Acoustic and electrical methods combined for

localizing partial discharge in power transformer

Stefan M. Hoek, Omicron Austria

Stefan Hoek is Product Manager for partial discharge measurement system and
joined OMICRON, Austria in 2008. He studied electrical engineering at the Uni-
versity Stuttgart (Germany) and worked there as research assistant with focus
on partial discharge detection and localization in GIS with help of measurements
in the UHF range. Stefan Hoek is member of VDE ETG and CIGRE working group
B3.24 and published several papers.

1 S12
Acoustic and electrical methods combined for
localizing partial discharge in power transformer

ACOUSTIC AND ELECTRICAL METHODS COMBINED FOR

LOCALIZING PARTIAL DISCHARGE IN POWER TRANSFORMER
1* 1 1 2 2
Stefan M. Hoek , Alexander Kraetge , Michael Krger , Stephan Krber , Ole Kessler ,
2 3
Rene Hummel , Ulrike Broniecki
1 2 3
OMICRON electronics, Austria, OMICRON energy solution, Berlin, Berlin University of
Technology, Germany,
*Email: [email protected]

Abstract: The reliability of electrical systems depends on the quality and availability of
the power apparatus. Power transformers are important nodes in the electrical power
grid. Common reasons for breakdown are problems in the insulation system. Detecting
partial discharges in the insulation system of a power transformer at an early stage
reduces the risk of total breakdown. One method to detect partial discharges is acoustic
measurement. With this technique detection and localization of partial discharge is
possible by placing acoustic sensors on the surface of the transformer tank. The low level
of electrical interferences from outside the measurement setup constitutes one of the
strengths of this method. A further advantage is the ability to identify the position of the
partial discharge source, which is needed to estimate the risk and to enable a fast and
effective repair.

1 INTRODUCTION More than one propagation path from the

PD source to a sensor is possible (direct oil,
Partial discharge (PD) measurements on reflection, steel path).
transformers are an accepted tool of quality control
The speed of sound depends on the
in factory and on site. Different PD measurement
propagation path (crossed medium),
techniques are using a variety of different physical
frequency and temperature.
characteristics of the PD phenomenon, e.g. electric
discharge currents (acc. to IEC 60270 [1]), gas Depending on the location of the source and
formation (DGA - dissolved gas analysis), the inner structure of the transformer, a
electromagnetic (UHF Measurement) or acoustic proper measurement of the direct oil path
radiation. The main benefits of an acoustic PD signal may be impossible due to high signal
measurement are the possibility to detect PD attenuation.
without a shut-down of the transformer, and the
ability of localizing a PD source with an accuracy of Taking into account the individual
a few centimeters to a few decimeters. In case of measurement setup and the inner structure
expected PD, an acoustic PD measurement can of the device under test (DUT), a cautious
help to verify it, while the location is important interpretation of the measurement results by
information for further risk assessment and repair an experienced person is essential.
planning. Speed and damping of the acoustic waves are
dependent on the transfer medium, frequency
range and temperature [3], [4]. For example, the
2 THE PROPAGATION BEHAVIOR OF propagation speed decreases during the heat-up
ACOUSTIC PD SIGNALS IN period of an transformer by approximately 15%,
TRANSFORMERS from about 1400 m/s at 20 C to 1200 m/s at 80 C.
This has to be considered to avoid miss-
The acoustic effect of PD inside a transformer is localizations since the signal speed directly
typically measured by piezo-electrical sensors in influences the calculated results.
the frequency range of tens of kHz up to hundreds The propagation path is often complex. Multiple
of kHz [2]. Using the different arrival times of the propagation paths of the emitted sound wave are
acoustic PD signal at multiple sensors, algorithms possible, as shown in Figure 1. Depending on
can compute the location of the PD source. sensor and PD location, multiple acoustic wave
The complex physical processes involved in sound components of the same PD event are potentially
propagation and the large structural variety of detected by one sensor and overlay the direct oil
different transformer may present a challenge signal as illustrated in Figure 2. The acoustic wave
during the measurement. The following parameters can be reflected by the tank wall, core, winding, flux
have to be considered: shields and other components.
The PD source position and the inner
structure of the transformer mainly influence
the propagation path of the acoustical
waves.

