 3132 Figure 10. Comparison of the residual voltage waveforms across the arrester in a 7case
of 13of lightn

impulse current injected.

Experiment
Simulation

Current (kA)

Energies 2021, 14, x FOR PEER REVIEW Time (µs) 8 o

Figure 11. Comparison of the simulated and experimental steep impulse current waveforms.
Figure 11. Comparison of the simulated and experimental steep impulse current waveforms.

Experiment
Simulation
Voltage (kV)

Time (µs)

Figure 12. Comparison of the residual voltage waveforms across the arrester in a case of steep
Figure
impulse12. Comparison
current injected. of the residual voltage waveforms across the arrester in a case of steep
impulse current injected.
It was confirmed that the proposed model is fairly highly accurate for the residual
voltage tests.
6. Applications of the Proposed Method in the Residual Voltage Test
It is hard for
6. Applications a test
of the engineer
Proposed to control
Method the current
in the Residual waveform
Voltage Test and peak without a
and error
It is hard for a test. engineer
approach Therefore, in this
to control thesection, an effective
current waveform technique
and peak withoutisa proposed
trial and to de
error approach.
mine Therefore,
the charging in this
voltage section,
for an effective
obtaining technique
the current is proposedaccording
waveform to determine tothe
the stand
charging voltage for obtaining the current waveform according to the standard requirement.
requirement. The technique starts from the determination of the unknown circuit par
The technique starts from the determination of the unknown circuit parameters of the genera-
eters of theand
tion circuit generation
the proposed circuit and
arrester the proposed
model as shown inarrester
Figure 6. model as shown
The current in Figure 6 .
and residual
current and residual
voltage waveforms fromvoltage waveforms
the preliminary from
test with thethe preliminary
lower current peaktest with the
specified lower cur
by the
standard were utilized to determine the circuit parameters. Then, the
peak specified by the standard were utilized to determine the circuit parameters. Thprevious determined
model and a genetic algorithm were employed to search the charging voltage to obtain the
the previous determined model and a genetic algorithm were employed to search
specified current peak.
charging voltage
To confirm theto obtain
validity ofthe specifiedtechnique,
the proposed current peaksome. experiments on the residual
Totests
voltage confirm the validity
were carried out on oftwothe proposed
surge arresterstechnique,
with the ratedsome experiments
voltage of 9 kV. The onfirstthe resid
voltage tests were carried out on two surge arresters with the rated voltage of 9 kV .
arrester had the nominal discharge current of 5 kA and the second one had the nominal
discharge
first current
arrester hadofthe2.5nominal
kA. discharge current of 5 kA and the second one had the n
According to the standard requirement [3] for the residual voltage test of the light-
inal
ningdischarge current
impulse current, 0.5,of 2.5and
1.0, . times the nominal discharge current are generated
kA2.0
According
through to the
the arrester and standard
the residual requirement [3] for theInresidual
voltages are examined. voltageexperi-
the preliminary test of the li
ment,impulse
ning the charging capacitance
current, 0.5, 1.0,andand voltage were the
2.0 times set to be 2 µF discharge
nominal and 30 kV, respectively,
current are genera
and the measured internal capacitance was 56 pF. From the known circuit parameters
through the arrester and the residual voltages are examined . In the preliminary exp
and to obtain the lightning impulse current waveform [18,19], the total inductance of
ment, the charging capacitance and voltage were set to be 2 µF and 30 kV, respectiv
and the measured internal capacitance was 56 pF. From the known circuit parameters
to obtain the lightning impulse current waveform [18,19], the total inductance of abou
µH was required, so the series inductor (L) with inductance of 30 µH was connecte
 Energies 2021, 14, 3132 8 of 13

