21, rue d’Artois, F-75008 PARIS B1-205 CIGRE 2016

http : //www.cigre.org

Learning from on-line monitoring of medium voltage power cables – with PD

and fault location
B.C. VAN MAANEN, DNV GL, The Netherlands
D. HARMSEN, Alliander, The Netherlands
P. BLEEKER, Locamation, The Netherlands
T. BROERSMA, Enexis, The Netherlands

SUMMARY

For several years, monitoring of MV power cables with a two-sensor based, partial discharge
monitoring system (2BPM) has been proven to be an effective instrument for diagnosing MV
underground power cables. Partial discharges (PD’s) from weak spots can be detected and located,
both continuously and on-line. Based on the same technology, the DNO’s Alliander and Enexis,
together with DNV GL and Locamation improved the system such that it can now also detect and
locate faults (breakdowns) in MV cables typically locating these to within 1 % of the cable length.
This is a unique new possibility to guard MV cables. Moreover, these results are obtained regardless
of the circuit’s grounding philosophy (impedance earthed, isolated neutral, etc.) or short circuit
current.

In this paper a short introduction will be given on the functionally of the 2BPM system. Different from
traditional monitoring equipment for MV cables, it works with two sensors placed at both cable ends.
Both sensors measure the arrival time of the travelling waves associated with PD’s or faults. For the
required time synchronization between sensors, it uses own injected pulses. Knowing the difference in
arrival time, any PD activity or fault can be pinpointed within 1 % of the cable length between the two
sensors. The maximum distance between two sensors is 4 to 12 km (depending on the cable type and
the wish to measure PD or faults or both). It is even allowed to have substations along the cable circuit
being monitored, since travelling waves from PD pulses and faults easily pass such substations.

Monitoring PD activity as a function of time for many different medium voltage power cable systems
in the recent years has delivered better insight in the actual behaviour of PD activity. Some of the
experiences are:
• PD activity from a weak spot is in many cases not constantly present. There can be minutes, hours
or days of ‘radio silence’ between burst of PD activity from a certain weak spot;
• Often PD magnitudes reach quite high levels (nC range) in the weeks or days before a failure;
• The duration between the start of measurable PD activity and its related failure is on average a few
months for polymer cable systems and a few years for paper insulated cable systems.

The monitoring of faults by this system has also resulted in some interesting findings. Especially the
fact that it detects the travelling waves associated with faults makes it possible to detect faults in any
situation, independent of the network grounding and independent whether a fault is self-healing or not
Bernd.vanMaanen@dnvgl.com
and how often it is repeating. In this paper, it will be shown that especially faults that happen in MV
power cable networks with an isolated neutral or impedance grounding often come with a row of self-
healing faults of short duration (1 ms for instance) before a major breakdown event occurs that triggers
the protection equipment (switches off). In one case 124 self-healing faults were detected over a
period of many days. This information offers the network owner the ability to repair such faulty spots
before complete outages happen.

KEYWORDS

Fault, breakdown, location, pinpoint, on-line, medium voltage power cable, PILC, XLPE, partial
discharges, monitoring, degradation.

BASICS OF THE 2BPM SYSTEM

For on-line PD monitoring, traditional PD detection devices use only one inductive sensor. Using only
one sensor has the disadvantage that it is difficult, or often even impossible, to locate PD’s and to
distinguish PD’s from noise.
The 2BPM system has two inductive sensors (for a general set-up, see Figure 1). These sensors can be
placed for instance around the earth lead(s) at each cable end of the cable system being guarded (see
Figure 2 for an example of a sensor placed at one cable end).
In Figure 1, the two inductive sensors are placed (for instance) around the earth leads (see Figure 2) in
the substations A and B. These positions can be several kilometres (km) away from each other, up to
12 km (depending on whether the focus is on “defect-locating” (DL),”fault locating” (FL), or both).
For safe installation the inductive sensors have a split core that allows for clamping around the earth
lead(s) without the need for disconnecting the earth leads.

A X B

Fig. 2: Typical set-up of a

2BPM sensor placed
Fig. 1: Typical 2BPM set-up with left (location A) and right around the earth leads of a
(location B) a sensor and a dedicated industrial computer three phase XLPE MV
with wireless or LAN internet connection. power cable.

From a weak spot or fault at position X (in Figure 1), electromagnetic waves travel along the cable in
two directions away from the spot X, at travelling speeds close to half the speed of light. Each of the
two inductive sensors detects the travelling waves passing. The travelling wave amplitude, together
with the time of arrival, is stored. The difference in arrival time tells where the source of the travelling
wave (PD or fault) is located.

