In service high-voltage (HV) substation equipment is exposed to many stresses, from the electrical, mechanical and thermal to the environmental. These stresses can act to accelerate the deterioration of the insulation and the electrical integrity of the HV equipment eventually leading to failure. Partial discharge (PD) is both a symptom and a cause of insulation deterioration, so the detection and measurement of PD phenomena can provide early warning signs of insulation failure.
Critical to this detection is the availability of accurate and cost effective surveillance tools, which, if non-invasive, can provide early recognition and location of possible sites of electrical degradation while components are still in service. Gathering and trending PD activity over time is essential to monitor the rate and severity of degradation. Maintenance can then be planned in an effort to avoid unplanned outages, interruptions and inevitable loss of revenue.
The use of radio frequency interference (RFI) measurement is an efficient, non-invasive surveillance technique to detect and locate partial discharges in individual HV apparatus. This article will look at the benefits of combining the assessment of RFI emissions with the targeted deployment of complementary, non-invasive electromagnetic interference (EMI) detection techniques. Specifically, frequency sweep data and time-resolved traces can be compared with follow up assessments using complementary EMI couplers such as high frequency CTs (HFCT) and transient earth voltage (TEV) couplers. This combination of tests provides an increased level of confidence in the location, identification and assessment of the severity of degradation and is beneficial when dealing with complex HV apparatus.
The detection and measurement of RFI emissions from PD phenomena involves the measurement of complex waveforms varying considerably and often erratically in amplitude and time. RFI signals from such phenomena are considered to be broadband and impulsive in nature with low repetition rates.
Measurements carried out on PD activity within oil-insulated HV equipment demonstrate that the discharges produce current pulses with rise times less than a nanosecond and therefore capable of exciting broadband signals in the VHF (30 to 300MHz) and UHF (300MHz to 3GHz) bands. Other investigations in open-air insulation substations show that signals from PD and flashover occupy a frequency range up to 300MHz.
When PD occurs inside a metal enclosure, such as in a transformer tank, the signal propagates within the structure, suffering frequency attenuation, reflection, etc. Detection of RFI emission relies on the placement of apertures in the tank walls and penetrating conductors to allow the RFI emissions to propagate and radiate externally.
In the following examples, the instrument used (Doble PDS100) has two different detection modes: spectrum analyser and time-resolved mode. Within spectrum analyser there are three separate detection techniques: peak detection, average detection and separated peak and average detection.
Case Study 1
Dissolved Gas Analysis (DGA) carried out on a South African, 275/88/11KV, 250MVA transformer showed signs of a discharge type fault. RFI measurements (using the Doble PDS100) and conducted EMI measurements (using a HFCT) were performed to establish correlation between the measurements. RFI measurements were taken around the periphery of the transformer. The frequency traces (FIGURE 2) exhibit a discrete appearance as pulses are accumulated. Short bursts of pulse accumulation were interspersed with long intervals of no or low energy activity. Triangulation based on signal intensity of the higher frequencies locates the source of propagation in the vicinity of the HV B-phase. Observing the RFI at 900MHz for a period of time in spot frequency mode sees the measured peak amplitude reaching -45dBm at that location. This mode also confirms the burst nature of the pulse sequence.
The conducted EMI was measured using a 300MHz split-core HFCT at the HV neutral connection to earth. The measured conducted EMI is subjected to significant attenuation through the HV neutral connection and requires an extended observation time. In time-resolved mode both the RFI and conducted EMI measurement confirm the measured pulse behaviour. However, the pulsed activity is more easily captured and more of the lower energy pulses are detected.
Partial discharge activity is indicated by both the RFI and EMI techniques. In each, the dynamic behaviour of the activity is characterised by very short burst activity interspersed by intervals of no or low energy activity. The sequence exhibits the characteristics of a floating type discharge. A secondary source of discharge is evident in the time-resolved traces. The results confirm the conclusions drawn from the DGA analysis. This study proves the use of RFI as an assessment tool while the use of an HFCT coupler provides increased sensitivity to internal PD activity, offering an increased level of confidence in the identification and assessment of PD activity.
Case Study 2
At a distribution substation, RFI measurements were undertaken to survey the condition of each of the oil-filled circuit breakers making up a typical 11kV distribution switchboard configuration commonly found in the UK. A high percentage of 11/33kV switchboards have an installed age of over 25 years. They are subjected to various types of duty plus a varied level of maintenance. The trend is to extend the maintenance period for medium-voltage (MV) switchgear, which in turn creates the need for interim non-intrusive condition monitoring techniques to offer confidence in the equipment’s safety and reliability.
