Practical Non-Invasive Partial Discharge Cable Testing

William G. Higinbotham Industry Topics, Spring 2021 Industry Topics

Partial discharge (PD) is a poorly understood phenomenon that can cause catastrophic failure of electrical assets. It can exist at initial commissioning or occur only after decades of use. The dictionary defines partial discharge as:

a flashover of part of the insulation system due to a localized electric field greater than the dielectric withstand capability of that part where the overall insulation system remains capable of withstanding the applied electrical field

Partial discharge has been shown to be a leading indicator of impending failure in cables, switchgear, generators, transformers, and insulators. The discharge has a number of negative side effects, the most troubling being heat and nitrous oxide generated when surface discharge ionizes the air. Nitrous oxide combines with humidity to form nitric acid, and nitric acid causes irreparable harm to electrical insulation that,  in turn, increases the discharge.

A number of discharge side effects are actually helpful. The discharge causes a high-frequency current pulse that is useful in detection along with ultrasonic emissions, radio emissions, and the presence of transient earth voltages.

Partial discharge can be caused by a variety of factors including age, insulation voids and contamination, physical damage, and workmanship. Each part of the electrical system suffers from discharge in different ways.

Partial Discharge in Cables

Medium- and high-voltage cables are key components in any electrical system, and their reliability is important to maintaining continuity of supply to industrial, commercial, and residential customers. Cable systems are owned and maintained by utilities, renewable energy providers, and industrial customers. It has long been established that failure rates of aged cables can be reduced through periodic testing, as discussed in Y. Wang. Unfortunately, it has traditionally been very difficult to assess cable condition and determine whether there are impending failures. Cables may be buried or otherwise inaccessible for most of their length, and terminations are either wholly contained inside a sealed metal enclosure or, if exposed, offer limited access. 

Testing cables for partial discharge in the field can be divided into on-line and off-line methods. Traditionally, off-line methods require long outages, invasive techniques, and large and expensive test equipment. This limits its usefulness as a condition monitoring method, and testing has usually been done only at commissioning and after repairs.

On-line cable partial discharge testing and monitoring has been increasingly deployed over the past 10 years, and with advances in technology, it has become effective, safe, and relatively inexpensive. This article highlights the basics of on-line partial discharge testing in cables, covers the non-invasive on-line techniques available today in addition to periodic surveying and full-time monitoring, and provides case studies of successful detection of partial discharge in cables.

Detecting Partial Discharge in Cables

Partial discharge has long been accepted as a major cause of failure in high- and medium-voltage switchgear, and non-intrusive instruments for PD detection are widely used by utilities with excellent outcomes. S. Shamuri provides details of results from a large national distribution company where the number of outages due to MV switchgear failure were reduced by 71% over a five-year period resulting in more than 470 fewer failures per year. A high proportion of these faults were identified on cable terminations (Figure 1). Whether cable termination problems are classed as cable issues or plant (switchgear/transformer) issues is an ongoing debate. However, what is apparent is that non-intrusive detection of these defects using transient earth voltage (TEV) and ultrasonic techniques is well-established and effective in detecting termination faults. The rest of this article will therefore concentrate on the detection of faults over the length of the cable.

Figure 1: Partial Discharge Damage on Cable Terminations

Detecting Partial Discharge with RFCT

Once a PD event has occurred within the electrical insulation of a shielded cable, a set of radio frequency current pulses equal in magnitude but opposite in polarity is seen on the line conductors and the earth conductor/shield. On-line PD detection utilizes this effect by measuring these pulses using radio frequency current transformers (RFCTs) placed on the cable’s earth sheath (Figure 2).

Figure 2: Radio Frequency Current Transformers Installed on Sheath Grounds

One very important and practical consideration must be taken into consideration when looking at the applicability of taking these on-line measurements: If the RFCT is placed over the line conductor and the earth cable at the same time, the discharge currents are cancelled. The connection must therefore always be made by monitoring the earth cable only (or the line conductor after the earth has been taken off, provided the line current is low).

