502 H. H. Kordkheili et al.

: Determining the Probability of Flashover Occurrence in Composite Insulators

Determining the Probability of Flashover Occurrence

in Composite Insulators by Using Leakage
Current Harmonic Components
Hadi Hosseini Kordkheili
Mazandaran Electric Power Distribution Co.
Sari, Mazandaran Province, Iran

Hassan Abravesh, Mehdi Tabasi, Marzieh Dakhem and Mohammad Mehdi Abravesh
Mazandaran Hadaf University Babol Noshirvani University of Technology
Sari, Mazandaran Province, Iran Department of Electrical & Computer Engineering, Iran

ABSTRACT
Flashover of contaminated insulators in polluted areas is one of the most important
factors influencing the operation of transmission and distribution lines and substations.
Hence predicting an approaching flashover is of a great importance. In this paper, a
new method and criteria have been proposed to predict the flashover occurrence of SIR
composite insulators and its probability by using leakage current harmonic component.
The tests were performed on various profiles of artificially polluted SIR insulators
under clean fog condition. It is proposed that an important index is the ratio of 5th
harmonic component to the 3rd for which the operational and critical limits have been
determined. The probability of flashover occurrence has also been calculated by
geometric distribution. Only the measurements of a minimal set of parameters are
required. The presented criteria can be used to optimally schedule overhead lines
maintenance teams and to develop proper Condition-Based Monitoring Systems
(CBMS). Experimental results illustrate high performance of this approach.
Index Terms — Composite insulator, silicone rubber, condition-based monitoring
systems (CBMS), leakage current, flashover, harmonic components.

1 INTRODUCTION approaching critical conditions can be anticipated. Hence

determination of proper criteria is important to find out the
CONTAMINATION caused insulator flashover has been condition of insulators, as quickly as possible.
studied by many researchers in the recent years [1-17]. This event
Various investigations have been performed to develop such
which is a critical condition for insulators operation, leads to a
criteria and systems [1-14]. Some presented criteria were based
line-to-ground fault and its corresponding outages. A large
percentage of outages in distribution and transmission lines, on measuring many different parameters and comparing them to
especially in polluted areas, are related to insulators flashovers. pre-determined quantities. In Florida, a remote monitoring system
The appropriate operation of composite insulators has reduced has been designed to measure the amount of contamination,
insulator-related events. Factors such as relative humidity and which could monitor contamination build-up by using several
contamination gradually affect the insulating properties of sensors. This system worked on the basis of measuring
insulators and cause surface discharges which finally lead to parameters such as leakage current, relative humidity,
flashover. Therefore, anticipating and preventing these events, temperature of insulator surface, velocity and direction of wind,
especially with the purpose of reducing non-supplied energy is of and the amount of rainfall. The liquid water sensor in the system
great importance. also measured the amount of condensation and liquid water in
Optimized scheduling of operations related to replacing and contamination layer on the surface of the insulator. Finally the
washing of insulators are employed by many utilities. Besides, system determined the condition of insulator by comparing
CBMS (Condition-Based Monitoring System) is considered to be measured parameters to pre-determined quantities and generated
an important approach for utilities. By using these systems and necessary alarms [1].
measuring parameters such as leakage currents and ambient Recent studies use leakage current parameters as signatures of
humidity, and comparing them with pre-determined criteria, insulators condition. In fact, leakage currents carry useful
information about insulator’s operating condition and
Manuscript received on 10 August 2009, in final form 13 December 2009. approaching flashovers.

1070-9878/10/$25.00 © 2010 IEEE

IEEE Transactions on Dielectrics and Electrical Insulation Vol. 17, No. 2; April 2010 503

