0% found this document useful (0 votes)
15 views4 pages

Sensitivity Check For UHF PD Detection On Power Transformers

Uploaded by

trang
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
15 views4 pages

Sensitivity Check For UHF PD Detection On Power Transformers

Uploaded by

trang
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Conference Record of the 2004 E E E International Symposium on Electrical Insulation, Indianapolis, IN USA, 19-22 September 2004

Sensitivity check for UHF PD detection on power transformers

Sander Meijer, Edward Gulski, Johan J. Smit and Huub F. Reiinders


Delft University of Technology Smit Transformatoren B.V.
High-voltage Technology & Management P.O. Box 9107
P.O. Box 5031,2600 GA Delft, The Netherlands 6500 HJ Nijmegen, The Netherlands
[email protected] [email protected]

Abstract: It is known that failure due to insulation defects are


often preceded by partial discharge activity. Over the last years, UHF DETECTION TECHNIQUE
on-line diagnostic methods for power transformers have been
investigated. In this paper the sensitivity of one of these methods,
the UHF PD detection technique, has been applied on five The UHF detection method is based on the displacement
merent high- and medium voltage transformers with power current that occnrs during a partial discharge. Even in oil
ratings in the range of 14 to 500 MVA. From these experiments, insulation, the discharge takes place in a gas-bubble under
the sensitivity of the UHF technique to detect partial discharges high pressure and results in a very steep current pulse [2].
in power transformers has been proven. Based on these results, P High-frequency signals can be excited in this way ranging up
seusitiniy check similar to GIs is proposed in this paper. to 3 GHz. These electromagnetic waves propagate through the
transformer and can be picked up by appropriate sensors.
INTRODUCTION
However, a power transformer is a totally encapsulated metal
Power and distribution transformers form important box and no electromagnetic signals will radiate out of the
components in the electricity grid. Failure of such important transformer. For that reason, electromagnetic-leaks have to be
components can give rise to high costs (such as costs for repair introduced at certain places and dielectric windows are used
and cost for non-delivery). Therefore means to determine the for this purpose. A simple solution is to make such.dielectric
risk of failure of transformers are of importance. It is h o w n windows in the always-present hatch covers, see figure 1.
that failure due to insulation defects are oflen preceded by
partial discharge activity. In order to assess the risk of Figure 2 shows the typical UHF partial discharge detection
insulation defects, the following procedure has to be followed: circuit. The UHF coupler is fitted against the dielectric
1) Sensitive detection ofpartial discharges window. The signals are amplified and then detected by a
2) Location of the insulation defect spectrum analyzer (SA) or digital storage oscilloscope @SO).
3) ldentification of the type of insulation defects All data are transferred to a computer for storage, post-
4) Criticality Analysis processing and analysis.

For detection of partial discharges, the ultra-high frequency


(UHF) PD detection technique has shown to be very sensitive
[I, 21. However, due to the complex propagation path between
the PD source and the UHF sensors, frequency dependent
attenuation and resonances, a calibration according to the IEC
60270 standard is not possible for the U H F detection
technique at this moment. In this paper a sensitivity check is
described which makes it possible to estimate the PD level.

Figure 2 - UHF PD detection circuit consis& of


1) a dielectric window with external UHF coupler
2) pre-amplifier
Fieure I - 3) Diclcctnc u i n d o u located m a hatch cover and hJ a 3) specmm analyzer or oscilloscope
U H F sensor &d pre-amplifier mounted against the window. 4) computer

0-7S03-8447-4/04/$20.00@2004IEEE. 5s

Authorized licensed use limited to: Hitachi Energy. Downloaded on May 31,2024 at 05:40:10 UTC from IEEE Xplore. Restrictions apply.
SENSITIVITY CHECK In the next step a simulation of the coupled frequency
spectrum from the spark gap to UHF-sensor 2 was performed.
An artificial pulse was injected in UHF-sensor 1. Changing
Description of the sensitivity check the voltage level to 14.2 V resulted in a comparable frequency
spectrum in the frequency range between 200 MHz up to 700
A procedure as described for the sensitivity check for GIS is
MHz, see the gray line in figure 4b. Thus the relation of 1.5
as follows [3]: V/pC can be used in further experiments to check the
1) In a laboratory setup use a PD-source with known sensitivity of the UHF-sensor 2 regarding a PD-source close to
magnitude (PC) and locate it close to sensor I .
UHF-sensor 1. If more than two sensors are present, this
Measure the frequency spectrum of the coupled
procedure is repeated for all sensors.
signal to sensor 2.
2) Use an artificial pulse generator to inject pulses at
sensor 1 and change the amplitude until a frequency
spectrum is detected at sensor 2, which is comparable
to the frequency spechum detected at sensor 2
originating from the PD source (see step 1).

