[[{“value”:”
Yesterday demonstrated some of the challenges of high wind generation on the GB system, with lots of activity in the Balancing Mechanism, particularly to curtail wind and CCGTa, as well as bringing up CCGTs, and some counter-intuitive interconnector activity. It also exposed issues with poor forecasting, which I have highlighted before.
Wind output was forecast to be above 14 GW for much of the day. This pushed GB power prices down, incentivising exports, however, due to grid constraints, it was impossible to deliver on these exports using wind. This resulted in significant swings on the interconnectors, significant curtailment of wind generation (up to 3 GW) and localised dispatch of CCGTs in the Balancing Mechanism (“BM”) to provide both inertia and generation to meet the export needs.
In effect, because the market does not have a means of pricing in the impact of grid constrains, some gas power stations had to run in order to export electricity to the Continent after high wind expectations depressed wholesale power prices.
While some with argue that this means we should move to locational pricing, the reality is that under a locational pricing model, wind would run even less and gas mode because suppliers with generation in the south-east where demand is most highly concentrated, will seek to procure electricity from the interconnectors and more proximate CCGTs than from Scottish windfarms whose deliverability is uncertain due to lack of transmission capacity.
Due to the difficulty in gathering the data, I am only looking at the first half of the day, up to 1:30pm. The main data used are: Generation by Fuel Type and Bid Offer Acceptances from BMRS and the wind output and curtailment data from Amira Technologies PowerPlus platform (paid subscription). I used ChatGPT to clean the data and carry out some of the analysis.
Problems with forecasting
First of all we have the wind and solar forecasts. As the sun rose and demand ramped to its daily peak (peak transmission system demand in the summer is in the morning as distribution-connected solar depresses the evening demand peak), it was clear that the day ahead wind forecasts were significantly off. Over the morning peak there as a 3 GW difference between the day-ahead and within-day wind forecasts. There was an even bigger difference between the within-day wind forecast which was higher than the day-ahead forecast, and the actual wind generation (before curtailment) going through the morning peak – up to a huge 4.5 GW (28% of actual wind output and 17% of actual demand).
Just after the morning peak there was a 1 GW difference between actual and forecast demand.
Between 06:00 and 09:00, actual wind output (before curtailment) exceeded both day-ahead and within-day forecasts, with a forecast error of up to 3 GW vs within-day forecast and even more vs day-ahead. The difference between actual and curtailed wind output (ie volume curtailed) widened rapidly after 06:30 and stabilised around 4–5 GW curtailed—likely due to grid export constraints or inertia issues, especially since demand was ramping at the same time.
The within-day forecast adjusted sharply upwards after 07:00, but still lagged behind actuals, highlighting the limitations of short-term forecasting when conditions change quickly (wind ramping faster than models expected).
This level of discrepancy (17% of actual demand) was incredibly difficult for the control room to manage and would typically trigger substantial re-dispatching, with wind curtailment, turning up CCGTs in the BM and changes to interconnector flows.
Interconnector flows were often contradictory
Interconnector flows demonstrated multiple changes of direction and contradictory flows. Interconnectors flow from the higher to lower priced market, and expected high wind output depressed GB power prices for today.
The day ahead and hourly prices for GB on N2EX are shown below. In the middle of the day, GB power prices fell close to zero:
NESO made 5 interconnector auction trades during the day up to 2pm, all of which were exports. All were at negative prices, meaning that although GB was exporting (selling) electricity, it incurred a cost of £53,775 to do so (for 3.9 GW):
Of particular significance is the fourth auction for 1.3 GW which was carried out under 2 hours before the start of the desired delivery period.
GB was paying France/Belgium/etc. to take its excess power. This is indicative of oversupply conditions, driven by: high wind output, limited domestic absorption and grid constraints preventing internal redistribution.
