Since Friday’s blackout, there have been press reports that National Grid has presided over three “near-miss” situations where system frequency fell close to the lower frequency tolerance of 49.5 Hz, 1% below the target level of 50 Hz. National Grid responded to these reports, denying them:
“Contrary to media reports there have been no near misses nor early warning signs of incidents similar to last Friday’s power cuts. This was a highly unusual event, without precedent in the past ten years. We work with Ofgem to set the agreed limits of frequency of electricity output to keep the whole system safe and the lights on. All the time until Friday’s events, the system has remained within safe limits. We are conducting a thorough internal investigation and will report our interim findings in detail to Ofgem by the end of this week. We can and must learn lessons from Friday’s events – however rare their occurrence – as National Grid and as an energy industry.”
Analysis of frequency data
However, analysis of the frequency data from BM Reports from 1 May to 13 August found the following occasions when frequency fell close to the lower end of the tolerance band (in addition to last Friday):
It’s possible there were other instances as there are some missing data from the series, specifically 09:26:00 – 13:49:45 on 22 May, 19:58:00 – 20:31:45 on 31 May, 09:26:00 – 11:05:45 on 6 June and 10:48:00 – 10:55:45 on 9 July. There were also several periods where the recorded frequency was zero, which is clearly erroneous.
National Grid may not consider being 0.92% below 50 Hz a “near miss”, and that might not be an unreasonable position – it would be difficult and time-consuming to try to unravel the exact state of the system at each of these times in order to assess whether the frequency drops observed were close to triggering load-shedding protocols.
What my analysis did show was that these dips were unusual – there were over 600,000 data points in my spreadsheet but only 17 where the frequency was more than 0.75% below the 50 Hz level:
The question is whether these “near-misses” are signs of a growing trend as the proportion of renewable generation increases at the expense of traditional, primarily thermal generation. As I explained in my previous post, an increase in the amount of renewable generation impacts grid security in two ways – firstly by introducing more intermittency making it harder to balance supply and demand, and secondly by displacing sources of inertia which resisted frequency changes. Renewables currently do not carry out this function, although National Grid is exploring ways in which the role of renewable generation can be extended beyond the provision of energy.
Unravelling the blackout sequence of events
John Pettigrew, National Grid’s chief executive told Sky News that reports of “near misses” are “scaremongering” and that in his 28 years at the company there has only been one other time that two large generators had failed simultaneously. National Grid has also been denying that the wind levels were at all responsible for the outage, rejecting the link with high wind penetration last week.
However, there’s something off about the timings of these events. After publishing my post, I responded to some comments from readers and was prompted to look again at the sequence of events. It’s not straightforward to piece this all together because data are held in different places and may not always be precise.
Assuming the frequency data are precise, the system frequency began to fall between 15:52:30 and 15:52:45 (frequency data are published at 15 second intervals).
I looked at the REMIT notices to see when Little Barford and Hornsea went offline. RWE reported at 15:55:37 that it would go offline at 15:57:40…after the frequency has returned to normal levels. Orsted did not issue its REMIT reports for Hornsea until 19 and 26 minutes after the 16:00:00 published start of their outages, and although market transparency rules allow asset operators up to an hour to make their REMIT announcements, the start time is an unusually round number.
BM Reports also contains MEL (Maximum Export Limit) and FPN (Final Physical Notification) data – these are values provided to the system operator in respect of each Balancing Mechanism Unit (“BMU”) (in this case Little Barford, Hornsea unit 2 and Hornsea unit 3 are all BMUs), normally by the operator of the BMU indicating the amount the expect to export to the grid in each settlement period, and the maximum amount they could export to the grid. These data are to be provided before gate closure which is one hour before the start of the settlement period, and are manually updated by the trading teams meaning the are not always accurate if some unplanned event occurs.
The charts for Little Barford and Hornsea show the MEL falling at Little Barford while the FPNs are unchanged, and vice versa for Hornsea.
Understanding this picture is also not helped by the different time values used, with frequency and REMIT data being in GMT ie 1 hour before British Summer Time, and the BMU data being in settlement periods, where the first settlement period begins at 00:00 and are always in local time, ie currently in BST.
The table accompanying the BMU data on BM Reports has the corresponding times in GMT, and shows Little Barford’s MEL falling to zero at 15:57:00 – close to the time in the REMIT Report. The corresponding tables for Hornsea show the FPNs for both units 2 and 3 falling to zero at 18:00 GMT which does not match up with REMIT or any other data source.
Might high renewables penetration have been to blame after all?
