A Versatile System for PD Diagnostics

L. V. Badicu, W. Koltunowicz, O. Krause

OMICRON Energy Solutions GmbH
Lorenzweg 5, Berlin, Germany
[email protected]

Abstract: The best practice for performing in-service PD

measurements with a versatile system that is applicable to
different high voltage equipment and can be connected to vast
range of sensors is presented in the paper. The successful
examples of application to assess the condition of the paper-oil,
gas and solid insulation of power transformer, gas insulated
switchgear and high voltage cable are described.

I. INTRODUCTION
Partial Discharge (PD) testing and monitoring is nowadays a
recognized technique for the condition based assessment of the
medium (MV) and high voltage (HV) apparatus [1-3]. Since its Figure 1. Frequency range of PD measurements for
massive introduction in the late eighties, several solutions for electrical and acoustic methods
sensors and acquisition techniques have been proposed, applied The best solution is the combination of different methods, as
and lately acknowledged by international standards. It happens recorded phase resolved PD patterns (PRPDs) are very similar
frequently that different electrical assets of one utility are in all type of the measurements and more accurate evaluation
equipped with different sensors and measuring systems of of the results can be drawn.
different technology, purchased from different vendors at An example of the measurements with different techniques
different time. There is a strong need of versatile PD system, and different PD sensors but with the same acquisition unit is
applicable to a vast range of equipment and existing sensors. shown in Fig. 2. The measurements were performed at the MV
This paper describes the best practice for performing PD switchgear and PD signals were acquired with HF current
measurements in service with such a versatile system to assess transformers (HFCT) mounted around the cable shield, with
the condition of different types of insulation. Transient Earth Voltage sensors (TEV1 and TEV2) and with
II. CONVENTIONAL AND UNCONVENTIONAL PD Voltage Detection Sensor (VDS) built in the MV switchgear.
MEASUREMENTS TECHNIQUES The similar PRPD patterns acquired with HFCT, TEV1 TEV2,
and with VDS are presented in Fig. 2.
In CIGRE Study Committee D1 “Materials and Emerging
Test Techniques” many technical brochures (TBs) were
recently published with the recommendations of PD test
procedures to be applied on different HV equipment like cables
(TB 662), power transformers (TB 662 and 676) and GIS (TB
654 and 525). Since conventional PD measuring systems used
in the controlled factory environment are not usually suitable
for on-site application, unconventional PD detection and
measurement methods have been introduced. PD signals can be
detected by methods and systems which utilize different
Figure 2. PRPD patterns measured with HFCT, VDS and TEV
physical characteristics and properties of the PD phenomena, sensors for the same PD defect
so different types of sensors can be used [4, 5].
In general, electrical methods are based on the measurement III. REQUIREMENTS FOR MODERN PD MEASURING SYSTEM
of electrical signals in the radio frequency (RF), high
frequency (HF), very high frequency (VHF) and ultra-high A modern, versatile PD measuring system should fulfil
frequency (UHF) ranges (Fig. 1). Acoustic methods are also different hardware and software requirements. A four-channel
commonly applied for defect´s recognition and localization. data acquisition unit and pre-processing unit, which is based on
It must be emphasized that there is no correlation between FPGA technology are preferable. A minimum 30 MHz high-
the apparent charge in pC values recorded from standardized sensitivity analogue-to-digital stage at each channel guarantee
conventional measurements and the values recorded by acquisition of the signals from most of the coupling capacitors
unconventional measurements [4]. and inductive sensors available in the market. Frequency