Up to 2pm today, NESO also entered into 1,400 SO-SO trades on the interconnectors – these are transactions with the system operator of the connected country – in this case Ireland: all the SO SO trades were over the Irish interconnectors. Each trade was for a volume of 25 MW and the net volume for the period was zero. However the value was £2,783,259. This imbalance between traded energy and financial outcome is a clear indication that these transactions were not to do with resolving an energy imbalance, but were more about managing grid constraints.
Ireland has very low inertia, and as a result, grid stability issues can arise quickly, especially under high renewable conditions – as wind output was high in GB it will also have been high in Ireland. NESO trades with the Irish TSOs (EirGrid or SONI) are usually bilateral emergency support trades (not market-based) and are used to resolve real-time system issues. Since the trade value was positive to NESO, this suggests that Ireland was having bigger issues with its grid stability today than GB was.
The chart shows the multiple changes in flow direction on the interconnectors with Europe. There were times when GB was importing from France over IFA2 while exporting over IFA 1. Similarly there were times when it imported from Scandinavia over NSL while exporting over Viking. For most of the day the trades with Benelux (Nemo and BritNed) were both exports.
The French cables land at different grid locations – IFA in Kent, IFA2 in Hampshire (Eleclink was on an outage yesterday). The opposing flow directions imply regional imbalances within GB, which is likely in the context of the limited ability to transmit wind from Scotland to the south of England.
For much of the day, the interconnector flows were significantly below the Day-Ahead nomination. This is consistent with wind generation being lower and demand being higher than forecast.
Just as the IFA interconnectors arrive in different regions of GB, so to do the Scandinavian ones: NSL lands in Northumberland and Viking lands in Lincolnshire. NSL is Britain’s most northerly interconnector, making it ideally placed to export wind from Scotland and northern England into southern Norway. So why, on a day of surplus wind and curtailments, was GB importing via NSL? To be fair, as the day progressed, NSL did switch to exports, but the net position for the day was an average of over 800 MW of imports.
GB imported over NSL in the early morning when wind was ramping. During the wind peak, exports occurred, but not enough to reverse the overall import picture. The timing of the wind ramp and the forecast errors described above meant NESO had to curtail output just as exports began. The position is worse given the fact that NSL does not re-trade within day unlike the other interconnectors. This means the flows are largely determined at the Day Ahead stage against wind forecasts that were higher than the out-turn even before curtailment.
This poses a real problem in the context of the fragile relationship between GB and Norway over the use of the electricity interconnector. That Norway does not receive net imports on a day of particularly high wind output in GB will strengthen the arguments of those in Norway who would like to re-negotiate the agreement between the two countries.
Lots of balancing trades on gas plant
Up until 2pm, NESO accepted 1,076 Bids and Offers in the BM for CCGTs. This activity was heavily shaped by a mixture of volatile wind forecasts, challenging interconnector behaviour, and resulting system balancing stress, particularly in managing locational constraints and system inertia.
In the South of England, the Marchwood CCGT (located near Southampton) was notably active in the BM, receiving multiple bid-up and bid-down instructions throughout the day. This frequent switching is unusual and suggests it was being finely controlled to support local voltage or inertia conditions—possibly due to heavy offshore wind generation in the southern North Sea region that could not be exported as forecast. Indeed, the data show that Marchwood was ramped up strongly in the morning and bid down in early afternoon, before being called up again later. This pattern coincides with shifts in wind output and changes in interconnector flows (IFA 2 lands nearby).
Further east, Grain also saw periods of ramp-up, consistent with the pattern of trying to support the southeast corner of the grid when Viking and IFA flows were shifting rapidly. Both IFA and Viking showed multiple direction changes, while IFA2 sometimes flowed opposite to IFA, as noted above, which suggests system constraints rather than pure market economics were at play.
In the Midlands and North, CCGTs such as Staythorpe, West Burton, and Keadby were largely bid down, which is consistent with high local generation from wind and solar and perhaps reduced demand relative to expectations. It indicates that these plants were not required for energy balancing but may have been held back to avoid exacerbating local constraints.
