Recent Developments For On Line Partial

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/253763684

RECENT DEVELOPMENTS FOR ON-LINE


PARTIAL DISCHARGE DETECTION IN CABLES

Article · January 2001

CITATION READS

1 34

4 authors, including:

Toan Phung Trevor Blackburn


UNSW Australia UNSW Australia
206 PUBLICATIONS 1,076 CITATIONS 195 PUBLICATIONS 1,188 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Trevor Blackburn on 03 April 2014.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
RECENT DEVELOPMENTS FOR
ON-LINE PARTIAL DISCHARGE DETECTION IN CABLES

B.T. Phung, Z. Liu, T.R. Blackburn and R.E. James

School of Electrical Engineering and Telecommunications


University of New South Wales, Australia

Abstract: The viability of on-line partial discharge (PD) monitoring on high-voltage power
cables is investigated with particular emphasis on methods to overcome the interference problem.
These include the use of a multi-sensor monitoring system, choice of detection frequency and
signal discrimination based on the waveshapes. To reduce the noise level, measurements are
carried out at high frequencies (above 2MHz). An added advantage is better pulse resolution that
enables more accurate locating of the PD source. Comparison is made between two different types
of sensors used for detecting the PD signals.

up a wide range of signals such as radio, television and


1. OVERVIEW mobile phone transmissions. The detection sensitivity
is thus limited by the interference. For example, the
High-voltage distribution and transmission cables are acceptable limit of PD level can be as low as few tens
an important component of the power network. pC in the case of XLPE cables whereas the
Following the deregulation of the power industry interference in a typical substation can easily exceed
which imposes tighter economic constraints on new 1000pC.
capital expenditure, the current capacity (or ampacity)
in existing cables is usually stretched to satisfy This paper is a continuation of work carried out by the
increasing loading demands. The resulting higher authors [4,5] over the last few years with support from
working temperature together with long-term Integral Energy, a major power utility in New South
operation can lead to insulation deterioration in cables Wales. The emphasis is on the application of various
and their accessories (terminations, joints). Partial techniques to reject interference. Many on-site
electrical discharges (PD) then occur. The impact of measurements in the substations were carried out on
high-energy electrons from these discharges gradually paper/oil and XLPE cables as well as testing in the
cause further damage and eventual breakdown. laboratory. The cable voltage ranges vary from 11kV
up to 132kV. A substantial database of useful
In addition to the so-called intrinsic ageing mentioned measurement results was obtained and some of the
above, PDs can also be caused by the presence of more interesting results are discussed in this paper.
contaminants, defects, protrusions or voids in the
insulation or at the interfaces. For this type of extrinsic 2. MEASUREMENT TECHNIQUES
ageing, the insulation degradation is initially localised
but gradually propagates throughout the cable. The basic configuration of the monitoring system is
shown in Fig.1. One PD sensor is attached to each
Measurement of partial discharges is considered to be cable. The output signal from the sensor is then fed via
one of the most effective methods for monitoring the a long 50Ω coaxial cable to the electronic circuitry for
integrity of the cable insulation [1]. Over the years, a further conditioning and processing.
number of investigations have been carried out on the
possibility of applying this technique for on-line As will be discussed in Section 3, the PD signals are
condition monitoring without isolating the cables from wideband and only their higher frequency components
service, e.g. [2-7]. However, the obstacle is the are monitored. A high-pass filter with a 2MHz cut-off
presence of substantial noise and interference on site. frequency is used to reject interference from AM radio
The main culprits are corona from sharp protrusions, broadcasts (540kHz-1600kHz). Other high-pass filters
switching transients and discharges from other HV with higher cutoff frequencies (50, 100, 150 and
equipment in the substation. Furthermore, cables are 200MHz) are also available for use to further reduce
usually terminated to overhead transmission lines the low-frequency noise.
which in turn act as a huge receiving antenna, picking
The raw signal from the sensor is small (mV) so HF electrical current pulses propagating through the
amplification is necessary. The broadband amplifier earth terminal and thus can be detected with the clip-
provides adjustable voltage gain in 10dB steps. Up to on CT.
40dB gain, the bandwidth is from 0.1MHz to 1GHz.
For higher gain (up to 60dB), the upper cutoff
frequency of the amplifier reduces to 500MHz.
Cable
sheath

To filters, amplifiers,
and CDA system 3 phase
HF-CT system

Fig.1: Basic configuration of the monitoring system.


