IET Science Measure Tech - 2024 - Safaei - A New Method For Surge Arrester Placement in High Voltage Substations

Received: 18 May 2023 Revised: 4 January 2024 Accepted: 5 June 2024 IET Science, Measurement & Technology
DOI: 10.1049/smt2.12208
ORIGINAL RESEARCH
A new method for surge arrester placement in high-voltage

substations considering environmental effects
Faridoddin Safaei Mohsen Niasati
Faculty of Electrical and Computer Engineering Abstract

(ECE), Semnan University, Semnan, Iran
Conventional approaches for determining the optimal locations of surge arresters (SAs)
can result in an unnecessary increase in the number of these devices in high-voltage
Correspondence
Mohsen Niasati, Central Administration of Semnan substations. This study presents an effective technique for determining the optimal loca-
University, Campus 1, Semnan, 35131-19111, I. R. of tions of SAs. The lightning back flashover (BF) and switching over-voltages are predicted
Iran.
with the accurate modeling of the transient behavior of all elements of high-voltage sub-
Email: [email protected]
stations and transmission lines, considering the effect of environmental conditions. Also,
the proposed limiting parameter Monte Carlo (MC-LP) method is utilized to correctly
select the probability distribution of the possible strokes for predicting the insulation risk
(IR) of a transformer based on transient over-voltage on the transformer end. Therefore,
the most appropriate location to install an SA can be determined with a minimum num-
ber of calculations using the structural data of the substation, lines connected to it, and the
transformer. Simulations are based on experimental results, and the number of calculations
significantly decreases using the proposed algorithm. Simulations of the sample network
and implementation of the proposed algorithm with MATLAB and EMTP-RV prove the
efficiency of the proposed method for optimum SA placement.
1 INTRODUCTION reasons for outages in UHV substations is the over-voltages

of lightning back flashover (BF) on the lines connected to the
Substations feed electricity to consumers by using overhead substations. The stresses due to these over-voltages in outdoor
transmission and distribution lines. An ultra-high voltage substations cause insulator flashovers, and then short circuits
(UHV) substation is an assemblage of electrical components, may cause power outages [8, 9].
including switchgear, busbar, and transformers [1]. Overvolt- It is necessary to use a procedure that presents random values
age protection is vital to protect these equipment and stability of the statistical distributions known as independent variables
purposes. Surge arresters (SAs) are protective devices used in to evaluate lightning over-voltages on the power network. The
substations to limit over-voltages [2]. The nonlinear resistance Monte Carlo (MC) method is the most popular procedure in
feature of SAs protects equipment from over-voltages that may this field [10].
exceed the basic lightning impulse insulation level (BIL) [3]. The Many studies [9, 11–13] have focused on the effects of SA
locations of SAs affect the equipment terminal voltage. The failures and SA placement on substations. In these studies, the
desirable reliability of UHV substations, insulator coordination, effective factors, such as the energy absorption capability of
and determination of the optimum distance between the SAs to the SAs due to direct lightning strokes and the effect of the SA
protect the equipment are very important issues [4]. The rapid installation location, have been investigated. Most papers have
response to over-voltages, high reliability, negligible losses, and neglected environmental effects like temperature, humidity,
appropriate nonlinear behavior are some advantages of ZnO or altitude [14]. The statistical distribution used in most of
SAs [5]. Several related studies have been devoted to optimiz- these studies is uniform or typical. However, it is essential
ing the number and location of substation SAs based on the to use accurate transient modelling to achieve a precise solu-
basic switching level (BSL) [5–7]. However, one of the main tion for optimum SA placement. Some references have used
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is
properly cited.
© 2024 The Author(s). IET Science, Measurement & Technology published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.
IET Sci. Meas. Technol. 2024;1–14. wileyonlinelibrary.com/iet-smt 1

17518830, 0, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/smt2.12208 by Cochrane Mexico, Wiley Online Library on [31/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 SAFAEI and NIASATI
mathematical techniques and fuzzy logic to solve this problem TABLE 1 Typical line surge impedance with single and bundled
[15]. In this case, the computational burden increases con- conductors [22].
siderably. In [16] and [17], the reliability indices in operating System rated No. of conductors Typical surge
conditions have been considered. Also, the number of SAs and voltage (kV) in bundle impedance (Ω)
their locations are determined based on reliability indices, but in 3–150 1 450
these papers, the effect of cumulative lightning stress on insu-
151–325 1 400
lation risk (IR) as a result of the non-self-restoring insulation
326–400 2 350
behavior of transformers has not been taken into account [18].
The Weibull distribution is the best statistical distribution that 400–800 3 or 4 320
can describe the stochastic behavior of the IR for a transformer > 800 6 or 8 300
under voltage stresses [19, 20]. Many studies have not consid-
ered Weibull distribution, and stochastic behavior has usually
been described using a normal or uniform distribution, as is
be used [22]:
done for line insulators [21].
dV D
In this study, the optimal placement of SAs is presented by VC .B = VSA + 2 × ×
modeling a real high-voltage substation. To this end, the effect dT v . (2)
VC .B ≤ VC .BW
of switching over-voltage is included using the hybrid Monte
Carlo-lightning limiting parameter (MC-LP) method for evalu- The following equation calculates the energy absorption of
ating all stresses that can affect high-voltage substations due to SA [16]:
lightning BF. Consequently, the computational burden remark-
(V − VP )
ably decreases since transient simulations are conducted only w = O.V × VP × 2𝜏 × N , (3)
for strikes within the risk zone. Structural data of the sub- Z
station and its connected lines are received, and an accurate
where 𝜏 is the travelling time of the wave, Vp is the protective
approximation of insulation risk is provided to determine the
level of the SA, VO.V is the over-voltage without an SA, N is
appropriate locations for the SAs. The proposed algorithm in
the number of discharge currents through the SA, and Z is the
the present study has been developed in the MATLAB environ-
surge impedance. The surge impedance of transmission lines is
ment, and the modeling and simulation of the studied substation
described in Table 1 [22].
and transmission lines (TLs) are implemented in EMTP-RV.
3 INSULATION FAILURE RISK OF

