1 An Improved Dynamic Multi-Arcs Modeling Approach For Pollution Flashover of Silicone Rubber Insulator
1 An Improved Dynamic Multi-Arcs Modeling Approach For Pollution Flashover of Silicone Rubber Insulator
1 An Improved Dynamic Multi-Arcs Modeling Approach For Pollution Flashover of Silicone Rubber Insulator
1, FEBRUARY 2022 77
Abstract — This article presents a dynamic modeling model discharge of the insulator. Recently, single-arc modeling
approach to analyze pollution flashover of silicone rubber approach is done by the use of circuit modeling [6], analytical
insulators. Pre-flashover conditions can be used to rep- modeling [2], [7], and artificial intelligence (AI) [8]. In addi-
resent analytical formulations for different stages of dry
band arcing activities on the surface of polluted insulator. tion, a recent study is done to describe the dynamic behavior of
In this article, effects of dust deposit as nonsoluble deposit FOV in polyvinyl chloride (PVC) insulators with continuous
density (NSDD) are investigated on the dynamic behavior and discontinuous pollution types. This research develops a
of partial arcs on the surface of insulators by means of a comprehensive analysis of arcing propagation [9].
proposed multi-arcs modeling approach. It is shown that Multi-arcs approach in polluted insulators originated from
analytical formulations of dynamic resistance of surface
pollution can be used to analyze different modes of arcing the work in [4], and it has attracted the interest from other
as unstable and stable. Real data of leakage current (LC) researches [10]–[12]. Recently, various methods based on
and flashover voltage (FOV) of polluted insulator are used multi-arcs modeling approach are developed such as the study
to make correlation with analytical multi-arcs formulations. on iced insulators with dc voltage input [13], performance
The proposed analytical multi-arcs model is based on an analysis of multi-arcs model with impulse voltage input [10],
artificial neural network (ANN) which is able to predict the LC
and FOV of the insulator in real-time monitoring. In addition, and other multi-arcs modeling to improve prediction of
effects of pollution resistance as variable surface conduc- dynamic behavior of ice-covered insulators [12]. Dynamic
tion and thickness of pollution layer are investigated on LC behavior of partial arcs is dependent on pollutant materials.
and FOV of the insulators. It is shown that the proposed A critical challenging service of over-head distribution lines
multi-arcs representation has a closer correlation with real in the south of Iran is the existence of dust storms and coastal
testing data in comparison to previous single-arc and multi-
arcs models. pollutants which are deposited on sheds of insulators during
time. When combined with water in wet conditions, this results
Index Terms — Dynamic modeling, flashover prediction, in flashover of insulators and outages of power distribution
flashover voltage (FOV), leakage current (LC), pollution
flashover, self-consistent multi-arcs dynamic model, sili- lines.
cone rubber insulator. The purpose of this article is to represent an improved
multi-arcs circuit model for silicone insulators which is able
to predict leakage current (LC) and flashover voltage (FOV)
I. I NTRODUCTION
of insulators according to the studied pollution. In addition,
1070-9878 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: VTU Consortium. Downloaded on December 12,2023 at 09:39:52 UTC from IEEE Xplore. Restrictions apply.
78 IEEE TRANSACTIONS ON DIELECTRICS AND ELECTRICAL INSULATION, VOL. 29, NO. 1, FEBRUARY 2022
Authorized licensed use limited to: VTU Consortium. Downloaded on December 12,2023 at 09:39:52 UTC from IEEE Xplore. Restrictions apply.
SEZAVAR et al.: IMPROVED DYNAMIC MULTI-ARCS MODELING APPROACH FOR POLLUTION FLASHOVER 79
Authorized licensed use limited to: VTU Consortium. Downloaded on December 12,2023 at 09:39:52 UTC from IEEE Xplore. Restrictions apply.
80 IEEE TRANSACTIONS ON DIELECTRICS AND ELECTRICAL INSULATION, VOL. 29, NO. 1, FEBRUARY 2022
Authorized licensed use limited to: VTU Consortium. Downloaded on December 12,2023 at 09:39:52 UTC from IEEE Xplore. Restrictions apply.
