Understanding of Partial discharge activity in

transformer oil under transient voltages adopting

Acoustic emission technique

A. S. Prasanna Venkatesh2, M. G. Danikas

3 and R. Sarathi

1

1Department of Electrical Engineering, IIT Madras, Chennai- 600 036

2Department of Electronics and Communication Engineering, SSNCE, Kalavakkam, Kancheepuram- 603 110

3Department of Electrical and Computer Engineering, Democritus University of Thrace, GR-671 00 Xanthi, Greece

Abstract-- One of the causes for the failure of power

transformers is due to formation of partial discharges. Partial

discharge taking place in transformer oil under AC voltages

were studied by researchers worldwide in detail. But for the PD

formation under transient voltages, the results are scanty. In

the present work, three different defects were considered viz.

corona discharge, discharge initiated due to floating particle

and discharge formation due to particle movement and the

discharges initiated by these defects under standard lightning

impulse were studied adopting acoustic emission technique. The

FFT analysis of AE signal obtained due to different defects has

close resemblance and hence the ternary plots were obtained

using the FFT output of the AE signal and classified. MRSD

technique was adopted to remove the noise signal in the AE

signal formed due to the discharges. The AE signal obtained

due to discharges under AC voltages were compared with the

signals obtained under standard lightning impulse voltage.

Index Terms— Acoustic Emission, corona, FFT, Lightning impulse, Partial Discharge, Transformer oil.

I. INTRODUCTION

Transformer is one of the most important and expensive

equipments in the electrical power system network. The

insulation of the transformers consists of the transformer oil

with solid insulating materials such as pressboard, paper

insulation, bakelite, etc. One of the major causes for failure

of transformer insulation is due to formation of partial

discharges (PD) [1]. These partial discharges can be formed

due to any conducting or non conducting particle present in

the transformer insulation, corona discharge from the

protrusions in the current carrying conductor or in the ground

electrode or by the discharges initiated due to floating

conducting particles. The floating conducting particle can be

due to any dropout of metal from the surface of the conductor

during manufacturing.

The presence of any defect in the insulation structure,

under normal operating voltages can cause local field

enhancement near the defect site initiating discharges,

thereby releasing certain amount of energy in the form of

burst/impulsive pulses (acoustic energy) that radiate in all

directions [1, 2]. The released energy can be detected by

mounting a transducer over the surface of the structure. This

process is known as “Acoustic Emission” (AE). The signals

detected are called acoustic signals, which are used for

diagnostic study [3]. Acoustic emission technique’s use for

identification of defects and its application in the high voltage

field is significant. Considerable research work was carried

out to understand the partial discharge activity in transformer

oil insulation under AC and DC voltages [6, 7]. In the

present work, in addition to the normal operating AC voltage,

the transformer insulation is subjected to transient voltage

formed due to lightning or due to any switching operation.

Most of the work carried out in transformer oil is to

understand the partial discharges under the AC/DC voltages.

The partial discharge formed due to lightning impulse

voltage, for various defects were studied.

Hence the author has carried out a methodical experimental

study to understand the partial discharge activity due to

various defects in transformer oil, under lightning impulse

voltage, adopting Acoustic emission technique. The

influences of polarity of the applied voltage on the PD

formation were studied. It is essential to classify the

discharges initiated due to different defects and in the present

study ternary plots were obtained by using the FFT output of

the AE signals measured during discharges, for classification.

II. EXPERIMENTAL STUDIES

The basic experimental setup used in the present study is

shown in Fig. 1. The experimental setup could be sectioned

into three parts. The first, second and third parts of the

experimental setup include a high voltage source, an oil test

chamber with an acoustic emission sensor mounting (Test

apparatus) and a pre-amplifier with a data acquisition system

for post analysis of the acquired acoustic signals.

A. High Voltage source

The high AC voltage is produced from a 100kV, 5kVA, 50

Hz test transformer. The applied voltage was measured using

a capacitance voltage divider with peak voltmeter. The

standard lightning impulse voltage was generated using single

step impulse generator (140kV) and the generated voltage

was measured using capacitance divider.