2 DETERMINING THE DISTANCE TRANSFORMER
BETWEEN EQUIPMENT AND SURGE
ARRESTER The characteristics of non-self-restoring insulating devices, such
as transformers, change over time if they are often under electri-
According to the IEC standard for determining the distance cal stresses [9]. The inverse power law describes the IR of power
of an SA in transformer protection (Figure 1), the following transformers [23]. Therefore, the Weibull statistical distribution
equation is used [22]: can be considered the most appropriate statistical distribution
to describe the stochastic behavior of the IR transformers under
dV D voltage stresses [20]. Many studies have not regarded as Weibull
VT = VSA + 2 × ×
dT v , (1) distribution, and stochastic behavior is usually described using
VT ≤ VTW a normal or uniform distribution, as is done for line insulators
[4, 5]. The Weibull distribution can be used to determine the
where v is the speed of the travelling wave, VSA is the residual transformer IR as follows:
voltage, VTW is the insulation strength of equipment, and D is
the installation distance of the SA. To determine the distance of P (V n t ) = 1 − exp(−KV m t a ). (4)
the SA for circuit breaker protection, the following equation can
In Equation (4), m and a can be determined in the laboratory
using the V-T characteristics based on the IEC standard, and K
is a constant [10]. The insulation response of the transformer
can be predicted during the service life by initially predicting
possible lightning overvoltages. When the potential stresses are
determined, suitable high-voltage experiments are performed
to formulate the proper relationship between the experimental
results and stresses applied during service life [24].
ln Rt
FIGURE 1 Protection zone of the SA [22]. Vt m Tt a = Vs m Ts m , (5)
ln Rs
SAFAEI and NIASATI 3
FIGURE 2 Transformer test conditions [24].
where subscript t is related to test time intervals, S corresponds

to service time, and R is defined by R = 1-P as a measure of
reliability. As shown in Figure 2, the experiment has three parts.
The value of V1 is 1.5 per unit, and the duration of t1 and t3 is
one hour, while t2 is one minute. Additionally, V2 is dependent
on the stresses on the transformer over its service life.
Stresses cause progressive degradation over operating life
and, therefore, result in IR as obtained below:
{ }
Pt = 1 − exp −AVt m Tt a
{ }
∑N , (6a) FIGURE 3 Tower model in the EMTP-RV software.
Ps = 1 − exp −A Vsim Tsia
i=1
where ZT denotes the surge impedance of the tower, H is
where Pt can be calculated by the voltages applied during t1 , t2 ,
the height of the tower, and r is the radius of the ground
and t3 , and N denotes the total possible stresses within the trans-
wire.
former service life. Generally, the transformer insulation levels
The CP impedance model of a 400 kV tower in EMTP-RV is
are designed with the critical concept of the desired level of reli-
depicted in Figure 3.
ability in mind. In the literature, the desired reliability levels have
been 95%−99.8% [24].
4.2 Grounding system model
4 MODELING THE TRANSIENT
STATE OF A HIGH-VOLTAGE The footing resistance of the line towers seriously affects the
SUBSTATION peak over-voltages caused by strikes to the guard wires. Since
footing impedance is nonlinear and frequency-dependent, an
4.1 Modeling TL tower accurate representation of this impedance is complicated. In
this paper, a nonlinear resistance, RT , is used to approximate
When lightning strikes a transmission tower, the current of the footing impedance as follows [27]:
lightning flows from the tower’s top to the ground and increases
the voltage at the top of the tower. Then, the increase in 1
voltage causes insulator string BF, which in turn results in an R (t ) = R0 . √ , (7)
I (t )
incoming lightning wave and damages the substation equipment 1+
Ig
[25]. Thus, choosing an accurate tower model to evaluate the
effects of lightning is very important in power systems. Towers
where R0 is the low-frequency, low-current footing resis-
are often denoted as a single conductor distributed-parameter
tance, Ig is the limiting current for the initiation of sufficient
line terminated by a resistance representing the tower footing
soil ionization, and I is the stroke current through the
impedance [12]. The line model used in EMTP-RV to simu-
resistance.
late the towers is the constant parameter (CP) line model. Wave
The limiting current can be calculated as follows [27]:
velocity and surge impedance are the parameters of the CP line
model. In this paper, wave velocity is equal to the speed of light.
E0 . 𝜌
For modeling the TL tower, if network voltages are less than Ig = , (8)
63 kV, the following equation is used [26]: 2 . 𝜋 . R02
( ) (r )
H where ρ is the soil resistivity (in Ω.m) and E0 is the soil
ZT = 60 ln + 90 − 60, (6b)
r H ionization gradient, which in the range of 300–400 kV/m.
FIGURE 5 Electro-geometric model representation of conductors and

guard wire [28].
The following equations can be used to calculate the values

of Dc and Dg [28]:
[ ]
Dc = rc cos(𝜃) − cos(𝛿 + 𝛽)
[ ]
Dg = rc cos(𝛿 − 𝛽) , (9)
where angle δ is shown in Figure 5, and 2β is measured at point

B. The relationship between striking distance rc and the lightning
current crest is given by [28]:
rc = 8 × I p0.65 . (10)
FIGURE 4 Circuit model of the lightning wave (CIGRE model) [28].
4.4 TL model
4.3 Lightning waveform model Three models, including π-model, distributed model, and
frequency-dependent model [27], can be used for modeling
The CIGRE current source model is very close to the real TLs in EMTP. Because of the distributed properties of TLs,
lightning waveform, considering the peak lightning current a frequency-dependent (FD) line is used to model the TLs to
and lightning wave shape [15]. Hence, we use a CIGRE cur- achieve actual simulation results. In this study, J. Marti’s FD
rent source in parallel with a resistor as the lightning channel model is used to model the TLs. As the input data, electrical
impedance. The impedance of the return-stroke channel is parameters of lines can be inserted into the line data of FD lines
assumed to be one kΩ [10]. The variation of the channel [28].
impedance from 400 Ω to 1 kΩ, often used in the sim- In this study, the origins of travelling waves on incoming lines
ulation, leads to little difference in the estimated flashover are back flashovers. When the transmission line is shielded and
rate. the resistance of its tower footing is high, the back flashover
Figure 4 shows a typical lightning current waveform. In this occurs. In such cases, the lightning strike to a shield wire or
figure, Ip , tf , th , and Sm are the peak amplitude, front time, tail tower may elevate the tower’s potential well beyond that of a
time and the maximum of the current steepness, respectively. phase conductor.
The value of Ip gives the maximum current attained in the wave-
form. The front time is the time a lightning current takes to
reach from zero to peak value. The tail time is the sum of the 4.5 SA modeling
front time and the time that the current falls to 50% of its peak
value. The lightning waveform based on the CIGRE current Because of the dynamic behavior of SAs, they cannot be sim-
source is modelled according to Figure 4. ulated using nonlinear resistors. The IEEE WG 3-4–11 group
The electro-geometric model (EGM) is applied to estimate proposed the model shown in Figure 6, including the nonlinear
the reliability of lightning protection systems on overhead resistances A0 and A1 , which are separated by an RL low-
power lines. Based on this theory, the required parameters are pass filter. This model is based on estimated SA height and the
as follows: alignment/height of phase conductors/guard wire, number of columns of ZnO disks [6].
dimensions of the ground structure, striking distances of nearby In this model, the nonlinear V-I characteristics (Figure 7) are
ground, guard wire, and phase conductors [28]. modelled with two nonlinear resistances A0 and A1 separated
TABLE 2 IEEE model parameters and characteristics of SAs.
Voltage level (kV)
SA parameters (TRIDELTA) 63 230 400
Arrester height (mm) 890 1980 4400

