Fault Identification of In-Service Power Transformer using Depolarization Current Analysis

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 7, No. 2, April 2017, pp. 559~567
ISSN: 2088-8708, DOI: 10.11591/ijece.v7i2.pp559-567  559
Journal homepage: http://iaesjournal.com/online/index.php/IJECE
Fault Identification of In-Service Power Transformer using
Depolarization Current Analysis
M.A. Talib1
, N.A. Muhamad2
, Z.A. Malek3
, B. T. Phung4
1,3
Institute of High Voltage & High Current, Fakulty of Electrical Engineering, Universiti Teknologi Malaysia,
81310 Johor Bahru, Johor, Malaysia
1
TNB Research Sdn Bhd, No 1 Jalan Air Itam, Kajang Selangor, Malaysia
2
School of Electrical and Electronic Engineering, Engineering Campus, University Sains Malaysia, 14300 Nibong Tebal,
Penang, Malaysia
4
School of Electrical Engineering & Telecommunications, University of New South Wales, Sydney, Australia
Article Info ABSTRACT
Article history:
Received Apr 7, 2016
Revised Oct 4, 2016
Accepted Oct 18, 2016
Preventive diagnostic testing of in-service power transformers require system
outage and expert‟s knowledge and experiences in interpreting the
measurement results. The chemical oil analysis may cause significant
variance to measurement results due to the different practices in oil sampling,
storage, handling and transportation. Thus, a cost effective measuring
technique by means of a simpler method that is able provide an accurate
measurement results is highly required. The extended application of
Polarization and Depolarization Current (PDC) measurement for
characterization of different faults conditions on in-service power
transformer has been presented in this paper. The oil sample from in-service
power transformers with normal and 3 different faults type conditions were
sampled and tested for Dissolved Gases Analysis (DGA) and PDC
measurement. The DGA results was used to confirm type of faults inside the
transformer while the PDC pattern of oil with normal, partial discharge,
overheating and arcing were correlated to the oil sample conditions. The
analysis result shows that depolarization current provides significant
information to defferenciate fault types in power transformer. Thus this
finding provides a new alternative in identifying incipient faults and such
knowledge can be used to avoid catastrophic failures of power transformers.
Keyword:
Depolarization current
Dielectric response
Fault diagnosis
Polarization current
Transformer fault
Copyright © 2017 Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
M. A. Talib
Institute of High Voltage & High Current,
Faculty of Electrical Engineering,
Universiti Teknologi Malaysia,
81310 Skudai Johor MALAYSIA
Email: mohdaizam@tnbr.com.my
1. INTRODUCTION
The continuous supply of electricity from generating stations to the customer is very much
dependent on reliability of power transformers installed in the substations [1]. The failures of this strategic
unit at any substation can cause interruption of power supply and substantial damage to other components. It
may also become potentially dangerous for utility personnel as failures may result to explosions and fire.
Therefore, determining their condition would be of tremendous importance to the power utilities. Presently,
the maintenance paradigm has moved from scheduled based maintenance to condition based maintenance
with the objective to minimize the operational cost and improve the reliability of transformers [2], [3]. A new
material and maintenance techniques mainly towards condition based maintenance thus continuously
developed to sustain high power reliability and availability [4].

