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

net/publication/233612387

Detection and Location of PD in MV Cables in Electrically Noisy Industrial

Environments

Conference Paper · June 2011

CITATION READS

1 734

3 authors, including:

Carl Eastham Colin Duncan Smith

IPEC Ltd IPEC Ltd
4 PUBLICATIONS 11 CITATIONS 19 PUBLICATIONS 140 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Colin Duncan Smith on 19 May 2014.

The user has requested enhancement of the downloaded file.

CIRED 21st International Conference on Electricity Distribution Frankfurt, 6-9 June 2011

Paper 1205

DETECTION AND LOCATION OF PD IN MV CABLES IN ELECTRICALLY NOISY

INDUSTRIAL ENVIRONMENTS
Carl EASTHAM Dr Colin SMITH Dr Fa-Chung CHEN
IPEC Ltd – UK IPEC Ltd – UK Wain-Tsiang Enterprise Co., Ltd – Taiwan
[email protected] [email protected] [email protected]

ABSTRACT
PD DETECTION IN MV CABLES
The use of on-line methods to detect and localise partial Partial Discharge in underground cables is very often
discharge in underground cables has become more difficult to accurately detect and locate than in other
increasingly common as coupling and signal processing electrical assets such as associated switchgear. This is
techniques have improved. However, the use of on-line because the source of the activity can be up to several
PD detection has often been limited in environments kilometres from the detecting sensor and high levels of
prone to high levels of conducted and radiated electrical noise can be induced in the cable earth.
noise such as industrial processing plants. The level of
electrical noise interference in these environments can Partial discharge activity in solid high voltage insulation
often exceed those of cable PD by more than an order of induces small high frequency currents in the earth of the
magnitude. The problem of reliable PD detection is electrical system. These impulses travel through the
compounded by the fact that the frequency components of equipment earth to the substation earth. In cables,
the interference are often similar to those of partial detection of these signals relies on the cable sheath being
discharge in cable. isolated from the substation switchgear, such that the PD
signals pass exclusively through the cable earth
In order to be effective, it is very important that a connection. Using a High Frequency Current
permanently installed PD monitoring system does not Transformer (HFCT) they can be detected as they pass
interpret such noise signals as discharge whilst reliably along the cable earth sheath where the HFCT is coupled.
detecting genuine discharge when it occurs.
Figure 1 below shows the coupling point of a HFCT
This paper describes the use of high resolution data sensor around the earth sheath of all three phases of an
acquisition and wavelet signal processing for the 11kV XLPE cable circuit.
automated detection and quantification of PD in
electrically noisy circuits.

A case study is presented from a steel processing plant in

Taiwan where PD was detected in a 33kV XLPE cable
and monitored for a period of several months until a
planned outage could be arranged. The study shows how
the PD wave shape can be analysed to accurately locate
its source despite being superimposed on noise levels that
are many times higher.

The discharge activity before and after a repair to the

cable was carried out will be presented along with the
results of a full analysis of the defect and discharge site.

INTRODUCTION
Partial Discharge (PD) monitoring and trending is
becoming of increasingly important to Utilities and
private networks needing to better understand the
condition of their assets, to target maintenance and to
increase network reliability. With advances in technology
constantly improving the accuracy of PD detection and
classification it is becoming an important tool for asset Figure 1: HFCT coupled to cable sheath
condition assessment in MV and HV cable systems. The HFCT is a passive transducer with a transfer function
of approximately 5.0V/A and a fairly flat frequency

Paper No 1205 1/4

CIRED 21st International Conference on Electricity Distribution Frankfurt, 6-9 June 2011

Paper 1205

response between 50kHz and 20Mhz. PD monitoring channels. This method is effective if a
suitable noise channel can be set up but this is difficult
Signals picked up by the HFCT are routed back to the PD and unreliable. It can be used to discriminate against
monitor via a series of switching multiplexers and low some types of noise interference during switchgear PD
loss coax cables, all shielded so as to minimise any detection but it is of very little value in cable PD
radiated noise from the environment. detection.
High resolution sampling and Signal Processing
NOISE INTERFERENCE
By digitally sampling and recording PD signals from the
Noise interference is induced in HV cables from radiated sensors, powerful processing algorithms can be applied to
and conducted sources. Long cables are particularly the data to identify the PD. In order to do this effectively
prone to radiated interference from RF radio and signals should be sampled at sufficiently high resolution
communication sources. These noise sources are often in and for PD in cables a sampling rate of at least
the form of modulated carrier waves. In industrial 100MSamples/sec is required. If small PD are to be
environments noise is very often introduced into the detected in high noise levels, a good dynamic range is
earthing system by large electrical plant, rotating necessary and the sampling voltage resolution should be
machines, thyristor switching etc. These noise sources are at least 12 bits.
generally in the form of transients often in regular
patterns. Sampling at such a high resolution gives rise to a lot of
sampled data for every power cycle and the challenge is
SIGNAL PROCESSING to reliably identify the very small portions of this data
that include partial discharge. Wavelet analysis is a
There are a number of techniques that can be employed to
powerful tool for identifying PD pulses in very large data
process the raw signal to identify partial discharge.
sets [1] and has been used by IPEC in the development of
Level Triggering a PD recognition and classification algorithm, DeCIFer
In quiet environments the background electrical noise for this purpose. The output of the wavelet analysis
induced in the earth sheath of the cable will be minimal includes transients resulting from PD and noise sources
and any Partial Discharge activity in the system will so a rule based expert system is then used to classify
create pulses that are higher than the noise. Level pulses as either noise or PD and quantify those that are
Triggering above the noise would then be the simplest identified as PD.
way to remove all signals other than the actual PD pulses.
CASE STUDY
In practice this approach is not practical as noise levels
Introduction
are rarely sufficiently low. If used the system will either
only detect very high levels of PD or it is likely to In 2009 IPEC installed a permanent substation
incorrectly classify noise as PD. monitoring system to monitor for Partial Discharge on 54
33kV feeder circuits of the largest steel manufacturing
Frequency Gating plant in Taiwan. The Advanced Substation Monitor
Although near its source discharge pulses are very high (ASM) installed was running the DeCIFer algorithm to
frequency, when detected at the end of an HV cable the eliminate background noise and identify and classify any
frequency components have are typically in the range PD on the monitored assets. The 33kV system consists of
0.5MHz and 5MHz. This makes it possible to use XLPE cables fed from GIS switchgear at the primary
Frequency Gating to eliminate signals that fall outside substation, where the monitoring system and transducers
this bracket and class those within it as PD. were installed.