In [2, 4, 7] ceramic suspension and cap and pin insulators have thresholds. The presented criteria can also aid in optimally
been studied in clean fog with artificial pollution. It was found scheduling power lines and substation maintenance teams and to
that leakage current waveforms become similar to symmetrical develop a proper CBMS.
wave when strong local arcs occurred. Hence, a generally higher
intensity of the odd order of harmonic component is observed. 2 EXPERIMENTAL TESTS
Thresholds were presented for the peak values of leakage currents
and prominent odd-order harmonic components (e.g. 50 Hz, 150 2.1 EXPERIMENTAL SETUP
Hz and 250 Hz). But more accurate criteria were still needed and In order to perform the tests on contaminated insulators and
they should be revised to be used in composite insulators. measure leakage current and save the data, the laboratory setup
As in [8, 12], other studies have been performed on ceramic was prepared as shown in Figures 1, 2 and 3.
and composite insulators with a different approach. It was
concluded that in a post-pin ceramic insulator, the magnitude as
well as the waveform and the THD of leakage current were
significantly correlated with the insulator surface conditions. The
THD of leakage current waveform and its thresholds in different
contaminations were used as a tool for diagnosis of the insulator
[8]. Similar tests on EPDM composite insulators showed that
under discharge condition, the leakage current THD criteria
should be combined with leakage current magnitude. The
presented diagnostic parameter was the product of THD and
Leakage current magnitude and its thresholds [12]. Yet, no
quantified method was presented. Figure 1. Experimental setup.
In [3, 5, 9-11, 13] various composite insulators in salt fog were
tested. It was found that the low frequency harmonic components
show the insulator condition much better than the peak values. In
general, harmonic components increased only when arcing
occurred; thus providing a signature for dry-band arcing
activities. The level of both fundamental and low frequency
harmonic components of leakage current as a signature for
detecting SIR insulators’ surface damage has been presented in
[3, 9, 10]. They also have suggested an auto-correlation function
(ACF) [5] and use of time series modeling [11] for the third
harmonic component of leakage current as an indication of dry-
band arcing rather than a simple threshold value. A neural
network based system has also been developed in these studies
[13].
Other studies developed systems based on theoretical models
such as dynamic arc modeling and their regression analysis. The
models’ correlation with experimental results has also been Figure 2. Data acquisition system and fog chamber.
studied. Critical values of insulator’s surface resistance were
determined. The main idea was contamination flashover voltage
prediction using surface resistance criteria [6, 14].
While previous studies offer useful methods for determination
of approaching flashover, but we didn’t find any method to
calculate the probability of contamination flashover occurrence,
especially in composite insulators [15-17].
In this paper, new criteria and thresholds based on the ratio of
5th to 3rd harmonic components (K5/3 index) have been proposed.
This allows the prediction of approaching flashovers in composite
insulators. The probability of flashover occurrence is also
calculated through geometric probability distribution.
The presented criteria minimize the number of required
parameters to be measured. It can also determine the approximate
contamination level of the insulator and the probability of an
approaching flashover occurrence. Our analysis may also be
applied to other types of insulators to find their corresponding Figure 3. A sample contaminated insulator under test in fog chamber.
504 H. H. Kordkheili et al.: Determining the Probability of Flashover Occurrence in Composite Insulators

This setup which was prepared according to IEC60507 [18], the surface humidity. Then the insulator was suspended to dry in
includes a clean fog chamber, a high voltage transformer, and a ambient temperature. After building a pre-contamination layer,
leakage current measuring system. High voltage transformer can contamination slurry was prepared according to IEC60507 to be
supply up to 300 kV. In leakage current measuring system, a applied to insulators.
protecting electronic circuit was used to isolate measuring The amount of salt has a direct effect on the electrical
instruments from high voltage. The system worked based on conductance of contamination slurry, so three different types of
measuring voltage drop across a specific resistor. Waveforms and solutions was prepared to contaminate the insulators in different
their FFT were stored by a digital oscilloscope in the form of stages of the tests based on Table 2.
.CSV files.
Table 2. The amount of salt and Kaolin in different contamination slurries.
No. of contamination slurry Kaolin (gr/lit) Salt (gr/lit)
2.2 TESTED INSULATORS 1 40 20
The tested samples were several SIR composite insulators with 2 40 30
3 40 40
different profiles. From each profile, proper number of identical
insulators (5 insulators from each profile) was chosen to apply
artificial contamination with different densities onto their surfaces After preparing the contamination slurry, it was poured in a
(Figure 4). Table 1 shows the insulators characteristics. proper container. Pre-contaminated insulators were submerged
into the container one by one. In the next stage, insulators were
suspended vertically to dry out. Figure 5 shows a contaminated
insulator.

Figure 4. Sample profiles of contaminated insulators.

Table 1. Characteristics of Tested Insulators

Insulator No. 1 2 3 4 5
Nominal voltage (KV) 33 24 24 24 24 Figure 5. Contaminated insulator.
Mechanical tension
70 70 70 70 70
strength(KN)
Height (mm) 720 520 520 449 449
Creepage distance
1070 770 590 674 630
2.4 DATA MEASUREMENT
(mm)
Shed diameter (mm) 100/86 98/73 70/35 100 100/86
In order to measure leakage current harmonics of composite
insulators with different profiles and contamination levels, every
contaminated insulator was tested in dry (ambient humidity) and
2.3 INSULATORS CONTAMINATION humid (clean fog) conditions. In dry condition, the insulators
In order to artificially contaminate the surface of composite were energized up to their nominal voltage in ambient humidity
insulators under test, solid layer method with pre-contamination without applying any fog into the chamber. In humid condition,
was chosen [18]. This method can produce an approximately they were energized in clean fog produced by fog generators.
uniform contamination layer and can also be performed rapidly, After measuring leakage current, ESDD was measured
so that several insulators can be contaminated with the same according to IEC60507 to determine the amount of insulators’
slurry. surface contamination level. Table 3 shows the results of
Contaminating of composite insulators is rather difficult measured electrical conductance and calculated ESDD in tested
because of their surface hydrophobicity. Therefore the insulators insulators.
should be pre-contaminated initially. The stored data were processed by a VBA macro under Ms-
To perform the pre-contamination stage, small droplets of Excel and the results were summarized in identical tables as
distilled water were sprayed on the surface and then, kaolin shown in Table 4. Figure 6 and Tables 5 to 9 show FFT and
powder is poured on it. Kaolin forms a thin layer by sticking on measured leakage current waveforms.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 17, No. 2; April 2010 505
o
Table 3. Electrical conductance and ESDD in 20 C for two profiles of tested
insulators.
Measured electrical Insulator
Kaolin /Salt conductance ESDD Pollution
Ins. No. surface 2
(gr/lit) ( P S/cm) 2 (mg/cm ) level
area (cm )
1 40/40 1343 2283 0.167 Heavy
2 40/40 719 1141.5 0.176 Heavy
3 40/40 431 695 0.170 Heavy
4 40/40 568 920 0.171 Heavy
5 40/40 526 890 0.163 Heavy
1 40/30 621 2283 0.075 Moderate
2 40/30 332 1141.5 0.079 Moderate
3 40/30 218 695 0.084 Moderate
4 40/30 279 920 0.082 Moderate
5 40/30 253 890 0.077 Moderate
1 40/20 375 2283 0.045 Light
2 40/20 205 1141.5 0.048 Light
3 40/20 127 695 0.048 Light
4 40/20 176 920 0.051 Light
5 40/20 164 890 0.049 Light
Figure 6. (Continue …)

