BY JASON SOUCHAK, Megger

Transcription

1 IMPROVING SYSTEM RELIABILITY WITH OFFLINE PD TESTING BY JASON SOUCHAK, Megger Partial discharge (PD) testing is a new generation of diagnostic testing for medium- and high-voltage underground power cable systems, offering a very powerful diagnostic technique. Partial discharge testing can identify problem spots in cables long before they cause an in-service outage or, if used during commissioning testing, can detect problems before the cable is even placed into service. Ideally, the cable should be tested as if it were in service, at 50 or 60 Hz. However, the power requirements to charge a cable at this frequency make this kind of field testing equipment large, expensive, and impractical. Because of this, cable testing is usually done at very low frequency (VLF). VLF tests are most commonly performed at 0.1 Hz but this 0.1 Hz voltage source is not in the same frequency range as 50 or 60 Hz, so will the results of PD testing be the same, or what kind of differences can be expected? ONLINE PARTIAL DISCHARGE Online PD can be an effective tool as well. If used properly, it can give advance warning prior to a cable failing in service. However, it has the limitation 58 SUMMER 2017

2 that it can only be used at operating voltage. Many times an outage is still required to install the highfrequency current transformer (HFCT) sensors before the system can be monitored online. To get accurate location information on where the PD is occurring, additional sensors will need to be installed and their measurements synched so that recorded pulses can be accurately compared from multiple locations. This contributes to the cost and complexity of an online PD measurement system. Online PD can either be continuously monitored or intermittently scanned. In the case of an installation that is intermittently scanned, think of scan results as a photograph: The results capture an instant in time and, in this instant, the cable can be either PD free or show signs of PD. No further analysis is possible. In the case of an installation that is continuously monitored with online PD measurements, the amount of useful information also increases, but so does the cost and complexity. This method retains the photograph quality of intermittent scans in other words, the results can indicate if the cable has PD or if it is currently PD free but it also gains the ability to provide an alert if the cable develops PD at any time at operating voltage. As an example, think of online PD as an alarm. Imagine a house, built near a river. When heavy rains come, the house could flood, so the occupants must leave for their own safety. The continuously monitoring PD method is like installing a flood alarm. This alarm goes off once the water reaches a certain height and, if the occupants were unaware of the approaching rains, they would have enough time after that alarm goes off to safely evacuate the house. The intermittent scan method of PD would be like polling this alarm only once a day. If the water is already at a dangerous level, it will provide an alert, but it has no way of indicating how fast the water would rise if it wasn t already a problem. This flood alarm for both the intermittent scan and continuous scan examples provides no advanced warning the occupants or the cable owners must react immediately. OFFLINE PARTIAL DISCHARGE Offline PD solves many of online PD s shortcomings, but requires the cable to be removed from service during a planned outage. With offline PD testing, a separate power source is connected to the cable. This allows the PD test to be performed at any desired voltage, allowing for diagnostic test results that indicate the quality of the cable before it fails. By varying the voltage from about 0.5 times rated voltage to 1.7 times rated voltage for aged cables or up to 2 times rated voltage for new cables, it is possible to find out at what voltage PD begins. If PD events don t occur until well above operating voltage (say 1.5 or 1.7 times operating voltage), the cable should be monitored, but it is not necessarily an immediate concern. However, if PD events begin much closer to operating voltage (say 1.1 or 1.2 times operating voltage), then that cable is of greater concern and should be monitored more frequently. Cables with a partial discharge inception voltage (PDIV) below operating voltage would be of immediate concern and repaired or replaced as necessary. Conversely, cables that do not exhibit PD up to maximum voltage can have a much longer testing interval. In the case of partial discharge starting at 1.1 or 1.2 times rated voltage, which is not uncommon, none of the online PD measurements would give an alert to the critical state of this cable. The cable would remain in service without any indication of the PD occurring in it until the PD became severe enough to occur at operating voltage. Once PD begins, it is a continuously degrading process; if the PD is occurring at operating voltage, that cable may only have a short time until failure. Imagine that same house on the same river. It may still have the flood alarm installed, but now the occupants watch the weather report. The report says that a hurricane is going to reach the area in three days. The occupants know that amount of rain will force them to evacuate, so they can be prepared when it becomes necessary to leave. In this case, the weather report is like an offline PD diagnostic test. The occupants (cable owners) received notice of incoming danger before it was critical and took planned action instead of taking unplanned emergency action. NETAWORLD 59

