Accurate predictions of electricity at the national or regional level are essential inputs for electricity power trading, transmission grid managements and production planning. The innovative part of this project has been to jointly predict electricity net-demand across the 14 regions forming the GB's transmission network, shown in Figure 1. In particular, the results show that the correlation between forecasting errors across the regions depends on factors such as the weather and the time of year, as show by plots (a)–(b) in Figure 2. Each node represents a region, the color of the nodes is the expected size of the errors, and the edges are the correlations between regions. Plot (a) corresponds to 7 a.m. on December 31, 2018 (a), a hard-to-predict period during which large and highly correlated errors are to be expect, while plot (b) corresponds to midnight on August 20, 2018,  a much quieter period with small and uncorrelated errors. Given that electricity production planning is based on net-demand forecasts, knowing the expected size and correlation of the errors is obviously important to determine production reserves.

The research behind this project started back in 2019, when Prof. Browell (University of Glasgow) visited Dr. Fasiolo (Bristol) under the Heilbronn Visitors scheme. In a 2021 article (https://arxiv.org/abs/2103.10335) they analysed the data separately for each region, while modelling the 14 regions jointly required new methods and software, that were developed during the PhD of Dr. Gioia (University of Trieste, IT). Dr. Gioia's work was co-supervised by Prof. Bellio (University of Udine, IT) and Dr. Fasiolo, and the work of the latter on this project was supported by Électricité de France.

Huge congratulations to Dr Fasiolo on this achievement. The paper is freely available online, while the software for building the models described in the paper is also available to view.