Sie sind auf Seite 1von 10

VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS

FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

IMPROVED PARTIAL DISCHARGE MEASUREMENT BY REAL-TIME IMPULE


WAFEFORM ANALYSIS

M. Krüger, A. Kraetge, K. Rethmeier


OMICRON Energy GmbH
Austria

SUMARY

Partial Discharge (PD) measurement is a worldwide accepted tool for quality control of high voltage
(HV) apparatus. Outside screened laboratories PD signals are very often superposed by noise pulses, a
fact that makes a PD data analysis more difficult for both human experts and software expert systems.
Therefore the handling of disturbances is one of the main tasks when measuring PD.

As differences in the frequency spectrum and therefore differences in the pulse shape do exist, these
can be used to distinguish between multiple PD sources and interference. A real-time data analyzing
algorithm evaluates the spectral response of the pulses for three selected PD filter settings
simultaneously. It allows a clustering of different pulse shape classes, visualized in a two-dimensional
color-coded diagram (3CFRD = 3 Center Frequency Relation Diagram). The multiple clusters in this
diagram stand for multiple PD or noise sources. By selecting single clusters, the phase-resolved PD
pattern of a single PD mechanism can be shown and therefore analyzed by the PD engineer.

As noise pulses are sorted out, the analysis becomes easier and more reliable. Instead of selecting
single PD sources for further investigation, also single sources of interference can be selected in order
to gate them out individually. PRPD patterns without this disturbance can be created. A time-
consuming physical elimination of the disturbance therefore becomes unnecessary.

The new technique doesn’t only enable effective noise elimination but also the separation of signals of
different PD sources. This way PD of an internal void can be separated from other voids or from
surface discharges. Also the different propagation of PD signals from the source to the PD
measurement instrument with the corresponding damping particularly of the high frequencies causes
different frequency spectra. This enables a proper separation, even if the frequency spectra of the PD
impulses are very similar. The efficiency of the described principle is demonstrated with real case
studies of sensitive on-site PD measurements on power transformers under difficult conditions.

KEYWORDS

Partial Discharge measurement, noise suppression, wave form analysis, spectrum analysis, three
center frequency relation diagram, 3CFRD

Michael.krueger@omicron.at
VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS
FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

INTRODUCTION

Partial Discharge (PD) measurement is a worldwide accepted tool for quality control of high voltage
(HV) apparatus [1] [2] [3]. Outside screened laboratories PD signals are very often superposed by
noise pulses, a fact that makes a PD data analysis more difficult for both human experts and software
expert systems [4]. Therefore the handling of disturbances is one of the main tasks when measuring
PD [1]. Additionally, with the ongoing development of permanently-installed PD monitoring systems,
the PD data analysis needs to become more effective to be done automatically. By using pulse-
waveform-analysis it is possible to gain de-noised PD data from separated PD and disturbance sources
in order to make PD measurements more reliable.

PD measurements are often conducted under noisy conditions. The PD signal is superposed by
stochastic noise pulses or even multiple PD sources, which will lead to a complex phase-resolved PD
pattern that is not easy to analyze [5]. For DC PD measurements where the expected PD rate might be
very low, even single disturbance pulses can influence the test result significantly [6].

Conventional frequency filters are not able to eliminate these pulse-shaped disturbances, as their
frequency spectrum is very similar to the spectrum of PD pulses. PD experts and automated computer
expert systems will have difficulties with the superposition of multiple PD faults and noise. Some
well-known evaluation techniques as pulse-sequence-analyses will even fail with non-correlated PD
pulses to be compared [7]. Separating PD from noise and separating multiple PD sources must be the
first step for de-noising PD data. In the future this will become even more important with the rising
number of automated PD monitoring systems installed at critical points of the grid.

MODERN METHODS OF DATA EVALUATION

A new field of evaluation methods is opened by fully synchronous multi-channel PD acquisition


in order to gain more reliable measuring results combined with effective noise suppression. The
OMICRON MPD 600 as a modern type of a fully digital PD measuring system is capable of
performing this kind of PD measurements. A technical overview of the system is given in [ 8] and
in the type test description [9].