Both inductive sensors measure the currents from the travelling waves passing by, whether it is from
PD’s (milli-Ampere range) or from faults (Ampere range).
In order to locate the PD’s and faults, the internal clocks of the two inductive sensors must be time
synchronized with an accuracy of 100 ns or better. The time synchronization is based on patented
pulse injection technology, being:
• economical: it is using the inductive sensors for regular pulse injection as well;

1
• practical: the pulse injection doesn’t use GPS, thus there is no GPS module needed nor a GPS
antenna on the roof of a building;
• reliable: if the sensor can inject on regular moments time sync pulses, being correctly measured at
the other cable end, then this sensor can also measure pulses / travelling waves from PD’s or faults
correctly. In other words, there is a regular self-check (each minute).

The fact that the time is synchronized at both cable ends with an accuracy of 100 ns or better makes it
possible to calculate the difference in arrival time of the travelling waves of PD’s or faults quite
accurately. Hence, a location accuracy of 1 % of the cable length between the two sensors is achieved.
It might be conflicting to identify an absolute value of 100 ns on one hand and 1 % accuracy on the
other. The 100 ns time synchronization accuracy is important to reach the 1 % location accuracy in
cables down to 200 m in length (i.e. 2 m). For longer cable lengths the absolute accuracy of 2 m will
not hold, because of the fact that the frequency content of the travelling waves reaching the sensors
will become less, based on which that is becoming the dominant factor in the location accuracy. As a
result, over the full range of cable lengths that can be monitored by the system (200 m - 13 km), the
location accuracy is always close to 1 % of the total cable length between the two sensors.
At both locations A and B, once per hour (for PD’s) or immediately (for faults) the stored data is sent
via internet to a server. Here all data is combined. Only travelling wave data is accepted of which the
difference in arrival time corresponds to a source (defect or fault) located inside the cable length under
test. The rest of the data is considered to be noise. This is a robust noise suppressing feature. The red
thin arrows at the locations A and B in Figure 1 represent such external noise, correctly discarded by
the system.

2BPM AS DEFECT LOCATOR (DL)

DL: partial discharge map

The DL functionality is fully based on PD detection and localization. Detection of PD’s is being used
by many network operators to trace weak spots successfully. Numerous papers were written about this
feature. An example of a PD map is shown in Figure 3. This PD map was measured in a network of
one of the network operators using the 2BPM system. The weak spot was replaced before the cable
failed. Investigation of the replaced component showed heavy degradation and the network operator
clearly stated that an important outage in a city was avoided [3].

time

cable location

Fig. 3: PD map as a function of time (y-axis) for the whole cable (x-axis). The PD intensity is
plotted on the vertical axis (z-axis). The insert shows the 2BPM colored warnings levels given as
a function of time [3]. The photograph on the right shows the defect causing the discharges that
was found upon dismantling the faulty component.

2
DL: development from PD-activity to fault (if the weak spot is not replaced)

Sometimes, network operators don’t replace a weak spot identified. Although often this results in a
fault, a network operator can have reasons to do so. This gives the development team the unique
opportunity to check how much time there is between a) the start of PD activity, b) a warning about a
critical weakness (warning level 1) and c) the actual fault moment for MV cables in service. Weibull
statistics were used to make a correct interpretation of all the data. Considered as a fault was all data
where indeed a fault happened after a certain period of PD activity. Also data was included from cases
where a weak spot showed such severe degradation that the network operators indicated that the weak
spot would have resulted in a fault in the near future and moreover that they were happy the defect
was replaced.

In this way, in total 29 cases could be used for the Weibull calculations (in addition to censored data).
Another paper [4] has treated all information in detail. Although it is impossible to predict accurately
the time until a fault occurs, one can estimate the chance on having a fault within a certain time period.
In Table 1 this period is identified for a 50 % change on having a fault. For example: after seeing for
the first time PD activity in an XLPE cable system it will take on average 2 months to have a chance
of 50 % on a cable fault (row 3; column 2). After a 2BPM warning of the highest level 1, this period is
reduced to only 10 days (row 3; column 3).

