Using PD Survey Tools to Improve Safety and Detect Incipient Failures

Presented By:
Dr. Tony McGrail
Doble Engineering
TechCon 2017


Insulation deterioration in substation equipment may lead to the catastrophic failure of equipment, consequent collateral damage and possible impacts to safety. Equipment failures may include metalclad device, insulators, bushings and other items. The symptoms of deteriorating insulation may include elevated temperature such as would be detected by an IR scan, while a common side effect of insulation deterioration is associated arcing, sparking, and discharge. These effects produce radio frequency interference and may be detected with an appropriate sensor during a partial discharge (PD) survey. IR scans are common, PD surveys less so, but they can be just as effective. A PD survey may be conducted alongside routine visual inspections of a substation. PD emits broadband radio energy which may be detected using several methods including CT’s, antennae and surface probes. Spectral analysis is a quick way to look for activity across a broad frequency range; statistical analysis to indicate sources associated with PD rather than communications or electronics; level analysis to look for activity at a given frequency; phase resolved analysis to relate PD pulses to the power system frequency. This presentation will look at PD survey and test techniques – how to start, what to look for and how to get value from the approach in terms of reliability and safety.


Electrical equipment needs to be insulated to ensure energy flows to where it is supposed to go. The insulation may deteriorate through a variety of physical phenomena, such as ageing or weathering and electrical stress. The breakdown of insulation may lead to equipment failure – which is a hazard to both equipment and personnel. Detecting advanced deterioration helps reduce likelihood of exposure to ‘suspect’ equipment. Regular testing/maintenance may indicate the presence of deterioration, but if the failure mode is such that the timescale from detectability through to failure is less than a test/maintenance cycle, failures may occur without warning. Regular InfraRed surveys help identify suspect equipment. In this paper we discuss the addition of regular Partial Discharge surveys to increase the likelihood of incipient failure detection and thus improve safety.

Figure 1 Bushing failure which may have been prevented with appropriate PD detection

Insulation breakdown may lead to catastrophic failure, and may be quite insidious in nature – developing over several months before leading to failure. PD detection is performed On-line and is non-invasive – sensors are used to pick up airborne or conducted signals; there are no outages required. There is a degree of knowledge required to interpret PD signals, but a basic skill level can be acquired in a few hours and successful surveys run immediately – numerous organizations have included PD surveys in their annual surveys for many years. The survey is low risk and may be performed at any time – allowing for review of equipment at operating temperatures, voltages and stress distribution.

PD Detection – history

In 1984 the Central Electricity Generating Board (CEGB), which ran the generation and transmission systems in the UK, had received a number of complaints about TV interference near a particular substation; the calls were identified in the general calls received, but no root cause was found. After 6 months of complaints, a bushing on a station transformer failed catastrophically; the resulting fire is shown in Figure 2.

Figure 2 Substation fire resulting from bushing failure with PD symptoms

A moment’s reflection: if untrained neighbors, using TV’s as sensors and detectors, could identify that there was something wrong, how much more detail could be obtained by trained operatives with more appropriate sensors? PD was not new to the industry or to the CEGB (1).

So the CEGB began surveying every station, twice yearly using antenna to detect radio frequency interference – RFI – which are broadband signals generated by PD sources. National Grid, which inherited the UK transmission system has continued those services, which are now undertaken by Doble PowerTest UK.

Have the surveys found anything?

So far the surveys have uncovered PD related issues in over 50 transformers which could have led to failure. The PD survey detected, amongst other things PD related to: 14 transformers with loose frame clamps or core bolts; 4 selector problems; 2 broken barrier boards for OLTC connections; 3 winding based discharges; multiple loose connections and floating components; several surge arresters.

An initial survey is an excellent way to detect a PD source – but further and more detailed investigation may require deeper analysis. Survey may lead to test and intervention and/or subsequent monitoring to track development of a source.

What is PD?

PD, in general, is a localized deterioration and breakdown in an electrical insulation under high voltage stress; more technically, it is the breakdown of a small portion of a solid or fluid electric insulation under high voltage stress, which does not bridge the space between two conductors. If it did bridge the space then we would have moved from partial discharge to complete discharge. Figure 3 shows some examples of sources of PD provided by voids in solid/fluid insulation or on surfaces/barriers. In a capacitive model C1 represents the void , C2 the capacitance of ‘good’ insulation in series with the void and C3 the ‘good’ insulation in parallel to the void.

