Draft, submitted to JICABLE Committee, Paris Versailles, France, June 24-27, 2007

ON-SITE PARTIAL DISCHARGE TESTING OF DISTRIBUTION CLASS

CABLES USING VLF AND POWER FREQUENCY EXCITATION

Detlev W. GROSS, Power Diagnostix Systems GmbH, (Germany), gross@pdix.com

Markus SÖLLER, Power Diagnostix Systems GmbH, (Germany), soeller@pdix.com

With the low-energy surface of an embedded spherical cavity

ABSTRACT in fresh polyethylene, no such free electron is available to
have the partial discharge incepting, although the internal
The ongoing deregulation of the energy market has had up electrical field of that gas bubble is larger than the inception
to now a strong impact on the maintenance strategies of field (EStr). Thus, no partial discharge occurs until such free
most utilities. Especially, the increasingly service aged electron is provided by de-trapping of space charges or by a
population of distribution class cables influences the reliabil- photon of the natural radioactivity. This natural radioactivity –
6
ity of the distribution network. Thus, field-testing is required radioactive soil, gasses, cosmic photons – cause about 2·10
-3 -1
to assess the severity of the degradation and to determine m s free electrons. Thus, the statistical delay to have a
the cables, which require urgent replacement. cavity discharge in fresh polymeric material incepting may
Different methods are available for the high voltage excita- reach several tens of minutes depending on the cavity's size.
tion of the cables under test. The fixed frequency resonant
test set running at 50/60Hz power-frequency offers a perfect
match of the service condition, but such test sets are
comparably heavy and costly. Other solutions, such as EMax
variable frequency resonant test sets, damped oscillating
wave, or very low frequency (VLF) excitation offer more cost-
effective testing solutions. EStr
ERes
KEYWORDS
ERes
Partial discharge location, PD mapping, VLF, resonant test EStr
set, tan delta, field-testing.

INTRODUCTION EMax
Partial discharge activity is a prominent indicator to assess the
grade of degradation of high voltage equipment. The mapping
of the partial discharge activity versus the cable length allows Figure 1: Internal E field with low electron availability
identifying cable partial discharge as well as weak accesso-
Once this initial partial discharge impulse has occurred, the
ries. However, strong differences exist between polymeric
cavity's surface is polluted with electrons being kept with traps
cables and mass-impregnated cables.
of comparably low energy levels. Thus, these electrons
Historically, partial discharge measurements on cables were
become available statistically depending on the material's
increasingly applied since the change to polyethylene (PE)
properties. With fresh polyethylene (polypropylene, epoxy
and later cross-linked polyethylene (XLPE) as insulation
resin, etc.), this de-trapping time constant is in the range of
material for the majority of power cables.
tens of ms and above.
Improved signal acquisition and processing made this tech-
Figure 1 shows how the electrical field in the cavity reacts on
nique also applicable to non-shielded field environments.
the statistical delay caused by the de-trapping. Figure 2 shows
However, care must be taken to adequately filter the high
a corresponding partial discharge pattern or so-called ϕ-q-n
voltage source and to reduce pickup of ambient noise.
pattern.

OCCURRENCE OF PARTIAL DISCHARGE

The solid and liquid insulation materials used for insulating
power cables can tolerate an electrical field that exceeds by
far the normal operational field strength typically applied with
cables and other high voltage equipment. Thus, to enable
partial discharge, it requires an imperfection that has a lower
inception field, such as a gas inclusion, or, alternatively, which
strongly increases the local electrical field within the insulation,
as a sharp metallic inclusion, for instance. Both types of
imperfections can happen during production, may remain
undetected during initial testing, or evolve during service.
Partial discharge, an electron avalanche, requires a free initial
electron that is being accelerated in the electrical field strongly
enough to trigger the avalanche.
Figure 2: ϕ-q-n pattern with low electron availability
Here, the discharge magnitude reflects the phase dependent
electrical field. Thus, if, for instance, the voltage shape would
be triangular, the pattern shown with Figure 2 would be
triangular as well. Other electrode configurations, such as a
flat delamination, show other correlation between electrical
field and discharge magnitude.

EMax

EStr
ERes
ERes Figure 5: Partial discharge with a cable termination
EStr
Figure 5, for instance, shows the discharge activity of several
flat cavities. Here, the symmetrical pattern indicates that the
EMax cavities are symmetrically within the insulation and not at-
tached to any conductor.

