Power outage strands trains

[whole diary]

[stuving]

The frequency disruption was for a relatively long duration, some put this down to the lack of inertia on the National Grid due to the loss of so much heavy rotating generating plant

There is a figure for "inertia" in those reports - 200GVAs, which I suspect is about the usual level at the moment. A useful number to tuck away in your toolbox (ideally with a proper definition). NG^▸ (Natural Gas) do have research going on on the use of batteries to provide "synthetic inertia", but I could see no mention of that in the reports. They do talk about campaign to reprogram small (embedded) generators to trip on a RoCoF (rate of change of frequency) of 1 Hz/s rather than the current value of 0.125 Hz/z. That would have kept a lot more of that generation on line. They say that will take three years, but some of the other evolutionary changes do seem to be going rather slowly.

On more point that I don't think was addressed directly was the cliff-edge effect of generators tripping due to RoCoF. The 1st July event (described in the main report) saw 1,000 Belgian MW lost, and the system coped - just.  The 9th August event was initially not much larger, but was just big enough that the reserve could not cope and, for want of enough power, the frequency fell. That led to enough extra generation disconnecting itself that the power shortfall suddenly became very large. Changing the RoCoF limit does address this, of course - eventually. 

[stuving]

The Times has pointed out that the makers of the turbines farmed at Hornsea - who presumably supplied the control software - were Siemens Gamesa (59% owned by Siemens). They didn't labour the point, but ... Before anyone starts on about "German software", Gamesa's engineering centres are in Denmark, Spain, and India.

Incidentally, the Hornsea system is new to the grid - the three modules went live on 1/2/19, 30/4/19, and 15/7/19 - so a certain amount of optimisation based on observation is to be expected. In which case, why had they not picked up the tendency for oscillatory power flows (as in the example ten minutes before the outage) already?

[stuving]

You may have seen that Ofgem have issued their final report on this event, and announced some voluntary fines to be paid by the operators of the two generators that disconnected and one DNO^▸ (Distribution Network Operator) (UKPN, for reconnecting before being asked to). They seem uncertain whether they have a clear right to impose such fines, but a bit of arm-twisting of the emabarrassed has had the same effect.

Their analysis of what happened changes NGET's original story very little. However, some more poor performance has shown up in the operation of the low-frequency response process (LFDD). Not only did less than 5% of the load get shed, but two DNOs reconnected some load before NG^▸ (Natural Gas) told them to. LFDD disconnected some of the generators contracted to supply reserve power too, which was unintended. And the total of distributed generation lost in the event is now estimated at 1300-1500 MW (and the exact figure isn't knowable).

ORR^» (Office of Rail and Road, formerly Office of Rail Regulation - about) have also published their report on ... well Siemens, mainly. They do document all the minor loss of power incidents too, none of which was serious.

The background at Siemens was much as reported earlier; a concern that some genuine failures produce a trip that could be reset and cause more damage led to some trip conditions being moved across to the list causing permanent lock-out. Low frequency was one of those, and should not have been. ORR's explantion is that, such events being rare, the risks of doing this were not considered - potentially the whole fleet could shut down out on the network and need a technician to attend.

Siemens admit this. A couple of quotes:

Most permanent lock-outs are triggered by events relating to the train itself, which are unlikely to arise simultaneously on multiple trains. Variations in the power supply frequency, however, affect many trains at the same time and result in the same response from all trains that have the same software. It appears therefore that the collective response of the Class 700 and 717 trains to the out-of-specification supply frequency was in accordance with the software design, but was not an explicit intention. Siemens accepts that the temporary reduction in frequency should not have been considered a situation that requires a permanent lock-out.

In an interim email to ORR, Siemens stated, “The original design of the class 700 is for the 4QCs to stop pulsing when the line frequency is out of range and automatically restart when the frequency come back in range. This does not require driver action. With the hysteresis implemented, the 4QCs lock when line frequency drops below 49Hz and they restart when the frequency rises above 49.5Hz.”