Assessing the performance of reanalysis models for wind power modelling in Germany

Country-wide energy systems analysis requires the simulation of wind power feed-in. This is usually derived using wind turbine data and wind speed time series from reanalysis models. As reanalysis models exhibit biases especially in complex terrain, the modelled feed-in deviates from the observed wind energy feed-in. In this work, we assess several state-of-the-art reanalysis models such as REA6, ERA5, CERRA and NEWA by simulating all current German windfarms and comparing the results with country-wide feed-in time series from ENTSO-E or SMARD. In a second step the wind turbine model parameters are calibrated for each reanalysis model to better match the observed capacity factors.

In order to simulate the wind power feed-in, we use a regionally smoothed power curve to model feed-in and an empirical shading curve to account for intra park turbine wakes. The location, hub height, rotor diameter and installed capacity are taken from the Marktstammdatenregister (MaStR) with missing parameters estimated from wind turbines with similar characteristics. The total installed capacity is corrected or scaled using the ENTSO-E and SMARD data. If available, manufacturer power curves are used; otherwise, synthetic power curves are applied. Additionally, other loss factors such as technical availability are included using constant factors.

In the first analysis the feed-in model is parameterized with standard values to assess the general performance of each reanalysis model. The obtained results are analyzed with respect to capacity factors, bias, and correlation coefficients, providing an indication of the overall performance of the reanalysis models in terms of wind energy generation.

In the second step, the analysis framework is used to derive calibration factors for the feed-in model, aiming to achieve more realistic results for each of the reanalysis models. By calibrating the wind turbine model parameters, we aim to reduce the difference between the modeled and the observed capacity factors.

The performed analysis and calibration highlight the need for an improved understanding of the performance of reanalysis data sets for wind energy system modelling and the development of more sophisticated corrections.