SEZAVAR et al.: IMPROVED DYNAMIC MULTI-ARCS MODELING APPROACH FOR POLLUTION FLASHOVER 81
E. Computation Steps
The proposed model is based on the equivalent circuit that
was shown in Fig. 3(b) in which Rdi includes the resistances
of dry band arcing which were represented in (3), (12), (19),
and (22). In this research, an ANN model is developed to
do required calculation of the circuit model. The designed
Fig. 7. Variation in voltage (green color) and current (dashed line with ANN has two main purposes: first is the calculation of circuit
blue color) during oscillations in stable mode.
parameters and second is the prediction of LC and FOV of
insulator.
To achieve these two purposes, a feed-forward ANN type is
5 kHz [based on Figs. 5(b) and 6]. As LC increases up to 5 mA developed which uses testing data of silicone insulators with
[average value based on Fig. 5(c)], it enters the stable mode different profiles as input values to train the ANN model [8].
and f gradually tends to be zero [20]. Hence, f can be The Levenberg–Marquardt algorithm is adopted and 38 hidden
expressed as follows: layers are used in the development of the designed ANN
f = 1250 ∗ (5 − Ia ) (18) model. Then, a genetic algorithm (GA) is used to optimize
obtained results and achieve method of least square (MLS).
which Ia is the LC of the insulator in mA. According to the The first purpose is to predict the circuit parameters according
calculated peak value of dynamic resistance from the Mayr to the following equations [10]:
equation as Rm and variation in the dynamic resistance in the Rd1 (x, t)
unstable arc mode close to this value as well, the maximum Vin (t) − V1 (t) = V4 (t)
Rd4 (x, t)
dynamic resistance can be approximated ten times of Rm [20];
hence, can be calculated according to Fig. 5. Therefore, the 4
d V i (t)
+ Rd1 (x, t) Ci (x, t) (23)
final expression for dynamic resistance is obtained as follows: dt
i=1
N R p1 (x, t)
1 1 V1 (t) − V2 (t) = V4 (t)
Rua (t, Ia ) = 2 2
. Rd4 (x, t)
10π Rm n=1
t − 1250∗(5−I
n
a)
+ 1
10Rm 4
d V i (t)
(19) + R p1 (x, t) Ci (x, t) (24)
i=2
dt
Rd2 (x, t) + Rd3 (x, t)
V2 (t) − V3 (t) = V4 (t)
Rd4 (x, t)
D. Dynamic Resistance Modeling in Stable Mode
4
According to Fig. 5(c), the dynamic resistance in stable + Rd2 (x, t) + Rd3 (x, t) Ci (x, t)
mode can be represented as sinusoidal resistance similar to i=3
Fig. 7. According to [20], [25], and Fig. 7, the arc voltage as d V i (t)
× (25)
u a (t) can be expressed by means of u t1 as arc ignition voltage dt
Authorized licensed use limited to: VTU Consortium. Downloaded on December 12,2023 at 09:39:52 UTC from IEEE Xplore. Restrictions apply.
82 IEEE TRANSACTIONS ON DIELECTRICS AND ELECTRICAL INSULATION, VOL. 29, NO. 1, FEBRUARY 2022
R p2 (x, t)
V3 (t) − V4 (t) = V4 (t)
Rd4 (x, t)
d V 4 (t)
+ R p2 (x, t)C4 (x, t) . (26)
dt
According to [10], when the arc is expanded on the surface
of the insulator, it can cause a “bridge mode” on the polluted
band; based on this explanation, the proposed circuit topology
is changed as shown in Fig. 8. The changes include either a
bridge in the first band as B1 [see Fig. 8(a)] or a bridge in the
second band as B2 [see Fig. 8(b)]. Then, the ANN model is
trained so as to calculate the critical LC and FOV according
to prediction of time series of LC. This training would be
achieved based on the circuit parameters and impedance crite-
rion in critical conditions which are illustrated in (27) and (28)
[according to Fig. 3(b)]: Fig. 9. Flowchart of computational steps based on the proposed ANN.