Being able to perform synchronous multi-channel PD measurements, the 3-Phase-Amplitude-


Relation-Diagram (3PARD) was introduced [10] as a new powerful analysis tool to distinguish
between different PD sources and noise pulses when measuring 3-phase high voltage equipment
such as power transformers [11], rotating machines [12] and cross-bonded cable systems [13]. As
an enhancement of 3PARD the 3-Center-Frequency-Relation-Diagram (3CFRD) is introduced as
an additional tool for PD data analysis and PD fault separation in real -time on single-phase test
objects. The synchronous consideration of three different frequency parts of the PD spectrum of a
single PD pulse provides information on its discharge nature and indicates its possible PD fault
location due to PD signal propagation and attenuation.

SPECTRAL PROPERTIES OF PULSES

The 3CFRD method requires 3 different PD band-pass filters, measuring every PD event
simultaneously at their predefined centre frequencies. Here the proper selection of these 3 band-pass
positions in the frequency domain is the key to get the optimum benefit from this method. These 3
filters have to be set in a way that the spectral differences of PD pulses and other pulses are at their
maximum.
VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS
FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

Figure 1 gives a graphical comparison of statistically firmed average curves, demonstrating the
spectral differences of pulses of various natures compared to a calibrator reference signal.

0
Calibrator 100pC (dBm)
Average Curve A - Touching Ground Electrode with Metal Tool (dBm)
-20 Average Curve B - Random Noise (dBm)
Average Curve C - Touching HV Electrode with Metal Tool (dBm)
Average Curve D - Switching lights on (dBm)
Amplitude [dBm]

-40 Average Curve E - Main switch on (dBm)


Average Curve F - Negative corona pulse (dBm)
-60

-80

-100

-120

-140
0 5 10 15 20 25 30 Figure 1: Comparison of averaged FFT-curves
Frequency [MHz]

This refers to the fact that, due to the discharge physics, different PD types or noise pulses generate
different but characteristic energy-frequency spectra. To come to an even more intuitive visualization
the off-set compensated amplitude differences of the pulse-spectra, basing on the calibrator reference
curve, are drawn into a diagram in figure 2.

40
Delta Calibrator - Corona Delta Calibrator - Ground
Delta Calibrator - HV Delta Calibrator - Light
30 Delta Calibrator - Main Switch Delta Calibrator - Random Noise

20
Amplitude [dBm]

10

-10

-20

-30

-40
0 5 10 15 20 25 30
Frequency [MHz]
Figure 2: Comparison of normalized differences of
averaged FFT-curves

Based on this comparison the frequency position of the 3 PD band-pass filters can be chosen (see blue
dotted rectangles in figure 2). As the first PD filter should always be selected in a range below 1 MHz
to ensure conformity with IEC 60270, the center frequency of the second filter should be placed at
about 5 MHz, because all curves show significantly different behavior at this spectral position. The
third band-pass filter should be located at about 9 MHz, 11 MHz or 24 MHz, depending on what
setting leads to the maximum separation of clusters in the 3CFRD view. Thus the signal output of
three filters with different center frequencies allows a pulse-waveform-analysis.

3-CENTER-FREQUENCY-RELATION-DIAGRAM (3CFRD)

The known 3PARD method generally requires PD pulse triplets. Usually these are gained from PD
amplitudes measured at three phases of a tested object. It must be noted that the a triplet of three
amplitudes represents one pulse only, meaning the original pulse in the effected phase and it's cross-
coupled appearances in the other phases of the device under test (DUT).These three PD amplitudes of
the same pulse are compared and plotted into one star diagram to form separable clusters [14, 15].
However, in principle this type of diagram can be created by any synchronously gained pulse triplets.
Even for a single phase or a single PD decoupling position, pulse triplets can be acquired by using
three different PD filter settings as described above. As a result, a 3CFRD graph can be generated. The
heart of this diagram is a two-dimensional star diagram.
VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS
FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

Every axis represents one center frequency. The following figures explain the construction of a
3CFRD graph. Figure 3 shows exemplarily the spectra of three measured pulses and the three filters as
set in the MPD software.

Figure 3: FFT of PD pulses, statistical view, 0-


20MHz with 2MHz/Div, -30dBm to -110dBm with
10dBm/Div

Based on the figure above figure 4 shows the spectra of three PD pulses and the three filters marked as
blue bars. The red arrows indicate the absolute charge values of PD pulse 1 (shown in red) at the
discrete filter frequencies.

Figure 4: FFT of three PD pulses, indication of the


measurement frequencies and arrows of the discrete
charge values of pulse 1

The charge values are drawn into the star diagram as shown in figure 5. The lengths of the vectors
represent the measured charge and the axes indicate the respective filter. By geometrically adding the
vectors one single dot is the final representation of the initial triplet.

