Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Procedia
Available Engineering
online 00 (2017) 000–000
at www.sciencedirect.com
Procedia Engineering 00 (2017) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ScienceDirect
Procedia Engineering 202 (2017) 189–201

4th International Colloquium "Transformer Research and Asset Management”

4th International Colloquium "Transformer Research and Asset Management”
Improvement of ultrasonic method for testing of power transformers
Improvement of ultrasonic method for testing of power transformers
Danijel Brezakaa, Dalibor Filipović-Grčićbb
Danijel
Končar Electrical Brezak
Engineering , Dalibor
Institute, Filipović-Grčić
Transformer Department, High Voltage Laboratory,
Končar Electrical Engineering Institute,
Fallerovo Transformer
šetalište Department,
22, 10002 Zagreb, High Voltage Laboratory,
Croatia
1 Fallerovo šetalište 22, 210002
dbrezak@koncar-institut.hr, Zagreb, Croatia
dfilipovic@koncar-institut.hr
1
dbrezak@koncar-institut.hr, 2dfilipovic@koncar-institut.hr

Abstract
Abstract
Ultrasonic method of detecting PD is based on the fact that the electrical energy of the PD transforms in a mechanical energy,
an Ultrasonic method of
ultrasonic acoustic detecting
wave PD is based
that spreads on the
through the fact that the electrical
transformer energy
to the tank wall.ofThese
the PD transformswaves
mechanical in a mechanical energy,
are detected with
an ultrasonic acoustic
piezzo-electric sensors wave that spreads into
and transformed through the transformer
electrical signals which to the
aretank wall. These
processed with amechanical waves arecalled
special instrument detected with
Acoustic
piezzo-electric
Emission System sensors
(AES).and transformed into electrical signals which are processed with a special instrument called Acoustic
Emission System
Since not only (AES).
defects inside the transformer can produce acoustic waves, during analysis there is a problem distinguishing the
Since
actual not of
origin onlythedefects inside
detected the transformer
acoustic waves. can produce acoustic waves, during analysis there is a problem distinguishing the
actual origin ofadditional
Therefore, the detected acoustic
sensors canwaves.
be added to the system which monitor some parameters of the transformer such as
Therefore,of additional
temperature sensors
the tank, load, OLTC canoperations,
be addedfan to and
the pump
systemoperations,
which monitor some
PD level parameters
during laboratoryoftesting
the transformer suchwith
(in accordance as
temperature
[1]) etc. of the tank, load, OLTC operations, fan and pump operations, PD level during laboratory testing (in accordance with
[1])These
etc. parameters can simplify analysis using a time correlation between detected hits (hit is a designation for event when
These
sensor parameters
detects can wave)
an acoustic simplify
andanalysis using
in-service a time correlation
parameters between detected hits (hit is a designation for event when
of the transformer.
sensor detects
Another an acoustic
problem wave) and
with location of in-service
the source parameters
in 3D system of the transformer.
is the lack of knowledge of the zero time of the hit.
Anothera problem
Using with location
high frequency of transformer
current the source in(Rogowski
3D system is the and
coil) lack the
of knowledge of the
fact that the zero time
electric of has
signal the hit.
a high propagation
Usingzero
velocity, a high
timefrequency current
can be added to thetransformer (Rogowski
system providing highercoil) and the fact that the electric signal has a high propagation
accuracy.
velocity, zero time can be added to the system providing higher accuracy.
©
© 2017
2017 The
The Authors.
Authors. Published
Published by
by Elsevier
Elsevier Ltd.
Ltd.
© 2017 The
Peer-review Authors. Published by
Peer-review under responsibility of the
under responsibility of Elsevier Ltd. committee
the organizing
organizing committee of of ICTRAM
ICTRAM 2017.2017.
Peer-review under responsibility of the organizing committee of ICTRAM 2017.
Keywords: Ultrasonic method; Power transformers; Partial discharge, Transformer diagnostics, Location of PD, Rogowski coil, Parametrics
Keywords: Ultrasonic method; Power transformers; Partial discharge, Transformer diagnostics, Location of PD, Rogowski coil, Parametrics

1. Introduction
1. Introduction
Ultrasonic method of detecting PD is based on the fact that the electrical energy of the PD transforms in a
Ultrasonicenergy,
mechanical methodanofultrasonic
detecting acoustic
PD is based
waveonthat
the spreads
fact thatthrough
the electrical energy of tothethePDtank
the transformer transforms in a
wall. These
mechanical waves
energy,are
an detected
ultrasonicwith
acoustic wave that spreads through the transformer to the tank wall.
piezzo-electric sensors and transformed in to electrical signals which are These
mechanical waves are detected with piezzo-electric sensors and transformed in to electrical signals which are
1877-7058 © 2017 The Authors. Published by Elsevier Ltd.
Peer-review
1877-7058 ©under
2017responsibility
The Authors. of the organizing
Published committee
by Elsevier Ltd. of ICTRAM 2017.
Peer-review under responsibility of the organizing committee of ICTRAM 2017.

