Available online at www.sciencedirect.

com

ScienceDirect
Procedia Technology 11 (2013) 411 – 418

The 4th International Conference on Electrical Engineering and Informatics (ICEEI 2013)

Effect of Test Method and Needle Plane Configuration on Partial

Discharge Inception Voltage Measurement of Mineral Oil
Based on Weibull Analysis
T.F. Sipahutara,c,A.A. Kemmaa,c,N. Pattanadechb,F. Pratomosiwib,Suwarnoa, M.Muhrb,*
a
School of Electrical Engineering and Informatics, Institut Teknologi Bandung, Jl. Ganesha 10, Bandung Indonesia
b
Institute of High Voltage Engineering and System Management, TU Graz, Inffeldgasse 18, A - 8010 Graz,Austria
c
PT PLN (Persero), Jakarta, Indonesia

Abstract

The discussion of test method and electrode configuration for Partial Discharge Inception Voltage (PDIV) measurement is
continuously developed because the PDIV can be used as indicator to determine condition of insulating liquid. However, many
researchers investigate the PDIV using different test method and electrode configurations. This paper discusses the effect of test
methods and needle to plane electrode configuration on the PDIV measurement of mineral oil as insulating liquid. The mineral
oil that used in experiment test was Nynas Nitro 4000x with water content less than 10 ppm. In experiment, the test circuit was
set up according to IEC 60270. The first test method that was carried out in this experiment according to IEC 61294 with ramp
rate of test voltage 1 kV/s until PDIV occurred. Meanwhile, a Combine Method as the second test method was used as
comparative PDIV test technique for this experiment. With this combine method, the test voltage was applied to the test object
with rate 1 kV/s until 70% of the PDIV value obtained from the first test method. Then the test voltage was increased in steps
with 1 kV/step with step duration of 1 minute until PDIV obtained with PD charge ≥ 100 pC. The rest time between consecutive
tests was 5 minutes. The needle tip radii that used as high voltage electrode were 10 μm and 20 μm, meanwhile brass plane
electrode with diameter 50 mm was used as the grounded electrode. In this paper, the electric field stress of the needle to plane
electrode was calculated using Finite Element (FEM) simulation with input voltage 100 volt. The test result was analyzed using
Weibull distribution to calculate the Cumulative Distribution Function (CDF) and Probability Density Function (PDF) of PDIV
value. The calculation results show that the test results agree with Weibull distribution.
©
© 2013
2013 The
TheAuthors.
Authors.Published
PublishedbybyElsevier
ElsevierB.V.
Ltd. Open access under CC BY-NC-ND license.
Selection
Selection and
andpeer-review
peer-reviewunder
underresponsibility
responsibilityofofthe
theFaculty
Facultyofof
Information Science
Information and&Technology,
Science Technology,Universiti
UniversitiKebangsaan
Kebangsaan
Malaysia.
Malaysia.

Keywords : Partial Discharge Inception Voltage ; Electrode ; Ramp Method ; Combine Method ; Weibull ; CDF ; PDF ; CV ; FEM

* Corresponding author. Tel.: +62 813 61 086 312

E-mail address: ferdi_nd@yahoo.com

2212-0173 © 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of the Faculty of Information Science & Technology, Universiti Kebangsaan Malaysia.
doi:10.1016/j.protcy.2013.12.210
412 T.F. Sipahutar et al. / Procedia Technology 11 (2013) 411 – 418

