As we wrap up 2024, let's take a look back at wind trends in Ireland. Wind generation in Ireland hit a record in August 2024. According to Wind Energy Ireland's (WEI) monthly wind energy report, wind power generation for the month totaled 1068GWh, a 3% rise when compared with the previous record set during the same month in 2023 (1042GWh).
Strong winds also meant that, at 34%, Irish wind farms met just over one-third of Ireland's electricity demand in August 2024, surpassing the previous record of 33% in August 2023. Solar power and other renewables accounted for 6% of Ireland's electricity during August 2024, meaning 40% of Ireland's electricity came from renewable sources.
Impacts on prices and forecasts
While this is great news for hitting sustainability goals, the intermittent nature of wind generation can cause fluctuating prices for wind power. The average wholesale price of electricity in Ireland per megawatt-hour during August 2024 was €100.04, down slightly from €106.46 in August 2023, according to WEI data.
Prices on days with the most wind power saw the average cost of a megawatt-hour of electricity decrease by nearly 10% to €90.67 per megawatt hour and rise to €125.96 on days when the country relied almost entirely on fossil fuels.
Many organizations rely on forecasts to manage their forward pricing, and from August 20 to August 26, wind generation forecasts were consistently hard to predict. This led to volatility in day-ahead auction prices and left some power market stakeholders badly exposed, putting millions of euros at risk.
During this period, Amperon significantly outperformed EirGrids and Soni's forecasts by 46%, as can be seen from the chart below.
Looking more closely at the relationship between day ahead and imbalance prices, one can see that on the evening of August 20 and the early morning of August 22, wind generation was significantly over forecast by EirGrid and Soni with actual generation coming in at nearly half of what was predicted. EirGrids and Soni were particularly off for the hours of HE8, with an average nMAE of 166% compared to Amperon's 21.5%.
August 22 saw the largest spread between day ahead and imbalance prices, as the wind underperformed EirGrid's forecast, and at times the price spread averaged more than €100. EirGrid forecasted wind that was higher than actual by more than 3GW, a key driver behind these price spikes. In addition, EirGrid under forecasted demand and higher-than-expected morning ramp-up further contributed to the large price spread.
The value of multiple models for accuracy
This raises the question of why these forecasts were so inaccurate. Many power stakeholders in Europe are reliant on a single weather forecasting model – the ECMWF. This is a medium-range weather forecast model that collects data from satellites, weather observations from the ground, and other meteorological data and converts them into a numerical weather prediction, working with 34 separate countries.
Amperon uses specific weather vendors depending on the type of forecasts required. Our weather vendors come with different strengths with their forecasts, so we'll use the best forecast for specific metrics depending on if we are forecasting for grid demand, renewables, or asset-level wind or solar. These are then pulled into an hourly granularity weather forecast going out 15 days. This information is overlaid on additional data sets that Amperon uses to serve its models, enhancing the overall accuracy of our predictions. An ensemble like this leads to better answers, as demonstrated in the charts.
Furthermore, Amperon uses more than 8,600 weather points across Europe to build specific models for demand and renewable supply. These weather points are then fed into demographic data to determine the density of demand and capacity of supply. From there, our models take over to deliver accurate forecasts, which are updated hourly.
Our forecasts use a variety of machine learning and statistical models, including linear regression and boosted tree models, which are then dynamically weighted and continuously recalibrated as new load and weather data becomes available. By using the most up-to-date data, our models are more capable of predicting demand, wind, and solar generation for better power forecast accuracy.
.png)
.png)