1877-7058 © 2017 The Authors. Published by Elsevier Ltd.

Peer-review under responsibility of the organizing committee of ICTRAM 2017.
10.1016/j.proeng.2017.09.707
190 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201
2 Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000

processed with a special instrument called Acoustic Emission System (AES) From time difference of wave detection
on different sensors, a possible location of the source can be estimated. Since the method detects ultrasonic acoustic
waves spreading through the tank it can detect other transformer deficiencies such as loose contacts and local
overheating of oil (T>200 °C) which gives this method a unique diagnostic status [2].

Since not only defects inside the transformer can produce acoustic waves, during analysis there is a problem
distinguishing the actual origin of the detected acoustic waves. Therefore, engineers should have as much data of the
transformer state as possible. Monitoring systems can provide that data but it is not always easy to implement that
data from another system in the acoustic system. Therefore, additional sensors can be added to the system which
monitor some parameters of the transformer such as temperature of the tank, load, OLTC operations, fan and pump
operations, PD level during laboratory testing etc.
Using the parametric inputs, data can be directly loaded in to the system avoiding an error of incorrect time
synchronization between the time of AES and monitoring system of the transformer in substation. These parameters
can simplify analysis using a time correlation between detected hits (hit is a designation for event when sensor
detects an acoustic wave) and in-service parameters of the transformer. Based on time correlation, hits that can
originate from mechanical sources such as OLTC operation can be excluded from the analysis. Of course, these
mechanical noises can be distinguished from the actual source from the characteristics of the recorded hits, but that
is not always the case. AES is equipped with analog inputs 0-10 V, so current and temperature transducers were
implemented in the circuit [3].
Another problem with location of the source in 3D system is the lack of knowledge of the zero time of the hit.
The system needs 4 sensors to detect the hit to find the source locations based on different arrival times to each
sensor. Using a high frequency current transformer (Rogowski coil) and the fact that the electric signal has a high
propagation velocity, zero time can be added to the system providing higher accuracy. Another possibility is to
discriminate the signals measured by acoustic method if there is no correlation with the signal on the Rogowski coil.
Laboratory tests were performed to investigate the advantage of using the method with zero-time as a parameter.

2. Implementation of new instrumentation parts

2.1. Parametric inputs

Current and temperature sensors are connected to the system using an analog input box (AIB) which is used to
transfer the 4-20 mA signal from the sensors to an accepted signal level on the parametric inputs of the AES, 0-10
V. Besides the signal conditioning it is used to power the sensors using a loop-power principle.
AIB consists of 8 input channels and an internal 499 Ω resistor for transforming 4-20 mA signals to a
0-10V signal [3].
Loop-powered sensors are used because there is no need for extra wiring that is usually used for external sensor
powering. On the same wiring, sensors are powered and the measured signal is sent.

a b

Fig. 1. (a) AIB and HSM-I temperature transducer; (b) Loop-powered sensor
 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201 191
Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000 3

2.1.1 Temperature as a parameter

Temperature parameter sensor consists of a magnetic Pt100 probe and a transducer which converts temperature
change in to a 4-20 mA signal. Output from the transducer is connected to a parametrical input on the AIB.

Fig. 2. Temperature sensor – connection principle

Because the AES does not measure actual temperature but the 0-10V signal on one of its parametrical inputs
implementation of the sensor was performed and the function T=f(U) was determined. The dependency between
temperature and voltage was approximated with a linear function. The function is later inputted in the system
enabling the voltage to be directly transformed to °C. Implementation was performed by using two of a kind Pt100
probe: one connected to the AES and the other to a referent measuring system that gives the temperature in °C. Both
probes were mounted on a metal plate which was heated.

Fig. 3. Implementation of the temperature sensor

Linear function T=f(U) was obtained by simultaneous measurement of the voltage given by the AES and the
temperature given by the referent measuring system sampling by discreet time intervals:

T = 12 ,412 ⋅ U − 23,852 (1)

In Fig. 4 and 5 and Table 1 it is shown that linearity is satisfying as well as the approximation with regards to the
real value (error. ±0,5°C).
192 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201
4 Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000

Table 1. Real and calculated temperatures.

