Another Interesting Event

This event occurred about a decade ago in one of our 230 kV substations. Motor operated disconnect switch 2D3 was accidentally opened under load and during this event, 2B8 and 2MB3 bus protections tripped simultaneously. The two buses do not share any common breakers and there are no connections between the two protections.

The protections on these two buses consisted of electromechanical relays, so there was no oscillography available from the two protections, only flags from the bus relays which indicated a C-G fault on 2B8 and an A-G fault on 2MB3. With only that to go on, you start making up stories (you can call them hypotheses, if you want to look smarter).

• Was there a breaker fail protection operation? The two zones don't share a breaker, so a failure of either 2CB8 or 2CB9 to trip would only trip the adjacent bus zone.
• Was there insufficient restraint or bad diff relay wiring in one of the two protections? It's hard to tell without oscillography or an inspection of the wiring, but these protections had behaved themselves for at least 25 years before this event.

At this point, you start looking for something which will give you an overview of the entire event. Luckily, we had a DFR in the station, which while not monitoring any of the currents involved in the event, was connected to a VT connected to the 230 kV bus. This was the record (please forgive the completely non-standard colour coding here. Phases are A - B - C from top to bottom).

This record clearly showed the sequence of events that day. There was a C-G fault first (C phase voltage flattened, with minimal reduction of A and B phase voltages, then around 15 cycles later, there is an A-G fault. Both of these faults show the "square wave" voltage signature typical of a very close-in arcing fault (arcs are non-linear and tend to act like a bidirectional constant voltage drop).

Therefore, it appears that 2B8 faulted first, and after it cleared, 2MB3 then faulted, seemingly independently. However, we all know that these kind of things typically don't happen independently 20 cycles apart. To complete the story, we need some pictures of the actual layout.

When you can see the physical layout of the bus, the sequence of events becomes clearer. 2D3 is opened under load and an arc is generated, which rises up and also drifts sideways due to the wind at the time. At some point, the arc drifts into the steel dead-end structure, causing the 2B8 fault. After the fault clears, there is still a cloud of ionized air floating between the partially open disconnect and the steel structure. This cloud rises and eventually intercepts the rightmost (A) phase of 2MB3, causing it to also fault to the dead-end structure. So there you have it - a double contingency in a single event.

Lessons learned here were:

1. Voltage oscillography can be very valuable in helping to get the big picture of what happened in a complex event. When untangling a large event, long duration voltage oscillography can help you to generate a sequence of events and also help to provide time tags for oscillography that has inaccurate time tags.
2. The physical layout of the system is sometimes very important.
3. Hot air rises...

We had a similar situation during a design review of a series capacitor bank. The station designer had put a shield wire for the station directly over a row of disconnect switches. During the design review, this was pointed out and the shield wire was eventually moved. I did argue to leave it there on the basis that if the disconnect were to be opened under load, having a grounded object directly overhead would ensure that the arcing across the open contacts of the disconnect switch would stop relatively quickly when the line protection cleared the ensuing line fault.