Dealing with Noise

When carrying out on-line testing with the cable in service, noise on the cable earth system will be encountered. Noise often occurs in the same frequency band as PD signals and will be detected by the RFCT sensor. Simply measuring amplitude can therefore provide false alarms of PD and will only have limited application. Simple devices such as a high-pass filter in the 1.4–1.8 MHz range will typically be utilized  to overcome the effect of noise. This blocks much of the noise on the earth system, but at the same time reduces the ability of the instrument to detect PD signals far down the cable. As the current pulses transmit down the conductor and earth cable from the discharge source, the signals attenuate and flatten, effectively cutting off the higher frequencies to the point that they are filtered out. Despite the use of high-pass filters, simple measurements can still be prone to false alarms and have not been found to be particularly effective for initial widespread screening of cables for PD.

Table 1 demonstrates this from the results of recent testing of a representative sample of circuits on a utility network. The circuits chosen were from a number of different substations in different geographic locations on primary and distribution networks using XLPE cables operating at 11 kV and 33 kV. The samples included a mixture of three-core cable measurements and single-phase cable measurements. Through more advanced analysis, it was determined that of the 14 sections tested, 12 (cable group A) had no PD, and 2 (cable group B) had moderate levels of PD. 

Table 1 shows the classifications using only amplitude as a measure. Without the application of any filtering, all 14 falsely indicated red. With a 500 KHz filter, 11 indicated red and 3 amber, still all incorrect. Even with the application of a 1.8 MHz filter, one sample was classified erroneously as serious red-level discharge and three erroneously as amber-level discharge. To restate, with moderate filtering (500 KHz), all 14 still showed incorrect indications. With the highest level of filtering (1.8 MHz), 4 out of 12 or 33% of this small sample incorrectly identified the level of PD.

To improve diagnostic capability from this simple amplitude-only measurement, information on the pulse shape and reference captured signals to the 50/60 Hz supply frequency must be obtained. This requires additional processing capability (Figure 3).

Figure 3: Equipment for Three-Phase On-line Measurement of Cable PD

To help with the analysis, the equipment uses a supply frequency reference; if one is not locally available, the instrument will detect the reference from the connected RFCTs. This particular instrument automatically applies a series of filters during the capture process as well as capturing raw, unfiltered data. In addition to enabling better subsequent interpretation of the data, applying an automatic filter makes the data-capture process simple, quick, accurate, and a non-specialized function.

Figure 4: On-line Measurement of Cable with Noise on the Earth System

The improved analysis is demonstrated in Figure 4 and Figure 5. The instrument recorded similar amplitudes of 3,000–4,000 pC, clearly in the red zone for XLPE cable. Looking at the waveform and phase-resolved plot in Figure 4, it is a relatively simple matter to conclude that there is no phase-related signal and the waveform is not consistent with partial discharge.  Figure 5, on the other hand, shows clear, phase-resolved activity with a pulse shape indicative of PD.

Figure 5: On-line Measurement of Cable with PD

Without the ability to look at the waveform or reference signals to the 50/60 Hz supply, both of these measurements would be classified as serious PD. Therefore, it is important to use equipment with this capability to screen the raw data. Without this, as demonstrated in Table I and Figure 4, there can be a high, potentially unacceptable number of false positives. Along with taking up valuable engineering resources and time to investigate false positives, it can quickly lead to loss of confidence in the use of on-line technology to detect PD and can become a barrier to adoption of a useful tool for the asset manager.

Another point to note about noise on the earth connection is that these measurements were taken on a utility distribution network. On industrial networks with higher numbers of potential interference sources such as variable speed drives, the instances of high noise interference have typically been found to be worse.

Locating PD in Cables

When PD has been successfully detected, knowing the exact location of the problem on the cable is highly valuable. With limited access to the span of the cable, even an approximate location of a defect has great benefit. The use of time domain reflectometry (TDR) is well established for determining a change in properties in a great variety of linear assets. For example, TDR can be used in concrete pilings to determine their depth and the depth of any cracks. This uses ultrasonic pulses reflected back from discontinuities. In fiber optics, TDR using laser light pulses can show the distance to fiber damage.