U (V) Tcalculated (°C) Treal (°C)
3,8795 24,3 24,4
4,3352 30,0 30,2
5,1891 40,6 40,2
5,9203 49,6 49,8
6,8273 60,9 60,8
8,4136 80,6 80,0
9,1870 90,2 90,4

100
T / °C
90 Referent MS

AES
80

70

60

50

40

30

20

10

0
0 1000 2000 3000 4000 5000 6000 t/s 7000

Fig. 4. Signals obtained by RMS and AES

100
T / °C y = 12,412x - 23,852
90

80

70

60

50

40

30

20

10 T=f(U)

lin[T=f(U)]
0
3 4 5 6 7 8 9 U/V 10

Fig. 5. Temperature vs. voltage – linear approximation

 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201 193
Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000 5

2.1.2. Current as a parameter

Loop-powered split-core current sensors 4-20 mA, AT 5 B420L (0-5A) and AT 50 B420L (0-50A) are used as a
current parametric sensors. Use of AT 50 B420L is intended for monitoring the OLTC position changing and the
cooling pumps/fans turning on/off. AT 5 B420L will be used for monitoring burden of the transformer placing it
around the bushings current transformer secondary lead in the control cabinet.

Fig. 6. Temperature sensor – connection principle

Current sensors implementation was performed and the linear function I=f(U) was determined. Implementation
of the 50 A sensor was performed using a current source, current transformer and an ammeter as showed on Fig. 7.

a b

RT – regulation transformer – 0-250 V

MT – matching transformer – 400/16V
CS – current source 0–250 A
CT – current transformer – 150/5 A

Fig. 7. (a) Current sensor - implementation; (b) Implementation connection

Table 2. Implementation of 50 A sensor results

I (A) UAES (V)
7,02 3,2185
9,96 3,5859
15,00 4,4861
19,80 5,2583
24,90 6,0669
30,00 6,7746
34,80 7,5278
39,00 8,2105
45,00 9,0826
48,00 9,6938

Fig. 8 shows the dependence I=f(U) and the function obtained using the implementation results lin[I=f(U)].
6194 Danijel Brezak andDanijel
DaliborBrezak et al. / Procedia
Filipović-Grčić Engineering
/ Procedia 202 (2017)
Engineering 189–201
00 (2017) 000–000

50
I/A
45
I=f(U)

40 lin[I=f(U)]

35

30

25

20

15

10

5
y = 6,3799x - 13,423

0
0 1 2 3 4 5 6 7 8 9 U / V 10

Fig. 8. Current vs. voltage – linear approximation for 50 A sensor

Implementation of the 5 A sensor was performed using a rheostat, regulation transformer and an ammeter. As
showed on Fig. 9.

a b

RT – regulation transformer – 0-250 V

R – rheostat 3,3 Ω, 10 A

Fig. 9. (a) Current sensor - implementation; (b) Implementation connection

Table 3. Implementation of 5 A sensor results

I (A) UAES (V)
0,5 2,7380
1 3,4702
1,5 4,3006
2 5,0836
2,5 5,9372
3 6,6698
3,5 7,4795
4 8,2776
4,5 9,1031

Fig. 10 shows the dependence I=f(U) and the function obtained using the implementation results
lin[I=f(U)].
 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201 195
Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000 7

I/A
4,5 I=f(U)

lin[I=f(U)]
4

3,5

2,5

1,5

0,5 y = 0,6272x - 1,1976

0
0 1 2 3 4 5 6 7 8 9 U / V 10

Fig. 10. Current vs. voltage – linear approximation for 5 A sensor

2.1.3. Test voltage and PD level as a parameter

Since AES is equipped with analog inputs 0-10 V it is possible to import signals from PD detectors that have
voltage outputs proportional to apparent charge and applied voltage.

AES voltage measurement was compared to a referent measuring system, and results are given in Table 4.

Table 4. Voltage measurement comparison

U (V) UAES (V) URMS (V)
1 0,994 1,000
2 2,044 2,040
3 3,031 3,027
4 3,996 3,992
5 4,990 4,990
6 6,058 6,055
7 7,050 7,047
8 8,035 8,030
9 9,025 9,015

The difference between the measured voltages is not greater than 10 mV (1 % of the full range of the AES).