about 30 µH was required, so the series inductor (L) with inductance of 30 µH was con-
nected to the generation circuit. As shown in Figure 13, the current waveform deviated
from the standard requirement due to effects of the parasitic impedance and the arrester.
From the experimental results of current and residual voltage waveforms as shown in
Figures 13 and 14, the circuit parameters were determined by using the equivalent circuit
in Figure 6. The equivalent inductance (Le ), the equivalent resistance (Re ), the internal
arrester inductance (L0 ), the internal arrester resistance (R0 ), and the damping resistance
(R2 ) were 41.2 µH, 0.1 Ω, 4.1 µH, and 3.7 kΩ, respectively. From the determined circuit
parameters, the total inductance in the circuit was 45.3 µH (Le + L0 ). Therefore, the series
additional inductance of 14.7 µH was required to obtain the lightning impulse current
Energies 2021, 14, x FOR PEER REVIEW
waveform specified by the standard. Another experiment with an additional inductance
Energies 2021, 14, x FOR PEER REVIEW
of 15 µH was carried out to confirm the validity of the proposed technique. Furthermore, 9
using the proposed technique, the calculated charging voltage of 34 kV associating with
the current peak of 2.5 kA was set in the experiment. It was found that as shown in
residual
Figure 13,voltage agreed
the current well according
waveform with thattofrom the experiment
the standard specificationaswas
shown in Figure 1
generated
residual
time
whentothe voltage
peak
corrected agreed
of the well with
calculated
circuit that
residual
parameter was from
voltage
applied thethe
in experiment
was almost the
generation as shown
same
circuit. asinthe
Moreover,Figure 14
experi
the
one. calculated
timeTheto peak residual voltage
of the peak
calculated agreed
calculated well
residual
current with that
voltage
and voltage from the
was almost
deviated experiment
fromthe
thesame as shown in
as the experim
experimental ones b
Figure
one. 14. calculated
The The time to peak
peak of the calculated
current and residual
voltage voltage was
deviated from almost
the the same as the ones by
experimental
0.55% and 0.35%, respectively.
experimental one. The calculated peak current and voltage deviated from the experimental
0.55% and 0.35%, respectively.
ones by only 0.55% and 0.35%, respectively.
Experiment
Simulation
Experiment
Simulation

Corrected test with 2 5 kA current peak

(kA)

Corrected test with 2 5 kA current peak

(kA)
Current
Current

Preliminary test
Preliminary test

Time (µs)
Time (µs)
Figure 13.Comparison
Figure 13. Comparison of generated
of the the generated current
currentcurrent waveforms
waveforms in
in cases ofin
thecases of thetest
preliminary preliminary test
and the test
Figure 13. Comparison of the generated waveforms cases of the preliminary an
corrected one.
corrected one.
corrected one.

Experiment
Experiment
Simulation
Simulation
Corrected test
Corrected test
Preliminary test
(kV)

Preliminary test
Voltage(kV)
Voltage

Time (µs)
Time (µs)

Figure 14. Comparison of the residual voltage waveforms across the arrester in cases of the prelimi-
Figure
Figure 14.
14.
nary test andComparison
Comparison of the residual
of the
the corrected one. residual voltage
voltagewaveforms
waveformsacross
acrossthe
thearrester
arresterinincases
casesofofthe
the
p
liminary test
liminary test andand the corrected one.
one.

In
In aa similar
similar manner, the proposed
proposed technique
techniquewith
withthe theequivalent
equivalentcircuit
circuitassoc
ass
with
with the
the current
current peak of 2.5 kA
kA was
was applied
applied to
tosearch
searchthetheappropriate
appropriatecharging
chargingvol
v
associated
associated with
with the current peaks
peaks of
of 55 kA
kA and
and10 kA. .The
10kA Thepiecewise
piecewiselinear
linearcurve
curveusu
 Energies 2021, 14, 3132 9 of 13

