Fast prediction of electricity supply shortages: A data-driven approach to select relevant scenarios for resource adequacy assessments

Kurzfassung

The increasing share of variable renewable energy sources in the European power system poses challenges for grid operators, that have to keep electricity supply and demand balanced at all times. To ensure the feasibility of this task in the future, decision-makers need reliable information on the implications of potential investments and network development plans. The annual European Resource Adequacy Assessment (ERAA) is a central source of such information and evaluates the risk of electricity supply shortages over a ten-year horizon. To appropriately account for the uncertainties involved, a range of scenarios should be considered. They are defined by assumptions about properties of the installed power generators as well as a variety of weather years from different climate models and greenhouse-gas emission scenarios. However, the amount of resulting scenarios can put a serious strain on computational resources if they are all analysed in detail. Therefore, a fast way to preselect the most relevant scenarios can be of great use. To address this challenge, we assessed established indicators of critical situations for power systems with regard to their capabilities to predict electricity supply shortages. Based on that, we evaluated the ability to identify the most relevant scenarios for the analysis of resource adequacy concerns. We found that widely studied "Dunkelflaute" or dark doldrum indicators, that focus on the supply side of the electricity balance, perform comparatively poorly, while a concept that additionally incorporates the electricity demand showed satisfactory results. However, the dependence on the demand data can be a disadvantage as it is not always available. As an alternative, we developed a data-driven indicator that does not rely on electricity demand data. Instead, we use air temperature and time data as proxies without significantly compromising predictive performance. As an exemplary application of our approach, we evaluated if the switch from reanalysis-based weather years to ones from climate projections in the 2024 ERAA has a relevant influence on the estimated resource adequacy risks.