 ISSN:2088-8708
IJECE Vol. 7, No. 2, April 2017 : 559–567
560
Well established chemical and electrical diagnostic testing methods have been used to assess the
condition of the transformer and classified them according to different fault and ageing mechanisms. For
example, the presence of moisture in insulation system can be determined by dielectric dissipation factor
(DDF) and insulation resistance (IR) measurement [5], while the mechanical integrity of winding and core
clamping structure due to short circuit forces can be confirmed by Frequency Response Analysis (FRA)
[6], [7] and the problems with bad joints or connections are identified through winding resistance
measurement [8]. For instance, decomposition of the hydrocarbon chain caused by thermal and electrical
faults will liberate small quantities of gases and can be classified and quantified by means of DGA [9-11].
Nevertheless, the existing diagnostic technique particularly electrical measurement require electrical
isolation or transformer outage before the measurement can be performed. Beside that, the duration for the
complete measurement on one transformer which can last for several hours has concerned the power utilities
which limit the measurement on basic routine testing and applied on the selected critical transformer due to
difficulties of long hour outages. In some situations, due to system overloading the outage of power
transformer is not permitted and caused the insulation ageing is not being assessed and determined.
Even though, the chemical analyses on transformer oil insulation provide direct information on
parameters like oil moisture content, sludge, acidity, Degree of Polymerization (DP) and DGA [12-14],
however, most of the chemical analyses are conducted under laboratory condition where sample sampling,
storage and transportation cause significant variance due to different practices and location of the site. All
these will incurred high cost to the power utilities as part of their condition based maintenance activities.
In order to avoid any unwanted circumstances, accurate assessment of transformer condition is
crucial so that replacement or refurbishment of the assets can be taken before failures take place. Thus, a cost
effective measuring technique, simpler but can provide an accurate measurement results is highly required.
In this paper, the PDC was investigated to be used as an alternative diagnostic technique to
determine the incipient fault occurs in power transformers. The oil samples was taken from various
transformers and tested for its condition. The analysis of PDC pattern for transformers with normal, partial
discharge, overheating and arcing faults conditions are presented in this paper. An Artificial Neural Network
(ANN) was used to identify each spectrum characteristic for faults identification and the results was
compared with other established diagnostic interpretation techniques.
2. POLARIZATION AND DEPOLARIZATION CURRENT MEASUREMENT
This section covers the details of experimental test set-up and the testing procedure of PDC
measurement.
2.1. Measuring Test Circuit
The polarization and depolarization current are measured using the commercial PDC known as
Analyzer-1MOD manufactured by Alff Engineering Switzerland. The analyzer can supply the excitation
voltage up to 2kV DC and measure the dielectric response in time domain up to 10000s [15]. The analyzer is
then connected to the test cell with electrode gap of 1.5mm and capacitance of 60pF. Two electrodes circuit
were used, one for application of the test voltage and the other is for current measurement. The amount of
insulating liquid required for the measurement is about 210ml. Figure 1 shows the measurement set-up used
in the laboratory experimental works. All the measurements were carried out at 1000V and polarization and
depolarization times of 10,000s each.
Figure 1. Laboratory Experimental Test Set-Up [15]
S1
S2
Control
voltage
Computer
with DSP
board A
+
-
Test VesselMeasured
current
PDC Analyzer 1MOD

IJECE ISSN: 2088-8708 
Fault Identification of In-Service Power Transformer Using Depolarization Current Analysis (M. A. Talib)
561
2.2. Oil Test Sample
The oil samples used in this experiment are hydrocarbon based mineral oil obtained from in-service
transformers operating at Malaysia power distribution network. The operating conditions of these
transformers had previously been identified based on DGA data of their respective oil samples. The test
samples were classified into four conditions; normal, partial discharge, overheating and arcing. Prior to PDC
measurement, the DGA was performed again on the oil samples to confirm the presence of combustible gases
and its nature of faults.
In each condition, the oil samples were taken from 10 units of power transformers and every unit
has 3 samples; i.e. there are 30 oil samples for each condition. Each PDC assessment was repeated for five
times on all oil samples. The aim of having such a large number of the same samples and repetition of PDC
assessment process was to ascertain the reproducibility of the test results.
To avoid the effect of moisture on polarization and depolarization current measurement, it was
crucial that the oil samples were dry. Thus, the selection of transformers are limited to the units with moisture
content less than 30ppm; considered dry according to IEC 60422-2013 [12].
2.3. Testing Procedure
The consistency of experimental test results is highly dependent on the PDC test procedure used.
The test vessel must be cleaned prior to each measurement process as contamination from previous oil
samples could affect the next test results. The residual charges can also affect the polarization current results,
thus before the excitation voltage is applied to the test sample, the test object was discharged for 500s. For
safety purposes, the test sample was placed in the shielded case. Figure 2 shows the flow chart of
experimental works.
Figure 2. Flow Chart of Experimental Works
3. RESULTS AND DISCUSSIONS
The polarization and depolarization current of oil with different fault conditions in this research are
analyzed based on its graphical pattern obtained from the experiments. Artificial Neutral Network (ANN)
was used to further analyze the measured currents for transformer fault identification.
3.1. Analysis of Polarization and Depolarization Current
Since oil was the only insulation tested in this research, it would be expected that the changes of
measured current is attributed to changes at the initial time response [16]. Figure 3(a) and (b) shows the
average of polarization and depolarization current on oil samples with normal, partial discharge, overheating
and arcing condition. The analysis of polarization current demonstrated that the measured current starts to
Start
Discharged 500s
Check residual
charges. Low and
constant reading
Measured polarization
current (1s -10000s)
Completed?
Measured depolarization
current (1s -10000s)
Completed?
Yes
No
Yes
No
No Yes Re-Test for
5 times.
Completed?
No
Test other
oil sample
Yes
End

