Grid infrastructure

A breaker loses SF6 for a fortnight before the next round finds it

IEC 61850 on the station bus, an RTU point list that is nothing but DNP3 indices, a SCADA mnemonic scheme cut to seven characters, and a PI historian at the regional centre.

132 kV BUS4 sectionsTRANSFORMERS6 bays, 1 running warmBREAKERS20, one losing SF6FEEDERS18 outgoingSTATION AUXbattery, cooling groupsIEC 61850 station bus · an RTU point list of bare DNP3 indicesstation SCADA mnemonics · a PI historian at the regional centre

One bay is running warm and one breaker is losing gas. Both are measured continuously, and both are normally found by a person on a round.

What actually goes wrong

A substation is quiet until it is not. The things that break are slow: gas leaking past a seal, a cooling group losing flow, an arrester ageing. All three are measured continuously and all three are usually found by a person on a round, at whatever interval the rounds happen to be.

The data problem here is sharper than in a works. An RTU point is AI:1043 and nothing else: no name, no unit, no equipment. An IEC 61850 tag has the meaning buried in a logical node. The station SCADA uses a mnemonic cut to seven characters. Three systems, three ideas of what a tag is, and the drawings that reconcile them are in a cabinet.

Indices are handed out in blocks, one block per bay, so a number that falls inside a block belongs to that bay. That is what an engineer does with the drawings on the table, and it is what the compiler does in a few milliseconds. On our substation it places the great majority and refuses the rest with a reason, which is the list you work through in week one.

What it looks like in the data

Drawn from the decoded values. Neither of these would trip an alarm limit on the day it started.

power_transformer-001 · temperatureiec61850:NO1TRBTRF001/STMP1.Tmp.mag.fday 1day 1461.45 to 77.72 degC

transformer oil temperature climbing through the second week

cooling_group-001 · volume_flowiec61850:NO1TRBCLG001/MFLW1.Flw.mag.fday 1day 14147.7 to 195.2 m3/h

oil flow through the cooler falling away

How far we have got

the label on this story

modelled

0.901 (0.894 to 0.907) of tags resolved on our substation over fifty unseen plants, with the core carrying no knowledge of substations at all: that sector is 121 lines of connector.

read next

The numbers, including the ones that missed

Sector two is scored in two industries, one core, where the compiler carried no knowledge of substations at all.