2 S12
Acoustic and electrical methods combined for
localizing partial discharge in power transformer

3 LOCALIZATION OF PD

Different algorithms can be used to perform a time-

based localization of PD. The input information
used by the algorithms is the time of arrival of the
signals propagating on direct oil paths to multiple
sensors. The exact time of arrival has to be
determined by evaluating the measured signal. A
criterion for the starting point can be found e.g. by
investigation of energy steps [7] or by threshold
criteria [8].
The relative arrival times at different sensor
positions lead to time differences (t1,2). These time
lags are the only available data in all-acoustic
measurements, when the data acquisition is
triggered by the acoustic signal at one of the
sensors. If the time delay between occurrence of a
PD and the arrival of the associated acoustic wave
is available, the absolute propagation times (t1, t2)
Figure 1: Possible propagation paths in the test from source to sensor can be used for localization.
object [5] Both principles are shown in Figure 3.
Components of the reflected wave will arrive at the Amplitude
t1,2
sensor position later than the signal travelling a
direct path. Furthermore, the acoustic wave can t1
couple into the transformer wall and propagate t2
through the steel of the tank. Due to the higher
propagation speed in steel of about 3.000 - Sensor S1
5.000 m/s [2], the so-called steel wave signal can
reach the sensor earlier than the waves following
the direct oil path. This effect complicates the Sensor S2
automated determination of the starting point of the Occurrence of PD time
direct oil signal.
The measurable direct oil signal at the sensor Figure 3: Absolute and relative times in a two-
position depends on the intensity of the causative sensor-setup
PD event [6] and on the damping on the
propagation path. Therefore, the attenuation by The exact timing of the emission of the PD signal
core, winding, transformer board, flux shielding etc. can be estimated e.g. by an electrical PD
should be as low as possible. For that reason, the measurement according IEC 60270 or a
search for sensor positions that ensure good signal measurement in the ultra-high frequency (UHF)
quality is essential during measurement procedure. range. In the latter case, sensors within the
The knowledge about the inner structure of the transformer walls can be used to receive the high
transformer is helpful for a good positioning and frequency electromagnetic wave that is emitted
repositioning of the sensors. during PD [9]. A measurement setup is shown in
Figure 4 [10].

Figure 2: Acoustic PD signal components

reflecting the propagation paths
Figure 4: Installed UHF probe

3 S12
Acoustic and electrical methods combined for
localizing partial discharge in power transformer

The distance between sensor and source is displayed for three or more acoustic sensors like in
calculated using the available absolute or relative Figure 15.
propagation times and an estimated average
propagation speed. With the determined distances In an all-acoustic measurement environment the
and the sensor positions a geometrical localization approach is in principle identical. In this case a
of the PD source can be performed in several steps fourth sensor delivers the necessary information to
(Figure 5) estimate a point representing the origin of the PD
signals.
The arrival time (t1) at a single sensor in relation to
the PD occurrence leads to a surface in the shape The depicted method is based on a direct
propagation path for the acoustic wave from source
of a sphere around the sensor position on which
the PD source is supposedly located. The radius r to sensor. As described above, the transformer
cannot be considered as an empty box and the
depends on the absolute propagation time (t1) and
propagation speed is highly dependent on the
the propagation speed (Figure 6 left)
signals travel path. For that reason, the model is
In all-acoustic measurements the data of a single always a simplification of the real setup inside the
sensor (without a relation to another sensor signal) tank. Thus, also an inaccurate localization of the
does not contain information leading to a source position is possible.
triangulation. In this case the data of two acoustical
sensors - the relative time t1,2 deliver a distance
difference (d1,2) and therefore a hyperbolic sphere
(Figure 6 right).

Figure 7: Source localization with three sensors

using absolute times

To ensure reliable measurement results, a workflow

Figure 5: Principle of acoustic localization [5] is proposed that is based on an iterative relocation
The position of the source can be specified with the of the sensors with the intention to find positions
information from a higher number of sensors. For with a minimal and undisturbed path between
this purpose several of the described geometrical sensor and source.
shapes are intersected. The absolute propagation
time of the signal at a second sensor leads to a
second sphere, the resulting intersection shape is 4 CASE STUDIES OF SUCCESSFUL
circular. In a further step the absolute coordinates ACOUSTIC PD FAULT LOCALIZATIONS
of the source can be estimated by intersecting the ON TRANSFORMERS
circulars of three sensors.
4.1 PD localization on a 16 MVA transformer
using electrical PD signals as trigger
source
The described investigation has been performed on
a 150 kV/20 kV three-phase power transformer
(Yyn0) with a nominal rating of 16 MVA. Figure 8
shows the electrically measured PD pattern of
phases V and W indicating a PD problem nearby
phase W. Consequently the acoustical sensors
were placed on the tank walls close around the
suspicious winding. After analyzing the first results,
the sensor positions have been optimized. Two
Figure 6: Spatial information from one absolute sensors had to be placed on top of the transformer
time (t1) and from one relative time (t1,2) housing because of limited accessibility due to the
assembled coolers. Figure 10 shows the tested
This procedure is shown in Figure 7. The figure transformer, modeled in the software of the location
shows the spheres around three acoustic sensors system and the finally used sensor positions.
(black, yellow and blue). The resulting intersection
circulars of the spheres are shown as blue rings.
The estimated point of the acoustic source is