Table 1: Time until 50 % chance on having a fault (failure or breakdown) from

the start of PD activity (column 2) and a warning level 1 (column 3)
start PD until fault warning level 1 until fault
PILC MV cable systems 16 years 3 years
XLPE MV cable systems 2 months 10 days

These results show several interesting things:

• it takes many years (on average) before weak spots in PILC cable systems result in a fault;
• it takes only a couple of days to months (on average) before weak spots in XLPE cable systems
result in a fault;

The warning levels, based upon the interpretation of the PD activity measured, is indeed able to come
much closer to the actual fault moment than without such interpretation, taking into account the
following:
• for XLPE cable systems the warning moment is considered to be adequate
• For PILC cable systems it is still difficult to come with a warning only a few days before a
reasonable chance on a fault.

So far, the above mentioned results were obtained for complete systems (PILC or XLPE based) and
not yet for their individual components because the number of cases is still too limited. It is the
intention of the development team to come in the future with similar accurate warnings for these
individual components such as cables, joints and terminations. Further refinements concerning voltage
level and types of components are also a challenge. The expected large scale application of the system
will for sure help to achieve these refinements.

2BPM AS FAULT LOCATOR (FL)

FL: application range

Traditional protection equipment and short circuit indicators are based on measuring the 50/60 Hz
short circuit currents. Not all networks with a fault come with a short circuit current large enough to be
detected by this equipment. Especially, faults in impedance grounded networks can stay undetected or
at least un-interrupted for a long time. As a further result, in case of an un-interrupted fault, a cross-

3
country fault can occur, leaving the network now with two faults to be found and repaired. This can
happen especially in cable networks or networks with a mix of overhead lines and cables.

The fault locator feature doesn’t detect the 50/60 Hz short circuit currents in case of a fault, but it
detects the travelling waves that are emitted from a fault. Such a travelling wave occurs at the very
start of the fault current and it is always present, independent of the network grounding. Like a PD, the
travelling waves from a fault travel in two directions away from the fault location, running through the
cable with about half of the speed of light, passing the sensors. The sensors detect these waves,
including their arrival times. Whether such travelling waves develop into a full and large 50 Hz short
circuit current or not doesn't matter, the traveling waves have been detected already. The arrival time
of the first slope is a very accurate indicator for the fault position. The arrival times are immediately
communicated to a server. And within minutes after the fault, the network operator gets a message
(text, email) with the fault’s exact time and precise location.

Due to the detection of these traveling waves, FL can be used successfully in any situation, for
instance in the following situations:
• in networks with low or high grounding impedance to find and locate any fault (because it
doesn’t matter whether a fault is switched off or not by the protection equipment for this
system);
• in a network where certain faults recover (self-restoring faults), only to appear again after
a few days, weeks or months; these faults are difficult to locate for network operators
because of their temporary recovery. 2BPM can also detect such faults, even the first one
occuring;
• faults with the unwanted characteristic to become high-ohmic after the fault is switched-
off (because the system is not influenced by the ohmic nature of the fault afterwards).
Any of such faults are found immediately locating these within 1 % of the cable length between the
two sensors [4].

FL: example of a fault located

After having been tested in the laboratory and the field [5], FL has been further applied in the field.

One cable circuit chosen to check the fault locator feature is shown in Figure 4. It is a 4.5 km long 10
kV PILC cable, with 5 substations spread along the total cable length. The cable circuit is running in
the Netherlands, from a substation called “Sneek” to a substation called “Clegauwe”. In this paper this
cable circuit is further identified as the “Sneek cable”. The network here is impedance grounded with a
short circuit current of 2.5 kA.

The “Sneek cable” was showing regular self-healing faults. Such self-healing faults are quite common
in a network with an isolated neutral as is the case here. The knowledge that there were self-healing
faults occurring indeed was obtained from a third party sensor system connected to this “Sneek cable”.
This sensor system was installed for regular protection service. An example of two short circuit
currents as measured by the sensor and related to the self-healing faults of the “Sneek cable” is shown
in Figure 5. From this result one can see that sometimes the self-healing fault repeated itself indeed,
here even after 1 s already! The bandwidth of the sensor system is not high enough and it also is
lacking a double sensor system with time synchronization to locate a fault accurately.
Therefore, in order to capture and locate the fault, a 2BPM system was installed at this “Sneek cable”
early February 2015.

4
Fig. 4: A 4.5 km long PILC cable where 2BPM was installed with the intention to capture a self-
healing fault. Apart from the two substations in which it was installed, there are 5 more
substations along this 4.5 km long power cable. Substations are identified as ∆. The joints are
colored dots (the color indicates the type of joint).

Fig. 5: The current as measured in a network close to substation Sneek with a third party sensor
system. With this sensor, several self-healing faults were seen, among which the two shown
happening within 1 s.