 ISSN:2088-8708
562
stabilize and saturated almost after 100s of being charged until the completion of polarization process.
Meanwhile, the depolarization current takes a longer time and only after 1000s the measured currents are
starts to stabilize until the end of discharging process.
(a) Polarization current (b) Depolarization current
Figure 3. Polarization and depolarization current pattern of normal, partial discharge, overheating and arcing
oil samples
Figure 4. Depolarization current pattern of oil samples at duration of 4 seconds to 100 seconds
The curves have shown that the oil with normal condition had the least value of polarization current,
while the values were progressively higher in partial discharge, overheating and arcing fault conditions. The
average magnitude of maximum polarization current recorded on normal oil sample is 50.23nA, partial
discharge is 118.33nA, overheating is 184.96nA and arcing is 226.45nA. The high polarization current of oil
samples with fault conditions can be attributed to the fact that faults in power transformer cause breakdown
of hydrocarbon structures of oil; thus leading to higher charge carrier mobility and high current magnitude
when any external electric field is applied. In addition, it can be seen that the polarization current pattern
10
1
10
2
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
x 10
-9
time,t,[s]
DepolarizationCurrent,Idepol,[A]
Normal
Partial discharge
Overheating
Arcing
DepolarizationCurrent,Idepol,[A]
time,t[s]
10
0
10
1
10
2
10
3
10
4
0
0.5
1
1.5
2
2.5
x 10
-7
time,t,[s]
AveragePolarizationCurrent,Ipol,[A]
Normal
Partial discharge
Overheating
Arcing
10
0
10
1
10
2
10
3
10
4
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
x 10
-9
time,t,[s]
AverageDepolarizationCurrent,Idepol,[A]
Normal
Partial discharge
Overheating
Arcing
AverageDepolarizationCurrent,Idepol,[A]
time,t[s]
AveragePolarizationCurrent,Ipol,[A]
time,t[s]

