Improving the Wind Power Forecast Quality by Combining Numerical Weather Models and Applying Scaling Techniques

Meteogroup has done research on how to optimize the wind power forecast quality for the medium range, using a mix of the operational model and the ensemble median forecast. For a selected set of stations in Europe, this has been investigated. Different experiments have been performed to include information from the EPS (Ensemble Prediction System) median in the later forecast range. Best results are obtained with a gradual transition from the deterministic forecast to the EPS median starting 72 hours after the start of the EPS forecast and with an e-folding timescale of 96 hours.

Information from the EPS median reduces the MAE for the forecast period between 3 and 10 days ahead. However, the bias increases when the EPS median is used, due to the differences in climatology between the operational model and EPS median. A Weibull correction can be applied to limit the increase in ME. The Weibull correction has a small beneficial effect on the MAE.

The main conclusions can be summarized as follows: Including a gradual transition towards the EPS median gives an improvement compared to the current forecast. A gradual transition from the deterministic forecast towards the EPS median gives better results than just using the EPS median value for the whole forecast period. Including a Weibull correction on top of the gradual transition to the EPS median further improves the forecast compared to the case without a Weibull correction.