Figure 3: Internal E field with high electron availability

The same setup (gas bubble in polymeric material) is exposed

to ionizing irradiation causing a much quicker de-trapping
(Figure 3). Thus, the electrons are available earlier and the
discharge magnitude is lower (Figure 4). Accordingly, material
properties, such as the ability to store surface charges or
volume and surface conductivity, for example, have a strong
impact on the partial discharge appearance. Humidity or
corrosion of polymeric surfaces by partial discharge, for
instance, as introduced during the service of the equipment,
has an additional impact on the material's ability to provide
free electrons and its time constant and, hence, strongly
influences the appearance of the discharge pattern.
Generally, with aging (corrosion) of polymeric surfaces chains Figure 6: EPR cable - delamination of semicon layer
are being cracked up and the time constant of providing free
electrons drops. With respect to power frequencies this time Figure 6, in contrast, shows a non-symmetrical pattern. Here,
constant is negligible for metallic surfaces. the delamination of semi-conductive material causes different
Based on such fundamental consequences of gas discharge starting conditions for the discharge activity for the positive
physics [1,2,3], the analysis of the ϕ-q-n pattern or phase and negative half-cycle.
resolved partial discharge pattern has proved to be an
excellent method to identify the properties of the discharge PARTIAL DISCHARGE MAPPING
activity, to distinguish different sources, and to draw conclu-
At their origin, partial discharge impulses have a rise of about
sions on the properties of the materials involved with the
1ns under Nitrogen atmosphere and ambient pressure. Thus,
discharge activity.
at this origin, the current impulse, i.e., the displacement
current caused by the electron avalanche, has an amplitude
spectrum that is flat from DC to a couple of hundred MHz. In
case of a power cable, the impedance of the cable transforms
the current impulse into a voltage impulse that travels the
cable in both directions. Generally, the original bandwidth
cannot be maintained over distance [4]. Technically, in order
to form an ideal cylindrical capacitor, a triple-extrude polymeric
cable consists of three layers – the main XLPE insulation is
terminated on both sides with a thin layer of semi-conductive
polymeric material. These layers inherently limit the high
frequency bandwidth of triple-extruded power cables. De-
pending predominantly on the thickness of the semi-
conductive layers the transmission bandwidth is limited to 5-
10MHz with a strongly increasing attenuation as a function of
the frequency and the distance, of course. This location
dependent bandwidth can be used for sensors embedded in
Figure 4: ϕ-q-n pattern with high electron availability accessories and for monitoring concepts [5,6].
Additionally, this layer structure causes dispersion, i.e., that HIGH VOLTAGE SOURCES
the travel time is not uniform over the frequency. Thus, the
originally steep front of the traveling partial discharge impulse Typically, in the test room of the cable manufacturer a trans-
is further smoothened by this effect and by the said attenua- former or a fixed-frequency resonant test-set is used. This
tion and, hence, the achievable location accuracy is limited. ideally matches the normal operation conditions of the cable.
Compared with polyethylene cables this attenuation of the When it comes to field-testing of laid cables, weight and cost
traveling high frequency impulses is substantially stronger for considerations demand more effective solutions and the
EPR cables. However, the strongest attenuation is found with heavy and costly transformer and fixed frequency test set are
mass-impregnated cables. Here, the reflection from the far not a practical option.
end increasingly becomes hidden in the noise floor for cables The variable frequency resonant test set, which uses a filter
longer than a couple of hundred meters. and coupling capacitor as base load, whereas the added
The shape or frequency of the high voltage source chosen capacitance of the cable to be tested then lowers the fre-
does of course not influence deriving the location of the partial quency, can be set to stay close to the nominal frequency
discharge activity within a cable, as long as the imperfection (25Hz-300Hz). However, a typical test set for distribution class
produces partial discharge. Thus, in principle any high voltage cables easily reaches a weight of 1000kg.
source can be used for partial discharge location. Using VLF (very low frequency) high voltage sources dramati-
cally reduces weight and cost, because the power require-
ment of the capacitive load drops by a factor of 500, when
comparing the commonly used 0.1Hz with the 50Hz. Both
electronic solutions as well as electro-mechanical solutions
are available. Such units are portable on a cart or can even be
carried.
Finally, oscillating wave test sets, whereas charging the cable
with a DC supply and discharging it into an inductor produces
a damped oscillating wave. Here, as well, the frequency
depends on the load, i.e. cable capacitance. Depending on
the inductor's quality and the cable's tan delta, the oscillating
wave more or less quickly decays.
Another special case is the VLF co-sine source. Here, a
square-wave signal is combined with a transition in the range
of the power frequency.
Figure 7: Reflection Pattern (blue: post processed). However, besides weight and cost also the impact of the high
Once a partial discharge occurs, it travels towards both ends, voltage frequency on the local electrical field and the occur-
becomes reflected and produces, if viewed from one end, a rence of partial discharge must be considered.
typical trace of three impulses. With such trace, the time As pointed out above, the partial discharge pattern reflects the
between first and second pulse corresponds to the distance different time constants of deriving the starting electron.
from the far end, whereas the time between first and third Therefore, the appearance of a partial discharge pattern and
impulse reflects the time to travel the entire cable twice. Figure in consequence the magnitude is influenced by the power
7 shows such a typical trace. Here, also the influence of frequency. This becomes evident, if the average de-trapping
attenuation and dispersion can be seen. time constant of the partial discharge activity comes in the
Simple calculation produces then the discharge position with order of magnitude of the test voltage cycle.
reference to the cable length [6,7,8]. Thus, for a partial discharge activity, which has a de-trapping
time constant of, say, 10ms, the measurement at 50Hz
produces a well-distributed pattern. If the same activity is
exposed to 0.1Hz, the 10ms time constant will be small
against the 10s cycle time and, hence, the partial discharge
pattern changes its appearance similar to a change as shown
with Figures 1 and 3. Thus, especially with fresh polymeric
surfaces, the magnitude of the partial discharge activity is not
comparable between 0.1Hz and 50/60Hz. However, if it
comes to service-aged, i.e. corroded surfaces, the de-trapping
time constant drops and becomes also short, when compared
with the 50Hz cycle. Thus, the magnitude is comparably low
with both excitation frequencies. Accordingly, an activity that
starts from metal surfaces with a very high availability of free
electrons, such as a point-plane discharge at a metallic tip,
Figure 8: Site Map of the partial discharge activity. shows similar activity with all testing frequencies (Figure 9,
next page).
Based on the valid location results the partial discharge With all discharge activities having a high availability of the
mapping is the built (Figure 8), which shows the activities starting electron, such as corroded surfaces of incorrectly
versus the cable length. Comparing this mapping with the mounted accessories, for instance, oscillating wave test sets
laying plan and the positions of the joints as verified during the will start the activity with the first half cycle or even during the
calibration identifies faulty cable accessories (shown as blue charging time. The chance to detect an embedded cavity in
squares with Figure 8) or weak cable sections. a freshly laid cable, however, is comparably low due to the
initial statistical delay as described above.
 The resonant test set comes close to the real-life conditions,
but it is comparably heavy and costly. VLF testing is a very
cost-effective alternative, but the effects of the large difference
in the testing frequencies on the partial discharge pattern
must be considered.
Thus, because of the individual weighting of the different
properties and requirements for each case, no general
recommendation can be given.