IJECE ISSN: 2088-8708 
563
seemingly correspond to degree of fault; the initial slopes of the oil samples with fault conditions are steeper
than the oil with normal condition.
The plot of depolarization current on oil samples with normal, partial discharge, overheating and
arcing fault conditions shows that the depolarization curves are almost identical. The average magnitude of
maximum depolarization current recorded is -7.12nA for normal oil samples, -7.06nA for partial discharge,
-8.31nA for overheating and -8.28nA for arcing samples.
An interesting point was observed on depolarization current pattern at duration of 4 seconds to 100
seconds as the responses are changes with different transformer conditions, as shown in Figure 4 It can be
seen that the normal oil condition has low current pattern and as the energy of the fault increases the current
pattern of oil samples with fault conditions are also increases.
3.2. Selection of Current for Fault Identification
According to [17], the polarization current is the combination of the absorption current due to
polarization and the conduction current. On the other hand, the depolarization current consists only of the
absorption current because conduction current exists only when the voltage is applied to the dielectric. From
the results, both polarization and depolarization currents were found to have correlation with faults in power
transformer. Nevertheless, it is important to identify and classify the currents that are effectively used to
characterize incipient faults in power transformers.
In this research, an Artificial Neural Network (ANN) was used to further analyze the currents for
transformer fault identification. The design and simulation of the ANN was done using Neural Network
Toolbox provided in MATLAB.
Initially, the input data was normalized with the objective of ensuring a uniform statistical
distribution of each input value. In this research, the maximum and minimum values of a group of data are
used to normalize each single input using the following equation
[ ] (1)
where I is the normalized value of a single data. Imin and Imax are the input range of the ANN and the
typical range is -1 to 1 or 0 to 1. D is the input value for each data while Dmin and Dmax are the minimum and
maximum values of the input data.
The ANN is a 3 layered structure comprised of input, hidden and output layer. In order to increase
the prediction capability of the network, the ANN utilized 2 hidden layers with 100 neurons on each layer.
Initially, the learning algorithm employed to train the ANN is Back Propagation (BP). However, the
algorithm is too slow for practical problem due to the fact that multilayer networks use sigmoid transfer
functions in the hidden layers. These functions, known as „squashing‟, compress infinite input range into a
finite output range. Sigmoid function is characterized by the fact that their slope must approach zero as the
input gets large. This causes a problem when using steepest descent to train a multilayer network with
sigmoid function because the gradient can have a very small magnitude; therefore causing small changes in
the weights and biases, even though the weights and biases are far from their optimal values. Thus, a faster
Back Propagation (BP) algorithm known as Resilient Back Propagation (Rprop) was used as learning
algorithm to train the ANN.
This algorithm eliminates the harmful effects of the magnitudes of the partial derivatives. Only the
sign of the derivative determines the direction of the weight update. The size of the weight change is
determined by a separate update value. The update value for each weight and bias is increased by a factor that
depends on the derivative of the performance function with respect to weight changes sign from the previous
iteration. If the derivative is zero, the update value remains the same. Whenever the weights are oscillating,
the weight change is reduced. If the weight continues to change in the same direction for several iterations,
the magnitude of the weight change increases.
Table 1. Relationship of ANN output value and type of fault
Output Neuron
Value
Type of Fault
1000 Normal
0100 Partial discharge
0010 Overheating
0001 Arcing

 ISSN:2088-8708
564
A PDC sample with a total of 21,900 input data were used to train the ANN and later was tested
with another PDC sample data that is separated from the earlier training data. The output will then refer to
different types of faults as shown in Table 1. The training procedure and the ANN structure with 2 hidden
layers are shown in Figure 5 and Figure 6 respectively.
Figure 5. ANN training procedure
Figure 6. Schematic diagram of the ANN
Finally, the accuracy and performance of the ANN output is determined by the following equation,
(2)
Start
Get the database
Matching?
No
YesReaching the
performance goal?
Test the ANN
Set input data of
training
Normalization of data
Set the target output
Start training the
ANN
Accept Denied
End
Yes No