Table 4. An example of stored and processed data of leakage currents in

insulator profile No.1 (mA-Heavy Pollution).
Table 5. Harmonic components and K5/3 index in various contamination levels
Humidity% > 67% 67% 67% 76% 82% 84% 87%
for insulator profile No.1.
rms 0.021 0.020 0.20 0.084 0.100 0.116 0.114
Harmonic
1st 0.017 0.016 0.016 0.068 0.081 0.090 0.087 K5/3%
Wave form component
3rd 0.001 0.001 0.001 0.007 0.014 0.028 0.031 index
(mA)
5th 0.002 0.002 0.002 0.003 0.003 0.002 0.006 3rd H 5th H K5/3%
Heavy 7th 0.001 0.001 0.001 0.002 0.003 0.007 0.006 (max) (max) (min)
9th 0.000 0.000 0.000 0.001 0.001 0.004 0.007
11th 0.001 0.001 0.001 0.001 0.001 0.003 0.006
<0.0005 <0.002 > 300%
13th 0.000 0.000 0.000 0.000 0.001 0.003 0.006
THD% 18.64% 19.47% 19.41% 11.87% 18.25% 32.84% 41.84%
3rd H 5th H
K5/3%
(avg) (avg)

0.0003 0.0017 567%

3rd H 5th H K5/3%

(max) (max) (min)

<0.019 <0.007 > 35%

3rd H 5th H
K5/3%
(avg) (avg)

0.009 0.004 44%

3rd H 5th H K5/3%

(max) (max) (min)

<0.020 <0.005 > 20%

3rd H 5th H
K5/3%
(avg) (avg)

Figure 6. Leakage current FFT spectrum in insulator No.1 with different 0.008 0.003 37.5%
contamination levels in ambient humidity.
506 H. H. Kordkheili et al.: Determining the Probability of Flashover Occurrence in Composite Insulators
Table 5. (Continue …) Table 7. Harmonic components and K5/3 index in various contamination levels
Harmonic for insulator profile No.3.
K5/3%
Wave form component Harmonic
index K5/3%
(mA) Wave form component
index
(mA)
3rd H 5th H K5/3%
(max) (max) (min) 3rd H 5th H K5/3%
(max) (max) (min)

<0.030 <0.006 > 8%

<0.0003 <0.003 > 800%

3rd H 5th H
K5/3% 3rd H 5th H
(avg) (avg) K5/3%
(avg) (avg)

0.012 0.003 25% 0.00026 0.0025 961%

3rd H 5th H K5/3%

(max) (max) (min)
Table 6. Harmonic components and K5/3 index in various contamination levels
for insulator profile No.2.
Harmonic <0.007 <0.004 > 55%
K5/3%
Wave form component
index
(mA)
3rd H 5th H K5/3% 3rd H 5th H
K5/3%
(max) (max) (min) (avg) (avg)

<0.0004 <0.002 > 500% 0.0032 0.0033 103%

3rd H 5th H
K5/3% 3rd H 5th H K5/3%
(avg) (avg)
(max) (max) (min)

0.0003 0.0020 667%

<0.013 <0.004 > 30%
rd th
3 H 5 H K5/3%
(max) (max) (min) 3rd H 5th H
K5/3%
(avg) (avg)

<0.011 <0.004 > 19%

0.009 0.003 33%

3rd H 5th H
K5/3% 3rd H 5th H K5/3%
(avg) (avg)
(max) (max) (min)

0.007 0.003 43%

<0.008 <0.002 > 20%

3rd H 5th H K5/3% 3rd H 5th H

(max) (max) (min) K5/3%
(avg) (avg)