Fig.2: High-frequency CT sensor.
There are a number of options to analyse the output
An example of HF-CT connection is shown in Fig.2.
signal from the amplifier. It can be fed to digital
An outage is necessary to allow modification of the
storage oscilloscope (DSO) for visual inspection. To
cable earthing termination. Such an outage is also
analyse the PD pulse shape, a DSO with a sampling
useful for carrying out calibration measurements. The
rate of at least 2Gs/s is required. To obtain the pulse
earthing conductor is cut into halves. The ends are
distribution over just one 50Hz cycle (phase-resolved
stripped back, fitted with lugs and then bolted
analysis), the memory needed is 40Mbytes. The DSO
together. This completes the modification. Subsequent
used in this work satisfies the sampling rate
measurements can then be carried out while the cable
requirement but its memory is insufficient.
is energised. The procedure to attach the sensor is as
follows. Firstly, a short length of conductor with the
To enable phase-resolved analysis, one solution is to
sensor fitted on is connected in parallel with the earth
limit the detection bandwidth to 2MHz-30MHz and
conductor. When this is fully secured, the main
use a pulse stretching circuit to process the output
connection is broken by removing the bolt. Thus at all
signal from the RF amplifier. It converts the fast PD
times, continuity between the cable fault screen and
pulse into a rectangular pulse. In other words, the
Earth is maintained. To remove the sensor, the above
magnitude of the original signal is retained but its
procedure is simply reversed. The technique is safe as
waveshape is lost. The required sampling rate is much
no direct connection with HV is required.
reduced and thus the DSO has enough memory to
record the full ac cycle. The data is then transferred to
the PC for further processing by software. This is the
so-called CDA2 system as described in [8].
Alternatively, the pulse stretcher output signal can be
fed to CDA3 data-acquisition system [8] without the
need of the DSO. This Computerised Discharge
Analyser provides a much faster data throughput for
phase-resolved analysis.

Two different types of sensors were used in the


investigation. The high-frequency current transformer
(HF-CT) sensor with a toroidal soft-ferrite core is
shown in Fig.2. It is clamped around the conductor
that connects the metallic fault screen of the cable to
the main Earth. This earthing usually is installed at one Fig.3: Clamping sensor.
end of the cable termination in the substation.
The second type of sensor used – the “clamping”
To understand the detection method, one can simplify sensor – is shown in Fig.3. It is a metallic ring
the whole cable as a single capacitor with the inner clamped around the cable itself. It can be easily
conductor as the HV terminal and the sheath as the attached to the cable and no modification to the cable
earth terminal. Discharges in the cable will result in is required. Here, the principle of signal detection is
somewhat different. It relies on the electromagnetic characteristic is better than that for the clamping
pulse that travels along the screen [2,3]. Near the PD sensor (Fig.5). Therefore, the HF-CT is better suited
source, the discharge pulse is carried over a small part for monitoring long cables where PDs may occur far
of the screen circumference. The pulse distribution away from the sensor. On the other hand, the
becomes uniform along the screen circumference at attenuation of HF components can be used to
some distance away from the source. This distance can advantage in monitoring cable accessories with the
be a few hundred metres for PD pulses in the VHF clamping sensor.
range. Over the section of the cable where the current
distribution in the screen is unbalanced, a magnetic signal attenuation of 11 kV XLPE cable
field outside the cable is produced and the clamping ( using clamping sensor and HFCT )
sensor acts as an inductive coupler.
1
Further reduction of the low-frequency interference 0.8
can be achieved by adding a coupling series capacitor

( ratio )
0.6 HFCT
to the clamping sensor [2]. This is done by breaking clamping
0.4
the continuity of the metal ring with a thin rubber
0.2
washer. Fig.3 shows the capacitively coupled inductor
sensor, the washer is where the two ends of the ring 0
1 10 100 1000
are joined. ( metres )

3. CHARACTERISTICS OF PD SIGNALS
Fig.5: Clamping sensor vs HF-CT.
The rise-times of the PD pulses vary in the order of One possible method for interference rejection is by
nano-seconds to several tens of ns. The pulsewidths examining the characteristics of the received signals. It
can extend from a few ns to some hundreds of ns. is expected that discharges would produce signals that
Thus, the bandwidth of the PD signal can extend up to have different waveshapes from those associated with
1 GHz. Conventional electrical PD detection is limited other types of disturbances. To this end, a test circuit
to the lower frequency range (<1MHz) where the noise was set up in the laboratory. As shown in Fig.6, the
level is relatively higher than that in the upper same PD detection circuit as used for on-line
frequency band. Therefore, the high frequency monitoring consists of the HF-CT, high-pass filter,
detection technique used here (>2MHz) would give a amplifier and digital storage oscilloscope.
better SNR.

signal attenuation
( using clamping sensor )

1
0.8
XLPE
( ratio )

0.6
0.4 paper
0.2
0
1 10 100
( metres ) Fig.6: Test circuit setup in the laboratory.