Starting February 25, the self-healing fault appeared again. Now, the fault locator could detect and
locate the faulty spot. It appeared to be a joint 2620 m away from substation Sneek.
The email message that was sent to the network operator within a few minutes after the first self-
healing fault appeared as is shown in Figure 5.

The network operator decided not to replace the joint in this case. That gave nice insight information
on the further development of this self-healing fault.
As it happened, there were a further 123 self-healing faults, all from the same spot. All were detected
by FL. The travelling waves from one of the faults as recorded by the left and right sensors of the
2BPM system in the “Sneek cable” are shown in Figure 6. They look almost identical, but that is
mainly because of the bandwidth of the sensors and the fact that the pulses are being clipped (on
purpose).

Fig. 5: Email message from the 2BPM server, telling the network operator that there was a fault,
including the moment the fault happened and the location of the fault (within 1 % of the cable
length).

5
These self-healing faults did spread over a period of a few days until a final and full breakdown
occurred in February 2015.

Fig. 6: Travelling waves from one of the faults in the Sneek cable, as received in the left and
right sensors.

It was further noted that the 2BPM system did not see partial discharges from this faulty spot. It is an
example of a situation where a serious defect is present, that is not producing partial discharges
preceding a fault and large enough to be detected by the system. After removal of the faulty joint, it
was identified as a bitumized joint and from other measurements in other cable circuits it was already
known that often a weakness in such a joint, preceding a fault, is not generating measurable partial
discharges indeed.

This case shows FL’s ability to detect and locate a fault in a network within 1 % of the cable length,
where it doesn’t matter whether the fault occurs in a network grounded via an impedance (2.5 kA short
circuit current), even if the fault is a self-healing fault. In other cases it was shown that the system is
also able to detect and locate faults in a network with an isolated neutral.

OPERATIONAL ADVANTAGES OF APPLYING THE SYSTEM

The Dutch distribution network operators Enexis and Alliander (together operating about 60.000 km
of MV cable network) evaluated the potential benefits of 2BPM from a business case point of view.
They took into account the combined possibilities to apply it for defect location (DL), fault location
(FL) or both.
In order to obtain this result, detailed information was needed about:
• the network (how many faults, related outages and present SAIDI and SIAFI, types of cables and
its components, background of the present faults, network grounding, etc….), per region or city;
• the effectiveness of the system to reduce the SAIDI and SIAFI, also in relation to the cable
component types and the network grounding.

6
An important starting point for the business case calculation was the assumption that 2BPM is able to:
• detect and locate weak spots from internal defects with an efficiency of 65 % in PILC cables up to
4 km and XLPE cables up to 8 km. These are weak spots with PD’s detectable by DL and that
would eventually result in a fault (if the weak spots are not being repaired prior to a fault). The
efficiency mentioned here is based on long term experience with DL
• detect and locate faults (those that can’t be avoided) in cables up to at least 8 km (at present > 13
km is reached) with an efficiency of at least 90 %. Such faults might result from weak spots that
don't generate PD’s detectable by DL (35 % of the weak spots related to internal defects) and all
other faults that are the result of external defects like those from digging activities.

It would require a full paper to discuss all further starting points and show all calculation results. But
in summary the conclusions of Enexis and Alliander are as follows:
• The application of 2BPM is a cost effective way to reduce the SAIDI and SAIFI:
• The defect locator (DL) is mainly of interest in areas which have a high density of customers and
where many failures happen;
• The fault locator (FL) is mainly of interest in networks where, because of the type of grounding,
first failures don’t result in a switching action, making such faults difficult to locate with
traditional equipment;

BIBLIOGRAPHY
[1] P.C.J.M. Van Der Wielen, 2011, “The need for smart diagnostics in future smart grids – a
practical example for MV cables”, Proceedings Jicable 2011 conference, France,
Versailles, paper B.9.2.
[2] S. Mousavi Gargari, 2012, “Pattern recognition and knowledge extraction for on-line
partial discharge monitoring with defect location”, PhD Thesis Technical University
Eindhoven, 2012, ISBN: 978-90-386-3209-4.
[3] R.P.Y. Mehairjan, 2014, “Experiences with the Introduction of Online Condition
Monitoring in Asset Management for Distribution Networks”, Proceedings CMD 2014
conference, Korea, Jeju.
[4] Yizhou Qian, 2015. “Risk on failure, based on PD measurements in actual MV PILC and
XLPE power cables”, Proceedings Jicable 2015 conference, France, Versailles, paper
C.1.2

7
8