2011 6th International Conference on Industrial and Information Systems, ICIIS 2011, Aug. 16-19, 2011, Sri Lanka

98

978-1-61284-0035-4/11/$26.00 ©2011 IEEE

Fig. 1 Experimental setup

B. Test Electrode Configuration

The stainless steel leak-proof test cell with dimensions of

12x12x12 cm of rectangular cross section, fitted with a high

voltage bushing at the top and bottom side of the chamber,

filled with transformer oil is used for the experiment. To

simulate the corona effect a needle-plane configuration was

used in the above mentioned test cell. To simulate the PD

activity due to the presence of floating particle, an

arrangement was made such that a conducting particle

(copper wire of dimension 5mm x 1mm) remains exactly

midway between the top electrode and the ground electrode.

For this purpose a pressboard (thickness 2mm) was used to

keep the particle midway and a small hole was drilled and the

wire was inserted. Thus the floating particle effect was

simulated. To simulate the PD activity due to movement of

the conducting particle (particle movement), a 2mm diameter

aluminium ball was placed on a slightly concave ground

electrode and the distance between the particle and the top

electrode was kept as 5 mm. The acoustic sensor is mounted

on the sidewalls of the test cell so that the longitudinal signal

energy of the acoustic signals produced due to the discharge

is transferred to the input of the AE sensor.

C. Acoustic Emission Instrumentation

The AE sensors are piezoelectric transducers, which

convert the acoustic signal into corresponding electric

signals. The partial discharge signals are wide band signals.

When a signal propagating in the medium hits the walls, it is

partly reflected and partly transmitted because of the low

absorption coefficient of the wall material. The partial

discharge inception voltage is identified as the voltage at

which the first acoustic emission signal is captured for the

defined conditions.

In the present work, a wide band sensor with frequency

response in the range 100KHz – 1MHz was used.

Optimization between the bandwidth and sensitivity is an

important factor. To get a maximum sensitivity, the sensor

must be attached to the test specimen in such a manner that

acoustic energy passes into the transducer with minimum loss

at the transducer material. The required contact was achieved

by applying a thin layer of gel between the sensor and the

surface of the chamber.

The AE signal generated by the sensor has to be amplified

to the required voltage magnitude. This is accomplished with

a pre-amplifier placed close to the sensor to minimize the

pickup of electromagnetic interference. The pre-amplifier has

a wide dynamic range and can drive the signal over a long

length of cable near the data acquisition system. Pre-

amplifiers inevitably generate electronic noise, and it is the

noise that sets the sensitivity of the acoustic emission system.

The gain of the integrated pre- amplifier is set to 40 dB with a

1 MHz bandwidth. In the present study, PCI-2, a 2 channel

acoustic emission system of Physical Acoustic Corporation

was used [8, 9]. The acquired AE signals were processed to

eliminate noise signal by adopting multi-resolution signal

decomposition technique (MRSD).

III. RESULTS AND DISCUSSION

Fig. 2 shows variation in discharge inception voltages due to

various defects (which includes corona activity, floating

particle and by the movement of conducting particle) under

different voltage profiles. The incipient discharges under LI

were generated by applying 80% of the breakdown voltage

that is calculated for the gap with the defect especially for the

floating conductor and the electrode gap with a spherical

particle. If the applied voltages were less than the 80% of the

breakdown voltage of the electrode gap, no discharges were

observed. Thus the inception voltage for the two defects

especially for the floating particle and the particle sitting on

the ground electrode were taken to be 80% of the breakdown

voltage. For each input voltage, the test was carried out eight

times and the AE signals were captured due to the discharge.