4 S12
Acoustic and electrical methods combined for
localizing partial discharge in power transformer

Acoustic source
Figure 8: Electrical PD pattern recorded at the X= 2.65 m; Y= 1.10 m; Z= 1.43 m
bushing taps of phases V and W

An example of the acoustically detected PD signals

is shown in Figure 9. The signal shape on all
sensors indicated propagation paths with just minor
attenuation. The yellow sensor detects the signal
first, shows no indication of an acoustical steel path
and the highest signal amplitude. These facts
indicate this sensor as being the one closest to the Figure 10: Computer model of the tested transformer,
PD defect. The black sensor seems to have a short sensor positions and located PD source
steel path while the sensors blue and red show an
increasing signal shape which usually can be
considered as an indication for signal damping due
to crossed solid insulation. This overall behavior
has been found as being reproducible and so a
localization could be carried out. The signal speed Source
used for the calculations was set to 1400 m/s and
the start points of the signals were determined
using the energy criterion and adjustment by hand
in some cases. The resulting circles indicating the
position of the PD source are automatically given
by the location system PDL 650 and are to be seen Figure 11: Finding: Bare Wire close to the main
in Figure 10. A brown dot marks the PD source at
insulation found as root cause of PD activity
the intersection of the circles.
4.2 PD location on a 100 MVA transformer by
using UHF signals as trigger source
The following investigation was performed on a
230 kV/20 kV three-phase power transformer
(100 MVA) in a manufacturers test field. The UHF
and conventional electrical PD measurement
according IEC 60270 [1] revealed the quite
comparable PD patterns shown in Figure 12. Both
signals were potentially qualified for triggering the
acoustical measurement system.

UHF

Phase U

Figure 9: Acoustical PD signals detected during Figure 12: Simultaneously measured PD pattern
the location procedure of the UHF system (top) and conventional PD
detector at the bushing tap
The subsequent inspection of the transformer
verified a close match of the defect (Figure 11) and
the estimated PD source.

5 S12
Acoustic and electrical methods combined for
localizing partial discharge in power transformer

The UHF measurement was performed by using surface of the tab winding between winding and tap
the narrowband acquisition method with a changer as shown in Figure 15.
bandwidth of 1.5 MHz. The center frequency of
159.5 MHz was chosen based on the evaluation of A.4
the frequency sweep shown in Figure 13. The
highest signal energy of the UHF spectrum is
shown as the upper line representing the pulses -
including the PD - while the sinusoidal or Sources
continuous wave (CW) disturbances are shown in
the lower curve. The measurement frequency can A.1
now be tuned into an area showing a large
distance between both lines. This method turned
out as being very effective to optimize the signal- U
to-noise ratio during a UHF PD measurement.

A.2 & Resulting

A.3 intersection
circulars

TOP view

Figure 15: Estimated PD source location and

some intersection circulars

The subsequent inspection of the transformer

Figure 13: UHF Frequency sweep of the spectrum verified a close match of the defect and the
inside of the transformer and the measurement estimated PD source.
filter used shown as grey bar at 159.5 MHz 4.3 PD location on a 500 MVA transformer in
Figure 14 shows the 3D computer model of the the field (all acoustic)
transformer with the final sensor positions and the A 500 MVA unit showed significant increase of
electric-acoustic test setup. hydrogen and methane a few months after
installation. Figure 16 shows the transformer and
its modelling in the localization software.
W
V
U

A.4

A.3
A.1

A.2
Figure 16: The unit under test and its
representation in the localization software.
Sensors are visible on the top.