A baseline RFI scan was captured in an adjacent room away from the surveyed switchboard. Measurements at the rear of each circuit breaker were captured and compared with the baseline. The observed uplift of frequencies indicated a nearby discharge source, which was eventually triangulated to one particular circuit breaker by comparing the uplift in higher frequencies while moving the receiving antenna along the rear of the switchboard. Further RFI measurements were captured at the front of the switchboard. A comparison of the front and rear RFI measurements shows that the uplift in the lower frequencies was strongest to the front of the unit. These tests were followed by complementary EMI measurements to gather more information.
RFI Peak Measurements: Front and Rear of Circuit Breaker
Legend: Front, Rear
The HFCT uses inductive coupling to detect PD pulses flowing to earth and is capable of picking up both local PD in the cable end box and also the lower frequency PD pulses coming from down the cable. The results of this method confirmed the observations from the RFI survey, with uplifts of up to approximately 50dB at 75MHz and 40dB at 200MHz. Time-resolved measurements also showed pulse behaviour is similar to those obtained from RFI measurements (FIGURE 6).
Lower Frequency ( circa 50MHz)
Legend: RFI, HFCT, TEV
Mid Frequency (circa 150MHz)
Legend: RFI, HFCT, TEV
Higher Frequency (circa 200MHz)
Legend: RFI, HFCT, TEV
The placement of HFCTs provides a means to trace the likely source of the signals by comparing the uplift in frequencies. The uplift reduces significantly as the location of the HFCT is moved away from the suspect circuit breaker. Repeated measurements on earth straps placed on adjacent circuit breakers indicate the circuit breaker identified is the source of the measured discharge activity.
The most advantageous setup for metal-clad switchgear is to use an HFCT sensor in conjunction with a TEV sensor. Transient Earth Voltage (TEV) measurements work on the principle that if a PD occurs within metal clad switchgear, electromagnetic waves escape through openings in the metal casing. The electromagnetic wave propagates along the outside of the casing generating a transient earth voltage on the metal surface. TEV sensors are “capacitive couplers”, which when placed on the surface of metal cladding can detect TEV pulses as a result of PD internal to the switchgear.
Observed peak TEV measurements on the main circuit breaker tank reach 0dBm at a frequency of 100MHz. Comparative measurements taken with the TEV sensor located on the cable end box show a reduction in uplift of approximately 20dB. The main circuit breaker tank is identified as the likely source of the discharge. Time-resolved measurements show pulse behaviour confirming the results obtained from both RFI and HFCT measurements.
The utility opened up the circuit breaker and found signs of carbon at the cable end in the main tank of the switchgear. Results of this study confirm that deploying frequency spectra measurements and time-resolved patterns from RFI, HFCT and TEV probes can be used to pinpoint PD issues within switchgear. Using TEV sensors in conjunction with RFI surveillance on metal clad switch gear offers an additional capability in confirming and localising the partial discharge source.
RFI monitoring offers, and has proven to be, a routine non-invasive and cost-effective surveillance technique that complements and provides added value to other well established HV asset monitoring techniques such as thermal imaging and DGA analysis. As the practical examples illustrate, measurements logged with an RFI instrument platform specifically designed for substation surveillance can assist in effective discrimination and recognition of the RFI emissions radiated from potential sites of insulation deterioration.
There are great benefits of combining the assessment of RFI emissions with targeted deployment of other complementary non-invasive electromagnetic interference (EMI) detection techniques using the same RFI instrument platform. The deployment of both ‘far field’ and ‘near field’ probes provide a diversity of sensors, which is of particular importance with complex HV apparatus such as transformers where the propagation paths for RFI are less well defined.
This article is based on the paper Substation Surveillance Using RFI and Complementary EMI Detection Techniques, which was recently presented at the 78th International Conference of Doble Clients in Boston, Massachusetts USA. The paper was written by Alan Nesbitt, Brian Stewart and Scott McMeekin of Glasgow Caledonian University and Kjetil Liebech-Lien and Hans Ove Kristiansen of Doble Engineering Company.
Part 2 of this article will appear in the June 2011 issue of Electrical Review.