Thus comparing, it is observed that irrespective of type of

defect causing discharge, the discharge inception voltage is

less under AC voltage compared with the lightning impulse

voltage. In the present study, the discharge inception voltage

Fig. 2 Inception Voltages in kV for different defects

under different voltage profiles; PM-Particle Movement;

CD-Corona Discharge; FP-Floating Particle

2011 6th International Conference on Industrial and Information Systems, ICIIS 2011, Aug. 16-19, 2011, Sri Lanka

99

is defined as the voltage at which the first AE signal is

acquired. In case of the lightning impulse voltage, the

inception voltage is nearly the same for both polarities.

Among the three types of defects, the discharge inception

voltage (AC/LI) is high for the floating particle followed by

corona discharge and then by the discharge initiated due to

particle movement. Fig.3 shows typical AE signal generated

due to corona discharge in transformer oil under different

voltage profiles. It is observed that burst type or impulsive

type discharges can occur. It is observed that under AC

voltages, for Corona discharge, the FFT analysis of the AE

signal indicates that the maximum energy of the signal lies in

the frequency range between 400 kHz and 600 kHz. This

characteristic is the same under lightning impulse voltages.

Fig. 4 shows the AE signal measured due to discharges

initiated due to floating conducting particle under different

voltages. It is observed that the magnitude of AE signal is

high under AC voltage compared with the AE signal

generated due to discharge initiated under lightning impulse

voltage. The FFT analysis of the AE signal generated due to

floating electrode discharges indicates the energy content

spreads in the range 100 kHz to 600 kHz. The frequency

contents in the AE signal generated due to discharges

initiated by floating particle under LI is much different from

the AC voltage and the frequency contents are different for

positive and negative LI voltages. In general impulsive type

discharge occurs with AE signal generated under AC voltage

and burst type signal occurs under transient voltages

especially with discharge initiated due to floating particle and

movement of conducting particle.

Particle initiated partial discharge is one of the major

causes for the failure of transformer insulation. The particle

levitates once the force exerted by the particle is much higher

than the applied electric field [4]. Fig. 5 shows typical AE

signal generated due to particle initiated discharges under AC

and LI voltages. It is observed that impulsive type discharges

occurs under AC voltage and burst type discharge occurs

under LI voltage. The FFT analysis of AE signal generated

due to particle movement under AC voltage, the frequency

content of the AE signal lies in the entire range of 100kHz to

1MHz. Under LI voltage, irrespective of polarity of LI

voltage, the frequency content of AE signal formed lies in the

range 200-600 kHz. The MRSD technique was used to

remove the noise content from the obtained AE signal in

order to obtain the required AE signal.

The AE time domain signals acquired for different type of

discharges and with its corresponding FFT patterns, it is

observed that the frequency contents of the AE signals are

nearly the same and so difficult to classify the defect causing

AE signal. To understand the intricate details further the

frequency contents in the AE signal formed due to different

defects, the partial power analysis to the AE signal generated

Fig. 3 Typical AE signals generated due to corona discharge under different

voltages (I) AC (II) +LI (III) –LI (a) Time domain (b) its corresponding FFT

output

Fig. 4 Typical AE signals generated due to discharge initiated by floating

particle under different voltages (I) AC (II) +LI (III) –LI (a) Time domain

(b) its corresponding FFT output

Fig. 5 Typical AE signals generated due to particle movement under

different voltages (I) AC (II) +LI (III) –LI (a) Time domain (b) its

corresponding FFT output

2011 6th International Conference on Industrial and Information Systems, ICIIS 2011, Aug. 16-19, 2011, Sri Lanka

100

Fig. 6 Ternary diagram classifying the defects under various voltages (i) AC (ii) +LI (iii) -LI

due to discharges was carried out, which will be the input

data for generating the Ternary plot.

Partial Power is calculated by summing the power

spectrum in a specified range of frequencies and dividing it

by the total power. The power spectrum was calculated up to

800 kHz is split equally and the partial powers were

calculated in three zones (x= (200–400 kHz), y= (400-600

kHz) and z= (600-800 kHz)). The Triangle coordinates

corresponding to energy content be calculated as follows.