In shielded cables, the actual live PD current pulse can be used to find the location of the disturbance. Current pulses reflect from any point where there is a change in impedance, providing reflected waveforms we can use. TDR has long been part of off-line PD testing. In an off-line test, the cable is removed from service, disconnected at both ends, and a clean, PD-free waveform is applied. If PD occurs, the initial pulse and its reflections are analyzed. Figure 6 shows a PD current waveform with multiple reflections.

Figure 6: Cable PD Current Showing Reflections

 

Figure 7: Cable Partial Discharge Reflection Timing

The waveforms represent current reflecting off the far end of the cable where a disconnected end results in a massive change in impedance. We can take a number of time measurements from the waveform and use them to calculate the location of the PD event. In Figure 7, T0 is the event start and T1 is the time from T0 to the first detected pulse. T1 cannot be measured directly. It can, however, be calculated by the following formulae:

  1. T3 = 2 * length of cable (L)
    propagation speed (Vp)
  2. T1 = (T3–T2)/2
  3. Distance to PD =  T1 * Vp Ergo
  4. Distance to PD = {(T3–T2)/2 * (2 * L)}/T3

When performing on-line testing, the goal is not necessarily to locate the distance to the PD event, but rather to identify at-risk cables for further study. Determining the location of PD in an on-line test is made more difficult by two factors:

  1. The reflection from the far end of the cable is not as significant because the impedance change in a connected cable is much less.
  2. The cable is likely to have higher noise level due to being in service as opposed to being off-line.

The ability to locate a smaller reflection among higher noise is not always possible. That said, when a reflection is detectable, the PD can be located precisely. It is our experience that on-line location is possible approximately 25% of the time. It should be noted that the waveform shown in Figure 6 and Figure 7 is actually from an on-line test. The noise floor is very low, and the reflection is very distinct. In this case, the PD was precisely located.

Field Results from Periodic Surveys

A major UK utility recently completed an evaluation of the technology with positive results. As this was a technology assessment, no attempt to address suspected cables was made during the trial. Over an 18-month period, 188 33 KV cables were tested with an on-line cable test device as described above. The cable results were classified into green, amber, and red as shown in Table 2. The classification was made using a combination of amplitude, phase-resolved plot, and waveform analysis.

Within the 18-month test period and six months afterward, 13 cables failed. The results below show that red and amber categories had significantly higher failure rates. This proved to the utility that this type of on-line testing was effective at identifying at-risk cables.

Table 3 shows that a very significant percentage of cables in the red and amber categories failed in this short time. Had the customer addressed only the red cables, they would have fixed less than 10% of the cables for a reduction in failures of over 50%. Had the user addressed all red and amber cables, they would have had to fix 16.5% of their cables and would have achieved a 94% improvement in failure rate.

The Case for Full-Time Monitoring

The use of periodic measurement to understand cable condition immediately brings business benefits that enable more effective management and inform replacement or expenditure decisions. Progressing on to permanent monitoring systems enables the same assets to be monitored under a variety of operating and environmental conditions and can be a less labor-intensive way of collecting condition data. On-line monitoring also provides benefits by warning of developing failure modes. Early warning can provide an opportunity to minimize the negative effects of failure and therefore reduces the potential for negative consequences, such as loss of supply, or costs, such as damage to equipment, and secondary consequences.

Permanent monitoring is able to bring additional benefits to an organization, but a calculation must always weigh the balance of additional benefit and the cost of implementation. One way to quantify the economic benefits of on-line condition monitoring is to consider the decision in terms of risk reduction. P. Blackmore et al demonstrates how using a means to quantitatively evaluate risk and understanding how on-line condition monitoring can be used to mitigate the risk associated with electrical assets can often warrant the installation of permanent monitoring solutions in a significant proportion of asset population.

One asset group where this approach often concludes in favor of permanent asset monitoring is  extra-high-voltage (EHV) cables because, for these assets, the high consequence of failure elevates the risk calculation (probability times consequence). Another factor adding to the justification of permanent monitoring is the difficulty of regularly accessing such installations.