Number of parallel columns 1 2 3
R0 (Ω) 89 99 146.66
R1 (Ω) 57.85 64.35 95.33
L0 (µH) 0.178 0.198 0.293
FIGURE 6 IEEE frequency-dependent model.
L1 (µH) 13.35 14.85 22
C (PF) 112.35 101.01 65.18
FIGURE 7 V-I characteristics of A0 and A1 [6] (the voltages in p.u. are

referred to as the residual voltage Vr8/20).
FIGURE 8 High-frequency transformer model [30].

by a RL low-pass filter. The parameters of the FD model of the
SA are given as follows [6, 7]:
TABLE 3 The capacitance values of the high-frequency model of the
0.2 d 15 d transformer [1, 30].
L0 = 𝜇H L1 = 𝜇H
n n Transformer
100 d 65 d power capacity
R0 = Ω R1 = Ω (MVA) CL (nF) CH (nF) CHL (nF)
n n
100 n 100 5–30 7–19 6–14
C = PF , (11) 200 5–27 8–25 4.5–18
d
300 20–40 11–22 5.5–21
where:
400 17–45 14–21 8–24
d: the height of the SA in meters

n: the number of parallel columns of ZnO discs.
Figure 8, where CHL is the capacitance between the high and
The datasheets provided by the manufacturers are used to low windings, CL is the capacitance between the low winding
determine the specifications of the three SAs in the IEEE and ground, and CH is the capacitance between the high wind-
model, as given in ref. [29] (see Table 2). ing and ground. Normally, CH is smaller than CL and CHL . This
paper considers these capacitances as given in Table 3 [1, 2].
4.6 FD model of transformer

4.7 Insulator string model
The use of the FD model of the transformer is essential for
determining the effect of a steep-fronted surge impinges on The airgap switch model is used to simulate an insulator string
the high-voltage terminals of the transformer. Since using stan- with a length of less than one meter (63 kV). For simplicity, it
dard transformer models available in EMTP is not suitable for can be assumed that the insulation withstands a sufficiently high
investigating high-frequency transient states, the high-frequency voltage such that the insulator does not break down. Volt-time
model of the transformer is used in this research. This model curves can be used to represent the BF or flashover mecha-
is suitable for a broad range of high frequencies. Transformer nism of insulators. Moreover, the volt-time characteristics of
capacitances are denoted by CHL , CL , and CH as shown in insulators can be shown as a function of insulator length [31].
TABLE 4 Circuit breaker modeling parameters [30].
Capacitance values (pF)
Equipment <765 kV <400 kV <115 kV
Dead tank circuit breaker 600 150 100

Disconnector 160 200 100
TABLE 5 Capacitance values of substation equipment [30].

FIGURE 9 Airgap leader model [31].
Capacitance values (pF)
Equipment <765 kV <400 kV <115 kV

Equation (3) can be used during the simulation to calculate the
insulator flashover. CVT 4000 5000 8000
PT 600 500 500
t
k CT 800 680 250
[|Vgap | − V0 ] d t ≥ D. (12)
∫ Autotransformer 5000 2700 3500
t0
As stated in reference [32], when the voltage is close to 0 V

(significantly lower than the critical flashover voltage, V50% ), the 4.9 Substation equipment model
parameter “k” assumes a higher value within the range of 3
to 5. Conversely, when the voltage approaches a high value of According to the IEEE and CIGRE workgroup proposals,
approximately 90% of V50%, “k” equals or less than 1. If the capacitive voltage transformers (CVTs), current transformers
above integral is greater than or equal to D, flashover occurs. (CTs), and autotransformers are modelled as capacitors in the
In this equation, t0 is the moment after which Vgap becomes simulations of transient over-voltages (Table 5) [30].
greater than V0 . When Vgap is lower than V0 , the integral is
reset, and the gap acts as an ideal open switch. The gap stays
closed after the flashover. 4.10 Lightning LP method
For modeling the insulator string with a length of more
than one meter, the leader development method is appro- Different experimental data have indicated that a log-normal
priate as shown in Figure 9. This method mainly considers function can be used to approximate the statistical behavior of
physical aspects and breakdown parameters related to the dis- each lightning parameter [8]:
charge mechanism. This method is approximated based on
experimental results. The differential Equation (13) is used to { ( )2 }
determine the leader length as a function of time. The break- 1 1 ln x − ln x
p (x ) = √ . exp − , (14)
down occurs if the gap length “g” and the leader length “L” are 2𝜋𝜎ln x x 2 𝜎ln x
equal.
[ ] where x̄ and σlnx are the mean and standard deviation of the
dL V (t ) standard logarithmic profile. The joint probability density func-
𝜈= = Ki × V (t ) − Eo , (13)
dt g−L tion of x and y is obtained by considering the log-normal
distribution for the lightning variable as follows:
where ν is leader velocity (m/s), L is the leader length (m), [ ( /( ( ))) ( )]
V(t) is the voltage across the gap (kV), Ki is leader coefficient exp − 1 2 1 − 𝜌2 f1 − f2 + f3
(m2 v−2 s−1 ), E0 critical leader inception gradient (kV/m) and g p (x, y ) = √
2𝜋 1 − 𝜌2 𝜎ln x 𝜎ln y xy
is the gap length (m) [31]. ( )2 ( )( )
ln x−ln x ln x−ln x ln y−ln y ,
f1 = , f2 = 2𝜌
𝜎ln x 𝜎ln x 𝜎ln y
( )2
4.8 Circuit breaker model ln y−ln y
f3 =
𝜎ln y
In this section, the circuit breaker (CB) modelling is proposed (15)
to study lightning and switching overvoltages. Generally, CBs where lnx̄ and ln̄y denote mean values; 𝜎ln x is the standard devi-
of high-voltage substations are divided into the dead tank ation of the random variable x and 𝜎ln y is the standard deviation
and live tank types. Based on experimental results, the capac- of the random variable y while ρ is the correlation coefficient.
itances of CBs and disconnectors of high-voltage substations In the present study, x is the peak amplitude and y is the current
are presented in Table 4 for different voltage levels [30]. rise rate, with the values shown in Table 6 [9].
TABLE 6 Current rise rate and peak amplitude of statistical distributions TABLE 7 IEC pollution severity classification [33].
[9].
Site severity based Leakage distance
σln(I) (pu) ρ (pu) σln(S) (pu) Sm (kA/µs) Im (kA) ESDD (mg/cm2 ) on pollution (inch/kV L-G)
0.68 0.36 0.55 14 30 <0.01 Very light 0.87