1. Introduction

Up to now, mineral oil is still widely used as an insulating material of transformer. There are many factors that
may affect the quality of the mineral oil such as moisture, electrical stress, particle, thermal stress and other factors.
These stresses generate Partial Discharge (PD) which may lead complete breakdown of insulation whilst the
degradation of insulations material occur [1-2]. Therefore, to maintain the condition of transformer oil, the
continuity monitoring is needed. One of the methods that used to monitor oil condition in transformer is PD
measurement. This method is an important diagnostic technique and used as non-destructive test for the insulation of
high voltage [1]. The occurrence of PD in mineral oil is strongly associated with the cavity formation within the
mineral oil. The evolution of these cavities will induce the rise of series PD pulses of ascending magnitude. Such PD
Pulses can be used as an indicator to determine the value of the Partial Discharge Inception Voltage (PDIV) in
insulating liquid [3]. In many case, PDIV is an alternatively important indicator that most researchers use for
representing the integrity of the liquid insulation [4].The definition of PDIV of an insulating liquid according to the
IEC 61294 is the lowest voltage at which an apparent charge occurs equal or exceeding 100 pC when the sample is
tested under the specified conditions [5]. Some researchers report that the characteristic features of partial discharge
phenomena greatly depend on experimental conditions such as electrode geometry, shape and amplitude of applied
voltage, liquid nature and purity [6]. The electrode geometry and shape have an important role for PDIV
measurement. It is associated with the deployment of electric field stress around the electrode and especially at the
tip. The needle electrode is widely used for PDIV research. The highest electric field was obtained at the tip of
needle electrode [7]. Besides of that, the test method also greatly affects the PDIV measurement of mineral oil.
There are doubts about the effectiveness of ramp method [8]. The advantage of ramp method is does not require
long test duration to obtain PDIV value. However, it may have a highly chance to lead breakdown events. The other
method that could be used for PDIV measurement is a combine method. This test method can more reduce the
chance of the breakdown events than ramp method. But the combine method requires long test duration to obtain
PDIV value.
The discussion about distribution pattern of partial discharge data is important to perform. This is done to
determine the characteristic of test data obtained by experiment. So the probability of data such as cumulative
distribution function (cdf) and probability density function (pdf) of PD can be known. Based on numerical analysis,
partial discharge fit to the two parameter of Weibull function [9]. This Weibull distribution is widely used for data
analysis due to its relative simplicity and flexibility. There are many formulas that can be used to calculate the CDF
and PDF of test data results. However, they are a lack of study for the implementation of Weibull analysis of PDIV
data.
The aim of this paper is to describe the effect of test method and needle tip radius for PDIV measurement of
mineral oil under room temperature. Besides of that, this paper proposes also the CDF and PDF of PDIV value so it
can be used to determine the occurrence probability of PDIV value. In this case, the PDIV data was analyzed by
using simple formula that used to determine the distribution of Breakdown Voltage (BDV) data [10].
2. Experiment setup and data analysis
The experiments were divided into two parts. First, we simulated the electric field simulation of the electrode
configuration system by FEM simulation. Second, we carried out laboratory scale experiment. The moisture content
in this measurement was kept less than 10 ppm which was measured by using Karl Fischer Coulometer.

2.1. Electrode configuration

The measurement of PDIV was performed in a test object. Tungsten needle electrode with tip radius 10 μm and
20 μm were used as high voltage electrode. The length of needle electrode is 45 ± 0.5 mm. Brass plane electrodes
with 50 mm diameter were used as the grounded electrode. The gap distance between the needles to plane electrode
was 50 mm. Fig.1.depicts the electrode configuration for PDIV measurement.
 T.F. Sipahutar et al. / Procedia Technology 11 (2013) 411 – 418 413

Tip radius Tip radius

Needle Electrode

10 μm 20 μm

50 mm 50 mm

Plane Electrode

Fig.1. Electrode configuration for Partial Discharge Inception Voltage (PDIV) measurement

2.2. Test circuit

The test circuit for PDIV measurement was set up according to IEC 60270 [11]. The discharge was measured with
Power Diagnostics ICM and Oscilloscope Yokogawa DLM series at the same time. With ICM, we recorded the
PDIV and PD patterns.The coupling device was connected to the coupling capacitor and to the ICM. This
configuration was chosen to denote the real condition of electrical system. In addition, this configuration was used
to avoid the possible destruction of measuring equipment in case of the breakdown event. Meanwhile with
oscilloscope, we used to recorded the PD pulse current signal. The shunt resistor was directly connected to the
oscilloscope via 50 ohm matching impedance. Fig.2. shows the test circuit of PDIV measurement.

Fig. 2. Test circuit for Partial Discharge Inception Voltage (PDIV) measurement
414 T.F. Sipahutar et al. / Procedia Technology 11 (2013) 411 – 418

2.3. Testing method

The Ramp Method (as specified in IEC 61294) and the Combine Method were employed to investigate PDIV
characteristic of mineral oil. According to IEC 61294, PDIV measurement is done by applying the ramp test voltage
with increasing rate 1 kV/s until the first discharge with its apparent charge ≥ 100 pC occur. The measurement of
PDIV was conducted ten times for every sample. The combine method is a procedure to generate PDIV by means of
increasing the voltage with the rate of 1 kV/s until 70% of average PDIV value that obtained from ramp method.
Then, the voltage was increased in steps with 1 kV/step until the PDIV value (PD charge 100 pC) occurred. After
the PDIV value was achieved, the test voltage was decreased until zero. The rest time between consecutive tests was
5 minutes. Fig.3.denotes the test procedure of PDIV measurement using ramp method and combine method.
PDIV (Charge ≥ 100 pC)
PDIV (Charge ≥ 100 pC)
a) b)