Field Results from Full-Time Monitoring

Figure 8 shows a 220 kV cable tunnel where a power authority chose to install a permanent PD monitoring system for its cables. Three circuits installed in two adjacent parallel tunnels were monitored. Each circuit was 3.4 km in length with a total of eight sets of grounding boxes for each circuit. Each grounding box required RFCT sensors to be installed on each side of the grounding point for each phase, as PD signals will not pass along the total length of the cable circuit due to the change in impedance at these points. It is worth noting that due to these complex earthing arrangements on higher-voltage installations, distributed monitoring enables easier analysis of data when compared with individual spot-check measurements.

Figure 8: 220 kV Cables in Tunnel Requiring On-line PD Measurement

Figure 9 shows the RFCTs installed onto the earthing cables linking the cable joints to the grounding boxes. Each sensor connects to a local node unit near each grounding box. The daisy-chained nodes communicate back to a hub server installed at the substation. The hub controls and monitors the system, collects data for analysis, and provides some automatic configurable alarms to provide early warning of degradation. Due to the distances involved between nodes, links are made using fiber-optic cable. Each local node obtains a phase reference source. 

Figure 9: RFCT Sensors on Earthing of 220 kV Cables in Tunnel

Despite these significant changes in hardware, the principle of the analysis remains effectively the same with detection of amplitude, phase patterns, and waveform analysis. The obvious benefit of early warning of unexpected degradation was realized soon after the monitoring system was commissioned on these circuits: Phase-resolved PD was detected on one of the cable joints (Figure 10).

Figure 10: Phase-Resolved Plot from Full-Time Monitor

Conclusion

The ability to monitor and effectively screen cable networks for PD activity on-line is available and shown to be effective. The work shows that effective screening of cables cannot simply rely on measurement of PD amplitude, as noise on the earthing system can cause a significant level of false positives. Determining whether PD is present and assessing its severity requires measurement of amplitude, analysis of the waveform, and phase-resolved analysis of the signals.

Automatic filtering and simple data capture to allow remote analysis can reduce the need for a specialist resource at site and make best use of this valuable time. Using the most appropriate resource to carry out the various elements of the task enables up-to-date condition data for cable networks to be quickly and cost-effectively gathered. This information allows better-informed asset management decisions to be made, may reduce or lessen the impact of failures, and improves network performance.

Finally, the article  provides an example where the criticality and assessment of risk on three parallel 220 kV cable circuits led to a decision to install a permanent partial discharge monitoring system and showed how immediate benefit was delivered through early detection of a defective joint.

Acknowledgements

The author would like to acknowledge Neil Davies of EA Technology Pty Ltd and Victor Chan of EA Technology Asset Management Pte Ltd for their contributions and encyclopedic knowledge of on-line PD testing.

References

Y. Wang. “Underground Cable Distribution System Management – Reduce Cable Failure Rate,” EEA Conference, Auckland, New Zealand, 2013.

S. Shamuri. “Enhancing TNB Medium Voltage Switchgear Condition Based Maintenance by Using Partial Discharge Application,” ISO 55000 and the Future of Asset Management, Asia Pacific Regional Conference, Bali, Indonesia, 2013.

P. Blackmore et al. “Quantifying the Economic Benefits of Online Monitoring,” Electricity Distribution Conference – South East Asia, KL, Malaysia, May 2014.

William G. Higinbotham has been president of EA Technology LLC since founding it in 2013. His responsibilities involve general management of the company, which is responsible for EA Technology activities in North and South America and the Caribbean/Bermuda. William is also responsible for sales, service, support, and training on partial discharge instruments and condition-based asset management. Prior to EA, William was Vice President of RFL Electronics Inc’s Research and Development Engineering group, where he led the development of numerous products in the areas of utility communications and system protection. Bill is an IEEE Senior Member and is active in the IEEE Power Systems Relaying Committee. He has authored or co-authored several industry paper, co-authored a number if IEEE standards in the field of power system protection and communications, and holds one patent in this area. Bill received his BS from Rutgers, The State University on NJ, School of Engineering in 1984 and worked in the biomedical engineering field for five years prior to joining RFL.

This work first appeared at the Doble International Conference of Doble Clients. Republication does not imply endorsement by Doble Engineering Company of any product or service other than its own. © 2016 Doble Engineering Company 83rd International Conference of Doble Clients. All Rights Reserved. Used with permission.