Comparison of AES input voltage with PD detector reading was made with different PD levels and different
sensitivity ranges, x1 from 0 to 1000pC, and x10 from 0 to 10000 pC. Charges were produced with PD calibrator
according to [4]. Readings of the instrument and voltage was compared for both cases and a linear approximation of
the function was determined.

PD ( x1) = 235 ,3 ⋅ U + 8,1 [2]

PD (x10 ) = 2353 ⋅ U + 81 [3]
196 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201
8 Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000

2.2. Rogowski coil

Rogowski coil is a high frequency current sensor used as an electrical trigger for detecting PD. Usually, it is
placed around the neutral-ground or ground connection of the transformer. It detects high frequency current pulses
which are produced by PD. Using the Rogowski coil, this method transforms from all-acoustic to acoustic system
with an electrical PD trigger [5]. Other than electrical detection of discharges it is used for triggering and calculating
zero time from discharge until the first acoustic sensor detects the acoustic wave. This should enhance location
possibility and accuracy.

Using the Rogowski coil we can also distinguish electrical discharges from mechanical occurrences (rain, hail,
OLTC operations etc.) or other noises which can be excluded from the location analysis.

Implementation of Rogowski coil was performed in High-voltage laboratory of Končar - Electrical Engineering
Institute. A PD source was placed in a tank of a transformer model (1.680x1.200x0.715 m).

Fig. 11. Implementation of Rogowski coil

Several test were made, some using the Rogowski coil (3D T0 Location algorithm), and some without the
Rogowski coil (3D Location algorithm) in order to compare the results and to determine influence on localization of
PD.

Rogowski coil, besides the filtering of non-electric emissions, enhances the accuracy calculating the zero-time
from the occurrence of the discharge until the first sensor detects the hit wave. Because of this feature it eliminates
some reflections of the wave because the reflected wave does not derive from an electric source and arrives to the
sensors too late to be considered as the same event. Reflections of the wave are often a problem, especially when
testing small transformers with short distances between inner parts as it was the case with the used transformer
model.

The difference in localization of PD source between the test with and without the Rogowski coil is showed on
Fig. 12. It is obvious how the Rogowski coil influences the obtained locations. Using a Rogowski coil fever
 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201 197
Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000 9

locations and only one cluster was obtained. In case with no Rogowski coil 3 clusters and more locations (influence
of wave reflection) were obtained. In further analysis only the cluster with the highest density of locations was taken
in consideration. Clusters are groups of locations with the highest density in a square of maximum 10% of the
transformer dimensions.

a b
Cluster z: Cluster z:
X axis – 0,50 – 0,62 m X axis – 0,63 – 0,64 m
Y axis – 0,63 – 0,65 m Y axis – 0,64 – 0,66 m
Z axis – 0,02 – 0,06 m Z axis – 0,09 – 0,18 m

Fig. 12. (a) Rogowski coil test results; (b) Test results without the Rogowski coil

Actual position of the PD source was:

X axis: 0,54 m
Y axis: 0,56 m
Z axis: 0,04 m
The comparison between the tests and the errors are given in Table 5.

Table 5. Comparison results

Test Axis Cluster Actual Error (m)
center (m) position (m)
X 0,56 0,54 0,02
3D T0 location Y 0,64 0,56 0,08
Z 0,04 0,04 0,00
X 0,54 0,54 0,09
3D location Y 0,56 0,56 0,09
Z 0,04 0,04 0,10

From the test results, Fig.12 and Table 5 it is obvious which method is more accurate. Taking a mean value of all
axis error it is obvious that using a Rogowski coil the obtained results have higher accuracy. The mean value of
error using a Rogowski coil is ±0,03 m and in case of test without it ±0,09 m so Rogowski coil improves accuracy
by approximately 3 times.
198 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201
10 Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000

3. On-site tests

3.1. Parametric inputs tests

Parametric inputs were tested on a 300 MVA transformer which showed higher amount of gases dissolved in oil.