From this test, the relation between calibrated PD magnitude


and artificial pulse is known for a certain propagation path.
This can be used on a power transformer to test the sensitivity
of the UHF coupler arrangement by the step 3.

3) Sensitivity check of the UHF coupler arrangement on ".


D a5 90 135 18D 225 270 315 360
the power transformer by injecting the pulse shape Ph. .no,.
found under point 2 on each of the placed couplers
.20,00 ,..................'.. . . . . . ,.... ............... , ................ ,
and detect at the other couplers. If the injected signal
is detected, this means that between the UHF-sensors
the sensitivity is high enough to detect partial
discharge activity with a certain magnitude in pC as
defined under step 1.

0 250 5M 750 IWO 1250 1-


Rewmney WHzI
Figure 4 - Measuring results coming from a spark plug and
simulated by artificial pulses:
a) phase-resolved PD pattern as measured using an IEC
60270 detection circuit
Figure 3 - Laboratory setup used to perform a sensitivity check for b) frequency specua as obtained on coupler 2 from the
UHF PDetection on power transformers. spark plug (black line) and from injected pulses with an
amplitude of 14.2V (gray line).

Practical example of the sensitivity check APPLICATION TO POWER TRANSFORMER


To simulate a discharging source inside the transformer, a Although the results of the sensitivity check previously
spark plug has been mounted on the top cover of the described have been obtained on a 14 MVA transformer in the
transformer tank, see figure 3. The spark plug is located next laboratory, the effect of the oil alone on the propagation of
to UHF-sensor 1 and gives a discharge level of approximately high-frequency signals is approximately similar for all types
10-15 pC when measured using a detection circuit according of transformer oil. However; the influence of the interior of
to the IEC 60270 see figure 4a. Simultaneously, the frequency the transformer cannot be neglected but it is impossible to
spectrum was detected at UHF-sensor 2, which is positioned at perform the first two steps of the sensitivitycheck for each
a distance of 2 meters of the sparkplug and at sensor 1, see the transformer separately. Therefore, 45V pulses (representing
black frequency spectrum in figure 4b. about 30 pC) have been injected in UHF-sensors distributed
on the tank of a 150 MVA transformer, see figure 5.

59

Authorized licensed use limited to: Hitachi Energy. Downloaded on May 31,2024 at 05:40:10 UTC from IEEE Xplore. Restrictions apply.
Figure 6 shows the detected frequency spectra at the three The sensitivity matrix shown in table 1 can he checked by
UHF-sensors W2, W3 and W4 (blue graphs) located on the cross-correlating the results: the value measured at sensor W2
LV-side, when 45V pulses are injected in UHF-sensor WI, from injected pulses in sensor WI should approximate the
located on the high-voltage side. The background noise at same value measured at sensor WI when pulses are injected in
UHF-sensor is also displayed for comparison reasons (purple sensor W2. This offers an additional verification of the
graphs). From the figures the following observations can be sensitivity check. As can be seen in the table 1, there is only
made: IO-20% deviation between the obtained results and this
I) the injected pulses can be detected at all three confirms the sensitivity check.
UHF-sensors located on the LV-side;
2) the amplitude of the spectrum is highest at UHF-
sensor W3. This can be explained by the fact that WfJW2
coupler WI and W3 are opposite from each other
and a direct coupling path is present;
3) Two frequency ranges with good signal-to-noise
ratio can be concluded from 750-850 MHz and
from 1100-1750 MHz

-1000 4
0 500 1003 1500 2000 290 30W
lMHzl

W12W3
Figure 5 - Location of the UHF couplers on a 150 MVA power ~ .. .......... " , .. ......... ,

transformer:
a) on the low-voltage side
b) on the high-voltage side

As can he seen in figure 6, the frequency spectra contain a lot


of different resonance peaks, which makes it difficult to
compare the obtained results. Therefore, the frequency spectra
are represented by a single value, which represents the energy 0 500 to03 1500 2000 2500 3003
content of the frequency spectra. This single value, or key- IMH4
value AREA,is calculated in the following way [4]: Wl->W4
1) Calculate the signal to noise ratio for the
obtained frequency spectra, resulting in an array
with N entries which are the processed
amplitudes SX;
2) All N processed amplitudes Si and Si+, of the . .
measured frequency spectrum are averaged,
summed and divided by the number of data
points and is converted in [a]:

,
0 500 lOW 1500 , 2000 2500 3003
IMHzl
Figure 6 - Examples of frequency spectra from injected pulses
in UHF sensor W1 detected at W2, W3 and W4. The purple line
The AREA of the measured frequency spectra are shown in is the background noise spectrum; the blue line is the signal
table 1. spectrum.

60

Authorized licensed use limited to: Hitachi Energy. Downloaded on May 31,2024 at 05:40:10 UTC from IEEE Xplore. Restrictions apply.
Table 1: Key-value AREA in (dB] as being processed Table 2 shows the processed key-value AREA for the
from frequency spectra measured from the sensitivity measured frequency spectra at all four UHF-sensors during
155 kV voltage application. A clear increase at couplers is
detected and is higher than the results fiom the sensitivity
check as shown in tahle 1. This means that the PD activity is
higher than 30 pC, which was unfortunately not confirmed by
a measurement because the noise level was above 150 pC.

Table 2: Key-value AREA in [dB] as being processed


from frequency spectra measured during voltage

Detection on UHF coupler

Area[&] 23.2 9.2 18.2 9.6


POWER TRANSFORMER UNDER TEST
During the factory test, the voltage was increased up to 155kV
in several steps. At a certain voltage level, a significant
increase in the frequency spectrum at all four UHF-sensors
CONCLUSIONS

_- Based on the results described in this paper the following can


7- I be concluded
1) The UHF technique is sensitive enough to detect
partial discharges of 25-30 pC;
2) A sensitivity check can he used to define the
sensitivity of the UHF-sensors installed on a
power transformer;
3) Noise-free phase-resolved PD pattems can be
obtained using the narrow band UHF technique;
.... .:. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..:
Future work will focus on investigation of the propagation
i .

reo
:m7 0 1 ’ . . . . . .
i m . I.,
:
.........................
. . . . : .;. ,;
: . . .;. . . . . . . . . . . . . . .j

. .
. . . . properties of the UHF-signals in the transformer to estimate
the location of the PD-source and to estimate more precisely
the magnitude of the discharge level.
0
0 CO 120 180 210 POD 360
.LI.

Figure 7 - Measuring results obtained during voltage application REFERENCES


of 155 k V
a) Frequency specl”: a significant increase can be [I] Rutgers, W.R., Fu, Y.H., “UHF PD-Detectdion in a
observed in the kqueucy range iiom 4M)-18M) MHz, Power Transformer”, Conf.Proc. IO” ISH, Montreal,
coming from an intemal PD source. Canada, 1997
b) Phase-resolved PD pattem measured at a center
frequency of 15M) MHz and a bandwidth of 3 M H z . [2] Judd, M.D., Cleary, G.P., Meijer, S., “Testing UHF
partial discharge detection on a laboratory based
power transformer”, Conf.Proc. 13” ISH, Delft,
Netherlands, 2003
was detected. Figure 7a shows as an example, only the result CIGRE TF 15/33.03.05, “PD Detection System for
measured at sensor W4 (low voltage side). Similar increases [3]
GIS: Sensitivity Verification for the UHF Method
were obtained at UHF sensors WI, W2 and W3. In particular and the Acoustic Method”, Electra No.183, April
the amplitude in the frequency range from 400 to 1800 M H z is 1999
increased. This can he explained by the presence of an intemal Meijer, S. Parfiul discharge Diugnosis of High-
[4]
PD source. Figure 7b shows a phase-resolved pattem, which Voltage Gus-Insulated Systems, Optima Grafische
was measured using the narrow-band UHF method. For this, Communicatie, ISBN 90-77017-23-2,200l
the spectrum analyzer was used as tunable filter, tuned at a
center frequency of 1500 M H z using a bandwidth of 3 MHz.

61

Authorized licensed use limited to: Hitachi Energy. Downloaded on May 31,2024 at 05:40:10 UTC from IEEE Xplore. Restrictions apply.

You might also like