<0.015 <0.005 > 18%

0.0023 0.0022 95%

3rd H 5th H
K5/3%
(avg) (avg)
Table 8. Harmonic components and K5/3 index in various contamination levels
0.006 0.003 50% for insulator profile No.4.
Harmonic
K5/3%
Wave form component
index
rd
3 H 5 Hth
K5/3% (mA)
(max) (max) (min) 3rd H 5th H K5/3%
(max) (max) (min)

<0.008 <0.005 > 50%

<0.0004 <0.0021 > 450%

rd th
3 H 5 H
K5/3% 3rd H 5th H
(avg) (avg) K5/3%
(avg) (avg)

0.003 0.002 67% 0.0003 0.0019 633%

IEEE Transactions on Dielectrics and Electrical Insulation Vol. 17, No. 2; April 2010 507
Table 8. (Continue…) Table 9. (Continue…)
Harmonic Harmonic
K5/3% K5/3%
Wave form component Wave form component
index index
(mA) (mA)
3rd H 5th H K5/3%
3rd H 5th H K5/3%
(max) (max) (min)
(max) (max) (min)

<0.022 <0.007 > 30%

<0.018 <0.006 > 30%

3rd H 5th H 3rd H 5th H

K5/3% K5/3%
(avg) (avg) (avg) (avg)

0.006 0.004 67% 0.011 0.004 36%

3rd H 5th H K5/3% 3rd H 5th H K5/3%

(max) (max) (min) (max) (max) (min)

<0.052 <0.011 > 8%

<0.022 <0.010 >30%

3rd H 5th H
3rd H 5th H K5/3%
K5/3% (avg) (avg)
(avg) (avg)

0.010 0.004 40% 0.018 0.004 22%

3rd H 5th H K5/3%

(max) (max) (min)
Figure 7 shows the variations of leakage current harmonic
<0.011 <0.008 > 35% components in various contamination levels when the relative
humidity increases gradually. From presented diagrams, it’s clear
3rd H 5th H that there are meaningful and noticeable values in the 5th and 3rd
K5/3%
(avg) (avg) harmonic components which will be discussed in the next part.

0.0047 0.0049 104%

Table 9. Harmonic components and K5/3 index in various contamination levels

for insulator profile No.5.
Harmonic
K5/3%
Wave form component
index
(mA)
3rd H 5th H K5/3%
(max) (max) (min)

<0.0003 <0.002 > 700%

3rd H 5th H
K5/3%
(avg) (avg)

0.0002 0.0016 800%

3rd H 5th H K5/3%

(max) (max) (min)

<0.030 <0.003 > 11%

3rd H 5th H
K5/3%
(avg) (avg)

0.007 0.002 29%

Figure 7. Variations of harmonic components in different contamination
levels and humidities for insulator No.1.
508 H. H. Kordkheili et al.: Determining the Probability of Flashover Occurrence in Composite Insulators

phenomenon finally leads to partial discharges on the

insulator surface due to leakage current flowing and
creation of dry bands. This caused flashovers in some
tests.
In the humidity saturation region, too much salt is
dissolved in the water droplets because of complete
wetting of surface. This humidity appears in the form of
small water droplets on the surface and drip down from
the edges of insulator sheds. This process reduces the
amount of salt on the surface layer and subsequently
reduces surface conductance and makes the operational
condition of the insulator more stable.
From the viewpoint of harmonic components of leakage
current in contaminated insulators, it was observed that in
no-fog condition, the 5th harmonic component was always
greater than the 3rd. That is, the insulator was in its
normal condition. But as soon as fog was applied,
previous condition changed and the 3rd harmonic
component increased and became greater than the 5th.
Again as a result, if the leakage current’s 5th harmonic
component of a contaminated composite insulator
becomes greater than the 3rd, the insulator is operating in
Figure 7. (Continue…) its normal condition.

