Seasonal prediction of renewable energy generation in Europe based on four teleconnection indices

With growing amounts of wind and solar power in the electricity mix of many European countries, understanding and predicting variations of renewable energy generation at multiple timescales is crucial to ensure reliable electricity systems. At seasonal scale, the balance between supply and demand is mostly determined by the large-scale atmospheric circulation, which is uncertain due to climate change and natural variability. Here we employ four teleconnection indices, which represent a linkage between atmospheric conditions at widely separated regions, to describe the large-scale circulation at seasonal scale over Europe. For the first time, we relate each of the teleconnections to the wind and solar generation anomalies at country and regional level and we show that dynamical forecasts of the teleconnection indices allow predicting renewable generation at country level with positive skill levels. This model unveils the co-variability of wind and solar generation in European countries through its common dependence on the general circulation and the state of the teleconnections.
The research leading to these results has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement nº 776787 (S2S4E) and nº 690462 (ERA-net ERA4CS INDECIS), and the MICINN grant BES-2017-082216 (”Ayudas para contratos predoctorales”). The authors acknowledge the Copernicus Climate Change Service (C3S) for providing seasonal predictions from several European meteorological centers and the ECMWF for producing the ERA5 reanalysis. We thank Stefan Pfenninger and Iain Staffell for providing the NINJA dataset, and Hannah Bloomfield, David Bryshaw and Andrew Charlton-Perez for producing the UREAD-ERA5 dataset. We acknowledge the Knowledge Management Unit, Directorate C Energy, Transport and Climate, Joint Research Centre, European Commission for the dissemination of EMHIRES. All the analyses have been done with the R language employing the packages s2dverification, CSTools, and SpecsVerification. We want to thank Pierre-Antoine Bretonni`ere and Margardia Sams´o for providing support with the download and formatting of the datasets
Peer Reviewed
Postprint (author's final draft)