In a similar manner, the proposed technique with the equivalent circuit associated
with the current peak of 2.5 kA was applied to search the appropriate charging volt-
ages associated with the current peaks of 5 kA and 10 kA. The piecewise linear curve
used to estimate the nonlinear characteristic of the nonlinear resistor (A0 ) is illustrated in
Figure 15. From the proposed technique, the equivalent inductance (Le ), the equivalent
resistance (Re ), the internal arrester inductance (L0 ), the internal arrester resistance (R0 ),
and the damping resistance (R2 ) were 27.1 µH, 0.1 Ω, 4.0 µH, and 3.6 kΩ, respectively. The
charging voltages to obtain the current peaks of 5 kA and 10 kA were 46 kV and 69 kV,
respectively. Comparisons of the experiment and simulated waveforms by the proposed
technique are shown in Figures 16 and 17. Good agreement was found in the case of
5 kA current peak but deviation occurred in the case of 10 kA current peak. In the case
of 5 kA current application, the calculated peak current and voltage deviated from the
experimental ones by only 0.73% and 0.88%, respectively. In the case of 10 kA current
Energies 2021, 14, x FOR PEER REVIEW
application, the calculated peak current and voltage deviated from the experimental ones
by 4.1% and 8.1%, respectively. However, in terms of current generation in the case of
Energies 2021, 14, x FOR PEER REVIEW
10 kA current peak, the deviation is acceptable because the current peak deviation of 4.1%
was less than 10% according to the standard tolerance.
(kV)

Estimated V I characteristic
(kV)
Voltage

Estimated V I characteristic
V I characteristic from the experiment
Voltage

V I characteristic from the experiment

Current (kA)
Current (kA)
Figure 15. Voltage and current characteristic of the arrester under test and estimated
Figure 15. Voltage and current characteristic of the arrester under test and estimated piecewise linear
Figure
linear 15. Voltage
characteristic
characteristic
and current
of the
of the nonlinear
characteristic
nonlinear
resistor. resistor. of the arrester under test and estimated
linear characteristic of the nonlinear resistor.

Experiment
Simulation
Experiment
Simulation
Current
Current (kA)
(kA)

Time (µs)
Time (µs)
Figure 16. Comparison of the generated current waveforms from the proposed model and experiment.
Figure 16. Comparison of the generated current waveforms from the proposed mode
Figure
ment. 16. Comparison of the generated current waveforms from the proposed mode
ment.

10 kA current Experiment
Simulation
Experiment
10 kA current
 Time (µs)

Energies 2021, 14, 3132 Figure 16. Comparison of the generated current waveforms from the proposed
10 of 13 mod
ment.

10 kA current Experiment
Simulation

2 5 kA current
5 kA current

Voltage (kV)

Time (µs)

Figure 17. Comparison of the residual voltage waveforms across the arrester from the proposed
Figure 17. Comparison of the residual voltage waveforms across the arrester from th
model and experiment.
model and experiment.
To obtain higher accuracy in the generation of the 10 kA current peak, the proposed
technique was applied to model the equivalent circuit using the waveforms in the case of
Energies 2021, 14, x FOR PEER REVIEW
To obtain higher accuracy in the generation of the 10 kA current peak,
the 5 kA current peak and it was employed to determine the charging voltage to obtain
technique wasofapplied
the current peak 10 kA. From to the
model the equivalent
experimental circuitand
results of current using thevoltage
residual waveforms
the 5 kA current
waveforms as shown peak and15itand
in Figures was 16,employed to determine
the circuit parameters the charging
were determined by vol
using the equivalent circuit in Figure 6. The equivalent inductance (Le ), the equivalent
the current peak of 10 kA. From the experimental results of current and res
resistance (Re ), the internal arrester inductance (L0 ), the internal arrester resistance (R0 ),
current
waveforms peakasof
and the damping
10 kAin
shown
resistance
waswere67.326.9
(R2 )Figures
kV 15 and
µH,and
a16,
0.1 Ω,
charging
2.4the
voltage
µH,circuit
and 3.1 kΩ,
of 67.5 was set
parameters
respectively.were d
Using .the
ment
using Comparison
themodel of the 5of
equivalent the
kAcircuitgenerated
current peak,
in the currents
Figure calculated by
6. Thechargingsimulation towith
voltageinductance
equivalent the
obtain thepropos
(Le), t
current peak of 10 kA was 67.3 kV and a charging voltage of 67.5 was set in the experiment.
that from the
resistance (Rthe experiment
e), generated
the internal is shown
arrester Figure 18. (In
ininductance L0)addition,
, the model the comparison
internal
Comparison of currents by simulation with the proposed and arrester
that re
imental
from the
and and simulated
theexperiment
damping is shown inresidual
resistanceFigure(R18.2) voltages
In 26.is
addition,
were theshown
9 µH, 0.1 in
comparison Ω,ofFigure
2the . A 3better
µH,19and
.4 experimental .1 kΩa
and simulated
tween the residual
the simulated voltages is shown
and5 experimental in Figure 19. A
was the better
found agreement between
. The calculated the
peak cur
Using
simulated andmodel of the
experimental kA current
was found. The calculatedpeak, calculated
peak current charging
and voltage deviatedvoltage
age
from deviated fromones
the experimental thebyexperimental
only 0.43% and 0.56%, ones by only 0.43% and 0.56%, respecti
respectively.

