A novel method for modeling renewable power production using ERA5: Spanish wind energy

In this work, a novel methodology for the estimation of wind energy resources at different regional levels was assessed. The method consists in employing a machine learning model applied to a virtual power plant at the center of the region, having as input ERA5-derived meteorological variables and real installed capacities and generation data. The method was found to provide enhanced wind capacity factors (CF) estimates in Spain, with RMSE < 0.18, R > 0.75 and negligible bias for all the analyzed regions. As an application of the proposed methodology, an enhanced open access database of Spanish wind energy resources (SHIRENDA_Wind) was built. This database consists of hourly values of wind CF for the Spanish NUTS 3 regions covering the period of 1990–2020. SHIRENDA revealed a marked interannual variability of the wind energy resources, the winter interannual CFs changes frequently reaching 30%. The spatial variability of the wind energy resources was found to be considerably high and partially driven by the NAO. • Introduces a novel approach for regional wind energy generation modeling. • ML conversion models based on ERA5, actual installed capacity and generation data. • A few meteorological variables may accurately account for the power generation. • Enhanced wind Capacity Factors database (NUTS 3, hourly, 1990–2020) for Spain. • Notable spatio-temporal variability of the wind energy resources, driven by the NAO.