FIELD TEST REQUIREMENTS VS. TEST ROOM

In a test room, typically a detector for partial discharge
measurement and location connects to a quadrupole in the
ground lead of a coupling capacitor. Preferably, a high voltage
filter is inserted between the transformer and the cable under
Figure 9: Point-plane discharge starting from a metallic tip test with the coupling capacitor. The partial discharge signal
as well as the voltage signal is fed via coaxial cable. The
The acquisition of a ϕ-q-n pattern with an oscillating wave test situation is similar, when it comes to the integration of the
system is not very helpful, since it combines discharge instrument into a test van, as it is required for the variable
activities, which are occurring under different voltage and, frequency resonant test set and the larger VLF test sets.
thus, do not belong together. Connecting the test set to the cable follows a fixed procedure
Another frequency dependent effect concerns the accesso- and the test van acts as reference potential.
ries. If, for instance, a stress cone has a week semi- Figure 10 shows the situation during a test with a portable
conductive path to ground the resulting voltage drop on that VLF unit. Here, the coupling capacitor, its quadrupole and the
path will cause surface discharge under 50/60Hz or variable required high voltage filter are combined into one unit. This
frequency testing. Since the currents are 500 times smaller combined filter and coupling unit, which is just inserted
with 0.1Hz testing this may remain undetected, when using between the high voltage source and the cable under test, is
VLF test sets. available for different voltage classes ranging up to 150kV.
Generally, when selecting a high voltage source for testing of Figure 10 shows a 65kV unit (TCC65) to suit a 90kVpeak VLF
laid cables the different properties (benefits and drawbacks) source.
must be compared. Figure 11 shows a setup together with a 60kVpeak VLF unit.
Here, the combined filter-coupling unit TCC45 is used to de-
noise the high voltage supply and to provide the partial
discharge signal as well as the signal for synchronization and
voltage measurement.
Combining on-site cable testing with partial discharge meas-
urement and location requires in general a high voltage filter,
since, regardless of the technical principle, all variable fre-
quency resonant test sets and all VLF test sets produce high
frequency noise that hamper the measurements, if not
adequately filtered. With the variable frequency resonant test
additionally gating techniques must be applied, because the
thyristor or IGBT switching impulses are too strong to be
solely filtered.

