A Guide For Partial Discharge Measurements On Medium Voltage (MV) and High Voltage (HV) Apparatus
A Guide For Partial Discharge Measurements On Medium Voltage (MV) and High Voltage (HV) Apparatus
A Guide For Partial Discharge Measurements On Medium Voltage (MV) and High Voltage (HV) Apparatus
Authors
November 2020
Version 1.0
© OMICRON Page 1 of 8
1 Introduction
Partial discharge (PD) measurement is a very sensitive test which requires measuring very small signals. As
an example, for factory acceptance tests, the pass/fail criteria of most apparatus are in the picocoulomb (pC)
range. Therefore, test results can be easily affected by interferences.
According to IEC60270, interferences are referred as signals detected during PD measurements, which do
not originate from the test object. The terms interference, disturbance and noise are commonly used
interchangeably and therefore, they all have the same meaning in this article.
There are different types of interferences, which can be coupled into the test circuit through either a radiated
propagation path or a conduction path. Examples of interferences are radiated electromagnetic interferences
(EMI), sparking (e.g. gas discharge lights or welding machines), electromagnetic waves from corona type PD
in nearby equipment and switching from power electronics. Figure 1 shows an example of different sources of
interferences frequently encountered.
Interferences can affect the results of PD measurements in different ways. For example, they can:
• Decrease the sensitivity;
• Increase the complexity of data interpretation;
• Superimpose to the actual PD pulses resulting in distorted magnitudes.
It is not feasible to completely suppress interferences. Even in well-controlled laboratories, a given background
noise level is usually observable. However, using efficient grounding design, Faraday cage and PD-free
apparatus, it is common to reach a sensitivity level of 1pC or below.
PD measurements are not always performed in such environment. Interferences can be measured in
laboratories, in factories and are prominently encountered during onsite PD measurements (offline & online).
In addition, EMI are present in most environments and are usually prominent in the allowable conventional
frequency range of IEC60270. The present chapter will provide a quick summary of different hardware or
software techniques for noise mitigation.
Figure 2 : left) Picture of high-voltage filters and right) typical test circuit including a HV filter
2.2 Gating
External disturbances that are visible in the PRPD diagram can be masked using the window gating function.
Interferences that are fixed in position with the test voltage frequency can be easily removed using this gating
function. These events will not be considered for the charge calculation. Figure 3 shows an example of the
application of manual gating.
Figure 3 : left) without manual gating and right) with manual gating
Another available method is the channel gating or antenna gating technique. It uses a second sensor and a
second PD input channel to measure interferences and remove them from the original measurement. This is
particularly useful when the external disturbances are not stable with the test voltage frequency. For example,
a high-frequency current transformer (HFCT) can be installed around the grounding cable of a nearby
equipment that is known to cause interferences. Figure 4 shows the results of an actual measurement,
respectively without and with channel gating.
If a conventional PD measurement is required, IEC60270 allows the use of signal gating to hide part of the
PRPD diagram for up to 10% of each test voltage period under certain conditions. For more information, please
refer to clause 10 of IEC60270.
One measuring impedance is connected in series with the test object (DUT 1) while the second measuring
impedance is connected in series with the sensor (Ck), or with a second test object (DUT 2) which has a similar
impedance as DUT 1. The common mode elimination is based on the principle that PD signals and common
mode disturbance signals show different polarities when measured in two different branches as described in
figure 5. When signals are captured by the measuring impedances, the differential processing unit (MBB1 in
figure 5) will display the difference between the two measured signals. If common mode noise signals (yellow
pulses in figure 5) are coupled into the test circuit and that both branches are balanced, the signals will
Figure 6 : left) without differential measurements and right) with differential measurements
Figure 10 : 3-PARD inverse gating function to remove one specific cluster (external interferences)
The recorded frequency content of the measured pulses can be influenced by several factors, including the
type of anomalies, the location, the traveling path, the insulation material and the sensors. Their distinctive
characteristics can therefore be measured and processed by different bandpass filters.
The visualization is done using a star diagram like the one used in 3PARD. The correlation between the PD
magnitude obtained from each frequency filter is plotted on a diagram where separate clusters are created
over many cycles. When selecting a specific cluster, it can be possible to isolate a PD source or other external
interferences. Figure 11 shows a simplified theoretical schematic of 3CFRD, Figure 12, 13 and 14 show a
practical example of the application of 3CFRD on a potential transformer (PT).
Figure 14 : 3CFRD inverse gating function to remove one specific cluster (ambient noise level / noise floor)
www.omicronenergy.com