0.01–0.04 Light 1.09
0.04–0.15 Medium 1.37
The MC method is not suitable for optimization when there 0.15–0.40 Heavy 1.70
are a large number of scenarios since this method imposes high >0.40 Very heavy 2.11
computational burden in dealing with a large number of simu-
lations. Hence, flashover predictions are obtained considering
the risk concept to reduce the computational burden in the
optimization method.
The incoming overhead lines are the lightning strike points
on the transformers in HV substations. This can be attributed
to the fact that the ground and protection systems are mainly
strengthened against lightning strikes. Thus, the parameters that
affect the IR of BF failure are: the protection of the lines con-
nected to the substation, insulation strength, and tower footing
resistance. In recent years, this field has extensively employed
the limiting parameter (LP) approach [9]. Unlike the methods
that require many iterations, LP can determine the variation
range of minimum lightning current in terms of the current rise
rate variation range. Accordingly, the peak amplitude of light-
ning current can be determined as the product of front time
FIGURE 10 The modified CFO of the vertical suspended insulating
and current rise rate. chain based on contamination level.
The EMTP software is a suite of tools for calculating the
over-voltage value according to the lightning current and cur-
rent rise. Insulation failure occurs if the resulting over-voltage 4.11.1 Contamination effect
amplitude is higher than the transformer insulation strength
level. The current amplitude is considered in the desired range Based on the equivalent salt deposition density (ESDD), the
if it leads to an overvoltage that causes an insulator flashover on IEC 60815 standard presents the pollution effect on the
the incoming line. If the current amplitude causes an overvolt- insulator leakage distance to avoid flashover (Table 7).
age that leads to flashover across the incoming line insulator, According to the IEEE standard, the critical voltage for every
its value is recorded in the range of interest. All procedures are centimeter of the vertical insulating chain under contamination
conducted from minimum to maximum in all current zones. In is determined based on the empirical relationship:
the end, the strength characteristics are used to calculate IR in
the desired zone as follows [10]: 1.64
CF On = 72.3 + , 0.02 < C < 0.04
C
. (17)
R = 0.6Ng L ∫ Dg f (I ) dI 1.96
RiskyArea , (16) CF On = 64.4 + , C > 0.04
C
Ng = 0.04(Td )1.25
The modified Critical flashover voltage (CFOn ) is given in
where L is the line length, Ng is the annual ground flash density mg/cm2 for contamination. Therefore, CFOn according to the
per square kilometre per year, f(I) is the probabilistic distribu- contamination level is illustrated in Figure 10 as below.
tion of lightning current, Td is the number of thunderstorm days From the findings presented in the IEEE Working Group
per year. Additionally, Dg can be determined using Equation (9) paper [1], Figure 10 provides a graphical representation of
and the multiplication factor “0.6″ is the effect of the distri- the CFO in kV from line to ground per meter of con-
bution of the lightning strikes along the TL spans on the BF nected length. This data is delineated concerning the density
rate [4]. of salt deposits in milligrams per square centimetre (mg/cm2 ),
specifically for standard insulators arranged in a vertical con-
figuration. The characteristics of the standard insulator are as
4.11 Environmental effects follows: spacing of 146 mm, diameter of 254 mm, equivalent
to 146 × 254 mm, and a leakage distance of 305 mm per
The effects of environmental conditions, such as high tem- insulator.
perature, rain, snow, salt, and dust, reduce the electrical Recent studies are shown that the morphology of discharges,
characteristics of HV insulators. the current and the voltage as well as the discharge velocity
depends on the polarity of voltage, the configuration and the where CFOs represents the standard insulation strength of the
conductivity of the pollution layer; Obenaus model with Rizk insulator, δ denotes the relative air density calculated P is the
ac reignition conditions has proved to be accurate and adapt- ambient pressure and h is absolute water vapor density. with
able to a reasonable range of station post insulators, where the respect to standard conditions of 101.3 kilopascals pressure and
assumption of uniform pollution resistance per unit length is standard ambient temperature in Kelvin. The correction fac-
more valid. The flashover on an insulator with a polluted sur- tor g for critical flashover V50% across the minimum discharge
face starts with the initiation of an arc at a region of high local path length (dry arc distance) L is [34]:
electrical stress and ends its propagation with increasing speed
along a layer of semiconducting pollution [33, 34]. V50%
g= . (24)
The ac flashover on a polluted surface is considered as an arc 500KV ⋅ L ⋅ 𝛿 ⋅ kov
in series with a residual resistance. The circuit equation for this
Empirical exponents, m and to describe the changing effect of
model is as follows [34]:
atmospheric corrections, based on the predischarge coefficient
g [34]:
Vm = A.x.Im−n + Im .R p (x), (18)
Δg = ||g − 1.1
(|
|
)
where Vm is the peak value of applied voltage (V), A and n are 0, 1.4279 + ,
wm = max
arc constants, x is the local arc length (cm), Im is the peak value Δg ⋅ (−4.0689 + Δg ⋅ (4.5352 − Δg ⋅ 1.9693 ))
of leakage current (A), and RP (x) is the residual resistance (Ω) (25)
of pollution layer from the arc root at x to the ground electrode. {
w, g < 1
w = min (wm , 1 ) m = . (26)
1, g ≥ 1
4.11.2 Altitude and humidity
Contamination test results are usually obtained at nearly
The IEC 60815-2 standard stipulates that the correction fac- 100% relative humidity with values of V 50% less than about
tor at higher altitudes is more critical for lightning impulse than half of the 540 kV/m impulse flashover level. Here, CFOs
pollution performance, so increasing the insulation strength represents the standard insulation strength of the insulator, δ
considering the altitude correction factor would suffice. The denotes the relative air density calculated P is the ambient pres-
IEC 60071–2 standard recommends the safety factor Ks = 1.05 sure and h is absolute water vapor density. with respect to
for external insulation [14]. The correction factor for height is standard conditions of 101.3 kilopascals pressure and standard
calculated in Equation (19). ambient temperature in Kelvin. The coefficient “m” equals 1
for dry conditions, and in humid environments, it ranges from
m(
H
)
0.29 to 0.5. Given the predominantly dry nature of the studied
Ka = e 8150 , (19) area, the value of “m” has been approximated to unity [34]. The
current knowledge of the air humidity influence shows a yet to
where m is 1 for coordination lightning impulse withstand volt- be researched interdependence to the leader length [14]. Thus,
age and H is the altitude above sea level (in meters), m is a factor in some publications, the influence of air humidity is neglected.
as: Hence, further aspects of air humidity will not be a topic in this
( ) paper. Albeit, the authors strongly endorse more precise studies
1.25CF Os CF Os and plan to conduct those in the course of the futures research.
m= − 0.2 , (20)
500S 500S The required rated withstand voltage Urw for equipment
under standard atmospheric conditions can be calculated by cor-
where S is Insulator strings length. recting the coordination withstand voltage of equipment Ucw
Humidity, relative to air density, has a more complex effect on from Equation (27) [35].
the breakdown voltage of air (flashover voltage of the insulator).
KS .Ucw
Specifically, after dew formation, it leads to an increase in the Urw =
insulation strength of the suspension chain. Whereas, before the (Kd .Ka )m , (27)
Kd ∼ 𝛿 m
dew point, which depends on air pressure, the coefficient for the
reduction of strength is calculated from the following relation where Kd is the air density correction function that is presup-
[14]: posed to be equal to the relative air density (δ) and the value of
m is 1.
CF OHumidity−Air Density = 𝛿 m kwov CF Os , (21)
P 293K 4.11.3 Rain effect

