A Characterization of Energy Drought Events in the Netherlands

Abstract

The increasing importance of renewable energy supply in the transition to carbon neutral energy systems highlights the need of an adequate understanding of the associated risks to energy supply. Variable renewable energy sources are inherently dependent on meteorological conditions, and as such could provide insufficient energy for certain periods of time. In this study, a characterization is performed on such energy drought events in the Netherlands. ERA-5 reanalysis and MERRA-2 climate data is used to model the performance of solar PV and wind power generation. Statistical approaches are used to investigate these events on provincial, sub-national and national scale. It has been found that the province least susceptible to energy drought occurrence is Zeeland, whereas Gelderland, Limburg, Noord-Brabant, Overijssel and Utrecht are most likely to experience energy droughts. Additionally, copula theory is used to investigate probabilities of the co-occurrence of energy drought events between regions and technologies. When provinces are aggregated to sub-national regions, this research shows that constructed copula models capture tail events in the distribution of power generation in the form of energy droughts well. A co-occurrence between 9.61% and 14.9% has been found for ED events between the two defined sub-national regions, with return periods between 67.7 and 117 days. Finally, extreme value theory is applied to investigate extremely long-lasting energy droughts. Using the Peak-OverThreshold method, a VaR-95% value between 27.2 and 46 hours and a CVaR-95% value between 41.6 and 58.1 hours has been found for extreme energy drought duration in the Netherlands. Furthermore, 50- year and 100-year ED durations are found to be between 84 and 99 hours and 91 and 107 hours, respectively. The results of this study can be incorporated in the planning of future renewable energy installations, and for grid operators to determine and manage the risk of black outs more accurately.