This approach can be effective in environments where the

Noise
only noise sources are either very low frequency or very
high frequency. However in practice many electrical As with many industrial sites, this steel plant has high
interference noise sources have high magnitude levels in levels of induced electrical noise, frequently saturating
this frequency band and these are not eliminated leading the signals coming from the sensors.
to false positives.
Some of this raw data was extracted as an example, and is
Noise Channel Gating shown in Figure 2 below.
If a sensor can be positioned such that it will detect
background noise but not PD on the plant being
monitored then it can be used as a „noise channel‟ to gate
out in the time domain all corresponding signals on the

Paper No 1205 2/4

CIRED 21st International Conference on Electricity Distribution Frankfurt, 6-9 June 2011

Paper 1205

It can be seen from the difference in scales above that the

Partial Discharge was hidden within the surrounding
prominent noise on the circuit. Also the main frequency
of the noise is similar to that of the PD detected.

By looking at this example it is clear that the alternative

methods of detection described earlier would have failed
to detect the source of discharge on this circuit.

Noise levels like this are often found in heavy industry

plants, where high electrical noise with a wide frequency
band will be present; emphasising the need for
Figure 2: High noise levels detected by HFCT sensor
sophisticated noise rejection techniques when conducting
(Vertical scale ±600mV) electrically sensitive testing or monitoring.

PD Detected Defect Location

Following installation one particular circuit was classified PD detected on this circuit was investigated by IPEC
as having high levels of Partial Discharge activity. Each Data Analysts. It was found the discharge had a regular
asset is automatically categorised as having a Criticality and clear reflection. An example of this reflection taken
rating between 0 and 100. Figure 3 below is the summary from the data analysis web front end is shown in figure 5
bar graph of the Criticality scores across the substation. below.

Figure 3: Feeder circuit Criticalities

This high level of PD detected prompted further

investigation. Figure 2, used to demonstrate the high
noise levels, was taken from the circuit with the high PD.
The DeCIFer algorithm had detected a significant PD that
was much smaller than the surrounding noise and of a Figure 5: PD as displayed on web front end
similar frequency (figure 3).
The wave shape above shows the original PD pulse,
followed by a reflection. By looking at many wave
16.67ms shapes acquired by the ASM system it was seen that this
reflection occurred in every segment.

Normally when a defect is an on-line cable map for PD

source location is conducted. Cable mapping uses Time
Domain Reflectometry (TDR) to locate the source of a
Partial Discharge based on the differences in arrival time
of a PD pulse and that of the reflected PD pulse from the
far end of the cable. As the PD reflections on cable
circuits are often not as strong as seen in this case, a
trigger unit is used at the far end of the cable to inject a
pulse when the PD is detected. In this case however, with
a clear reflection, the cable mapping technique can be
used without the need to install extra equipment in the
customer site.

High resolution data was downloaded from the ASM

monitoring device and processed through on-line cable
mapping software.

20μs The PD was identified as being within a section of cable

0.7% of the overall cable length which was 700m.
Figure 4: PD in relation to surrounding noise Customer records showed a straight joint within this 5m

Paper No 1205 3/4

 CIRED 21st International Conference on Electricity Distribution Frankfurt, 6-9 June 2011

Paper 1205

region and, as joints are commonly the weak points in Upon re-energisation of the circuit no Partial Discharge
MV circuits, this was suspected to be the source of the was detected, closing the loop and confirming that the
discharge. Partial Discharge had been detected and removed. In
industrial networks the cost of an outage can be
Planned replacement work was conducted to remove the significant, it was confirmed by the customer that this
suspect joint. This was shipped to an independent failure could have cost hundreds of thousands of dollars
laboratory in the UK for analysis. in lost production.

The investigation into the condition of the cable joint CONCLUSION

found evidence of significant PD activity. This was seen
on the inner wall of the screen insulating sleeve, adjacent The range of techniques available to effectively detect
to the change in the diameter of the conductor. partial discharge in cables depends on the network
voltage, structure and environment in which it is to
operate. Networks with high electrically induced noise,
such as process plant networks, make detection of PD
significantly more difficult.

Wavelet analysis techniques used in the DeCIFer

algorithm have proven to be effective in these
environments where other techniques may prove
unsuccessful.

REFERENCES
[1] Mingyou Hu, 1998, “A new technique for extracting
partial discharge signals in on-line monitoring with
wavelet analysis”, Proc of 1998 Int Sym on Elec Ins
Figure 6: Evidence of PD found Matls, 677-680.

“Due to poor workmanship during the making of the [2] ERA Technology, Cobham Technical Services, 2009
joint, when the red and black insulating sleeves were “Examination of 33kV Cable Joint”, Report
shrunk over the insulating tubing, excessive use of mastic Number:2009-0614, Project Number: 7F0574101
prevented full insulator shrinkage leaving a small gap at
the position at the change in diameter. At working
voltage this gap allowed discharge to occur which would
have continued until failure” [2]

Paper No 1205 4/4

View publication stats