𝛿= ⋅ , (22)
101.3kPa 273K + Tambient
( ) Rain reduces the disruptive discharge voltage very consider-
h ably for power frequency and switching impulses but only a
kov = 1 + 0.012 − 11g∕m3 , (23)
𝛿 little for lightning impulses. The power frequency and switching
impulse values for outdoor insulation are accordingly measured TABLE 8 The sample empirical results of icing conditions.
in wet tests, while for lightning impulses only dry tests are Approximate
made [36]. Ice ground-level
Rain is very effective for cleaning the upward-facing surfaces accretion ice Sample value
of insulators but the efficiency for the bottom surface varies level accumulation B in (22) of a in (22) c in (22)
with intensity and quantity [34].The dc breakdown strength of Very light <2 mm ESSD 12.7 −0.36
insulators in wet conditions can be considerably lower than the Light <12 mm CFO decreased 14%
minimum observed flashover gradient of 400 kV/m for either
Moderate 12–20 mm CFO depends on icicle bridging formation
polarity of insulators in dry conditions. Under rain conditions,
Heavy >20 mm ISP 1196 −0.37
the negative polarity flashover of 170–400 kV/m tends to be
lower than positive polarity at 300–500 kV/m, depending on
the insulator type and orientation, rain conductivity, intensity,
and duration. A factor of 0.87 ± 0.12 was suggested by the
CIGRE Working Group 33.07 for the ratio between dry and
5 PROPOSED METHOD
wet ac flashover of insulator strings. The IEC wet test gives a
The algorithm proposed for optimum SA placement is shown in
ratio of dry to wet flashover that is relatively constant at about
Figure 11. As shown in this figure, at first, the structural data of
0.7 for longer string lengths [34].
the substation are received and implemented in EMTP. In the
next step, MC simulation is started by applying random sam-
ples of lightning current to the closet tower to the substation.
4.11.4 Icing effect
A total of 30 N flashovers are generated in each iteration (N
is the average yearly flashovers), and for transient state simula-
The impact of ice accretion on insulation strength and flashover
tion, only flashovers in the S-I range are applied to EMTP. For
voltage depends on various factors, including the distribution of
the last five towers of the substation, simulations are repeated
the electric field, wetness and contamination level of the ice sur-
and the value of IR is recorded. Also, for the SAs with differ-
face, the number and location of air gaps formed by ice icicles
between the insulating chains, environmental conditions, and
the shape and dimensions of the ice layer on the insulation. In
standard environmental conditions, insulators are designed not
only to provide adequate clearance between parts at different
potentials but also to create the maximum creepage distance for
the passage of the minimum leakage current. However, the for-
mation of a standard ice layer on the suspension chain alters its
insulation strength. The standard insulation strength equation
under icing conditions is as follows:
CFO = a B c , (28)
where CFO, with units of kV/mleakage , represents the flashover

voltage along the leakage current path. The coefficients “a”
and “c,” as well as the quantity “B,” depend on the thickness
of the ice accretion. When light ice accretions occur, leakage
currents pass over the surface of the insulator, thereby weak-
ening its insulation strength due to the contamination present
in the ice layer. Consequently, the quantity “B” in this case will
be related to the ice contamination or ESSD (Equivalent Salt
Deposit) with units of (gr/mm2 ). However, in the case of heavy
ice accretions, the form of the insulation’s creepage distance will
be influenced by the formation of ice. In this situation, the leak-
age current path is closed by the ice, and the quantity “B” is
replaced by icing stress product (ISP). This quantity is the prod-
uct of accumulated ice on every centimeter of insulation length
in its electrical conductivity. The coefficients “a” and “c” also
depend on the specific conditions. Table 8 presents some values
obtained from standard tests.
Of course, due to the placement of the studied post in a warm
and dry region, the influence of ice accretion on the insulation
strength of the suspension chain has been disregarded. FIGURE 11 Flowchart of the proposed method.
ent distances from the transformer (maximum up to 10 m), the TABLE 10 Coefficients m and n of the transformer [19].
algorithm is repeated and finally, the location for installing the Coefficient m a k
SA with the minimum IR is chosen.
Value 12 0.39 30
6 SIMULATION RESULTS AND