To make validation of the proposed model, in the first which I Lpeaki is the maximum instantaneous value of LC.
step, the whole process of the developed multi-arcs model A regression model based on MLS is used to quantify the
Authorized licensed use limited to: VTU Consortium. Downloaded on December 12,2023 at 09:39:52 UTC from IEEE Xplore. Restrictions apply.
SEZAVAR et al.: IMPROVED DYNAMIC MULTI-ARCS MODELING APPROACH FOR POLLUTION FLASHOVER 83
Fig. 10. Variation in LC during test. The curves are displayed for the
single-arc model, multi arc model, proposed model, and experimental
values, respectively, from top to bottom. Fig. 11. Variation in critical LCs for real testing data and different models
according to ESDD.
TABLE I
G EOMETRICAL PARAMETERS OF THE S AMPLE OF VALIDATION T ESTS
TABLE II
C ORRELATION P ERCENTAGE B ETWEEN R EAL T ESTING
D ATA AND S IMULATION M ODELS
Fig. 12. Variation in critical LCs for real testing data and different models
according to the thickness of pollution layer.
with each other. The physical reason for this point can be
correlation performance of all the investigated models with explained that because in low-level surface pollution, dynamic
real testing data according to the following equation: behavior of arcs includes discharge activities in low LCs, and
N thus it hardly shows dry band arcing oscillations. Therefore,
(yi − f (t))2
R 2 = 100% − ∗ 100 (30) discharge resistance is dominantly the same as the resistance
i=1
σ2 which is represented in multi-arcs modeling of those in [10]
which f (t), yi , and σ are the real testing data, numerical sim- and [13]. However, for high-level pollution, it can be seen
ulation data, and variance of data, respectively. The correlation that the correlation of critical LC for the proposed model is
percentage according to R 2 between real testing data and all higher than those of common multi-arcs and single-arc. This
the investigated models is shown in Table II. As can be seen obviously shows an improved modeling in the prediction of
from the table, the correlation of LC prediction during the critical LC in high-level pollution.
testing time for the proposed model is obtained more than Another important point is the sensitivity of critical LC
other models. according to the thickness of pollution layer. In this matter,
related modeling for the resistance of pollution layer according
to the represented models and real tests can be considered.
B. Critical LC The variation in critical LCs for real test and different models
Critical LC is defined as the maximum LC of the insulator according to the thickness of pollution layer is shown in
on the verge of flashover or the corresponding LC when FOV Fig. 12. As can be seen from the figure, with the increase
is recorded. Since the condition of critical LC for testing in the thickness of pollution layer, the measurements for
the insulator occurs just for once in each test, experiments the proposed model, common multi-arcs model, and single-
are performed for several ESDD and several thickness of arc model show higher deviation from real testing data. The
pollution layer as variables of the tests. The results of the
real main reason for this matter is the differences between real
tests and numerical simulations according to ESDD (mg cm2 ) resistances of dynamic pollution layer and the simulated
are shown in Fig. 11. As can be seen from the figure, the resistances of pollution layer which cause some error in
results of the proposed multi-arcs model have close correlation calculation of numerical simulation according to the variation
with real testing data. In addition, it can be observed that in the thickness of pollution layer. Comparison of the results
for low values of ESDD, the proposed multi-arcs model and of numerical modeling and real tests shows that real testing LC
previously presented multi-arcs model have high correlation is always lower than the modeling LC which is obvious from
Authorized licensed use limited to: VTU Consortium. Downloaded on December 12,2023 at 09:39:52 UTC from IEEE Xplore. Restrictions apply.
84 IEEE TRANSACTIONS ON DIELECTRICS AND ELECTRICAL INSULATION, VOL. 29, NO. 1, FEBRUARY 2022
Fig. 13. Correlations between flashover voltages for real testing data Fig. 14. Correlations between flashover voltages for real testing data
and all models according to the variation in ESDD. and all models according to the variation in thickness of pollution layer.