Figure 5: Representation of pulse number 1 in the


3CFRD

The same procedure is applied to the pulses two and three. Figure 6 shows the blue and green arrows
which have to be inserted in the star diagram, too.
VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS
FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

Figure 6: Arrows of the charge values for all three pulses

After vector addition clearly separated clusters in the 3CFRD view are visible (see figure 7). This has
been expected due to the different spectral appearance of the compared pulses. With this method a
separation of different PD sources and different noise sources is possible by focusing on single
3CFRD clusters. It has to be mentioned that the software of the MPD 600 system provides the
described procedure in real-time.

Figure 7: Representation of pulse number 1 in the 3CFRD

Based on the resulting diagram it is possible to evaluate every cluster separately. Therefore a real-time
back-transformation of a chosen cluster is performed by the system. Now the phase-resolved partial
discharge (PRPD) pattern contains only pulses with the same spectral behavior. Finally this leads to
clear and de-noised PD results. In the following chapter practical application examples of the 3CFRD
method are given

CASE STUDIES

PD measurement on a power cable with a defective joint

This example shows that even three different PD sources can be separated from each other and from
interference by using 3CFRD technique. Figure 8 shows a cable with a defective joint, a needle
electrode to generate corona and an aluminum electrode which is put on top of an epoxy insulator,
generating surface PD, figure 9 the PRPD without 3CFRD filtering. The three PD sources can’t be
seen clearly due to the overlaying and the interference.
VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS
FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

Figure 8: Cable with defective joint, needle


electrode

Figure 9: Unfiltered PRPD

The time domain and corresponding frequency domain signals of two different PD sources are shown
in figure 10.

Figure 10: Time domain and frequency domain signals of two different PD sources

The filtered PRPD signals can be seen in figure 11.


VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS
FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

Figure 11: 3PARD and filtered of corona (green window), PD in the joint (blue window) and surface discharges
(red window)

On-site PD measurement on a transformer

An offline PD measurement has been carried out on-site at an 110kV / 21 kV transformer using a
diesel generator and an according step-up transformer after a DGA (Dissolved Gas Analysis) indicated
a partial discharge problem. Coupling capacitors were used since the bushings had no measuring taps.
Figure 12 shows the measurement situation in a life substation.

Figure 12:Transformer under test and diesel


generator (left)

A synchronous three-phase measurement identified the A-phase as cause of the problem. The 3CFRD
helped in the further evaluation. The possibility to measure at three center frequencies at the same time
delivered additional information about the PD location. Figure 13 shows the PD signals at all three
measured center frequencies and the according charge readings in the right corners, without 3CFRD
filtering (left) and with 3CFRD filtering (right). The reading at 300 kHz is significantly higher than the
same readings at 3 MHz and 10 MHz. Obviously the higher frequencies are strongly attenuated.
VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS
FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

3MHz 3MHz

300kHz 300kHz

10MHz 10MHz

Figure 13: Synchronous PD measurement at three


center frequencies, without 3CFRD filtering (left) and
with 3CFRD filtering (right).

Figure 14: 3CFRD with selection of the fault signals

Due to the fact that the PD amplitude decreased significantly with the measurement frequency, it can
be excluded that the PD fault is nearby the point of measurement.

PD test on MV switchgear

After discharge problems with VT's a PD measurement has been part of the final acceptance testing of
a newly-build medium voltage (MV) switchgear.

The circuit was calibrated according to IEC 60270, figure 16 shows calibrator signals of 100pC.

Figure 16: Calibrator impulses of 100pC


VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS
FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

The ground noise level increased after connecting the HV source to the mains but still remained
acceptable below 10pC (see figure 17).

Figure 17: PRPD at appr. 40 kV – no PD visible

Despite that, the 3CFRD showed two clearly separable clusters indicating a "hidden" PD signal inside
the noise floor. Focusing on the area inside the square (fig 18) delivered a de-noised PRPD pattern
showing obvious PD activity in the measurement circuit (see figure 16).