3.2 K5/3 INDEX AND ITS OPERATIONAL LIMITS

3 DATA ANALYSIS AND EXPERIMENTAL As discussed above, variations of 3rd and 5th harmonics
RESULTS with respect to each other are a criterion to judge the tested
insulators’ service condition. The increment of 3rd harmonic
3.1 HARMONIC VARIATIONS AND INSULATORS
component of the leakage current is originally caused by
CONDITION
surface contamination and humidity. The surface
Under clean conditions, no flashover occurred during contamination has nonlinear impedance. Surface
the tests. It is also obvious from clean condition diagrams contamination and humidity increases the conductance and
that the value of main harmonic component increases therefore there will be an increase in leakage current. The
sensibly by humidity increment, but this variation is not increased current does not have a sinusoidal form because of
as evident in other harmonics. nonlinear impedance. On the other hand, increase of leakage
There are two sharp increments of leakage current current forms dry regions and partial discharges which
(main component) around 75% and 90% humidity. This deform the waveform of the current. And finally we will
phenomenon is true for both clean and contaminated have a leakage current with 3rd harmonic content. The
insulators. Under clean condition, the amount of 5th increase of partial discharges deforms the wave form more
harmonic component is greater than the one of the 3rd. and causes the 3rd harmonic component to be increased [19,
Evidently, when the 5th harmonic component is greater 20]. The experimental measurements also show this
than the 3rd’s, the insulators are either in normal phenomenon.
condition or they have no or very light contamination. In practice, measuring the amount of insulators’
In light contamination level, the main harmonic contamination in hot lines is a difficult and costly operation.
component initially decreased, but when the fog was In addition the amount of humidity and fog are imprecise
applied, it increased gradually. After a while, it started to factors which depend on different elements. Therefore we
decrease and at the beginning of humidity saturation, it define K5/3 index as:
increased again.
As a matter of fact, if the insulator passes the beginning
of moistening region without flashover, the flashover will 5th Harmonic component
not occur in the humidity saturation region. Moistening
K5 1
3 3rd Harmonic component
region is the most critical condition for the insulator. The
reason is a sudden increment in the surface conductance
of the insulator caused moistening of some parts of Tables 10 to 14 shows calculated amounts for K5/3 index in
contamination layer and dissolving soluble salts. This tested insulators.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 17, No. 2; April 2010 509
Table 10. K5/3 Index for insulator No.1 with different contamination levels. Table 12. K5/3 Index for Insulator No.3 With different Contamination levels
Clean Clean

Humidity
Humidity

%
79% 79% 79% 79% 81% 85% 89% 91% 93% 95%
%

65% 65% 65% 66% 74% 81% 86% 90% 93% 94%

5th/3rd 5044% 1936% 860% 1224% 906% 823% 997% 792% 567% 417% 5th/3rd 1102% 866% 1037% 1056% 901% 932% 889% 932% 997% 867%

Humidity
Humidity

%
96% 97% 98% 98% 99%
%

96% 97% 98% 98%

5th/3rd 654% 444% 443% 393% 5th/3rd 1045% 842% 883% 893% 1121%
Light Light

Humidity
Humidity

%
56% 56% 56% 58% 61% 70% 80% 85% 89% 91%
%

74% 74% 74% 75% 76% 83% 89% 91% 92% 94%

5th/3rd 106% 133% 184% 209% 48% 40% 36% 41% 42% 35% 5th/3rd 785% 859% 922% 892% 525% 269% 146% 111% 61% 93%
Moderate

Humidity
%
93% 95% 96% 97% 98%
Humidity
%

75% 75% 75% 77% 80% 82% 85% 89% 93%

5th/3rd 79% 59% 60% 72% 83%

5th/3rd 532% 881% 320% 85% 68% 85% 32% 23% 21% Moderate
Heavy

Humidity
%
58% 58% 58% 62% 69% 75% 88% 92% 94% 95% 96%
Humidity
%

67% 67% 67% 76% 82% 84% 87%

5th/3rd 431% 455% 419% 354% 276% 74% 55% 43% 49% 41% 33%
5th/3rd 305% 292% 331% 47% 25% 8.5% 20% Heavy
Humidity
%

73% 73% 73% 77% 81% 88% 91% 93%

5th/3rd 593% 655% 735% 595% 224% 119% 40% 23%

Table 11. K5/3 index for insulator No.2 with different contamination levels.
Clean Table 13. K5/3 Index for Insulator No.4 With different Contamination levels
Clean
Humidity

Humidity
%

76% 76% 76% 77% 79% 85% 90% 92% 93% 95%
%

67% 67% 67% 67% 72% 80% 86% 89% 90% 92%

5th/3rd 1109% 1118% 1164% 942% 715% 670% 739% 734% 808% 657%
5th/3rd 697% 636% 572% 632% 608% 532% 594% 616% 676% 602%
Humidity

Humidity
%

96% 97% 98%

%

93% 94% 96%

5th/3rd 727% 562% 531%

5th/3rd 577% 478% 498%
Light
Light
Humidity

Humidity
%

58% 58% 58% 60% 63% 72% 80% 85% 88% 90%
%

56% 56% 56% 57% 60% 69% 79% 85%

5th/3rd 327% 1047% 1013% 1149% 133% 58% 34% 25% 25% 23% 5th/3rd 1175% 978% 1059% 945% 282% 72% 38% 16%
Moderate
Humidity
%

92% 94% 95% 96% 97%

Humidity
%

68% 68% 68% 71% 76% 81% 85% 88% 91% 93% 94% 96%
5th/3rd 19% 21% 32% 30% 29%
Moderate 5th/3rd 351% 355% 329% 284% 126% 88% 75% 48% 52% 41% 36% 33%
Heavy
Humidity
%

58% 58% 58% 68% 76% 82% 85% 89% 93% 95% 97%
Humidity
%

72% 72% 72% 73% 78% 85% 90% 93% 95% 96%

5th/3rd 224% 158% 205% 224% 83% 52% 50% 62% 27% 18% 28%
Heavy 5th/3rd 1228% 1020% 1427% 1111% 417% 186% 177% 102% 98% 122%
Humidity
Humidity