TABLE III
Fig. 12; this is because of the difference between the resistance P ROFILE PARAMETERS OF THE S AMPLES
of pollution layer in real case and numerical modeling. In real
case, the thickness of pollution layer results in a higher
resistance than that of numerical modeling in which as the
thickness of pollution layer increases, its effect also more
remarkably increases. Therefore, real testing LC is lower than
modeling LC. This is obviously demonstrated in Fig. 12 as
the thickness of pollution layer increases.
Authorized licensed use limited to: VTU Consortium. Downloaded on December 12,2023 at 09:39:52 UTC from IEEE Xplore. Restrictions apply.
SEZAVAR et al.: IMPROVED DYNAMIC MULTI-ARCS MODELING APPROACH FOR POLLUTION FLASHOVER 85
insulators are used to make correlation with the proposed [10] N. Dhahbi-Megriche and A. Beroual, “Self-consistent multi-arcs
analytical circuit-based formulations. In addition, correlations dynamic model for high voltage polluted insulators,” IEEE Trans.
Dielectr. Electr. Insul., vol. 23, no. 5, pp. 2899–2907, Oct. 2016.
between the proposed analytical arcing model with the varia- [11] M. Farzaneh and J. Zhang, “A multi-arc model for predicting AC critical
tion on surface conduction and thickness of pollution layer flashover voltage of ice-covered insulators,” IEEE Trans. Dielectr. Electr.
are investigated. The entire process of dynamic modeling Insul., vol. 14, no. 6, pp. 1401–1409, Dec. 2007.
[12] S. Farokhi and M. Farzaneh, “Multiple-arc propagation over an ice
analysis is proposed as a flowchart. Validation of the proposed surface,” in Proc. Electr. Insul. Conf. (EIC)., Jun. 2011, pp. 309–312.
multi-arcs approach is investigated through comparison with [13] S. Taheri, M. Farzaneh, and I. Fofana, “Improved dynamic model of
previously presented works as single-arc model and common DC arc discharge on ice-covered post insulator surfaces,” IEEE Trans.
Dielectr. Electr. Insul., vol. 21, no. 2, pp. 729–739, Apr. 2014.
multi-arcs model. It is shown that the proposed model has [14] I. Ahmadi-Joneidi, A. Majzoobi, A. A. Shayegani-Akmal, H. Mohseni,
a closer correlation with real testing data in comparison to and J. Jadidian, “Aging evaluation of silicone rubber insulators using
previous works. leakage current and flashover voltage analysis,” IEEE Trans. Dielectr.
Electr. Insul., vol. 20, no. 1, pp. 212–220, Feb. 2013.
[15] M. E.-A. Slama, A. Beroual, and H. Hadi, “The effect of discontinuous
R EFERENCES non uniform pollution on the flashover of polluted insulators under
lightning impulse voltage,” in Proc. Int. Conf. High Voltage Eng. Appl.,
[1] N. Dhahbi-Megriche and A. Beroual, “Flashover dynamic model of Sep. 2012, pp. 246–249.
polluted insulators under AC voltage,” IEEE Trans. Dielectr. Electr. [16] O. Mayr, “Beiträge zur Theorie des statischen und des dynamischen
Insul., vol. 7, no. 2, pp. 283–289, Apr. 2000. lichtbogens,” Archiv Elektrotechnik, vol. 37, no. 12, pp. 588–608, 1943.