Current sensors were place around the fan supply lead to monitor fans switching on and off (Fig. 13 a), around
the supply lead of OLTC to monitor the change of taps (Fig. 13 b) and around the secondary lead of a bushings CT
(1500 /2A) (Fig. 14 a) to monitor the load. Temperature sensor was mounted close to the top of the tank to monitor
temperature changes (Fig. 14 b).

a b

Fig. 13. (a) Fan supply monitoring; (b) OLTC supply monitoring

a b

Fig. 14. (a) Load monitoring; (b) Tank temperature monitoring

Direct loading of the monitored parameters was successfully achieved. On Fig. 15 all the above monitored
parameters are showed. As it can be observed that there is a correlation of the tank temperature and the load of the
transformer. Influence of OLTC tap change on the acquired acoustic data is showed on Fig. 16. A correlation is
observed between the acoustic activity and the tap change. Since the transformer was monitored during the night,
without the parametric input the only way to know when there was a change of tap is to take that data form the
monitoring system in a substation or from the substation personnel. In that case there is a problem of time
synchronization. Manual inputting of parameters, if they are not loaded directly, is only possible for one parameter
at the time which makes analysis complicated and time consuming.
 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201 199
Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000 11

Fig. 15. Parametrics monitored during the test

Fig. 16. shows the acoustic activity before and after the analysis when the acoustic activity during the tap change
was excluded from the data. If the tap change time is unknown, obtained location during the tap changing can easily
be mistaken for a possible fault inside the transformer.

Fig. 16. Tap change effect on acquired acoustic signal

200 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201
12 Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000

Fig. 17 shows that higher acoustic activity was recorded at the beginning of the test which is in correlation with
the temperature rise of oil. Since the acoustic activity was recorded after putting the transformer in on-load mode it
can be assumed that this activity is a result of material dilatation and friction inside the transformer as it heats up.
Acoustic activity is in correlation with the fastest rise of tank temperature. After it was heated, even at higher load
there was practically no acoustic activity recorded. This is an example how monitoring of load and tank temperature
can help in analysis and better understanding of processes inside the transformer.

Fig. 17. Correlation between sensor activity, load and tank temperature

3.2. Rogovski coil tests

One transformer was tested using the Rogowski coil.

During testing of transformer 250 MVA, there was no acoustic emission recorded during the tests. Obtained
results showed a significant activity of the Rogowski coil in all tests. The Rogowski coil activity was in correlation
with the measured PD level which was higher than 10000 pC during the whole time. No locations were obtained.
Several test were made and even thou there was electrical activity in all tests, only one test showed some acoustic
activity on one channel, channel 14. On Fig. 18 the correlation between acoustic activity on sensor 14 and Rogowski
coil (connected as sensor 23) can be observed.
Since there was some acoustic activity on only one channel, in case where the Rogowski coil signal was higher
(more than 80dB) than in other tests, we can assume that there was a PD source inside the tank highly damped and
impossible to be detected by the sensors in other cases.
Since no locations were obtained during the tests with Rogowski coil we can’t analyse the influence on
localization. The only thing that can be concluded is that the Rogowski coil signal was in correlation with the PD
level change over the time.
 Danijel Brezak et al. / Procedia Engineering 202 (2017) 189–201 201
Danijel Brezak and Dalibor Filipović-Grčić / Procedia Engineering 00 (2017) 000–000 13

23

14

Fig. 18. Correlation between Rogowski coil and channel 14

4. Conclusion

Usage of new instrumentation parts should enhance the method, simplify the analysis and improve the precision
of discharge localization.

The dependency between the measured parameters (current and temperature) and AES input voltage was
determined and approximated with a linear function.

Direct loading of the PD level from the PD detector was achieved and the dependency between PD level and
voltage was determined.

Input of the PD level and test voltage level into AES will simplify the analysis and the use of Rogowski coil will
improve the precision of localization and exclude any collected data which does not derive from an electric source
from location analysis.

Parametric sensors used on-site will simplify the analysis and enhance the interpretation of data. Parametric
values gained from the AES using the linear function transformations should not be taken as exact values but as an
approximate value. More important are their changes over the time and correlation of that change with the detected
acoustic emission.

References

[1] IEC 60076-3, „Power transformers - Part 3: Insulation levels, dielectric tests and external clearances in air“, Edition 2.0, 2000.
[2] Arturo Nunez, Samuel J. Ternowchek, Ronnie K. Miller, Barry Ward, „Detection, location, assessment and classification of faults in
energized power transformers using acoustic emission“, CIGRÉ Transformer Colloquium, 2003
[3] Physical Acoustic Corporation, „AewinTM Software Users Manual, Rev. 1.90“, 2004
[4] IEC 60270, „High-voltage test techniques - Partial discharge measurements“, Edition 3.0, 2000.
[5] IEEE PC57.127/D10.0, „IEEE Guide for the Detection and Location of Acoustic Emission from Partial Discharges in Oil-Immersed Power
Transformers and Reactors“, December 29, 2006