Fig.4: Attenuation in paper and XLPE cables. To obtain the discharge magnitude in pC as specified
in the IEC-60270 or AS1018 Standards, the
Frequency response measurement of the HF-CT sensor conventional PD measurement circuit was also
shows an upper limit of ~300MHz. The upper cut-off included where the signal was picked up through a HV
of the clamping sensor is much higher, at least coupling capacitor Cb. Note that the conventional PD
>500MHz. However, the high frequencies in both detector has a detection bandwidth of 10kHz-300kHz.
paper/oil and extruded cables are quickly attenuated.
This is shown in Fig.4 for the clamping sensor. The Different types of faults were artificially created in the
attenuation is much higher in paper/oil cable. It should cable samples and a collection of PD characteristics in
be noted that the allowable PD level in paper cables is the form of time-domain waveforms and frequency
also much higher. Also, since the HF-CT can measure spectra was obtained. In Fig.7 are shown some of the
the lower frequency components, its signal attenuation typical examples: (a) external corona, (b) external
surface discharges, (c) discharges in cavity in cable the cables. To demonstrate the effect, a detachable
and (d) discharges at the cable termination. discharge source was attached to a live 66kV Yellow-
phase bus bar. Fig.9 shows the disturbances were
picked up by the other two phases. Note the longer
time-base for displaying the pulses in relation to the
50Hz cycle and the relative magnitudes of the signals.
(a)

(b)

(a)

(c)

(d)
0.2µs/hor. div 10MHz/hor. div
50.0mV/vert.div 100mV/vert. div (b)
Fig.7: Typical PD signal characteristics. Fig.8: Waveshape and spectrum of signal recorded (a)
on-site and (b) in laboratory.
There are numerous factors that can influence the
signal characteristics. For example even for a simple Simultaneous recording of signals from sensors
void defect, the PD signal is dependent on the void attached to the three phases can facilitate the
shape, its size, electrical stress, dielectric composition, recognition of external interferences. Fig.10 is an
ageing stage, etc. Also if the PD source is remote, the example of different signal patterns recorded during a
received signal will be distorted since the attenuation site visit. In Fig.10(a), the signals occur on all three
varies with frequency. Therefore, it would be channels exactly at the same time. Furthermore, the
practically very difficult to identify the type of defects waveshapes and their relative magnitudes are very
based on the signal waveshapes. This type of pattern similar. Thus, it can be reasonably concluded that this
recognition is mainly helpful in discriminating against is an external radiative interference, equally picked up
the interference. Even in this case, it will fail if the by the three cables. Fig.10(b) also corresponds to
interference is a real external PD signal from other another interference. Note the difference in the
connecting HV components. waveforms between these two interference signals and
also against those shown in Fig.7.
4. ON-SITE MEASUREMENT RESULTS

Interference rejection based on signal waveshapes: 20mV


Red 5ms
To a certain extent, discharge recognition can be
achieved by matching signals recorded on-line with
those stored in the database. An example is shown in 20mV
Yellow 5ms
Fig.8. The on-line signal corresponds to surface
discharges on the HV bushing at the sealing end of a
132kV XLPE cable. The location of the discharge
source was confirmed using an ultrasonic detector. 20mV
Blue 5ms
Rejection of radiative interferences:
0.5V
Radiative interferences can be picked up by the cable 50Hz 5ms
sensors, either directly through the air or indirectly via
the overhead transmission lines or busbars attached to Fig.9: On-line injection of PD source on Y Phase.
magnitudes (coupling factor) is dependent on the
particular physical setting of the site.

Rejection using phase-position windowing:

Another method based on 'phase position windowing'


can be included to improve the interference rejection.
Its basis is that PDs normally occur on certain parts of
(a) the 50Hz ac cycle (see Fig.9) and there is a 120o shift
between the three phase supplies. Therefore any
signals outside the ‘normal windows’ must be either
noise or pick-ups from the other phases and thus
should be rejected.

Rejection of external discharges from same phase:

Rejection of external PDs originating from the same


(b) phase is a difficult problem. One possible solution is
the use of a pair of clamping sensors - positioned near
the cable termination - to detect the direction of signal
propagation. This, however, will not be able to
distinguish between external PDs and PDs at the
termination.

To detect PDs at the terminations or joints, one can


increase the lower cut-off frequency of the clamping
(c)
sensor so that it is responsive to local signals.