Figure 10: Setup in a substation with a 90kVpeak VLF unit Figure 11: TCC45 - setup with a 60kVpeak VLF unit
 CONCLUSION
Partial discharge measurement and location can be combined
with portable high voltage sources of different principles using
modern signal acquisition and processing techniques.
However, most of the different excitation methods produce as
a side effect high frequency noise that hampers the sensitivity
of the partial discharge detection and mapping and, hence,
requires adequate filtering and matching. Additionally, inte-
grating all required electronics into a combined filter, coupling
and acquisition unit simplifies the setup and improves operator
safety by using wireless Bluetooth® communication.
Variable frequency resonant test sets allow using excitation
frequencies close to the power frequency, but are typically
mounted in test vans. Further cost and weight considerations
lead to the application of VLF high voltage sources. However,
the partly strong deviation from the normal 50/60Hz service
Figure 12: TCC25 together with a 30kVpeak VLF unit conditions has an impact on the detectability of some defect
types and on the appearance of their partial discharge pattern.
Figure 12 shows such combined unit (TCC25) for a lower
voltage with another VLF source (30kVpeak). At this voltage
level the cage around the high voltage electrode of the filter
unit is not required to prevent discharge activity. REFERENCES
Integrating the T-filter and the coupling capacitor into one unit
greatly simplifies the setup. However, especially with portable [1] Niemeyer, L. " A Generalized Approach to Partial
high voltage sources, care must be taken to properly ground Discharge Modeling," IEEE Trans DEI, Vol. 2, No. 4,
the equipment and to double-check every connection. August 1995, pp. 519-528.
An instrument to even further simplify the application of partial [2] Heitz, C. " A Generalized Model for Partial Discharge
discharge measurements as well as tan delta analysis is Processes Based on a Stochastic Process Approach,"
currently in its final stage of development. Here, additionally, J. Phys. D, 9/1999, pp. 1012-1023.
the partial discharge detector is integrated into the top box of [3] Fruth, B., Gross, D. " Modeling of streamer discharges
th
the combined filter and coupling unit. The instrument has a between insulating and conducting surfaces," IEEE 5
multi-processor concept, is battery operated, and communi- ICSD, Leicester, UK, July 10-13, 1995, pp. 350-355.
cates via Bluetooth® with the notebook computer used for [4] Gross, D., " Signal Transmission and Calibration of on-
visualization. Thus, everything needed for the partial discharge line Partial Discharge Measurements," ICPADM, Na-
measurement and location is built into this unit and the goya, Japan, June 1-5, 2003, 03CH37417, pp. 335-
required remaining connections are limited to the high voltage 338.
and ground leads between high voltage source, filter- [5] Gross, D. and Comanns, R. " TE-Messung und Fehler-
instrument-unit, and cable under test. diagnose an Kabeln und Garnituren", ETG
st st
As with the passive combined filter and coupling units, this Fachtagung, Esslingen, Germany, Mai 31 - June 1 ,
new system comes in different voltage classes to suit the 1995, ISBN 3-8007-2089-2, pp. 213-217.
typical application. [6] Gross, D. "On-site partial discharge diagnosis and
Besides partial discharge testing and location, this new monitoring on HV power cables," Conf. Proc. of Jica-
instrument offers also convenient tan delta analysis using the ble, June, 1999, Versailles, France, pp. 509-514.
coupling capacitor as reference capacitor. No modification of [7] Gross, D. and Herbig, J. "Partial Discharge Fault
the setup is needed to change between partial discharge Location and Diagnosis on HV Power Cables," CEIDP,
measurement and tan delta mode and the operating software Oct. 15-18, 2000, 00CH37132, pp. 630-633.
presents the results in parallel. [8] Gross, D. and Hesse W. "On-Site Partial Discharge
Additionally, based on the same concept, also an instrument Diagnosis and Location on Power Cables," ETG, "Di-
that covers just tan delta analysis will be introduced for VLF agnostik Elektrischer Betriebsmittel," Berlin, Germany,
th th
testing. February 26 -27 , 2002, ISBN 3-8007-2671-8, pp.
The instruments generally offer a wide synchronization range 293-296.
from VLF to >300Hz allowing a combination with all relevant
high voltage sources for field-testing.