Figure 3 PD Sources, models and signals

As shown in Figure 3, PD pulses can be related to the phase of the applied voltage; the resulting pattern may help in diagnoses (1).

Much time and effort can be applied to measuring the various physical parameters which are affected or generated by PD – light emission, acoustic noise, chemical changes, thermal uplift, elctrical signals conducted, radio and othe electromagnetic (EM) waves generated.

A simple approach is to record a spectrum – see how strong the radio signals are across a frequency range which is known to be affected by PD, say 50 kHz up to 1,000 MHz.

Detection is using a tunable radio across the spectrum of interest and an antenna. If a TV can detect the signal we probably don’t need to be too precise; that isn’t to say that when we need to do more detasiled analysis we don’t use appropriate tools and techniques. Figure 4 shows a typical ‘background’ scan which is used as a reference for future measurements – taken either away from the station, or at a time when there is no suspected or actucal PD present.

Figure 4 Background PD Spectrum

The background has clear peaks at known energy sources – such as cell phone comms or radio/TV signals; such signals are usually narrow band as energy outside of a specific frequency is, in simple terms, a waste of money for the signal generator. The analysis of a subsequent scan, in the first instance, is to look for an increase in energy of the signal across a wide frequency band – PD being generally broadband. Figure 5 shows the overlay of a baseline (green) and subsequent survey (blue).

Figure 5 Baseline scan and subsequent PD Survey scan

The overlay in Figure 5 shows a positive indication of PD activity. The closer to the source, the ‘louder’ the signal will be, and the higher up the page the chart line will appear.

More detailed surveying may be performed by using directional RFI antennae, or by using high frequency CT’s to detect signals in conductors such as neutrals or ground straps.

Survey Cases

We will give here some practical examples of successful PD detection which led to subsequent intervention and likely failure prevention.

Surge Arrester

Figure 6 shows the arrangement of a variety of equipment on a station. Surveys were taken at the base of each arrester and at each CCVT, using a standard antenna; a background survey was taken some distance away from the suspect equipment.

Figure 6 Arrangement of Insulators, Arrestors and CCVT’s

Figure 7 shows the resulting 7 spectra overlaid in one chart. The background spectrum is the red trace – the lowest trace; 5 traces near arresters and CCVT’s are just above background; one trace, near to the ‘south’ surge arrester is much higher in energy, and is broadband.

Figure 7 PD nearest one arrester is higher than other units and background

As a result of the survey an outage was taken and the CCVT addressed; a catastrophic failure was likely prevented.

CVT Survey

In a similar manner to the previous case, Figure 8 shows several surveys of CVT’s at a transmission station.

Figure 8 CVT Spectra showing PD at one location

As with the CCVT’s one CVT shows a higher PD signal and an outage was taken to undertake a replacement.

Power Transformer

In this case, as part of a PD survey, signals on a transformer neutral were taken using a PD Spectrum device and a high frequency CT, as shown in Figure 9.

Figure 9 Deploying a 300 MHz Radio Frequency Current Transformer (RFCT)

The resulting spectra show a clear uplift when related to a background signal, as shown in Figure 10.

Figure 10 Transformer PD Spectra showing PD

The generation of the spectral data is sufficient to warrant further investigation. At this point we do no know of the PD signal is coming from the transfoirmer to be conducted to ground, or coming from the ground into the transformer! Some simple tests and analyses can help differentiate between the two possibilities.

Discussion & Conclusion

PD surveys are relatively inexpensive to perform and can be of great value. Detection of substation related issues has been carried out successfully in a number of locations and possibly catastrophic failures prevented. PD surveying is a relatively easy skill to acquire, and the tools are readily available. False positives may be recorded, but some appropriate training and support helps reduce those.


[1] ‘Discharge Detection under Noisy Conditions’, A. Wilson, Proceedings of the IEEE, Vol 121, No. 9, September 1974.

Join our email list

We use cookies to give you the best online experience. By using this website you agree with our cookie policy.