In Wales and the Northwest, Deeside and Rocksavage were bid up more consistently, again pointing to local balancing requirements or constraints in exporting Welsh wind generation eastwards. Deeside’s activity is particularly interesting as it straddles both North Wales and North West balancing zones, a region often constrained by transmission bottlenecks.
In Scotland, as is often the case, CCGTs were mostly inactive, consistent with the region being dominated by wind and hydro and typically subject to curtailment rather than thermal ramp-up.
The overall pattern is consistent with a day of complex system balancing. Poor wind forecasts, particularly in the early morning where forecast overestimates were up to 3 GW, left the system short. As curtailment increased, the control room was forced to rapidly call up thermal capacity. However, with significant export commitments locked in by Day-Ahead interconnector nominations (particularly via NSL, which does not re-nominate intra-day), these thermal units were required not just to meet internal demand but to fulfil export obligations. Most of the rest of the interconnectors did re-trade within day, with overall flows being significantly below the Day-Ahead nominations for much of the day.
The BOAs reflect this challenge – CCGTs were flexed up in regions near interconnectors and where voltage support or inertia was needed (South, Southeast). Elsewhere, where demand was better met locally by wind, CCGTs were bid down. The bid-up activity of Marchwood, Grain, and Deeside, interspersed with down instructions, shows how tightly the system was being tuned. Meanwhile, interconnector flow volatility reflects either forecast error management or attempts to re-optimise export paths intra-day. The lack of STOR dispatch in the data suggests the situation, while strained, remained within manageable operational limits, but only just.
.
In summary, yesterday was windy, but less so than expected, and a combination of high wind, poor forecasts and interconnector trading led to a very high level of activity in the BM with significant curtailment of both wind and thermal generation, but also significant local bidding up of CCGTs in order to meet export requirements and maintain grid stability.
.
.
Note on market data
This analysis has taken far too many hours to carry out, indicating the difficulties of understanding the market from official data. The first time I downloaded the BOA data, there were missing entries, so several hours of work were wasted (I realised there were missing data because I checked some individual BMUs and found discrepancies). Official data on BMRS do not include interconnector nominations at all, so I had to use Amira Technologies for these data. Also, the presentation of a lot of data on BMRS is unwieldy. For example, the generation by fuel data puts each different fuel into rows rather than columns, so in order to create charts like the ones that can be seen on screen at BMRS, it is necessary to re-arrange the data into columns. In general, data downloads from BMRS require some degree of processing before they can be used.
I have also been trying to analyse the frequency data over the bank holiday weekend but found junk data which Elexon is investigating with NESO. For example on 27 May there is a couple of hours where the frequency was reported as exactly 50.0 Hz for an entire hour, except for a few zero values. Neither 50.0 Hz nor 0 Hz are credible, particularly for multiple consecutive settlement periods.
I have also been trying to analyse daily balancing costs from the NESO Portal. However the most recent data are from 20 May, and in several of the historic data sets there are missing settlement periods (not related to daylight saving). I have asked NESO why this is and have yet to receive a reply.
There are very many different sources of GB market data, on BMRS, the NESO Portal and various dashboards such as Amira Technologies. However in order to properly understand the market it is necessary to use data from multiple sources, clean and re-format them and try to pull them together into something coherent. This is extremely time consuming. It is very difficult to quickly understand what is happening in the market – the picture yesterday was relatively simple: a mixture of high wind, poor forecasts and grid constraints. But there is no single place where this story can be easily seen, it has to be pieced together from different sources and inferred (for example there is no real way to view the constraints, only the impact of the constraints).
As the complexity of the market increases, it is more important that data are published in a timely fashion, that they are correct and complete, and that more is done to aid transparency
Source: Watt-Logic
The post High wind and forecasting errors cause havoc on the GB grid appeared first on Energy News Beat.
“}]]
Energy News Beat