97% 98% 98% 99%

%

73% 73% 73% 74% 80% 88% 91% 94% 95% 96%

5th/3rd 417% 390% 376% 403% 367% 51% 57% 68% 50% 57% 5th/3rd 120% 85% 38% 63%
510 H. H. Kordkheili et al.: Determining the Probability of Flashover Occurrence in Composite Insulators
Table 14. K5/3 index for insulator No.5 with different contamination levels. By using these criteria, insulator’s operational condition may
Clean
be studied and the approximate levels of surface contamination
Humidity

level can be determined. So prevention and maintenance

%

69% 69% 69% 70% 74% 81% 87% 90% 93% 94%
operations can be scheduled properly.
5th/3rd 1083% 1089% 1159% 1280% 876% 1367% 1035% 1387% 869% 923% When K5/3 < 30%, a geometric distribution statistical model
was used to calculate flashover occurrence probability. The
Humidity

analyses show that flashover probability in this circumstance is

%

95% 96% 97% 98%

more than 90%.
5th/3rd 961% 994% 822% 728%
Light
4 STATISTICAL MODELING
Humidity
%

56% 56% 56% 64% 65% 72% 81% 88% 90% 93% 4.1 PROPER STATISTICAL MODEL
5th/3rd 758% 785% 734% 645% 841% 133% 45% 27% 11% 25%
According to measurements, when K5/3 > 30%, there is no
flashover occurrence, so we consider K5/3 values which satisfy the
Humidity

condition K5/3 < 30% as our statistical population.

%

95% 96% 97% 98%

In 41 cases, K5/3 was less than 30% of which seven cases led to
5th/3rd 20% 16% 12% 13% a flashover. So 7/41 of studied periods with the condition of
Moderate K5/3<30% led to flashover. The 7 to 41 ratio proved too large to
assume a Poisson distribution, so we considered the problem as
Humidity
%

69% 69% 69% 74% 82% 89% 92% 94% 95% 96% 97%
below:
“As soon as the condition of K5/3 < 30% becomes true, the test
5th/3rd 443% 452% 438% 249% 182% 55% 43% 39% 48% 35% 33%
Heavy is similar to consecutive coin tosses, which has one of the two
outcomes: a flashover occurrence with the probability of (P)
Humidity

(success condition) and no flashover with the probability of (1-P)

%

77% 77% 77% 78% 82% 89% 90%

(failure condition). Each test lasts two minutes.”
5th/3rd 941% 99% 114% 143% 23% 8% 22% The processes can be modeled by using geometric distribution.
In clean condition, K5/3 index was always larger than 100% and If we repeat a Bernoulli test to gain the first success, then the
flashover never occurred. So if K5/3 • 100% then the composite test is a “geometric test” and the random variable X which shows
insulator is in its normal condition. the number of Bernoulli tests is called “geometric random
In light contamination condition the insulator continued variable”. The probability distribution of geometric random
working properly in all but two cases where flashover occurred. variable is called “geometric distribution” and is shown by g(x;p):
Insulators continued working even when K5/3 value sometimes
dropped below 10%. However, humidity saturation of fog x 1
chamber elevated K5/3 value to more than 100%. This is due to the
g (x ; p ) p q (x 1, 2, 3,...)  2
fact that in light contamination level the amount of salt in
contamination layer was less than moderate and heavy where x is the number of Bernoulli tests, p is the probability of
contamination levels. Therefore, the salt is mainly dissolved at the success and q=1-p is the probability of failure in each Bernoulli
beginning of humidity saturation region. This caused an increase test.
of surface conductance. Subsequently the 3rd harmonic 4.2 MODELING BY GEOMETRIC DISTRIBUTION
component increased in comparison with the 5th. So K5/3 reduced First, the probability of a flashover (p) should be calculated for
markedly. Continuation of insulator surface wetting process led to each measurement. In fact, f(x) is the probability insulator
dripping down salt-containing water droplets from the edge of flashover after passing x number of two minute measurement
sheds resulting in a reduction of salt and a change of condition to periods from the occurrence of K5/3 < 30%.
normal again. “Notably K5/3 was always less than 30% when According to the tests, in seven cases the flashover occurred
flashovers occurred in”. and there were eighteen measurements before the flashovers in
Moderate and heavy contamination levels contained more salt which the condition K5/3 < 30% was true. Hence, we expected the
than the light contamination condition. Although salty water flashover after 2.57 (=18÷7) measurement periods. Table 15
droplets dripped down from the edge of sheds, the resulting shows the cases of flashover occurrence in the tested composite
increase of electric conductance caused K5/3 to continue its insulators and the number of K5/3 < 30% occurrence before each
reduction to the point of flashover. Again in all cases that flashover.
flashover occurred, K5/3 was less than 30%.
As a result, the necessary condition to recognize the insulators
Table 15. Cases of flashover occurrence in tested composite insulators and
critical circumstance is K5/3 < 30%. This condition is necessary the number of K5/3 < 30% before each flashover.
but not sufficient. In some cases of K5/3 < 30%, the flashover did Xi
not occur. Generally, for K5/3 < 30%, an insulator flashover would The event of flashover X1 X2 X3 X4 X5 X6 X7
occur within two minutes with high probability. If K5/3 value occurrence
recovers, the insulator would pass the critical phase. The number of K5/3 < 30% 2 3 7 1 1 3 1
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 17, No. 2; April 2010 511