1st 2nd 3rd 10th

PDIV PDIV PDIV PDIV 70 % of
average PDIV
1kV/s
1 kV/s 1 kV/s 1kV/s
1kV/s
zero volt zero volt zerovolt PDIV pulse zerovolt
1 minute
1 minute 5 minute

Fig. 3. Test procedure of PDIV measurement (a) Ramp Method; (b) Combine Method

2.4. Data analysis

The PDIV test results were analyzed to obtain the pattern of distribution data, the cumulative distribution
function (CDF) and probability density function (PDF) of PDIV. The formula for PDIV data analysis is complied
with IEEE 930-2004 that used also to obtain the breakdown voltage distribution [10]. The formula used is as
follows:

F(t;α,β) = 1 – exp (-t/α)^ β (1)

In this formula, F(t;α,β) is defined as the cumulative density function, α is defined as a scale parameter, β is a shape
parameter that related to the range of PDIV, whilst t is defined as a measured variable and in this case is PDIV
value.

Simple approximation : F(i,n) ≈ ((i – 0.44)/ (n + 0.25)) x 100% (2)

The F(i;n) is defined as an estimation of CDF, i is defined as the rank order of data, and n is defined as weights of
small data sets. The expressions for the CDF is present the probability of PDIV occurring with the rising of test
voltage. The probability density function ݂ሺ‫ݔ‬ሻof Weibull distribution can be calculated by using the following
equation [12].
ߚ ‫ ݐ‬ሺߚ െͳሻ ‫ݐ‬
݂ሺ‫ݔ‬ሻ ൌ  Ǥ ቀ ቁ Ǥ ‡š’༌ሺെሺ ሻߚ ሻ (3)
ߙ ߙ ߙ

The f(x) is defined as probability density function, α is defined as scale parameter, ߚ is defined as shape parameter, t
is defined as measured variable. For fitting data, the Kolmogorof-Simirnov method is introduced because is
appropriate for small samples [13].
P0 (Xi) = P0 (X ≤ Xi) = CDF (Xi) (4)
 T.F. Sipahutar et al. / Procedia Technology 11 (2013) 411 – 418 415

ܺ൑ܺ݅ ݅
‫݊ܨ‬ሺܺ݅ሻ ൌ  ൌ  Ǣ ݅ ൌ ͳ ǥ ݊ (5)
݊ ݊

D+ = Fn – Fo and D- = Fo – Fn-1 (6)

D = Maximum of all D+ and D- ( ≥ 0) ; for i = 1, ….. n (7)

According to the Kolmogorof-Simirnov method, Fo(Xi) is defined as the CDF evaluated at Xi, Fn(Xi) is defined as
a empirical distribution function, D or ks is defined as the distance test, n is defined as the number of data, and X is
defined as the data value. The test error α = 0.05, CV(0.05), was used for fitting data. Table 1. shows the critical
value of Kolmogorof-Simirnov error.

Table 1. Kolmogorov-Simirnov Critical Value (CV) for α (0.20 and 0.05)

Number of sample CV (0.20) CV (0.05)

4 0.494 0.624
5 0.446 0.564
6 0.411 0.521
7 0.381 0.486
8 0.358 0.457
9 0.339 0.432
10 0.322 0.411