978-1-5386-5086-8/18/$31.00 ©2018 IEEE

shifters or down converters will extend the measurement to
VHF and UHF frequency range. The digitized signals from
sensors are sent to real-time FPGA-based system that digitally
filters the pulses within a freely-selectable bandwidth, to
reduce measurement noise. For separation of noise and PD,
different methods can be applied. The most common are time-
frequency classification maps and synchronous three
channel/multi-spectral techniques. In case of three channel
technique, the system involves the synchronous processing of
PD signals from sensors on three phases. Multi-spectral PD Figure 3. PRPD patterns of the PD signal detected at 1.0 Un
measurements are based on pulse spectra correlation by and fc = 250 kHz
applying simultaneously three different band-pass filters
(different centre frequencies and bandwidth). Finally, the During the visual inspection, a mechanical defect related to
the on-load tap changer (OLTC) was identified. The coupling
PRPD diagrams are generated for PD signals from separated
between the motor drive axis and the mechanical transmission
insulation defects and free from noise. The introduction of
on the top cover was found to have been not connected
fully digital systems that includes the synchronous multi-
properly and this has created a malfunction of the OLTC (Fig.
channel PD measurement technique, allows sensitive 4– left). The mobile contacts at tap selector side was blocked in
measurements and accurate separation and classification of PD an intermediate position between two fixed contacts (Fig. 4 –
sources in a wide range of electrical equipment [2,6]. The right).
examples of such application are shown in the next chapters.