3 OFFLINE PD POWER REQUIREMENTS Ideally, cables would be tested at 50/60 Hz to simulate what occurs in service. However, to do this as an offline test would require very large power sources. For example, a 1,000-foot 15 kv class cable tested per IEEE Table 3 specified maintenance test value of 16 kv would require 9,650 VA, nearly 10 kw capacitive charging power. And a 10,000-foot 15 kv class cable tested at the acceptance test value of 21 kv would require 166,250 VA, over 166 kw capacitive charging power: Power = 2 * π * Frequency * Cable Capacitance * Voltage 2, assuming nominal value of 100 pf/ft for capacitance. Since the power requirements are too large for practical field portable test equipment, what are the other options? DC testing used to be the go-to standard for cable testing. However, it has long since fallen out of favor for extruded cables. Based on the EPRI report V2, published in December 1995, and many others, it was found that dc testing can induce faults onto cables that they would not have otherwise seen due to trapped space charges. Additionally, studies have shown that dc is blind to gross defects in the cable insulation. As a result, Figure 1: Scale Diagram of Voltage Waveforms: One Cycle of VLF Sine, One Cycle of VLF CR, One Shot of DAC, and 600 Cycles of Power Frequency no IEEE standard exists for dc testing extruded power cables. So, if dc isn t a good choice and power frequency testing takes too much power, what other options are available? The problem with power frequency is that it takes too much power. If it was possible to lower the power requirements, then ac would be the best choice. The formula for calculating the power requirements is shown in the equation. In the equation, the numeral 2 and π are constants and cannot be changed. The cable capacitance is a function of the cable construction and length and cannot be changed at will. The voltage is dictated by the specific testing standard. This leaves only frequency available to be changed. If the frequency is lowered from 60 Hz to 0.1 Hz, there is a huge power savings. In fact, 0.1 Hz testing uses 600 times less power than the equivalent power frequency test. Those same cables that used 10 kw and 166 kw at power frequency would only need 16 W and 277 W, respectively, when tested at 0.1 Hz (very low frequency or VLF). A field test set using 300 W is a reasonable amount of power and is state of the art in cable testing today. OFFLINE PD WAVE SHAPES The term VLF covers several waveforms. In cable testing, these are most often one or more of the following: VLF sinusoidal, VLF cosine rectangular (CR), or damped ac (DAC). VLF sinusoidal and VLF cosine rectangular are continuous waves, while damped ac consists of discreet pulses that may have significant time between pulses. Figure 1 shows a visual comparison of these wave shapes. VLF Sinusoidal VLF sinusoidal is the easiest to understand. It is simply the same wave shape as power frequency, but it is slowed down to 0.1z. This wave is recognized by IEEE for VLF withstand testing, and test values are conveniently given in RMS voltage and peak voltage in standard Table 3. VLF sine is a very slow-changing wave, having a polarity crossover (from positive to negative, or vice versa) of 5,000 ms or 5 seconds. When zoomed in close to the zero crossing, as shown in Figure 2, the VLF sine wave looks almost flat with a very slow rate of voltage change. 60 SUMMER 2017

4 VLF Cosine Rectangular VLF cosine rectangular is more difficult to describe. It looks similar to a square wave, but when zoomed in close (Figure 2), it clearly is not a simple square wave. It has a 5-second dc hold period, followed by a sinusoidal transition, followed by a 5-second dc hold in the opposite polarity, and a sinusoidal transition back to the original polarity. These very short dc hold periods do not damage the cable as continuous dc testing does, since the duration in each polarity is kept short. The transition period is what makes this wave special regarding PD testing. Using the capacitance in the cable and a fixed inductor in the equipment, a resonance circuit is set up. When it is time to switch polarities, the circuit resonates one-half cycle, stopped by a diode. This results in a polarity reversal in the range of 16 ms to 1.6 ms (corresponding to a frequency range or 30 to 300 Hz, respectively). This is very similar to the polarity crossover of 60 Hz at 8.3 ms, or 50 Hz at 10 ms, and much closer to power frequency than VLF sine at 5,000 ms, which is approximately 1,000 times slower. DAC Similar to cosine rectangular, DAC sets up a resonance circuit between the capacitance of the cable and a fixed inductor in the test equipment (see circuit diagram, Figure 3). However, instead of limiting the resonance to a single half cycle, DAC allows the voltage to exponentially decay through the resistive losses in the circuit. This means that DAC has a frequency in the power frequency range of 30 to 300 Hz, and voltage is only on the cable for a very short period of time, oscillating for only a few hundred milliseconds. This makes DAC an ideal voltage for PD diagnostics. Because voltage is on the cable for such a short period, this waveform is very gentle on the cable and is very unlikely (although not impossible) to cause a weak spot in the cable to fail during testing. IEEE has recognized this waveform as viable for PD testing in standard 400.3, specifically clause Damped ac is also recognized by IEEE as appropriate for withstand testing, and commericaly available options up to 300 kv exist. Figure 2: Zoomed-in View of Figure 1 Showing Three Cycles of Power Frequency and the Different Polarity Reversal Times of Various VLF Waveforms (Note: DAC and CR have approximately the same polarity crossover speed as power frequency.) Figure 3: Diagram of Circuit Used to Create DAC Waveform (Note: Cosine rectangular circuit is similar, but includes a blocking diode to allow only a half cycle of resonance.) Why is all this important? The initiation and the amount of PD measured in a cable are related to the dv/dt or the rate of change in voltage over time. The faster the voltage changes (the higher the frequency), the more PD will be measured and the lower the PD inception voltage will be. Power frequency changes from one peak of voltage to the other peak of voltage in 8.3 ms or 10 ms (60 or 50 Hz), so the testing should be done to mimic this rate of voltage change. Both cosine rectangular and damped ac change polarity in this same few millisecond range, from 5 times faster to 2 times slower than power frequency, on the extreme ends of the scale, for very short or very long cables. However, VLF sinusoidal changes polarity 600 NETAWORLD 61