IJECE ISSN: 2088-8708 
565
The results of ANN analysis on polarization and depolarization current are shown in Table 2.0 and
Table 3.0 respectively.
Table 2. ANN result of polarization current
Condition
Correct output
recognised
Wrong output
recognised
% ANN
Accuracy
Normal 87 3 97
Partial discharge 58 7 80
Overheating 15 5 75
Arcing 36 9 90
Table 3. ANN result of depolarization current
Condition
Correct output
recognised
Wrong output
recognised
% ANN
Accuracy
Normal 86 4 96
Partial discharge 60 5 92
Overheating 19 1 95
Arcing 43 2 96
The ANN analysis of polarization current demonstrated that higher accuracy of more than 90% was
observed for oil samples on transformer with normal and arcing fault conditions. On the other hand, lower
accuracy was observed for identification and classification of partial discharge and thermal overheating faults
as the accuracy recorded is only 80% and 75% respectively.
Meanwhile, higher accuracy of more than 90% was recorded for depolarization current on all
conditions. Both normal and arcing condition has ANN accuracy of 96%, whereas thermal overheating fault
has accuracy of 95%. The partial discharge fault was observed slightly lower with ANN accuracy of 92%.
Based on this ANN simulation results, depolarization current yielded a better accuracy in identifying and
classifying different transformer conditions as compared to polarization current and identified as the current
of main interest that can be uniquely used for fault identification of in-service power transformer.
The results has confirmed the initial research done by Bhumiwat [18] on overheating of transformer
solid insulation stated that the condition can be identified by the prominently crooked shape of the
depolarization current.
4. CONCLUSION
In this paper, the extended application of PDC measurement for characterization of different faults
conditions of in-service power transformer was investigated. The oil sample with normal, partial discharge,
overheating and arcing condition were taken from respective units and tested for its combustible fault gases,
polarization and depolarization current.
The changes of the molecular properties of the insulating oil material due to different fault
conditions was found influences the initial time response of both measured currents. The sensitivity analysis
using Artificial Neural Network (ANN) shows that depolarization current gives a better accuracy in
identifying and classifying different transformer conditions as compared to polarization current. Thus,
depolarization current measurements provide a new alternative in identifying incipient faults and such
knowledge can be used to avoid catastrophic failures of power transformers.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the support of TNB and Malaysia Ministry of Higher
Education, Universiti Teknologi Malaysia (Vot 05J82, 04H67, 4L014, and 4L607) for technical and financial
support.
REFERENCES
[1] Sunil M, Jaralikar, Aruna M, “Study of Performance of Power Station for Operational Optimization”, International
Journal of Electrical and Computer Engineering (IJECE),Vol. 2, No. 1, February 2012, pp. 17-25.
[2] J.A. Lapworth, P.N. Jarman and I.R. Funnell, “Condition Assessment Techniques For Large Power Transformers”,
2nd International Conference on the Reliability of Transmission and Distribution Equipment, pp 85-90, 1995.
[3] R. Schwarz, M. Muhr, “Diagnostic Methods for Transformer”, IEEE International Conference on Condition
Monitoring and Diagnosis, pp 974-977, 2008.