Normalized energy content in the range 200-400 kHz = x/(x +

y + z); Normalized energy content in the range 400-600 kHz

= y /(x +y +z) and the normalized energy content in the range

600-800 kHz = z/(x + y + z). This normalisation technique is

identical to the process used to generate the gas in oil ratio

used to plot the Duval’s triangle. Fig. 6 shows the ternary

plot obtained for various defects under AC and lightning

impulse voltage. It is observed that the location in the ternary

plot is slightly different for the corona discharge under AC

and LI voltages. The location of discharge due to floating

particle and particle movement are nearly same under LI

voltages of both polarities. Further analysis is required for its

implementation in practice.

IV. CONCLUSIONS

The important conclusions acquired based on the present

study are the following.

(i) AE sensors could indentify PD generated under transient

voltages. Burst type and impulsive type AE signals are

generated due to partial discharges in transformer insulation.

(ii) The magnitude of AE signal is always high with AE

signals generated by discharges under AC voltages. It is also

noticed that impulsive type discharges occurs under AC

voltages and burst type discharge occurs under lightning

impulse voltage. Irrespective of the type of defect, the

duration of AE signal formed is high under negative LI

voltage compared with positive LI voltage.

(iii) The Frequency domain analysis of AE signal could help

one to identify the dominant frequency contents. But the

ternary plot obtained based on FFT output of the AE signal

helps one to classify the type of discharges. The results of

ternary plot indicates that the frequency content of AE signals

generated due to discharges under positive and negative

lightning impulse and AC voltages are different, indicating

that the mechanism of discharge formation under various

voltages are different.

(iv) Ternary diagram provides major location for variety of

discharges and is a simple visual technique for identification.

REFERENCES

[1] G. Koperundevi , M. K. Goyal, Sunil Das, N. K. Roy, R.

Sarathi, ” Classification of Incipient discharges in Transformer

Insulation using Acoustic Emission Signatures”, 2010 Annual

IEEE India Conference (INDICON).

[2] Ramanujam Sarathi, Prathap D. Singh, Michail G. Danikas,

“Characterization of Partial Discharges in Transformer oil

insulation under AC and DC voltage using Acoustic Emission

Technique”, Journal of ELECTRICAL ENGINEERING ,VOL 58,

NO. 2, 2007, pp.91-97

[3] Prasantha kundu, N. K. Kishore, A. K. Sinha,” Behavior of

Acoustic Partial Discharge In Oil-Pressboard Insulation System”,

2008 IEEE Region 10 and the third ICIIS, Kharagpur, INDIA

December 8-10, Paper Identification No: 88

[4] Boczar T.:” Identification of a Specific Type of PD from

Acoustic Emission Frequency Spectra”, IEEE Trans. Diel.

Electr. Insul. DEI-8 (2001), 598–606.

[5] Kennedy, W.: “Recommended Dielectric Tests and Test Proc

edures for Converter Transformers and Smoothing Reactors,

IEEE Trans. Power Deliv. PD-1 (1986), 161–166.

[6] Pompili, M, Mazzetti, C. Barnikas, R.: “Partial Discharge Pulse

Sequence Patterns and Cavity Development Times in Transformer

Oils under ac Conditions” , IEEE Trans. Diel. Electr. Insul. DEI-

12 (2005), 395–403.

[7] Cavallini A. Montanari,G.C.—CianiI,F.: “Analysis Of Partial

Discharge Phenomena in Paper Oil Insulation Systems as a Basis

for Risk Assessment Evaluation”, Proc. IEEE Intern. Conf. Diel.

Liquids, Coimbra, Portugal, 26 June–1 July 2005, 241–244.

[8] Pollock A. A.: ACOUSTIC Emission Inspection , Physical

Acoustic Corporation ,Technical Report TR-103-96-12/89.

[9] PCI-2 Based User’s Manual, Physical Acoustic Corporation.

(i) (ii) (A)-Corona Discharge; (B)-Floating Particle; (C)-Particle movement

(i) (ii) (iii)

2011 6th International Conference on Industrial and Information Systems, ICIIS 2011, Aug. 16-19, 2011, Sri Lanka

101