In this case the internal structures shown are the

tap changer compartments. During the initial PD
location electrical triggering could not be utilized.
Furthermore a detection of acoustic signals on the
tank walls was not possible, presumably due to the
Figure 14: Test setup (UHF, el. and acoustic PD) magnetic shunts which are known to have a
significant sound damping characteristic. Thus, the
After optimizing the UHF signal quality the pulses acoustic sensors had to be attached on top of the
detected by the UHF 620 system were used to transformer. The results measured with the first
trigger the acoustic PD localization. The evaluated sensor arrangement indicated a PD location at the
time delay led to PD source positions on the outer middle phase, near the tap changer or the 220 kV
bushing (see Figure 17, left). Therefore the

6 S12
Acoustic and electrical methods combined for
localizing partial discharge in power transformer

sensors were re-positioned closer to the expected 6 REFERENCES

PD spot, what led to improved signal quality and
higher sound levels. The resulting surfaces [1] IEC 60270 (2000) High-voltage test
representing the mathematical solutions of the techniques - Partial discharge measurements
localizing equations can be seen on the right-hand International Electrotechnical Commission,
side of Figure 17 and confirm the initial expectation Publication 60270, 2000
about the PD location. The internal inspection of [2] C57.127 (2007) IEEE Guide for the Detection
the tab changer connection via a man-hole and Location of Acoustic Emissions from
discovered defects of different insulation elements Partial Discharges in Oil-Immersed Power
which have been replaced. The transformer has Transformers and Reactors, The Institute of
been put back into service again and is in Electrical and Electronics Engineers, Inc. New
operation now without any indication of remaining York, USA, 2007
PD problems. [3] Howells, E. ; Norton, E.T.: Parameters
affecting the velocity of sound in transformer
oil IEEE Transactions on Power Apparatus
and Systems (1984)
[4] Lundgaard, L.E. Partial Discharge Part XIV:
Acoustic Partial Discharge Detection
Practical Application, IEEE El. Insulation
Magazine Sep 1992 Vol.8,No.5
[5] Hoek, S.M.; Kstner, B; Kraetge, A; Intro-
duction to performing acoustic partial
discharge measurements and localizations
using the PDL 650 OMICRON electronics,
Application Note ANP_12004_ENG,
Figure 17: Results with initial sensor placement www.omicron.at (2013)
(left) and after sensor-repositioning (right). PD
source indicated to be close to the tap changer or [6] Bengtsson, T.; Leijon, M.; Ming, L. Acoustic
Frequencies Emitted by Partial Discharges in
MV bushing of the middle phase. Oil, 8th International Symposium on High
Voltage Engineering, 1993, p.113-116

5 CONCLUSIONS [7] Groe, C.U.; Reinhardt, H.-W.; (1999) Schall-

emissionsquellen automatisch lokalisieren
MP Materialprfung, Jahrg. 41, pp.342, Carl
This paper describes the basic idea of time Hanser Verlag, Mnchen, Germany, 1999
based acoustical localization of PD faults in power
transformers and similar equipment. The PD [8] Wagenaars, P.; Wouters, P.A.A.F.;
signals are captured using three or more van der Wielen, P.C.J.M.; Steennis, E. F.
Algorithms for Arrival Time Estimation of
piezoelectric acoustic sensors, magnetically Partial Discharge Pulses in Cable Systems
mounted on the tank at different locations. For IEEE Vancouver, Canada, 2008
localizing the source, the time delays between the
recorded acoustic signals or between an electrical [9] Coenen, S.; Tenbohlen, S.; Markalous, S.M.;
signal and the acoustic signals are used to get Strehl, T. (2008) Sensitivity of UHF PD
information about the propagation of the acoustic Measurements in Power Transformers IEEE
Trans. on Dielectrics and Electrical Insulation,
signal inside the transformer tank and the distances Vol. 15, No. 6, pp. 1553-1558
between signal source and sensors. The received
sensor signals are processed to obtain the [10] Kraetge, A.; Rethmeier, K.; Hoek, S.M.;
difference between the signal arrival times at each Krueger, M.; Modern de-noising strategies for
sensor. PD measurements on transformers under
challenging on-site conditions, CIGRE SC
In addition to the acoustical measurement, a A2&D1 Colloquium 2011, Kyoto, Japan,
parallel electrical PD measurement can be used to Paper # PS1-O-14
obtain a trigger signal. This can be essential for the
success of the localization. Alternative to the
electrical measurement through the bushing taps or
with external coupling capacitors, unconventional
measurement techniques, e.g. in the UHF range,
can be used to gain a trigger source for an acoustic
measurement. Case studies of successful PD
localization of a significant PD source have been
shown. The procedure and successes of an
acoustical measurement with electrical triggering
has been shown. The results are analyzed and
visualized by using a 3D model.

7 S12