Figure 18: 3CFRD showing two sources of pulse signals

Figure 16: De-noised PRPD showing Pd activity


below noise level

An investigation was started to find out the cause of this PD activity. A measurement without the
tested switchgear showed, after performing a 3CFRD analysis, the same kind of partial discharges as
the tests before. Thus it could be concluded that the PD was generated by the test circuit, finally the
HV source turned out to be the reason for the PD appearance. Figure 17 shows the de-noised PD
signals measured without the connection to the tested switchgear.
VI WORKSPOT- INTERNATIONAL WORKSHOP ON POWER TRANSFORMERS
FOZ DO IGUAÇÚ, PR – 25 A 28 DE ABRIL DE 2010

Figure 17: De-noised PRPD showing Pd activity in


the measuring circuit without the test object

CONCLUSIONS

The separation of multiple PD sources and noise is mandatory for a clear PD data analysis. By
handling single PD faults, PRPD pattern and pulse-sequence analysis will work best in order to deliver
reliable results. Synchronous multi-channel PD measurements provide new and advanced options of
PD evaluation such as 3PARD and 3CFRD. Therefore a powerful hardware is needed to allow
synchronous PD acquisition in a time frame of a few nanoseconds and real-time data-handling. The
presented digital PD measuring system is capable of performing this kind of measurement, as proven
in many PD tests in the laboratory and especially under noisy on-site conditions world-wide.

REFERENCES:
[1] J. Smit: "Trends in PD-diagnostics for Asset Management of Aging HV Infrastructures" -14th International Symposium on High
Voltage Engineering, Bejing/P.R.China, 25-29th August 2005
[2] D. König and Y. N. Rao: "Partial Discharges in Electrical Power Apparatus", VDE 1993
[3] J. C. Montanari: ―Insulation diagnosis of high voltage apparatus by partial discharge investigation‖, Liu-Yeda Memorial Lecture, in
Proc. IEEE ICPADM, Bali, Indonesia, 2006
[4] E. Gulski and F. H. Kreuger, ―Computer-aided recognition of discharge sources,‖ IEEE Trans. Elect. Insul, vol. 27, no. I, pp. 469—
479, Apr. 1992
[5] CIGRE WG 21.03: ―Recognition of discharges‖, Electra, no. 11, pp. 61–98, 1969.
[6] K. Rethmeier, A. Küchler et al.: "Effective PD Measurements under DC Test Voltage by the Use of Synchronous Measurement
Methods", Nordic Insulation Symposium, NORD-IS 2009, Gothenburg, Sweden
[7] M. Hoof and R. Patsch: ―Pulse-sequence analysis: a new method for investigation the physics of PD-induced ageing,‖ IEEE Proc. Sci.,
Meas. Technol., vol. 142, no. I, pp. 95—101, Jan. 1995.
[8] W. Koltunowicz, R. Plath, "Synchronous Multi-channel PD measurements", IEEE Transactions on Dielectrics and Electrical Insulation,
Vol. 15, Nr. 6, December 2008
[9] R. Holle, R. Plath, K. Schon and W. Lucas: "Type testing a digital PD measuring system according to IEC 60270", ETG conference on
Diagnostic, Kassel, Germany, 2006
[10] K.-D. Plath, R.Plath, H. Emanuel and W. Kalkner: "Synchrone dreiphasige Teilentladungsmessung an Leistungstransformatoren vor
Ort und im Labor", ETG conference on Diagnosic, Berlin, Germany, 2002
[11] S. Schaper, W. Kalkner, R. Plath. "Synchronous multi-terminal on-site PD measurements on power transformers", 14th International
Symposium on High Voltage Engineering, Bejing/P.R.China, 2005
[12] A. Obralic, W. Kalkner et al. ―Verbesserte Zustandsbewertung durch neue Auswerteverfahren bei der Synchronen Mehrstellen-TE-
Messung an Hochspannungsmaschinen,‖ ETG conference on Diagnosic, Kassel, Germany, 2006
[13] S. Sutton, R. Plath and G. Schroeder: "The St. Johns Wood – Elstree Experience – Testing a 20km long 400kV XLPE-insulated Cable
System after installation", 7th International Conference on Insulated Power Cable Jicable07, France, Versailles, 2007
[14] K. Rethmeier, M. Krüger, A. Kraetge, R. Plath, W. Koltunowicz, A. Obralic, W. Kalkner, Experiences in On-site Partial
Discharge Measurements and Prospects for PD Monitoring, CMD Beijing 2008
[15] K. Rethmeier, A. Obralic, A. Kraetge, M. Krüger , W. Kalkner , R. Plath. "Improved Noise Suppression by real-time pulse-waveform
analysis of PD pulses and pulse-shaped disturbances", International Symposium on High Voltage on High Voltage Engineering (ISH),
Cape Town, August 2009
[15] Omicron electronics: "MPD600 - Product brief and specification", Austria, 2007