Experiment
Experiment
Simulation
Simulation
Current (kA)

Time (µs)

Figure 18. Comparison of the generated current waveforms from the proposed model and experiment.
Figure 18. Comparison of the generated current waveforms from the proposed mod
ment.

10 kA current
Experiment
Simulation

5 kA current
)
 Time (µs)

Energies 2021, 14, 3132 Figure 18. Comparison of the generated current waveforms from the proposed
11 of 13 mod
ment.

10 kA current
Experiment
Simulation

5 kA current

Voltage (kV)

Time (µs)

Figure 19. Comparison of the residual voltage waveforms across the arrester from the proposed
Figure 19. Comparison of the residual voltage waveforms across the arrester from t
model and experiment.
model and experiment.
In a similar manner, the proposed technique was applied for the residual voltage test
on the arrester with the nominal discharge current of 2.5 kA. The preliminary test with a
In a similar manner, the proposed technique was applied for the residu
steep impulse current was carried out at the current peak of 1.25 kA. For the steep impulse
on the generation,
current arrester with the nominal
the charging discharge
capacitance, current
the charging .5 kA
of 2and
voltage, the .additional
The prelimina
resistance
steep were set
impulse to 2 µF,was
current 78 kV, carried Ω, respectively,
and 38 out at the currentand the measured
peak of 1.25internal
kA. For the
capacitance was 50 pF. From the calculation using the proposed technique, the equivalent
current
inductance generation, the charging
(Le ), the equivalent resistance capacitance,
(Re ), the internalthe charging
arrester voltage,
inductance and the
(L0 ), the
sistance wereresistance
internal arrester set to 2(RµF, 78the
0 ), and kV, and 38
damping Ω, respectively,
resistance Ω, 2.8
and0.1the
(R2 ) were 14 µH, measured
µH, i
and 198 Ω, respectively. Using the proposed technique to obtain the current peak of 2.5 kA,
itance was 50 pF. From the calculation using the proposed technique, the
the calculated charging voltages was 134.5 kV and the setting charging voltage in the
ductance
experiment to (Lconfirm
e), the the
equivalent the proposed (technique
validity of resistance Re), the wasinternal
135 kV.arrester
Comparisons inductanc
of the experiment
ternal and simulated
arrester resistance (Rwaveforms by the proposed technique are shown in
0), and the damping resistance (R2) were 14 µH,
Figures 20–22, in which good agreement is found. It was confirmed that the proposed
and
Energies 2021, 14, x FOR PEER REVIEW 198 Ω,
technique is very effective in .the
respectively Using thevoltage
residual proposed technique
tests. In the case ofto2.5obtain
kA currentthe curre
application,
kA, the calculated
the calculated peak current
charging deviated was
voltages from the
134experimental
.5 kV andone thebysetting
7.8%, whichcharging
is still in the standard tolerance.
experiment to confirm the validity of the proposed technique was 135 kV.
of the experiment and simulated waveforms by the proposed technique are
ures 20–22, in which good agreement is foundExperiment
. It was confirmed that the p
Simulation
nique is very effective in the residual voltage tests. In the case of 2.5 kA cu
tion, the calculated peak current deviated from the experimental one by 7
Current (kA)

still in the standard tolerance.

