A Very Interesting Event

Here is an interesting event that happened a while ago. With the benefit of hindsight, I could probably write this up in such a way as to make it appear that the analysis of the event records was trivial and that when the e-mail came in asking what I thought had happened, I jumped straight to the correct answer. It doesn't work that way... Event analysis is often an iterative process. You look at the data, you form a hypothesis, predict what other things should have happened based on the hypothesis and then look for evidence of those other events to support your hypothesis. At some point, successive hypotheses converge and you stop and claim to know what happened. Other times, you're just left with a mystery that can annoy you for years.

This is the configuration of the system in question. Line A connects two sources and was tapped for a load interconnection. A single breaker substation was built near the tap and a separate line protection was supplied for Line B, so that faults on the tap line would not affect reliability of power transfers between Source 1 and Source 2. The existing permissive overreaching transfer trip scheme on Line A was modified to accept a blocking signal from Line B such that detection of a fault on Line B would prevent Line A from tripping, even if both ends of Line A saw the fault with their forward elements. This allows the overreaching scheme on Line A to remain in place with a minimal performance hit and without the expense of integrating a third terminal into the permissive scheme. All event records came from the distance relay shown below, which is protecting Line B. Currents are from bushing CTs on the source side and voltages are from the CVT also on the source side.

The weather that week was rather miserable, with heavy snowfall in the region. Fault rates on all lines in the region were much higher than normal. This was the first event record in the sequence.

The load at the end of Line B is delta connected on the high side, with no generation behind it. In cases like this, fault type identification is better done visually through the voltages, rather than the currents. C phase voltage is highly depressed, so we call this one a C-G fault. The currents are less than obvious, as they are a result of the loads responding to the now unbalanced phase to phase voltages across the delta primary of the transformer. The negative sequence reverse directional element 32QR correctly identifies the fault as being behind the relay (even though there are no sources within the load) and the current seen by the relay stops when the Line A protection clears the fault. Line B relay also receives a transfer trip from the Line A protection, but the load connected to Line B does not contribute to faults on Line A - the transfer trip is there solely for operational reasons.

Some time after the first event, but before the breaker was closed, a second event occurred. This time, it was a B-G fault in front of the relay (evidenced by high B phase current and the Z2G element picking up). The relay tripped via the Switch On To Fault (SOTF) protection, which enables sensitive elements for a short time period after the breaker is closed (and for the duration that the breaker is open). Approximately 8 cycles later, the breaker failure protection times out and the Line B protection keys a transfer trip to Line A protections to clear the fault.

At this point, from an analysis point of view, you come to a dead stop and stare at everything one more time to confirm that:

• The fault is in the forward direction (so in front of the breaker)
• The breaker indicates OPEN for the duration of the event
• There was no load current before the fault occurred

before you come to the conclusion that the breaker has faulted internally and the breaker needs to be repaired or replaced (neither of which is going to be fast or easy). While you are coming to this conclusion, there is another event, and this time you get higher resolution event records, which allows you see the non-60 Hz behaviour of the system. I had been working with a different client who had showed me an event where the fault type had evolved, but the difference in fault levels between the initial fault and the evolved fault was so high that you didn't see the original fault when the software autoranged the Y axis for you. I zoomed in on this record and was pleasantly surprised to see the fault attempting to ignite, but failing to do so for at least five times before successfully "taking off" on the next try. Some insights from this event record are:

• The fault occurs almost precisely at the phase to ground voltage peak, which is indicative of insulation failure.
• You can see the arc extinction occurring during the initial pre-arcing. There is some sort of high impedance path (maybe around 100 kohms) which does not create enough plasma to sustain the arc at the next zero crossing.
• The fault must be very close to the VT. Note that the faulted phase voltage is very nearly a square wave. Arcs tend to be constant voltage elements to a first approximation, and the relay voltage becomes more sinusoidal as it has to travel through more fixed impedances (like lines and transformers). In this case, we have about 4 kV arc voltage drop.

At this point, things are looking a bit grim, but staff are finally able to make it out to the substation (which is in a relatively remote area, and did I mention, the weather was terrible). The plan was to check the breaker for SF6 low pressure alarms (one reason for an internal flashover), verify that the mechanical semaphores for breaker position matched the indication from the auxiliary switches (perhaps the mechanism was jammed by ice), then isolate the breaker and try a couple of test operations before calling in the breaker experts from afar.

The staff on site were able to verify that no SF6 low pressure alarms had come in and the semaphores appeared to be in the right place, but they also sent this picture of the breaker (did I mention that it had been snowing?).

At this point, a reasonable hypothesis was that the fault was actually from the breaker terminal to ground through the snow, however, the oscillography clearly showed (twice, and on both the A and B relays!) that current had flowed through the protection CTs, which are at the bottom of the bushings, and the only way for this to have happened would have been for the breaker to have been closed (which it wasn't) and for the fault to have been on the load side of the breaker.

Luckily, while staring at this photo and talking to the field staff, I had an epiphany. I had a copy of the breaker manual and was able to determine that the design of this breaker was such that the CT covers at the bottom of the bushing actually connect to ground through the throat of the bushing, which sits INSIDE of the CTs, so a fault fully outside of the breaker can result in current through the CTs.

In the diagram above, the CT cover (I), bolts onto the bushing flange (F), which is at the end of the tank nozzle. The CTs (J) slide over the tank nozzle, so that a fault which ends on the CT cover will result in current which can be measured by the CTs at the bottom of that bushing (see the red line on the left side for the current path).

With that hypothesis in hand, we started looking for signs of arcing on the B phase source bushing, ideally on the CT cover. This had to wait until a whole lot of snow was removed first, but eventually we got some proof of the hypothesis.

With that, the mystery was solved, the protection engineer (me) withdraw his concerns about reenergizing the breaker, and the load customer got their utility feed back.