DISCUSSION flash density) is 1.08 per square kilometre per year. The length
of tower spans and other data are given in Table 10. The val-
6.1 Substation design ues resulting from experiments of the transformer tests are as
follows:
In this paper, the high-frequency model of the main equip-
ment of a high-voltage substation is used to study lightning and
switching transients. Therefore, the detailed substation structure
6.2 Lightning BF
is considered in this study. The high-voltage substation of “ALI-
A “back flashover” will occur from tower to phase conductor
ABAD” (in the north of Iran) has four 400 kV, eight 230 kV, and
[37]. Three test points are selected on the 63, 230 and 400 kV
six 63 kV lines. The structure of this substation and the lines
transformers to investigate and evaluate the over-voltage ampli-
connected to it are implemented in the EMTP-RV software
tude of the lightning BF. Moreover, for exploring the lightning
environment to carry out lightning studies.
over-voltage amplitude on high-voltage substations caused by
The substation has transformers with 1050 kV impulse
direct lightning flashovers, it is assumed that the locations of
withstand level (BIL) on an effectively earthed system. The
flashovers are on the guard wires of 230 kV, 63 kV, and 400 kV
specifications of the line connecting (Shahid Salimi) to the
TLs. The over-voltages on bushing terminals of the 400 kV
high-voltage substation are shown in Table 9:
(T3 transformer), 230 kV (T2 transformer), and 63 kV (T6
Figure 12 shows a part of this simulation. The chosen time
transformer) are caused by 96 kA current on the TL guard
step for simulation was 0.01 microseconds (µs), with a total
wire (second tower) with the SA installation at 5 m distance
simulation time of 200 µs. The value of Ng (average ground
from the transformers, as shown in Figure 13. It is obvious
that all current surges due to lightning do not cause BF in
TABLE 9 The specifications of the line (Shahid Salimi) connected to the power networks. Therefore, the over-voltages that cause light-
UHV substation. ning BF are evaluated using the finite-area MC method in the
Line specifications Unit Value following.
Rated voltage kV 400
Maximum voltage kV 420 6.3 Lightning LP method
Conductor type – ACSR (Canary)
For obtaining I-S characteristics, the simulations of 10,000 sam-
Guard wire – ACS (Curlew Core)
ples produced using the MC method are used, as shown in
Number of circuits number 1 Figure 14. Results of different simulations indicate a maximum
Line length km 140 of three flashovers in the shaded area for 233 strokes (seven
Pollution class of the region – Light Medium data sets) over the service life. Strokes are extracted from the
Minimum/maximum degrees centigrade −20/40 log-normal distribution with parameter values shown in Table 6,
temperature of the region and in this distribution, data points are denser around the region
Tower height m 30 with lower amplitude. The lightning flashovers that reach the
transformer after over-voltage correspond to the first to fifth
Height from sea level m 140
towers close to the substation. Considering the lightning wave-
The annual number of days days 14
length, line wave impedance, and tower resistance, other towers
with thunderstorms
(corresponding to Keraunic do not significantly contribute to the final risk value.
level) In the end, the efficiency of the proposed method is con-
Relative humidity % 90 firmed by comparing the results of the present study with those
in the literature. Reference [11] addresses the SA placement in
Wind speed m/s 30
Iran’s 400 kV sample network. This network has 60 nodes at
Line economic span m 300
a distance of 15 km from each other. As expected, the risk
(average)
decreases at further distances from the substation. In ref. [11],
Surge impedance of the ohms 130
the effect of stroke accumulation is not included, and the time-
tower
dependent BIL curve is used to calculate the IR of failure.
Maximum footing resistance ohms 30
According to Table 11, the method proposed in this study is
(ohms)
more efficient than the method presented in Ref. [11]. In this
Approximate length of the mm 3264
regard, the predicted over-voltage amplitude with optimal SA
insulator
placement based on our proposed method is lower than that
FIGURE 12 A part of the EMTP-RV environment; (a) substation and (b) transmission line (Shahid Salimi).
FIGURE 13 Transformer transient voltage waveform on phase A protected by SA at a distance of three meters away from the occurrence of a BF failure on
the 400 kV line (I = 150 kA, S = 40 kA/µs).
based on the method presented in ref. [11]. At the same time, the obtain the V-T characteristic. This figure is plotted using the
number of simulations is significantly reduced in our method. transient simulation based on EMTP in MATLAB.
Figure 15 illustrates the instantaneous values of Vm Ta in each As shown in Figure 16, the value of the Vm Ta index for
time step from Figure 13; these values are then integrated to the over-voltage stresses of different lines is extracted. Then,
FIGURE 14 Seven groups of random values were generated in the MC

simulation for I and S using the log-normal distribution.
TABLE 11 The proposed method with the method presented in Ref. [11]
for optimal SA placement.
Percentage
reduction in
over-voltage
amplitude using the Proposed
proposed method method Method in ref. [11] Line
1.9 Node 12 Nodes 12,13 SL102

1.3 Node 36 Nodes 36, 42 SA913
2.1 Node 53 Node 53 SA912
TABLE 12 Optimum SA placement.
Bus voltage (kV) 63 230 400

IR 9.23*10−9 4.84*10−7 1.6941*10−5
Optimum SA distance from 0.5 2 2.5
the autotransformer (m)
using the proposed method, after calculating the IR accord-

ing to Table 12, the optimum location for installing the SA is
determined.
FIGURE 15 Vm Ta based on Figure 13 for 63 kV (blue), 230 kV (red),

