Assessing the performance of reanalysis and meso-scale model datasets for onshore wind power modelling in Germany

This study evaluates the performance of several reanalysis and meso-scale datasets (ERA5, CERRA, COSMO-REA6, COSMO-R6G2 and NEWA) in modelling wind power generation in Germany and two of its grid control zones. It is the first detailed analysis of CERRA and COSMO-R6G2 for modelling wind energy generation. For this study, wind speeds from several datasets are used to simulate wind power generation for the years 2017 and 2018. The simulations are then compared to observed power generation data provided by the German transmission system operators (TSOs). The study shows that all investigated datasets overestimate the wind energy production in Germany, with overestimation ranging from 5 % to 45 %. As wind turbines are often placed in particularly windy locations within an area (e.g. ridge of a hill or hilltops), this indicates a significant overestimation of the average wind conditions by the reanalysis datasets. When looking at regional variations between the grid control zones, regional differences were observed. In the TransnetBW control zone, characterised by lower mountain ranges, the overestimation was lower. Correlation was generally high with CERRA and ERA5 showing the highest correlations. In general higher-resolution datasets did not perform better than the lower-resolution ERA5 dataset and in many cases showed a weaker agreement with the observed power generation data. Only in the diurnal cycle CERRA reproduced the observed pattern slightly better than ERA5. The study highlights the importance of considering regional and potentially topography-dependent calibrations of wind speed from reanalysis and meso-scale model datasets for power generation modelling. Among the datasets explored, CERRA and ERA5 provide the best wind speed data for wind power simulations, provided that a (regional) bias correction is applied. Moreover, artificial spikes in the diurnal cycle need to be addressed in both datasets. Due to its slightly better diurnal cycle CERRA has an advantage in capturing the temporal variability.