In geometric distribution we have:

which is a very low percent. In fact, by occurring K5/3 < 30%, the
E ( x ) 1/ p  3 insulator flashover will occur in the next 15 measurements with
the probability of (1 í 0.062 = 99.94%).
Figure 8 shows the insulator flashover probability variations
where E(x) is the expected value of X. We have a limited number
diagram, before r number of measurements for q=0.611. It’s
of tests, so we have an estimate of the whole statistical population
obvious that after almost 5 measurement periods in which K5/3 is
and we can put estimates of E(x) and p in (3) as follow:
less than 30%, the probability of flashover occurrence becomes
more than 90%. This high probability percent were clear in the
X 1/ pˆ  4 tests performed.
Modeling with geometric distribution has been checked
where X and p̂ are estimates of E(x) and p respectively. X has with “Goodness of Fit” statistical test.
been calculated previously, therefore ( p̂ & q̂ ) can be obtained
from (5) and (6):

pˆ 1/ X 1/ 2.56 0.389  5
And

qˆ 1 pˆ 1 0.389 0.611  6
( p̂ & q̂ ) are estimates of (p and q) in the statistical population,
respectively.
From the time K5/3 < 30% becomes true, we wait for flashover.
This may occur in the first 2 minute period or in the second or in
the third and so on. The number of periods to reach the flashover
is a random number that its probability distribution follows the
Figure 8. Diagram of flashover probability variations for q = 0.611.
geometric distribution. For example if x=5, it means that after K5/3
< 30% becomes true, in 4 initial measurement intervals, flashover
will not occur, but in the 5th measurement interval it will occur.
The probability of this event is: 5 CONCLUSION
This paper has proposed a new approach in determining the
51 4 operating condition of SIR composite insulators. Useful criteria
P (X 5) pq (0.389)(0.611) 5.42%
have been presented in order to anticipate the SIR insulators’
approaching flashover by using leakage current harmonic
According to geometric distribution properties, we have the components. The presented K5/3 index (the ratio of 5th to 3rd
following relation: harmonic component of leakage current) and its operational
thresholds have been determined and a statistical modeling using
r
P (X ! r ) q 7 geometric distribution has been proposed. The probability of
distribution occurrence and its mathematical relation have also
or been calculated.
r
The proposed methods and criteria can be used for any types of
P (X d r ) 1 q 8 insulators with their own operational thresholds and can
approximately recognize the insulators pollution level.
One of the important features is minimized required
P(X ” r) is the probability of flashover occurrence before r parameters to be measured. They can be used to optimally
number of measurements, and P(X ” r) is the probability of the schedule overhead lines maintenance teams and to develop proper
event that the insulator flashover does not occur up to next r CBM systems.
number of measurements and after that, the flashover takes place.
For example, we suppose K5/3 < 30% becomes true, so the
probability of the event that the flashover doesn’t occur up to next ACKNOWLEDGMENT
15 measurements and occurs after that is equal to: The authors are grateful to Dr. Edris Pouresmaeil for his
useful suggestions and editing points and Mr. Pooya
15 Moradinezhad for his assistance in statistical parts.
P ( X ! 15) 0.611 0.062%
512 H. H. Kordkheili et al.: Determining the Probability of Flashover Occurrence in Composite Insulators