3. Test results and discussion

Fig 4. Shows that the needle electrode with tip radius 10 μm generates lower PDIV than needle electrode with tip
radius 20 μm. These results are in accordance with the maximum electric field stress simulated for both tip radii.
The simulation results show that the value of electric field stress of needle electrode with tip radius 10 μm is nearly
two times of needle electrode with tip radius 20 μm as stated in the Table 2. However, the PDIV values differ 7.8
kV for ramp test method and 5.3 kV for combine method. The different of PDIV level may cause by the randomness
evolving of cavity within mineral oil and local heating process influenced by the electric field stress of different
needle tip dimensions. The electric field stress affects also the aggressiveness of ions movement within mineral oil.
This movement caused by the high electric field stresses which could evoke the kinetic energy for ions. While the
movement of ions toward area near the tip of needle electrode, the possibility of discharge is high. Fig 4 also shows
the standard deviation of PDIV value of needle electrode with tip radius 20 μm that higher than needle tip radius 10
μm by using ramp method and combine method. This result indicates that the use of needle tip radius 10 μm for
PDIV measurement of mineral oil under room temperature is better than using needle electrode with tip radius 20
μm. In addition, the combine method procedure is more sensitive to detect PDIV than ramp method. The reason of
this sensitivity is related to the development of cavities that requires sufficient energy to discharge and influenced by
local heating due to electric field stress. Under the constant test voltage for 1 minute (in accordance with combine
method), the discharge possibly occur whilst cavities are already formed. Unlike the case of ramp method, with
increasing rate of 1 kV/s, the cavity establishment is not progressing well. Therefore, the PDIV with charge ≥ 100
pC will be only detected when the electric field stress is higher than combine method and occasionally occur near to
the breakdown voltage level. Hence, the measurement of PDIV using the combine method is more secure from
breakdown event than ramp method. The quantity of charge can be seen on Table 3. The test results show that the
average PDIV charge (QIEC) of needle electrode with tip radius 20 μm is higher than PDIV charge of needle tip
radius 10 μm and this related to the increasing of PDIV value.
416 T.F. Sipahutar et al. / Procedia Technology 11 (2013) 411 – 418

Table 2. Simulation results of electric field stress at the needle tip using 100 volt input

Needle Tip Radius Electric Field Strength (kV)

10 μm 13.7
20 μm 8.1

Table 3.The average of PDIV charge (QIEC) in accordance to the test method and needle electrode tip radius

Needle Tip Radius QIEC - Ramp Method QIEC - Combine Method

10 μm 259 pC 136 pC

20 μm 407 pC 213 pC

45 9
Partial Discharge Inception Voltage (kV)

40.2
40 8

Standard Deviation (kV)

35 32.4 32.3 7
30 27.0 6
25 5
20 3.6 4
15 2.1 3
2.1
10 1.3 2
5 1
0 0
Ramp Method Test Method Combine Method
Needle Tip Radius 10 μm Needle Tip Radius 20 μm

Fig. 4.Test results of average PDIV and Standard Deviation from different test methods and needle tip radius

Fig 5 and 6 represent the cumulative distribution function and probability density function of PDIV value of ramp
method and combine method respectively. The numerical analysis of PDIV value of ramp method and combine
method is in accordance with the Weibull distribution. The compliance of data to the Weibull distribution for each
test method is proved by using Kolmogorov-Simirnov (KS) method. The congruence of PDIV data to the Weibull
distribution is indicated by D+ and D- parameter. While the maximum value of D+ or D- less than critical value of
KS that presented in the table 1, then hypothesis of data agrees with the Weibull distribution is accepted. The
Confidence Interval (CI) for lower and upper bounds is 90% confidence limit for the percentiles 0.1%, 1%, 5%,
10%, 30%, and 95%. The higher frequently density function of PDIV value using ramp method is in the 33.2-34.2
kV for needle tip 10 μm whilst 39.6-41.8 kV for needle tip 20 μm. Meanwhile, for combine method, the higher
frequently density function lies in the 27.4-28.2 kV for needle tip 10 μm whilst 34-34.1 kV for needle tip 20 μm.
The probability density function of PDIV value can be used as information to determine the frequently density of the
occurrence of PDIV.
 T.F. Sipahutar et al. / Procedia Technology 11 (2013) 411 – 418 417

100% 18%
Cumulative distribution function (%)

Probability Density Function (%)

90% 16% lower confidence bound [needle 10 μm]
80% 14% cdf data points [needle 10 μm]
70% upper confidence bound [needle 10 μm]
12%
60% lower confidence bound [needle 20 μm]
10% cdf data points [needle 20 μm]
50%
8% upper confidence bound [needle 20 μm]
40%
6% pdf data points [needle 10 μm]
30% pdf data points [needle 20 μm]
20% 4%
10% 2%
0% 0% 10 μm [α = 34.1715, β = 15.1088, D+ = 0.1764]
14.00 24.00 34.00 44.00 54.00 20 μm [α = 42.2330, β = 10.4032, D+ = 0.1830]
Partial Discharge Inception Voltage (kV) - Ramp Method [Confidence Interval (CI) =90%]

100% 40%
Cumulative distribution function (%)

lower confidence bound [needle 10 μm]