7 CONCLUSIONS and 400 kV (green) lines.
This research aimed to present a method to determine the

optimum SA placement in high-voltage substations. The pro- uses the Weibull distribution to increase accuracy. To prove the
posed algorithms optimized the SA location considering IR. proper performance of the proposed method, a sample UHV
Previous methods have not considered the effects of the accu- substation (ALIABAD high-voltage substation) was studied.
mulation of lightning strokes on the transformer insulation. Simulations were performed using special software (EMTP-RV)
However, while investigating the simultaneous effect of possi- to analyse transient over-voltages.
ble stresses based on the switching and lightning over-voltages, The results of simulating the real UHV substation showed
the proposed method uses a more appropriate statistical distri- the efficiency of the proposed method. Also, by presenting the
bution to determine possible flashovers. Furthermore, it uses LP method while maintaining accuracy, the computational bur-
accurate modeling for transients of equipment to achieve accu- den was reduced significantly compared to other methods. To
rate results. Contrary to the existing techniques, this method increase the accuracy of the results, in contrast to other existing
4. Savić, M.S.: Engineering method for high-voltage substations lightning

performance estimation. IEE Proc. C (Gener. Transm. Distrib.) 136 (4),
222–229 (1989)
5. Wang, L., Wan, K., Chen, L., Qian, Q., Huang, J.: Analysis about poten-
tial distribution of the Ur 780 kV metal-oxide surge arrester for 1000 kV
substation. In: 2020 IEEE 4th Conference on Energy Internet and Energy
System Integration (EI2), pp. 1125–1130. IEEE, Piscataway (2020)
6. Nafar, M., Solookinejad, G., Jabbari, M.: Comparison of IEEE and Pinceti
models of surge arresters. Res. J. Eng. Sci. (5), 32–34 (2014)
7. Li, Z., Lou, Y., Cui, T., Zha, Z.: Study on the Effect of Optimizing the
Configuration of Metal-oxide Arresters in UHV Engineering on Suppress-
ing Slow-front overvoltage. In: 2022 7th Asia Conference on Power and
Electrical Engineering (ACPEE), pp. 1522–1527. IEEE, Piscataway (2022)
8. Zalhaf, A.S., Han, Y., Yang, P., Wang, C., Khan, M.A.: Analysis of lightning
transient performance of 132 kV transmission line connected to Miramar
wind farm: A case study. Energy Rep. 8, 257–265 (2022)
FIGURE 16 Vm Ta index and IR profile using the proposed method. 9. Savic, M.S., Savic, A.M.: Substation lightning performance estimation due
to strikes into connected overhead lines. IEEE Trans. Power Delivery
30(4), 1752–1760 (2015)
10. IEC 62305-3: Protection against lightning—Part 3: Physical damage to
techniques, voltage-time characteristics of the transformer were structures and life hazard (2010)
obtained from experiments. Regional electricity companies 11. Shariatinasab, R., Vahidi, B., Hosseinian, S.H., Ametani, A.: Optimiza-
can use the results of the conducted investigation because of tion of surge arrester’s location on EHV and UHV power networks
the comprehensiveness of modeling and the accuracy of the using simulation optimization method. IEEJ Trans. Power Energy 128(12),
1465–1472 (2008)
solutions. 12. Castro, W.S., Lopes, I.J., Missé, S.L., Vasconcelos, J.A.: Optimal place-
ment of surge arresters for transmission lines lightning performance
AUTHOR CONTRIBUTIONS improvement. Electr. Power Syst. Res. 202, 107583 (2022)
Faridoddin Safaei: Conceptualization; methodology; soft- 13. Rao, M.M., Lanjewar, A., Tiwari, N.: Analytical and experimental studies
on 245 kV gas insulated surge arrester. Electr. Power Syst. Res. 204, 107713
ware; data curation; software; investigation; writing—original (2022)
draft preparation. Mohsen Niasati: Methodology; concep- 14. Schubert, U., Shirvani, A., Schmidt, U., Kornhuber, S., Kynast, E.: Pro-
tualization; investigation; supervision; project administration; posal for a general atmospheric correction method of breakdown and
validation; writing—reviewing and editing. withstand voltages of airgap insulated configurations based on a streamer–
leader differentiated model of the breakdown process. Energies 11(4), 776
(2018)
CONFLICT OF INTEREST STATEMENT 15. Orille, A.L., Bogarra, S., Grau, M.A., Iglesias, J.: Fuzzy logic techniques
The authors whose names are listed at the top certify that they to limit lightning surges in a power transformer. In: 2003 IEEE Bologna
have no affiliations with or involvement in any organization or Power Tech Conference Proceedings, vol. 2, pp. 6–pp. IEEE, Piscataway
entity with any financial interest (such as honoraria; educational (2003)
16. Vilar, P.B., Lira, G.R.S., Costa, E.G., Brito, V.S., de Albuquerque Maia, M.J.:
grants; participation in speakers’ bureaus; membership, employ- Development of a model for calculation of energy absorbed by a ZnO
ment, consultancies, stock ownership, or other equity interest; surge arrester during transients. IEEE Trans. Power Delivery 37(1), 136–
and expert testimony or patent-licensing arrangements), or 145 (2021)
non-financial interest (such as personal or professional relation- 17. Ahmadi Kamarposhti, M., Hejri, A.: Optimal location of surge arresters
ships, affiliations, knowledge or beliefs) in the subject matter or in distribution network considering reliability and technical and economic
factors to reduce costs using ICA algorithm. Int. J. Smart Electr. Eng.
materials discussed in this manuscript. 10(01), 7–16 (2021)
18. Thomas, B., Savadamuthu, U.: Impulse breakdown characteristics of aged
DATA AVAILABILITY STATEMENT oil impregnated paper. IEEE Trans. Dielectr. Electr. Insul. 24(4), 2354–
Data sharing not applicable to this article as all data is available 2361 (2017)
19. Ikeda, M., Okubo, H., Yanari, T.: PD and BD Probability Distribution And
in the article. Equiprobabilistic VT Characteristic of Oil-Filled Transformer Insulation.
Toshiba Corp., Kawasaki (1982)
ORCID 20. Shindo, T., Suzuki, T.: A new calculation method of breakdown voltage-
Faridoddin Safaei https://orcid.org/0009-0005-3193-4812 time characteristics of long air gaps. IEEE Trans. Power Appar. Syst. PAS-
104(6), 1556–1563 (1985)
21. Safaei, F., Niasati, M.: Calculation of UHV transformer insulation risk by
REFERENCES extended Monte Carlo method. Electr. Power Syst. Res. 218, 109219 (2023)
1. Hileman, A.R.: Insulation Coordination for Power Systems. CRC Press, 22. IEC 60099-5: Surge arresters—Part 5: Selection and application recom-
Boca Raton, FL (2018) mendations (2018)
2. Sima, W., Lan, X., Yang, Q., Yuan, T.: Statistical analysis on measured light- 23. IEC 60076-3: Power Transformer—Part 3: Insulation levels, dielectric tests
ning overvoltage surges in a 110 kV air-insulated substation. IET Sci. Meas. and external clearances in air (2000)
Technol. 9(1), 28–36 (2015) 24. Kawamura, T.: Temporary overvoltages and AC test voltage in 550 kV
3. Huang, L., Zhou, L., Chen, W., Wei, R., Zhang, D., Wang, D., … Ma, Y.: system with reduced insulation level CIGRE33-203 (1990)
Analysis of thermal effects for polymer-housed metal-oxide surge arrester 25. Khodsuz, M.: Externally gapped line arrester performance in high voltage
under multiple strokes. IEEE Trans. Power Delivery 37(5), 3917–3927 transmission line using frequency grounding system: Absorbed energy and
(2022) expected life assessment. IET Sci. Meas. Technol. 16(7), 426–440 (2022)
26. Gutierrez, J.A.R., Moreno, P., Guardado, L., Naredo, J.L.: Comparison of 33. Amrani, M.L., Bouazabia, S., Fofana, I., Meghnefi, F., Jabbari, M., Khelil,
transmission tower models for evaluating lightning performance. In: 2003 D., Boudiaf, A.: Modelling surface electric discharge propagation on
IEEE Bologna Power Tech Conference Proceedings, vol. 4, p. 6. IEEE, polluted insulators under AC voltage. Energies 14(20), 6653 (2021)
Piscataway (2003) 34. Farzaneh, M., Chisholm, W.A.: Insulators for Icing and Polluted Environ-
27. Martinez, J.A., Castro-Aranda, F.: Lightning performance analysis of over- ments. Wiley-IEEE Press, Piscataway (2009)
head transmission lines using the EMTP. IEEE Trans. Power Delivery 35. IEC 60071-2: Insulation Coordination Part 2: Application Guidelines
20(3), 2200–2210 (2005) (2018)
28. CIGRE: Guide to procedures for estimating the lightning performance 36. Diesendorf, W.: Insulation Co-ordination in High-Voltage Electric Power
of transmission lines, p. 63. CIGRE Technical Brochure, CIGRE, Paris Systems. Butterworth and Co (Publishers) Ltd., London (1974)
(1991) 37. CIGRE Working Group: Substations. International Council on Large
29. TRIDELTA Group: Home page. http://www.tridelta.de Electric Systems (CIGRE) Study Committee B3 (2019)
30. Imece, A.F., Durbak, D.W., Elahi, H.: Modeling guidelines for fast front
transients. IEEE Trans. Power Delivery 11(1), 493–506 (1996)
31. Datsios, Z.G., Mikropoulos, P.N.: Modeling of lightning impulse behavior
of long air gaps and insulators including predischarge current: Implications How to cite this article: Safaei, F., Niasati, M.: A new
on insulation coordination of overhead transmission lines and substations. method for surge arrester placement in high-voltage
Electr. Power Syst. Res. 139, 37–46 (2016) substations considering environmental effects. IET Sci.
32. Inthoulath, K., Banmongkol, C.: A simulation of breakdown characteris-
Meas. Technol. 1–14 (2024).
tics of suspension insulators under standard lightning impulse voltages.
In TENCON 2014-2014 IEEE Region 10 Conference, pp. 1–4. IEEE, https://doi.org/10.1049/smt2.12208
Piscataway (2014)