Location of discharge sources – A case study:

Sensor 1
Black Sensor 2
Blue
Joint

(d)
Fig.10: Rejection of external interference.
Sensor 3
Red
Fig.10(c) shows a pulse detected by the Yellow-phase
sensor (top trace) and at that instant nothing occurs on
the other two sensors. This suggests an internal
discharge in the Yellow-phase cable. Fig.10(d) is
another example of an internal discharge pulse on the Fig.11: Location of the three clamping sensors.
Blue-phase cable (middle trace).
In this case study, the subject is a three-core XLPE
Thus, a possible solution for interference rejection is 11kV cable which is approximately 1km long and has
the use of a multi-channel monitoring system that can a joint in the middle. Separate source tests gave an
record simultaneously signals from the three phases. unacceptable high reading of 3000-8000pC at 2Uo
Rejection is achieved by comparison of the recorded (12.7kV). In order to determine whether the cable joint
signals. A pulse is a valid internal PD if it occurs on is sound, three capacitively coupled clamping sensors
one channel only. However, if the cable discharges were positioned near the joint as shown in Fig.11.
near the termination, the signal can be picked up by Sensors 1 and 2 are 2.05m and 0.65m, respectively, to
other channels but the coupling is somewhat reduced. the left of the joint whereas sensor 3 is 2.15m to the
In other words, the pulse on the source cable is much right. The signals recorded simultaneously from each
larger than the pick-up signals on the neighbouring sensor pair are shown in Fig.12. The sampling rate of
cables. Thus the largest signal should be kept and the the DSO was set at 10Gs/s. It can be seen that the PD
others rejected. The criterion for comparison of the pulsewidths are <5ns and the clamping sensors are
able to capture these very fast PD signals. The time
delay between each pair of sensor signals can be 5. CONCLUSIONS
measured with accuracy in the order of 1ns. Based on
the propagation velocity of 0.2m/ns, the PD location The following are main points covered in the paper:
can be resolved to within 20cm.
¾ The PD signatures (time-domain waveforms and
spectra) of different types of defects on cable
samples were obtained from laboratory tests. Such
a database can facilitate PD recognition.
¾ Examples of interference and PD waveforms as
recorded on-site at a number of substations were
presented. A case of matching between laboratory
and on-line signal waveforms was given.
(a)
¾ Comparisons were made between attenuation
characteristics of paper and XLPE cables and also
between the clamping sensor and the HF-CT.
¾ The clamping sensor enables measurement at even
higher frequencies and thus it gives a much better
time resolution for discharge location. This was
demonstrated with a case study where a fault was
successfully located at the joint of an 11kV cable.

(b) 6. ACKNOWLEDGMENTS

The continued supports from Integral Energy, in


particular P. Taylor and J. Hickey, are gratefully
acknowledged.

7. REFERENCES

[1] J.C. Chan, P. Duffy, L.J. Hiivala and J. Wasik,


(c) "Partial discharge - Part VIII: PD testing of solid
Fig.12: Time delay between sensor signals. dielectric cable", IEEE Elec. Insul. Mag., 7, 1991,
pp.9-20.
Fig.12(a) indicates that the Blue sensor receives the [2] P.A.A.F. Wouters, van der Laan P.C.T., Hetzel E.
signal 10.3ns before the Red sensor. Thus the PD and Steennis E.F., "New on-line PD measurement
source is closer to the Blue sensor. Also, the source technique for polymer insulated cables and
must be in between the Blue and Red sensors because accessories", Proc. 8th ISH, Paper 63.08, 1993.
otherwise the delay should be 14ns which is the time [3] N.H. Ahmed and N.N. Srinivas, “On-line partial
for the signal to travel from the Blue to the Red sensor. discharge detection in cables”, IEEE-DEI Trans.,
Hence the solution narrows down to one point which is Vol.5, No.2, Apr.1998, pp.181-188.
18cm to the left of the centre of the joint. Using [4] B.T. Phung, Z. Liu, T.R. Blackburn and R.E.
similar arguments, Fig.12(b) simply confirms that the James, “On-line PD measurement on high voltage
source is to the right of both the Black and Blue power cables”, 11th Int. Symp. On High-Voltage
sensors. Fig.12(c) indicates the source is 15cm to the Engineering (ISH), London, Aug.23-27, 1999.
left of the joint centre which agrees well with the result [5] Z. Liu, B.T. Phung, T.R. Blackburn and R.E.
from Fig.12(a). Note that the joint itself is at least James, "The propagation of partial discharge
60cm long and thus this location of the PD source is pulses in a HV cable", Proc. AUPEC'99, Northern
still part of the joint. Territory University, Sep.26-29, 1999, pp.287-292.
[6] E.F. Steennis et al, “Learning from high-voltage
The joint was subsequently replaced with a short XLPE cable system testing and monitoring”,
length of spare cable and two joints. The cable passed CIGRE Paper 21-203, 2000.
the PD test carried out by the utility after the repair. [7] W. Wei, Z. Sen, C. Bin and L.En-heng, “The study
The removed faulty joint was opened up later for of on-line PD detector in power cable”, Proc. 6th
visual examination and the fault was confirmed. It ICPADM, Jun.21-26, 2000, pp.177-180.
happened when the semicon layer was stripped back [8] B.T. Phung, "Computer-based partial discharge
and accidental overcutting had caused damage to the detection and characterisation", PhD thesis,
insulation underneath. University of NSW, 1997.

View publication stats

You might also like