IV. CONVENTIONAL PD TESTING OF 16/400 KV GSU

The decision to perform PD measurements at a seven-year-

old 16/400 kV, 422 MVA generator step-up transformer was a
consequence of a fast increase of acetylene and hydrogen gas
concentrations in the transformer oil. In order to compare the
on-site PD test results with those obtained during the factory
tests, it was recommended to supply the transformer from the
turbo generator side and to synchronously measure the PD Figure 4. Faulty coupling between the motor drive axis and the
apparent charge level at the test taps of the three bushings. transmission mechanism of the tap changer
The PD measurements were performed using a synchronous
The manufacturer realigned the coupling between the motor
three-channel detection system. The calibration was performed
drive axis and the rest of the transmission mechanism. The PD
at central frequencies fc = 250 kHz and 2 MHz. The filter
tests were repeated at the same voltage levels and the PRPD
bandwidth was equal in both cases to f = 300 kHz. The lower
patterns are presented in Fig. 5 and Table I. As it can be seen
central frequency is recommended in IEC [7] and IEEE [8]
from Fig. 5, the high amplitude signal is no longer visible.
standards to be applied during routine tests in the factory. The
higher measurement frequency value was selected to get a
better signal-to-noise ratio (SNR) in case of high noise levels.
The PD measurements were performed at 0.8Un and 1.0 Un.
Two PD sources were identified and separated using a multi-
channel synchronous technique – 3PARD [6]. The charge
values recorded during the tests are presented in Table I and
the PRPD patterns of the signal acquired at 1.0 Un (fc = 250
kHz) are shown in Fig. 3. Very high PD levels were recorded
for both PD sources. Both PRPD patterns show a symmetrical
shape in both half voltage waves (Fig. 3 – bottom part), typical Figure 5. PRPD patterns of the signal acquired after
for a loose contact or floating potential type of the defect. transformer repair
Afterwards, a detailed investigation was carried out by the Table I
manufacturer in order to find the root cause of the problems. A RESULTS OF PD MEASUREMENTS AT THE BUSHING TEST TAPS
BEFORE AND AFTER REPAIR
series of tests were performed, such as winding resistance,
short circuit impedance, capacitance and dissipation factor for Voltage Red Yellow Blue
windings and bushings, insulation resistance, frequency Before 0.8Un 1.2C 360nC 1.4C
response analysis and dielectric response analysis. The results repair 1.0Un 1.2C 354nC 1.4nC
of the tests were comparable to those obtained during the After 0.8Un 1.6nC 3.3nC 7.9nC
routine tests. repair 1.0Un 14.7nC 18.9nC 15nC
 V. UHF TESTING OF 400 KV GIS The signal was equal to the background noise level at all other
The 420 kV GIS was equipped with 158 internal UHF partial measuring points/UHF sensors. The location of the PD source
discharge couplers/sensors (PDC). The feeder sensors are in Bay 06 by performing a time-of-flight measurement at three
located in each of 33 bays and bus bar sensors are placed in the UHF sensors: internal feeder sensor and two temporary
main and reserve bus bars. assembled barrier UHF sensors on both sides of the feeder
The scope of the in-service measurements was to check the sensor was recommended.
proper location and sensitivity of the UHF couplers using an
impulse generator and the PD measurement device of the
service company and to perform later measurements at all
installed UHF couplers. The sensitivity check was performed
in the following way. The signal from the impulse generator
was injected in one UHF coupler and was measured at the
neighbouring coupler. Measurements in three different
configurations were considered:
¾ Impulse signal injected between two bus bar sensors
located close by on the same section of the bus-bar.
¾ Impulse signal injected between two sensors located
close by on different bus-bars.
¾ Impulse injected between bus bar sensors and feeder
sensor located close by.
The measurements were made with the amplitude of the
impulse set to 10V - the voltage level determined in the factory
for a specific defect type and PD level according to the CIGRE
requirements [5]. The amplitude of the UHF signals was
measured with a broadband detector working in the frequency
range from 100 MHz to 2 GHz. At each measuring point, the
UHF signal was observed and the characteristic frequencies of
the spectra were checked for possible PD activity.
The following results were obtained:
¾ A 10 V signal injected to any of the bus bar sensors
was detected at the neighboring sensor located on the bas bars.
Figure 6. Results of the sensitivity check performed in-service
¾ A 10 V signal injected to any of the bus bar sensors
was not detected at the feeder sensor located in the bay close
by. The amplitude of the signal was increased to find the
minimum amplitude value being detected in this case. A 40 V
signal fulfilled this requirement.
The results of the sensitivity check agree with the results
obtained during routine tests and during on-site tests. The
distance between the sensors located on the main and reserve
bus bar is a proper one and critical defects generating
equivalent PD signals of 5 pC of apparent charge can be Figure 7. Example of the spectra (left) and PRPD (right)
detected by PD measurements at two neighboring UHF sensors. registered at UHF feeder sensor in Bay 06 of Red Phase
The distance between the bus bar UHF sensors and feeder UHF
sensors is too big to detect critical defects by these two close- VI. COMBINATION OF PD MEASUREMENTS WITH HFCT AND
UHF SENSORS ON 110 KV INTERNAL GAS PRESSURE
by UHF sensors. Each of them can detect the PD defect from
CABLE
its location up to the circuit breaker which is responsible for
the significant attenuation of the UHF signal. The example of The test object was a 110 kV three-phase internal gas
the measurements during the sensitivity check is shown in Fig. pressure cable already in service for 42 years. The cable has a
6. length of 1914 m and connects a GIS with an air-insulated
The PD-like signal was detected only at one UHF sensor, substation. The PD measurements were performed in sequence
located in the Red Phase of Bay 06. The frequency spectra of at both ends of the cable, at GIS and later at overhead line
the signal and PRPD patterns are shown in Fig. 7. No UHF PD (OHL) terminations. The HFCT sensors were installed at the
activity was detected at the neighboring sensors. PRPD GIS termination. At the OHL location, a six-channel
patterns of similar shape were detected at the other measuring synchronous PD measuring system was used and three HFCT
frequency (1092 MHz). To localize the PD source, acoustic and three UHF sensors were connected to the grounding shield
measurements were performed in the compartments of Bay 06 of the cables (Fig. 8). PD measurement with HFCTs is the
on both sides of feeder sensor but no PD activity was detected. most commonly applied method for on-line PD detection at
cables and their accessories [9]. In this particular case, the signals can be removed by gating or filtering techniques. On
HFCTs working at the characteristic frequency range from 50 the other hand, the PD activity in the cable or in the joints
kHz to 14 MHz, were used. For comparison and better data cannot be detected through UHF measurements at the cable
evaluation, the measurements with UHF sensors working in the termination due to the high signal attenuation. Internal PD
frequency range from 100 MHz to 1 GHz were performed in activity was detected by both types of sensors at the OHL
parallel at the OHL terminations. termination. The highest amplitude was measured at phase S of
the cable.
Detecting the PD activity in the UHF range is an important
indication that the PD sources are located inside the cable
termination. The patterns of the signal are surrounded by the
red rectangles in Fig. 9. It can be observed that their shape is
similar to patterns acquired with HFCT sensors. However, due
to the different way of signal coupling, the amplitude of the
pulses cannot be compared.
VII. CONCLUSIONS
Figure 8. Sensors installation at the GIS and OHL cable ¾ The recommendations from the CIGRE TBs were
terminations successfully considered during on-line PD tests.
No outage was required to install the HFCTs at the GIS side ¾ A versatile PD acquisition unit was successfully used for
of the cable. However, for safety reasons, a short outage was PD measurements on power transformers, gas insulated
necessary for the installation of the UHF sensor at the OHL substation and HV cables with connections to different
terminations. No critical internal PD activity was detected at types of PD sensors such as coupling capacitors, UHF or
the GIS end of the cable. inductive HFCT sensors.
¾ The main key to perform appropriate diagnosis of the
At the OHL location, a “wire test” was performed to simulate insulation state is to accurately separate and identify
the external noise source in order to check the sensitivity of the different PD sources from noise and disturbances. To
PD measuring system. A short wire was attached to the achieve this, state-of-the-art synchronous, three channel
conductor of phase R of the cable. It is known that during the evaluation techniques, the 3PARD separation technique
operation, a corona discharge is generated at the sharp tip of was successfully applied.
the wire. As expected, a discharge was detected with the ¾ The synchronous measurement with freely adjustable and
highest amplitude at phase R where the wire was located. A properly selected measuring frequencies allowed distinction
cross coupling to the other phases was also observed. The six- between internal PD and noise and increase the sensitivity
channel PRPD patterns of the signal are presented in Fig. 9. of PD measurements and the accuracy of PD localization.