5 times slower than 60 Hz power frequency does. Due to this much slower polarity change, PD test results collected with a VLF sinusoidal wave may not represent what is occurring in the cable under operating conditions. As seen in the following case study, VLF sinusoidal did not detect the presence of PD on the splice at 750 meters at all, but DAC and cosine rectangular were both able to easily identify it. PD TESTNG RESULTS According to IEEE 400.3, clause 7.4: In summary, it is not possible to standardize a specific test protocol at the current time for either online or off-line tests. This may become possible as more data are collected. This standard is unable to provide limits to what would be considered good or bad cables, due to the wide variety of cables and defects encountered in PD testing. However, through experience, certain parameters are clear indicators of problems on the cable system. These parameters include intensity, position, and PD inception voltage. It is obvious that when numerous events occur at a specific location, that location is weaker than the surrounding material. The intensity is a measure of the strength of the discharge. Again, it is obvious that a stronger discharge indicates a more severe condition. Finally, the PD inception voltage (PDIV) is the minimum voltage needed to trigger partial discharge events. This is a critical measurement, as a cable that exhibits PD at or below operating voltage is much more critical than one that only exhibits PD at 1.5 or 1.7 times operating voltage. In the first case, the cable will experience PD continuously until finally failing in service or during a testing period. In the second case, the cable will operate without PD during day-to-day operation and only experience PD events when exposed to a high voltage, such as a switching transient or a lightning strike. This all may sound intimidating, but cables in very good condition are easy to identify, as are cables in very bad condition. It is the cables in between that require additional analysis, and experience quickly makes this simple as well. Some PD threshold levels exist from research and field use to help guide users of PD equipment, even though no standard has been adopted. CASE STUDY In the following example, a single section of cable was tested with each of the three voltage waveforms discussed previously. The tests were performed in this order: DAC, VLF sine, and VLF CR. All three tests were performed within about an hour. The cable sample as shown in Figure 4 was a service-aged mix of XLPE and oil impregnated paper insulation rated at 12/20 kv, and had a total length of 1,335 meters. It had 10 splices, including four transition splices between XLPE and PILC cable. The results are presented in the order the tests were performed. For simplicity, only the results from phase L1 are shown, but the tests were performed on all three phases with similar results across all phases. A summary of all three phases is in Figure 8. DAC Results The DAC results show two clear weak points in the cable. The splices at 750 m and at 1,000 m both display large amounts of PD activity. These also correspond to transition splices in the cable system, from XLPE to PILC cable. The DAC results show up to 1,500 picocoulombs (pc) of PD activity on the splice at 750 m and up to 2,000 pc of PD activity on the splice at 1,000 m. The PD activity shown in Figure 5 is up to Figure 4: Overview of Test Cable (Note: Cable is a total of 1,335 meters and contains 10 splices in 11 sections. Green sections are XLPE cable, and blue sections are PILC cable. Splices are represented by small black segments.) 62 SUMMER 2017

6 Figure 5: Case Study DAC Results Showing PD Measurements Up to 1.7 Times Rated Voltage Figure 6: Case Study Sinusoidal Results Showing PD Measurements Up to 1.7 Times Rated Voltage a voltage of 1.7 times operating voltage, but the PD activity started at a much lower voltage. In this case (Figure 5), the PDIV was 0.4 times operating voltage for the weak spot at 750 m and 0.6 times operating voltage for the weak spot at 1,000 m. Both are well below operating voltage and are cause for significant concern. VLF Sinusoidal VLF sinusoidal results as shown in Figure 6 indicated only one of the weak points detected by DAC. The splice at 1,000 m is identified again as a site of potential weakness, but it was unable to detect the weak spot at 750 m. Additionally, the weak spot identified at 1,000 m shows a PD activity of only 250 pc at 1.7 times rated voltage. This is a significantly different result from the DAC test. Although 250 pc in a splice may not be critical, it should be monitored, while the 2,000 pc detected during the DAC test is a critical level. Furthermore, the PDIV in this sample was 1.3 times operating voltage, meaning that this result is far less critical than the 0.6 times operating voltage result observed in the DAC test. NETAWORLD 63