REFERENCES [16] S.H. H. Kordkheili, H. Abravesh, H. Yousefi and M.Dakhem, “New

diagnostic method for determining composite insulators’ service
[1] C. N. Richard and J. D. Renowden, “Development of a remote insulator condition”, Electric Power Distribution Conference (EPDC-13th), Iran,
contamination monitoring system”, IEEE Trans. Power Delivery, Vol. 2008 (in Persian).
12, pp. 389-397, 1997. [17] S.H. H. Kordkheili, Harmonic Analysis of Leakage Current on the
[2] T. Suda, “Frequency characteristics of leakage current waveforms of an Surface of Contaminated Wetted Composite Insulators”, M.Sc
artificially polluted suspension insulator”, IEEE Trans. Dielectr. Electr. dissertation, Dept. Elect. Comp. Eng., Univ. Mazandaran, Babol, Iran,
Insul., Vol. 8, pp. 705-709,. 2001. 2008, (in Persian).
[3] A. H. El-Hag, S. Jayaram, and E. A. Cherney, “Fundamental and low [18] Artificial pollution tests on high voltage insulators to be used on A.C.
frequency harmonic components of leakage current as a diagnostic tool systems, IEC Standard 60507, 1991.
to study aging of RTV and HTV silicone rubber in salt-fog”, IEEE [19] M.A.R.M. Fernando and S.M. Gubanski, “Analysis of Leakage Current
Trans. Dielectr. Electr. Insul., Vol. 10, pp. 128-136, 2003. Wave Forms for Field-aged and New Composite Insulators”, IEEE
[4] T. Suda, “Frequency characteristics of leakage current waveforms of a Conf. Electr. Insul. Dielectr. Phenomena (CEIDP), Minneapolis, pp.
string of suspension insulators”, IEEE Trans. Power Delivery, Vol. 20, 350-353, 1997.
pp. 481-487, 2005. [20] M.A.R.M. Fernando and S.M. Gubanski, “Leakage Current Patterns on
[5] A. H. El-Hag, “Leakage current characterization for estimating the Contaminated Polymeric Surfaces”; IEEE Trans. Dielectr. Electr. Insul.,
conditions of non-ceramic insulators”, Electric Power Syst. Res. J. Vol. 6, pp.688-694, 1999.
(Elsevier), Vol. 77, Nos. 3-4, pp. 379-384, 2007.
[6] S. Venkataraman and R. S. Gorur, “Prediction of flashover voltage of
non-ceramic insulators under contaminated conditions”, IEEE Trans.
Dielectr. Electr. Insul., Vol. 13, pp. 862-869, 2006.
[7] T. Suda, “Study on the frequency characteristics of leakage current
waveforms of artificially polluted 12t suspension insulators by the clean Hadi Hosseini Kordkheili was born in Tehran, Iran
fog method”, 9th Int. Symp. High Voltage Engineering, pp. 115-118, in 1980. He received the B.Sc. and M.Sc. degrees in
1999. electrical engineering from Mazandaran University,
[8] Suwarno, “Study on the waveform of leakage current on the 20kV post- Iran in 2002 and 2007, respectively. He joined
pin ceramic insulators under various conditions”, IEEE Intern. Sympos. Mazandaran Electric Distribution Co. in 2006. Now he
Electrical Insulating Materials, pp. 387-390, 2001. is the Chief of Planning and Control Department in the
[9] A. H. El-Hag, S. Jayaram, and E. A. Cherney, “Low frequency harmonic corporation. He is also the Learning Chief of Andishe
components of leakage current as a diagnostic tool to study aging of Pars Educational and Research Institute, Sari, Iran. He
silicone rubber insulators”, in Proc. 2001 Annu. Report Conf. Electrical teaches electrical engineering courses in Hadaf
Insulation & Dielectric Phenomena, pp. 597–600. University, Sari, Iran and Andishe Pars Educational
[10] A. H. El-Hag, S. Jayaram, and E. A. Cherney, “Use of Low frequency and Research Institute. His research interests are electrical insulation, HV
harmonic components as a signature to detect arcing on HTV silicon systems and distribution engineering.
rubber in salt fog”, IEEE Intern. Symp. Electr. Insul. (ISEI), pp. 212–
215, 2002.
[11] A. H. El-Hag, “Detection of dry-band arcing using time series
modeling”, IEEE Conf. Electr. Insul. Dielectr. Phenomena (CEIDP), pp. Hassan Abravesh was born in Kerman, Iran in 1953.
615–618, 2004. He received the B.Sc. degree in electrical engineering
[12] Suwarno, “Diagnostics of outdoor insulators using leakage current from Mazandaran University, Iran in 1979, the M.Sc.
waveform parameters”, IEEE Intern. Symp. Electrical Insulating and Ph.D. degrees in high voltage engineering from
Materials, Kitakyushu, Japan, pp. 111-114, 2005. Strathclyde University, Glasgow, U.K in 1987 and
[13] A. N. Jahromi, A. H. El-Hag, E. A. Cherney, S. Jayaram, M. Sanaye- 1991, respectively. Following graduation, he was
Pasand, H. Mohseni, “Prediction of leakage current of composite appointed as a Lecturer in the Dept. of Electrical
insulators in salt fog test using neural network”, IEEE Conf. Electr. engineering at Mazandaran university, with lecturing
Insul. Dielectr. Phenomena (CEIDP), pp. 309–312, 2005. duties in high voltage engineering, power system
[14] S. Venkataraman and R. S. Gorur, “Flashover voltage prediction of protection and electrical circuits, in both
outdoor insulators subjected to road salt contamination”, IEEE Conf. undergraduate and postgraduate levels. Dr. Abravesh has completed several
Electr. Insul. Dielectr. Phenomena (CEIDP), pp. 293–296, 2005. research and consultancy projects in Porcelain, glass and composite insulators.
[15] H. Abravesh and S.H. H. Kordkheili, “Insulators aging and related He has authored over 20 publications in this area. He is a member of National
events in Mazandaran region”, Electric Power Distribution Conf. Iranian Electrical Insulation Studies Committee.
(EPDC-11th), Iran, , 2006 (in Persian).