Grid frequency drifting outside operational limits implies the system is running at the margins

Britain’s national energy system operator (“NESO”) has a statutory duty to maintain the frequency with which current and voltage varies on the GB grid at 50 Hz +/- 1%. It also has an operational range which it may only exceed for 1500 minutes per year of +/- 0.4%. I have been surprised to discover just how often the operational limits have been breached. And in December 2023, the statutory limit was breached with little fanfare.

Typically the market notices frequency excursions that result from major outages. So far this winter, there were major trips of the NSL and Viking interconnectors that sent frequency tumbling, and a near simultaneous outage of IFA2 and the Langage CCGT which had a similar impact. However, most excursions are not caused by trips but by a gradual upward or downward drift in grid frequency, which is arguably more concerning.

In this post I will look at frequency data and the regularity with which the operational limit is breached. I will look at whether there are trends in these data, and consider what this means for the power system in the context of the Government’s widely criticised clean power 2030 ambitions.

Grid frequency frequently breaches operational limits

Having noticed the number of sudden drops in frequency this winter I began to look more carefully at the data. I downloaded the 15-second frequency data from the BMRS portal for the past four years (which is all that’s available on the new version of BMRS). I also obtained the 1-second data from the NESO portal and then aggregated back to the 15-second level. The reason for expanding the time interval is that the 1-second data is unwieldy and in leap years, exceeds the number of rows in an Excel sheet! It took about 5 hours to gather data for one season, so I made full use of the 15-second data available through BMRS.

I should also mention that the data are very flaky, particularly more recently. There are a lot of data outages ie gaps in the reporting, and there are also multiple occurrences of zero values appearing. In order to smooth these out, I replace missing or zero values with either data the previous interval, or, if there are multiple zero or missing data points, I took the average of the previous and subsequent real values and populated the missing spaces with those. This is somewhat arbitrary, but Excel does not deal well with gaps in time series, and empty or spurious zero valued cells would skew the data, implying there were limit breaches when probably there were not. This means I essentially assume that at none of the times where data were missing or zero there was an actual limit breach, so the results tend towards being on the conservative side. This may or may not be legitimate, but even a few hours of missing data over a six-month period is not that material.

I then analysed the number of 15-second intervals in each winter season for which the operation or statutory limits were breached. The results are in the table below. As can be seen, in the past 4 years, there have been c 500 times the upper operational limit was breached, but in 2017 – 2019 the number was higher. The lower limit has tended to be breached less often – 2022 was an outlier with a very high number of breaches. There has also been a consistent trend of lower limit breaches increasing in frequency, which is consistent with falling grid inertia (see chart).

What is also interesting is that Winter 2024 is on track to be one of the worst years for operational limit breaches, despite the fact that so far this winter, grid inertia has actually been increasing.

This fits with the expected picture that the grid is becoming less reliable. It is also interesting that when individual occurrences are inspected, it is more often the case that frequency has drifted outside the operational range rather than suddenly falling out as would be expected from an outage. This is more worrying as it points to a general difficulty in maintaining stable frequency – things will always break and trip, and the grid is designed to deal with that, but these drifts outside the range speak more to a wider reduction in stability versus what is expected.

I looked at the size of frequency variations over 5-second intervals. A 0.3 Hz variation is considered significant in the control room – these are actually quite rare, 0.2 Hz moves are more common, but still not that typical – smaller moves are much more likely, as can be seen from the chart above. There is a small trend of increasing larger jumps, but the numbers are too low for this to be statistically significant. What is interesting is that in the absence of large jumps, the times that frequency moves outside operational limits is more typically due to drift rather than jumps that can be linked to the loss of large units of generation or interconnectors.

Statutory breach, December 2023

According to NESO, the reasons for this event were trips on IFA bipole 1 (1000 MW) and the Cottam CCGT (383 MW). There was a simultaneous, co-incidental outage on the Caithness-Moray HVDC link.

According to the National Electricity Transmission System Performance Report 2023 – 24 there were no reportable frequency excursions in 2023-24.When asked, NESO told me this is because The Electricity Safety, Quality and Continuity Regulations 2002 requires and frequency excursions 1% above or below 50Hz (ie outside 49.5 t0 50.5Hz) that last for 60 seconds or more to be reported. The excursion on 22 December 2023 “only” lasted for 59.3 seconds so just under the 1-minute level. However, when I read the Regulations, I see no mention of a 60 second grace period, and my request for further clarification from NESO has gone un-answered. If any of my readers can shed light on this I would be grateful.

.

Pulling all of this together, what can we conclude? I would say that the grid is significantly less robust than most people think, and that NESO is doing a surprisingly good job of maintaining grid frequency in the face of falling inertia. But I would also say that operating at the margins as we seem to be, is a recipe for disaster – a lot of things have to keep going right for the grid to be maintained in stable operation, and it seems that this is becoming harder over time. There is griwing concern in the market that things will not keep going right and that at some point, something will go wrong and the grid will fail, triggering blackouts. And as I have often warned in the past, blackouts in winter would likely result in fatalities due to road accidents and accidents in the home, particularly among the elderly.

This may well become a limiting factor for CP2030, as there are real questions about how grid stability can be maintained if the use of conventional generation is reduced so quickly. The deployment of things like synchronous condensers would have to be accelerated to support the grid, and it’s unclear that these plans are sufficiently well developed to guarantee grid performance in a low inertia environment.

If the power grid is operating at its margins today, this does not bode well for the accelerated transition the Government is pushing for, with potentially risky consequences.

.

Note: I wanted to use an image of conductor galloping for my blog image, but there weren’t any good ones, so I thought it would be a good opportunity to post this video from 1986. The quality isn’t great but it’s interesting to see the extent of wind impact on pylons and power lines which are not actually that high in absolute terms – GB itself being relatively flat. An early Christmas present for the energy nerds out there!