 ISSN:2088-8708
566
[4] H. Borsi, E. Gockenbach, “Comparison between Oil Immersed and SF6 Gas Power Transformers Ratings”,
TechCon Asia-Pacific Conference, 1-2 May 2007, Sydney, Australia.
[5] Wang, M.; Vandermaar, A.J.; Srivastava, K.D., “Review Of Condition Assessment Of Power Transformer In
Service”, IEEE Electrical Insulation Magazine, Vol 18, Issue 6, pp 12-25, 2002.
[6] Liu Yong, Xiaochen ; Ji, Shengchang “Study On Sweep Frequency Impedance Method For Detecting Internal
Winding Short Circuit Fault In Power Transformer” IEEE International Conference on Power Modulator and High
Voltage (IPMHVC), pp 137-140, 2014.
[7] Aljohani, O.; Abu-Siada, A. “Application Of FRA Polar Plot Technique To Diagnose Internal Faults In Power
Transformers”, IEEE PES General Meeting Conference & Exposition, pp 1-5, 2014.
[8] IEEE Guide for Diagnostic Field Testing of Fluid-Filled Power Transformers, Regulators, and Reactors, IEEE
C57.152-2013.
[9] IEC Standard, Mineral Oil-impregnated Electrical Equipment in Service - Guide to the Interpretation of Dissolved
and Free Gases Analysis, IEC 60599:2015.
[10] Duval. M, dePablo. A, “Interpretation of Gas-in-oil Analysis Using New IEC Publication 60599 and IEC TC10
Database”, IEEE Electrical Insulation Magazine, Vol 17(2) March/April, pp 31-41, 2001.
[11] Sherif S. M. Ghoneim, Sayed A. Ward, “Dissolved Gas Analysis as a Diagnostic Tools for Early Detection of
Transformer Faults”, Advances in Electrical Engineering Systems (AEES) 152 Vol. 1, No. 3, 2012.
[12] IEC Standard, Mineral Insulating Oils in Electrical Equipment – Supervision and Maintenance Guidance, IEC
60422:2013.
[13] Priyesh Kumar Pandey, Zakir Husain, R.K. Jarial, “ANFIS Based Approach to Estimate Remnant Life of Power
Transformer by Predicting Furan Contents”, International Journal of Electrical and Computer Engineering
(IJECE),Vol. 4, No. 4, August 2014, pp. 463-470.
[14] A.A. Suleiman, N.A. Muhamad, N. Bashir, N.S. Murad and Y.Z. Arief. “Effect of Moisture on Breakdown Voltage
and Structure of Palm Based Insulation Oils”. IEEE Transactions on Dielectric and Electrical Insulation (TDEI),
21(5) October 2014.
[15] PDC-Analyzer 1MOD User Guide Hardware, Alff Engineering, Switzerland.
[16] A. Küchler, T. Leibfried, B. Breitenbauch, J.J. Alff, V. Der Houhanessian, W.S. Zaengl, “Transformer Insulation
Diagnosis by Polarization and Depolarization Current Analysis”, Proceeding of 48. Internationales
Wissenschaftliches Kolloquium, Technische Universität Ilmenau, 2003.
[17] S.A. Bhumiwat, “The Latest On-Site Non-Destructive Technique for Insulation Analysis of Electrical Power
Apparatus”, Weidman ACTI Annual Technical Conference, 2004.
[18] S.A. Bhumiwat, “Identification of Overheating in Transformer Solid Insulation by Polarization Depolarization
Current Analysis”, IEEE Electrical Insulation Conference (EIC2013), pp 449-453, 2013.
BIOGRAPHIES OF AUTHORS
Mohd Aizam Talib received his Bachelor in Electrical Engineering from the University of
Portsmouth, UK in 1997, and Master degree in Electrical Engineering from Universiti Tenaga
Nasional (UNITEN), Malaysia in 2001. Upon graduation in 1997, he worked with ABB
Transmission and Distribution Sdn Bhd as a Design Engineer. Since 1998, he has been employed
by TNB Research Sdn Bhd as a Research Engineer. His research interests are in transformer
condition monitoring, insulation diagnostic and dielectric measurements. He is currently a Ph.D.
degree student at Universiti Teknologi Malaysia.
Nor Asiah Muhamad is senior lectures in School of Electrical and Electronic Engineering
Universiti Sains Malaysia since Sep 2016. She has over 13 experince working with Institute of
High Voltage and High Current (IVAT) Universiti Teknologi Malaysia as lecturer as well as
researcher. Received her PhD in 2009 from University of New South Wales, Australia in area oil
insulation monitoring and diagnosis. Received Bachelor degree in Electrical and Electronic
Engineering from Universiti Teknologi PETRONAS, Malaysia in 2002 and Master degree in
Electrical Power Engineering from University of South Australia in 2006. Start doing research in
power system equipment monitoring since 2005 and the main topics related to her work are
insulation diagnosis, new material for HV insulation and condition monitoring systems.

IJECE ISSN: 2088-8708 
567
Zulkurnain Abdul-Malek was born in Kuantan, Malaysia in 1965. He obtained his B.E. from
Monash University (Melbourne) in 1989 and his MSc and PhD degrees from University of
Wales Cardiff in 1995 and 1999 respectively. Since 1989 he has been a member of the Electrical
Engineering Faculty at UTM. He is currently an Associate Professor and the Director of UTM
Institute of High Voltage and High Current. His research interests include intelligent condition
monitoring of HV equipment, lightning detection and warning systems, nanodielectrics and fast
transient response of high voltage surge arresters.
B.T. Phung Gained a Ph.D. in Electrical Engineering in 1998 and is currently a Senior Lecturer
in the School of Electrical Engineering at the University of New South Wales, Sydney,
Australia. He has over 30 years of practical research/development experience in partial discharge
measurement, and in on-line condition monitoring of high- voltage equipment. His research
interests include electrical insulation (dielectric materials and diagnostic methods), high-voltage
engineering (generation, testing and measurement techniques), electromagnetic transients in
power systems, and power system equipment (design and condition monitoring methods).

Fault Identification of In-Service Power Transformer using Depolarization Current Analysis

More Related Content

What's hot (19)

Similar to Fault Identification of In-Service Power Transformer using Depolarization Current Analysis (20)

More from IJECEIAES (20)

Recently uploaded (20)

Fault Identification of In-Service Power Transformer using Depolarization Current Analysis