[2] M. Slama, A. Beroual, and H. Hadi, “Analytical computation of dis- [17] N. Dhahbi-Megriche, A. Beroual, and L. Krähenbähl, “A new proposal
charge characteristic constants and critical parameters of flashover of model for flashover of polluted insulators,” J. Phys., vol. 30, no. 5,
polluted insulators,” IEEE Trans. Dielectr. Electr. Insul., vol. 17, no. 6, p. 889, 1997.
pp. 1764–1771, Dec. 2010. [18] R. Wilkins, “Flashover voltage of high-voltage insulators with uniform
[3] M. E. Ibrahim, A. M. Abd-Elhady, N. A. Sabiha, and M. A. Izzularab, surface-pollution films,” IET Proc. Inst. Electr. Eng., vol. 116, no. 3,
“Dynamic flashover model considering pollution layer resistance vari- pp. 457–465, 1969.
ation for fixed washed high voltage insulators,” IEEE Trans. Dielectr. [19] R. Wilkins and A. Al-Baghdadi, “Arc propagation along an electrolyte
Electr. Insul., vol. 24, no. 5, pp. 2960–2967, Oct. 2017. surface,” IET Proc. Inst. Electr. Eng., vol. 118, no. 12, pp. 1886–1892,
[4] T. Cheng, C. Wu, and H. Nour, “DC interfacial breakdown on contami- 1971.
nated electrolytic surfaces,” IEEE Trans. Electr. Insul., vol. EI-19, no. 6, [20] X. Zhang, A. Bruce, S. Rowland, V. Terzija, and S. Bu, “Modeling the
pp. 536–542, Dec. 1984. development of low current arcs and arc resistance simulation,” IEEE
[5] L. Lan and R. Gorur, “Computation of AC wet flashover voltage of Trans. Dielectr. Electr. Insul., vol. 25, no. 6, pp. 2049–2057, Dec. 2018.
ceramic and composite insulators,” IEEE Trans. Dielectr. Electr. Insul., [21] T. Yamashita, R. Ishimoto, and T. Furusato, “Influence of series resis-
vol. 15, no. 5, pp. 1346–1352, Oct. 2008. tance on dry-band discharge characteristics on wet polluted insula-
[6] T. Chihani, A. Mekhaldi, A. Beroual, M. Teguar, and D. Madjoudj, tors,” IEEE Trans. Dielectr. Electr. Insul., vol. 25, no. 1, pp. 154–161,
“Model for polluted insulator flashover under AC or DC voltage,” Feb. 2018.
IEEE Trans. Dielectr. Electr. Insul., vol. 25, no. 2, pp. 614–622, [22] M. E. A. Slama et al., “Monitoring of dry bands and discharge activities
Apr. 2018. at the surface of textured insulators with AC clean fog test conditions,”
[7] F. Hadjrioua, D. Mahi, and M. E. Slama, “Electro-thermal dynamic Energies, vol. 14, no. 10, p. 2914, May 2021.
model using the analytical arc parameters for the prediction of the [23] M. E. A. Slama and A. Beroual, “Behavior of AC high voltage
critical flashover condition on the HVDC polluted insulator,” IET Gener., polyamide insulators: Evolution of leakage current in different surface
Transmiss. Distrib., vol. 11, no. 2, pp. 427–436, Jan. 2017. conditions,” Adv. Electr. Electron. Eng., vol. 13, no. 2, pp. 74–80,
[8] K. Belhouchet, A. Bayadi, and M. E. Bendib, “Artificial neural networks Jun. 2015.
(ANN) and genetic algorithm modeling and identification of arc para- [24] T. E. Browne, “Practical modeling of the circuit breaker ARC as a short
meter in insulators flashover voltage and leakage current,” in Proc. 4th line fault interrupter,” IEEE Trans. Power App. Syst., vol. PAS-97, no. 3,
Int. Conf. Electr. Eng. (ICEE), Dec. 2015, pp. 1–6. pp. 838–847, May 1978.
[9] M. E. A. Slama and A. Beroual, “Flashover discharges dynamic with [25] X. Zhang and S. M. Rowland, “Modelling of dry-band discharge events
continuous and discontinuous pollution layer under lightning impulse on insulation surfaces,” in Proc. IEEE Int. Symp. Electr. Insul., Jun. 2010,
stress,” Electr. Eng., vol. 103, pp. 2887–2895, Apr. 2021. pp. 1–5.
Authorized licensed use limited to: VTU Consortium. Downloaded on December 12,2023 at 09:39:52 UTC from IEEE Xplore. Restrictions apply.