Time (µs)

Figure 20. Comparison of the generated current waveforms from the proposed model and experiment.

Figure 20. Comparison of the generated current waveforms from the proposed mod
ment.

Experiment
Simulation
 Time (µs)
Figure 20. Comparison of the generated current waveforms from the proposed mod
ment.
Energies 2021, 14, 3132 Figure 20. Comparison of the generated current waveforms from the proposed
12 of 13 mod
ment.
Experiment
Simulation

Experiment
Simulation

(kV) (kV)
VoltageVoltage

Time (µs)

Time (µs)
Figure 21. Comparison of the residual voltage waveforms across the arrester from t
Figure 21. Comparison of the residual voltage waveforms across the arrester from the proposed
model and experiment in a case of 1.25 kA current peak.
model and experiment in a case of 1.25 kA current peak.
Figure 21. Comparison of the residual voltage waveforms across the arrester from t
model and experiment in a case of 1.25 kA current peak.
Experiment
Simulation

Experiment
Simulation
(kV) (kV)
VoltageVoltage

Time (µs)

Figure 22. Comparison of the residual voltage waveforms across the arrester from the proposed
Figure
model and22. Comparison
experiment in a caseofofthe Time (µs) peak.
residual
2.5 kA current voltage waveforms across the arrester from t
model and experiment in a case of 2.5 kA current peak.
7. Conclusions
Figure 22. Comparison
A simplified of the
and accurate residual
model voltage
of a surge waveforms
arrester across
for the residual the tests
voltage arrester
was from t
7. Conclusions
model
proposedand experiment
in this paper. Thein
validation .5 kA
a case ofof2the current
proposed peak
model was. verified in comparison
with some experiments. Furthermore, the technique for the estimation of setting circuit
A simplified
parameters and accurate
and the charging modeltheofspecified
voltage to obtain a surgecurrent
arrester for the
waveform andresidual
peak vo
7.
wasConclusions
proposed
proposed. inInthis paper
addition, . The
the validation
residual of the proposed
voltage associated model
with the specified waswas
current verified
estimated
A precisely.
simplified The validity of
and accurate the proposed technique
model ofthe was
a surge verified
arrester by the experiments.
with some
It was confirmed experiments . Furthermore,
that the proposed technique is very technique
useful forfor
in the residual the the residual vol
estimation
voltage tests of
proposed
parameters inand
of arresters and this paper
helpsthe . The validation
testcharging
engineers involtage toofobtain
the selection the proposed model
the specified
of appropriate circuit was verified
current
components wavef
with
withoutsome
a trial experiments
was proposed and.error
In addition, . Furthermore,
approach. the residualthe technique
voltage for the
associated withestimation of
the specifie
parameters
estimated
Author and the
precisely
Contributions: .charging
The validity
Conceptualization, voltage
P.Y. and tomethodology,
ofS.L.;
the obtain
proposed the specified
S.L.technique current
was S.L.;
and P.Y.; validation, wavef
verified
was
formalproposed
ments . investigation,
analysis, N.T.; In addition,
. It was confirmed P.Y.the
that theresidual
and S.L.; voltage
writing—original
proposed associated
draft
technique preparation,with
is very P.Y. the
and
useful specifie
S.L.;
in the re
writing—review and editing, P.Y. and S.L.; supervision, P.Y. All authors have read and agreed to the
estimated precisely . The
published version of the manuscript. validity of the proposed technique was verified
Funding:. This
ments It was confirmed
project that theResearch
is funded by National proposedCounciltechnique
of Thailand. is very useful in the re
Acknowledgments: The authors would like to give special acknowledgement to the School of
Engineering, King Mongkut’s Institute of Technology Ladkrabang for providing the facility in this
research work. In addition, this project is funded by National Research Council of Thailand.
Energies 2021, 14, 3132 13 of 13

Conflicts of Interest: The authors declare no conflict of interest.