The Little Barford data appear more credible to me – the REMIT notice was issued promptly and the BMU data are consistent with the REMIT data. However these show the plant coming off as frequency returns to normal levels after the drop. The Hornsea data on the other hand are not consistent and there was some delay in the publication of the REMIT notice (which still reports the events as being under investigation). While all the Hornsea data also show the plant coming off after the frequency recovers, these inconsistencies and delays make me more doubtful of the veracity of the outage timings.
There are 4 possible explanations of these inconsistencies:
The frequency data are wrong and the drop in frequency occurred a few minutes later than shown in BM Reports, after Little Barford and Hornsea both went offline. I find this difficult to believe as frequency monitoring is entirely the role of National Grid and does not rely on inputs from third parties. It should be correct.
Both Little Barford and Hornsea came offline earlier than reported, before 15:53, with their combined loss causing the drop in frequency. This would mean two different parties submitting inaccurate plant data and REMIT reports.
One of the two power stations went offline before 15:53 and on its own caused the frequency to drop. Little Barford is too small to have this effect by itself, and normally, the loss of 800 MW of Hornsea would also not be enough, however on a day where there is a high percentage of renewables running and relatively low amount of thermal plant on the system, it might be the case that the size of loss that would cause a major frequency drop would be smaller.
All of the data described here are correct and something else caused the drop in frequency. This is also unlikely, as it would mean some other large power station tripped, and someone would have reported this by now.
If option 3 is correct, ie Hornsea came off first and Little Barford only went offline after the frequency began to recover, this would mean the loss of 800 MW on a high renewables/low thermal day was enough to cause the frequency to drop outside operational tolerances. It is not unreasonable to suppose that the grid would become more sensitive to the loss of a generating unit when there is a high proportion of renewables on the system, and low levels of thermal generation which provides inertia.
Interestingly, analysis from LCP indicates that system inertia was well above the levels (130 GVAs) that National Grid considers to be the minimum required for system stability. In fact, there was a period overnight when inertia fell below the safe level, but this was not the case at the time of the blackout. LCP suggested that the problems were compounded by embedded generators tripping off once the frequency fell, which might explain the secondary drop in frequency. LCP hypothesises there was a secondary event, but does not suggest what that might have been.
“National Grid should hold enough operating reserve to cover the “largest in-feed loss”, which is predominantly driven by large generators and interconnectors. However, with Little Barford’s 664MW trip and Hornsea’s 756MW trip followed by a secondary event there wasn’t enough time to catch the combined drop in frequency to prevent demand disconnections,” – Kyle Martin, head of market insight at LCP
Historically, the grid protection measures have been sized based on the loss of the largest available units, eg the 1200 GW Sizewell B, but if the growing use of renewables is making the grid more sensitive, smaller losses might trigger the same load-shedding response. This would increase the risk of blackouts since there are many more generating units similar in size to the 800 MW lost at Hornsea on Friday, which underlines the need to ensure there is sufficient inertia on the system, alongside fast-acting frequency response services, able to react to such events.
Of course, this is just conjecture, because the official report into the blackout is not yet available, and what data are publicly available are inconsistent. It is to be hoped that the report provides a full explanation of the incident and the exact timing of events.
The sharp end of the energy transition
Earlier this year National Grid announced that it was ready to operate a 100% renewables system if enough renewable generation was connected. National Grid regularly and enthusiastically reports on the amount of zero-carbon generation running and recently started reporting on the carbon intensity of the electricity system. However, the security of supply obligation must not be neglected, and it is important that the system operator does not allow ideology to win out over physical realities. It is interesting that while trumpeting the high levels of zero carbon generation on Friday, National Grid did not draw attention to the fact that there was a coal plant also running during most of the day.
The speed of the energy transition has been unprecedented, and it would not be at all surprising if the rapid change in the fundamental shape of the grid did not present some difficult challenges around system stability. The analysis of frequency data between May and August this year, while a relatively short timespan (albeit 600,000 data points!) indicates that in fact National Grid is doing a good job of keeping things stable. It should not be afraid of being transparent about the problems on Friday, and if they are to do with the high levels of renewables and low system inertia, it should set out very clearly, the short and longer term options to avoid a repeat.
Greg Clarke recently announced the end of the trilemma – he was wrong – the trilemma is very much still with us, and National Grid is at the centre of it, trying to balance increasing amounts of renewable generation, without increasing its costs to unacceptable levels, whilst maintaining system stability. If Friday was a sign of the limits for renewable generation in the market as it currently stands, then the system operator needs to take a step back and focus on building the frameworks for running high levels of renewables safely even if that means curtailments and keeping more coal on.