REFERENCES
[1] G. Stone, “Partial discharge diagnostics and electrical equipment
insulation condition assessment”, IEEE Transaction on Dielectrics and
Electrical Insulation, Vol. 12, No.5, October 2005.
[2] W. Koltunowicz, L.V. Badicu, U. Broniecki, A. Belkov, “Increased
Operation Reliability of HV Apparatus through PD Monitoring”, IEEE
Transactions on Dielectrics and Electrical Insulation, No. 3; June 2016.
[3] L. Testa, A. Cavallini, G.C. Montanari, A. Makovoz, “On-line partial
discharges monitoring of high voltage XLPE / fluid-filled transition
joints”, in proceedings of Electrical Insulation Conference (EIC), pp. 354
– 358, Annapolis, MD, 2011.
Figure 9. Six-channel synchronous PRPD patters acquired [4] CIGRE Technical Brochure 662 “Guidelines for Partial Discharge
with HFCT and UHF sensors Detection Using Conventional (IEC 60270) and Unconventional
Methods”, CIGRE Working Group D1.37, August 2016. 2
It can be observed that the pulses generated by the simulated [5] CIGRE Technical Brochure 654 “UHF Partial Discharge Detection
corona discharge were only detected by the HFCT sensors. System for GIS: Application Guide for Sensitivity Verification”, CIGRE
Working Group D1.25, April 2016.
They are not observed in the UHF frequency range, as the [6] W. Koltunowicz, R. Plath, “Synchronous multi-channel PD
frequency spectrum of the corona discharges does not overlap measurements”, IEEE Transactions on Dielectrics and Electrical
that of the UHF sensor. Insulation Vol. 15, No. 6, December 2008.
[7] IEC 60270 High-voltage test techniques - Partial discharge
Signals from radar or mobile telephones in the frequency measurements, 2015.
range between 300 MHz and 2 GHz are the main interference [8] IEEEC57.113-2010, IEE Recommended Practice for Partial Discharge
for the UHF measurements. These noise signals are present at Measurements in Liquid-Filled Power Transformers and Shunt
discrete frequencies in the frequency domain and can be easily Reactors”.
[9] W. Koltunowicz, L.V. Badicu, R. Humel, U. Broniecki, D. Gebhardt,
distinguished from PD signals. PD single resonance “PD Testing and Monitoring of HV Cable Systems”, in proceedings of
frequencies in the spectra are used for narrow-band UHF 9th Jicable International Conference on Insulated Power Cables,
detection. In case of broad-band measurements, the noise Versailles 2015.