References
1. Eddie, S.; Innocent, E.D.; Sindisiwe, C.M. Lightning Performance and Economic Analysis of an Overhead 88 kV Power Delivery
Network with Enhanced Protective Systems. Energies 2020, 13, 6519.
2. IEC 61641-1. Low-Voltage Surge Protective Devices—Part 1; IEC: Geneva, Switzerland, 2014.
3. IEC 60099-4. Surge Arresters. Part 4: Metal-Oxide Surge Arresters without Gaps for A.C. System; IEC: Geneva, Switzerland, 2014.
4. IEEE Standard 4TM-2013. IEEE Standard for High-Voltage Testing Techniques; IEEE SA: Piscataway, NJ, USA, 2013.
5. IEC 62475. High-Current Test Techniques: Definitions and Requirements for Test Currents and Measuring Systems; IEC: Geneva,
Switzerland, 2010.
6. IEEE.W.3.4.11. Of Surge Protection Pevices Committee. Modeling of metal oxide surge arresters. IEEE Trans. Power Deliv. 1992, 7,
301–309.
7. Montano, R.; Edirisinghe, M.; Cooray, V.; Roman, F. Behavior of low-voltage surge protective devices under high-current
derivative impulses. IEEE Trans. Power Deliv. 2007, 22, 2185–2190. [CrossRef]
8. Zeinoddini-Meymand, H.; Vahidi, B.; Naghizadeh, R.A.; Moghimi-Haji, M. Optimal surge arrester parameter estimation using a
PSO-based multiobjective approach. IEEE Trans. Power Deliv. 2013, 28, 1758–1769. [CrossRef]
9. Tuczek, M.N.; Hinrichsen, V. Recent experimental findings on the single and multi-impulse energy handling capability of
metal-oxide varistors for use in high-voltage surge arresters. IEEE Trans. Power Deliv. 2014, 29, 2197–2205. [CrossRef]
10. Sato, S.; Okamoto, Y.; Nishimura, S. Impulse current waveform parameters determination with ZnO load. In Proceedings of the
2010 IPEC Conference, Sapporo, Japan, 21–24 June 2010; pp. 78–82.
11. Sabiha, N.A.; Mahmood, F.; Abd-Elhady, A.M. Failure risk assessment of surge arrester using paralleled spark gap. IEEE Access
2020, 8, 217098–217107. [CrossRef]
12. Yvonne, S.L.; Erion, G.; Herbert, D.G. Electrothermal optimization of field grading systems of station class surge arresters. IEEE J.
Multiscale Multiphys. Comput. Tech. 2019, 4, 29–35.
13. Reza, S.; Javad, G.; Keyhan, S. Estimation of energy stress of surge arresters considering the high-frequency behavior of grounding
systems. IEEE Trans. Electromagn. Compat. 2018, 6, 917–925.
14. Sato, S.; Nishimura, S.; Seki, S. Parameters determination of oscillatory impulse current waveform. IEEJ Trans. Fundam. Mater.
2010, 130, 265–271. [CrossRef]
15. Lu, J.; Xie, P.; Fang, Z.; Hu, J. Electro-thermal modeling of metal-oxide arrester under power frequency applied voltages. Energies
2018, 11, 1610. [CrossRef]
16. Yang, X. Engineering Optimization: An Introduction with Metahueristic Applications; John Wiley & Sons: Hoboken, NJ, USA, 2010;
pp. 173–176.
17. Mehdi, B.; Davood, A.; Hassan, B.; Ebrahim, R. Identification of transient model parameters of transformer using genetic
algorithm. In Proceedings of the International Conference on Power System Technology, Hangzhou, China, 24–28 October 2010;
pp. 1–6.
18. Yutthagowith, P. Effective technique for circuit and waveform parameter extraction in impulse current tests and its application.
IEEE Trans. Power Deliv. 2015, 31, 704–710. [CrossRef]
19. Yutthagowith, P.; Triruttanapiruk, N. A program for design of passive circuits of impulse current generation. In Proceedings of
the ISET, Chiang Mai, Thailand, 24–26 April 2013; pp. 1–4.