Sometimes it can be difficult to explain a power quality problem and arrive at an effective solution without the benefit of a little back-and-forth discussion, which is one of the few shortcomings of Ask the Experts. And it always helps to have multiple perspectives to get a clearer picture of what may be causing your system distress. With that in mind, we posed this month's question to both Ask The Experts veteran Mark McGranaghan, vice president of consulting services for EPRI-PEAC, and newcomer Tom Short, senior engineer, EPRI-PEAC, early enough to give them a chance to pose follow-up questions to the submitter.
Q. We recently installed a transient recorder on the 138kV bus at one of our substations. Since the installation, a number of events have occurred that I'm not sure I can explain, but I've included plots of a few taken by the recorder. One plot shows a typical event on a 3-phase, 138kV bus (Fig. 1 above). Another plot is a close-up of Phase A for the same event (Fig. 2 at right). My question to you is, What is this? I think it's a lightning hit somewhere on the system, but I'm not sure. This event is happening when we have bad weather in the area. If it is lightning, why is it always on all three of the phases? I realize that the transients do go through possibly several delta configurations, but why always on all three phases? I have recorded 18 of these events since May 2003. There are a couple other interesting facts. For example, the initial hit will increase the voltage on each phase from 200V to 500V for one cycle and then drop to the original voltage. We've also seen no feeder breaker operations or any lateral fuses blown in any of these events. Finally, on a plot of all three phases, there are a couple more disturbances at 72 milliseconds and 250 milliseconds past the original hit. Each is one cycle in duration. Is this a reflected wave, assuming of course that this is a lightning hit? At this time I'm not seeing any detrimental harm to the system. If it's not lightning, what is it?
McGranaghan's answer: The plots you provided appear to be rms plots vs. time, rather than actual waveforms. Any short transient change in the waveform will show up in a plot of rms vs. time as affecting a whole cycle, assuming that the rms values are calculated as the rms value of the preceding cycle at each calculation point. Sometimes instruments calculate the rms value based on the previous half cycle. However, this half-cycle method isn't recommended.
A lightning transient is very short in duration and won't really affect the rms value to the amount shown on these charts. However, there may be something associated with a limit on the input circuits to the instrument that make it show up this way. The sampling rate of the instrument will also be an important factor in the correct calculation of the rms voltage, including a lightning transient.
Lightning appears to be a good theory. The follow-on events could be follow-on strokes associated with the overall lightning surge, which is a common occurrence. They aren't reflections of traveling waves; those happen in tens and hundreds of microseconds, not tens of milliseconds.
The fact that all three phases are being affected leads me to think that these could be strokes to ground rather than strokes to the line and that we're seeing the effect of some kind of neutral shift that affects all three phases at the monitoring point. It could just be related to the way the monitoring instrument is connected and the location of the ground connection point of the monitoring equipment. You're correct that a lightning stroke to the phase would be more likely to affect only one phase.
It would be more helpful to look at the actual waveforms associated with these events than plots of the rms voltage vs. time.
Short's answer: It looks like lightning to me, mainly because of timing. Each “event” is separated by 50 milliseconds to 200 milliseconds, which matches the characteristics of lightning. Lightning is often composed of an initial stroke followed by one or more subsequent strokes that follow nearly the same channel. That's why lightning flashes sometimes appear to have a strobe effect.
Another concurring element is that the first event is larger than the subsequent events. This again matches lightning, because the first stroke is generally larger than subsequent strokes.
Lightning can raise the voltage on all three phases. If lightning hits the shield wire of a 138kV line, it will raise the voltage potential of ground and also increase the voltage of all of the phase conductors relative to the shield wire.
What's interesting about the recordings is the one-cycle, square-wave like wave shape on the rms traces. My guess is it has something to do with the transducers and/or how your recorder calculates rms. If it calculates rms with a one-cycle rolling window, you might get something like that. Do you have any of the “raw” waveform traces (i.e. the actually sampled voltages, not rms)? Recordings of current might also reveal something.
After McGranaghan and Short's request for more information, the reader sent in additional plots (Fig. 3, Fig. 4, and Fig. 5) for review.
McGranaghan's response: As noted earlier, lightning can raise the voltage on all three phases. Florida transmission lines designed at 138kV will have a shield wire above the phase conductors. Lightning usually hits the shield wire rather than one of the phase conductors — that's why it's there. The current then flows to ground through the impedance of the tower(s) and its ground connection impedance, resulting in a transient voltage that affects all three phases simultaneously. If this transient voltage is greater than the insulation strength of the insulators on that tower, then the insulator will flash over. This is known as a back flash. If that happens, a short circuit will occur on one or more phases and the event would look much different than the one provided in the question.
The question about the recordings was with respect to the one-cycle, square-wave-like wave shape on the rms traces. The clarification provided here explains this is due to the one-cycle sliding window used to calculate the rms values. A three-cycle sliding window results in a different characteristic. The effect on the rms voltage isn't very significant, but it's enough to show that a transient event occurred.