Fig. 5. Weibull CDF and PDF of PDIV value Ramp Method

Probability Density Function (%)

cdf data points [needle 10 μm]
80%
30% upper confidence bound [needle 10 μm]
lower confidence bound [needle 20 μm]
60%
cdf data points [needle 20 μm]
20% upper confidence bound [needle 20 μm]
40% pdf data points [needle 10 μm]
pdf data points [needle 20 μm]
10%
20%

0% 0% 10 μm [α = 28.3766, β = 27.1657, D+ = 0.2796]

14.00 24.00 34.00 44.00 20 μm [α = 33.7061, β = 21.1148, D+ = 0.2991]
[Confidence Interval (CI) =90%]
Partial Discharge Inception Voltage (kV) - Combine Method

Fig. 6. Weibull CDF and PDF of PDIV value Combine Method

4. Conclusion
The effect of test method and needle tip radius for PDIV measurement is an important parameter to consider.
According to the test results, the combine method procedure is more sensitive because detect PDIV at low voltage,
secure from breakdown event than ramp method. Meanwhile, the needle tip electrode influences the test results, and
needle electrode with tip radius 10 μm is recommended for PDIV measurement of the mineral oil. Finally, the test
results for PDIV obtained from experiment test agree with Weibull distribution and by using the cdf and pdf value,
the frequency density of the occurrence of PDIV can be known.

Acknowledgements
The author would like to thank PT. PLN (Persero) for granting a scholarship and great gratitude for the facility
support by Institute of High Voltage Engineering and System Management, Inffeldgasse 18, 8010 Graz, Austria
Graz in the implementation of this project.

References
[1] Muhr.M, Schwarz.R, Pack.S, Koerbler.B. Unconventional Partial Discharge Measurement, proceedings of The Conference on Electrical
Insulation and Dielectric Phenomena; 2004. p 430-433.
[2] N.Rotby. Sally, H.El-Hag.Ayman, Salama. M.M.A, Bartnikas. R. Partial Discharge Detection and Location Inside The Winding of Power
Transformer”, proceedings of Electrical Insulation and Dielectric Phenomena (CEIDP);2012. p. 84 - 87.
418 T.F. Sipahutar et al. / Procedia Technology 11 (2013) 411 – 418

[3] Pompili. M, Mazetti. C, Bartnikas.R. Partial Discharge Inception Voltage Measurements in Dielectric Liquids, proceedings of International
Conference on Dielectric Liquids (ICDL); 2008. p. 1 - 4.
[4] Pattanadech. N, Pratomosiwi.F, Wieser.B, Baur.M, Muhr.M. The Study of Partial Discharge Inception Voltage of Mineral Oil Using Needle-
Plane Electrode Configuration, proceedings of International Conference on High Voltage Engineering and Application, Shanghai, China;
September 17-20 2012. p. 174 – 177.
[5] IEC 61294, Insulating Liquids – Determination of the Partial Discharge Inception Voltage (PDIV) – Test Procedure.
[6] Sima Wenxia, Jiang Chilong, Lewin Paul, Yang Qing, Yuan Tao. Modeling of Partial Discharge Process in a Liquid Dielectric: Effect of
Applied Voltage, Gap Distance, and Electrode Type, proceedings of Energies 2013 Journal, ISSN 1996-1073; 2013. p. 934 – 952
[7] Cisse. L, Bamji. S.S, Bulinski. A.T. Electric Field Calculations for Needle-Plane Geometry and Space Charge in Polyethylene, IEEE
Transactions on Journal on Dielectric and Electrical Insulation, 2003, Volume 10.
[8] X.Wang and Z.D. Wang.Discussion on the effectiveness of IEC 1294:1993, proceedings of the 11th International Electrical Insulation
Conference, Birmingham, UK; 26-28 May 2009. p. 27 – 33.
[9] Contin. A, Cacciari. M, Montanari. G.C. Estimation of Weibull Distribution Parameters For Partial Discharge Inference”, proceedings of
International Conference on Electrical Insulation and Dielectric Phenomena;1994. p. 71 - 78
[10] IEEEStd 930TM . IEEE Guide for the Statistical Analysis of Electrical Insulation Breakdown Data.
[11] IEC 60270. High Voltage Test Technique-Partial Discharge Measurement 2002-12; 2002.
[12] Jye-Wu Shuo. Estimation of the Parameters of the Weibull Distribution with Progressively Censored Data ,J.Japan Statistic Society;2002.
[13] Kolmogorof –Simirnov. A Goodness of Fit Test for Small Samples, A Publication of Reliability Analysis Centre, START 6, 2003,Vol.10.