7 Figure 7: Case Study VLF CR Results Showing PD Measurements Up to 1.7 Times Rated Voltage Cosine Rectangular The cosine rectangular results were very similar to the DAC results. This test (Figure 7) located the weak spots at 750 m and at 1,000 m, and identified a PD activity of about 2,000 pc at the 750 m point and about 4,000 pc at the 1,000 m point. While these are higher than the PD activity recorded by the DAC wave typically due to the continuing voltage application during the voltage hold phase the ultimate course of action for both tests would be the same. This cable shows a critical amount of PD in both these points for the CR and the DAC tests. The PDIV values for the CR test were approximately 0.5 times operating voltage for both weak spots detected. This test also revealed an additional weak spot. At approximately 1,050 meters, another PD weak spot emerges. This is different from the other points observed, as it is in the cable insulation itself and not in a splice. This is not unusual to see in an oil-impregnated insulation system and could indicate that the oil has leaked out at that point creating a dry spot, or possibly that a small rise in cable elevation resulted in a similar dry spot. At about 500 pc, this would not be an immediate concern for laminated insulation, but should be monitored closely. The results from this case study demonstrate the frequency-dependent nature of PD testing. In this example, the cable tested with VLF sine would be considered acceptable since the PDIV was above operating voltage, and the level of PD recorded was low. This cable would be placed back into service and would likely soon suffer an unplanned outage. However, by using waveforms that more closely simulate power frequency, the true health of the cable could be assessed. Both damped ac and cosine rectangular identified a weak spot at 750 m that VLF sine was not able to detect, and they detected higher levels of PD activity. The result for both VLF CR and DAC would mean that the cable would not be placed back into service until the splices were repaired or replaced. VLF CR was also able to identify a weak spot in the laminated insulation. At the time of the test, this weak spot did not exhibit enough PD to be of immediate concern, but it should be monitored closely. The summary in Figure 8 shows a side-by side comparison of the voltage waveforms. The results for DAC and CR are similar, easily locating the weak spots at 750 m and at 1,000 m at both operating voltage and at 1.7 times operating voltage. Both the DAC and CR waveforms had a resonant frequency of 280 Hz, similar in dv/ 64 SUMMER 2017

8 dt to 50/60 Hz and much faster than the 0.1 Hz from the VLF sine wave. The VLF sinusoidal waveform was just able to detect the weak spot at 750 m, with an intensity of less than 80 pc, compared to greater than 1,000 pc for DAC and 1,500 pc for CR. It is entirely possible that in an electromagnetically noisy environment, this amount of partial discharge may not be detectable because it is obscured by background noise. But the 1,000 pc level from DAC would be enough to detect above the background noise. Typical field background noise levels fall in the 20 to 150 pc range. Looking at Figure 8, it is obvious that DAC and CR have similar results, including a much more detectable PD compared to sinusoidal voltage. Even though DAC and CR detected more PD, it does not mean they were any more stressful to the cable. These tests were performed in the order DAC, VLF sinusoidal, and cosine rectangular. The sine wave was in between the other two wave forms, but did not produce the same results. CONCLUSION It is difficult to overstate the potential power of offline partial discharge diagnostic testing. From IEEE 400.3, clause 1.3: The PD measurement can, at times, predict with a high level of confidence that a given cable is in very poor condition and is likely to fail in the near future. This knowledge can then be used to improve system reliability, reduce unplanned outages, reduce emergency call-outs, and help target the most vulnerable cables for replacement while allowing good cables to remain in service. However, for accurate results, it is necessary to use a voltage source with characteristics similar to those experienced by the cable in operation, particularly the voltage rate of change from one polarity to the other. Practical, field-portable test sets are available that achieve results comparable to in-service conditions, allowing for the best and most accurate cable maintenance strategy. Figure 8: Summary of Test Results DAC (Top), VLF Sine (Middle), and CR (Bottom) Waveforms at Operating Voltage (Left Column) and 1.7 Times Operating Voltage (Right Column) Jason Souchak is an Application Engineer for Megger, specializing in cable testing and diagnostic and fault location techniques. He trains end users on the theory and proper use of various pieces of high-voltage testing equipment. Jason is an active member of the IEEE Insulated Conductors Committee (ICC), helping to develop the latest cable testing standards. He graduated from Drexel University in Philadelphia with a B.S. in Electrical Engineering (BSEE), specializing in the power track. He has worked with protective relay manufacturers, utilities, and development and certification testing labs in the United States and in Europe. REFERENCES R. Bach, P. Craatz, W.Kalkner, K. Hrefter, H. Oldenhoff, G. Ritter. Voltage tests to assess medium voltage cable systems, Elektrizitaetswirtschaft 92.17/18 (1993): NETAWORLD 65