IET Science Measure Tech - 2024 - Safaei - A New Method For Surge Arrester Placement in High Voltage Substations

Uploaded by

Copyright:

Available Formats

IET Science Measure Tech - 2024 - Safaei - A New Method For Surge Arrester Placement in High Voltage Substations

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

IET Science Measure Tech - 2024 - Safaei - A New Method For Surge Arrester Placement in High Voltage Substations

Uploaded by

Copyright:

Available Formats

Received: 18 May 2023 Revised: 4 January 2024 Accepted: 5 June 2024 IET Science, Measurement & Technology

A new method for surge arrester placement in high-voltage

Faridoddin Safaei Mohsen Niasati

Faculty of Electrical and Computer Engineering Abstract

1 INTRODUCTION reasons for outages in UHV substations is the over-voltages

IET Sci. Meas. Technol. 2024;1–14. wileyonlinelibrary.com/iet-smt 1

3 INSULATION FAILURE RISK OF

FIGURE 2 Transformer test conditions [24].

where subscript t is related to test time intervals, S corresponds

FIGURE 5 Electro-geometric model representation of conductors and

The following equations can be used to calculate the values

where angle δ is shown in Figure 5, and 2β is measured at point

TABLE 2 IEEE model parameters and characteristics of SAs.

Voltage level (kV)

SA parameters (TRIDELTA) 63 230 400

Arrester height (mm) 890 1980 4400

FIGURE 7 V-I characteristics of A0 and A1 [6] (the voltages in p.u. are

FIGURE 8 High-frequency transformer model [30].

d: the height of the SA in meters

4.6 FD model of transformer

TABLE 4 Circuit breaker modeling parameters [30].

Capacitance values (pF)

Equipment <765 kV <400 kV <115 kV

Dead tank circuit breaker 600 150 100

TABLE 5 Capacitance values of substation equipment [30].

Equipment <765 kV <400 kV <115 kV

As stated in reference [32], when the voltage is close to 0 V

0.68 0.36 0.55 14 30 <0.01 Very light 0.87

P 293K 4.11.3 Rain effect

where CFO, with units of kV/mleakage , represents the flashover

6 SIMULATION RESULTS AND

FIGURE 14 Seven groups of random values were generated in the MC

1.9 Node 12 Nodes 12,13 SL102

TABLE 12 Optimum SA placement.

Bus voltage (kV) 63 230 400

using the proposed method, after calculating the IR accord-

FIGURE 15 Vm Ta based on Figure 13 for 63 kV (blue), 230 kV (red),

This research aimed to present a method to determine the

4. Savić, M.S.: Engineering method for high-voltage substations lightning

You might also like