Reduction of Wind Power Prediction Error by Spatial Smoothing Effects

We discuss the statistical effects of predicting the power output of spatially distributed wind farms. Our forecasting procedure provides the expected power output for a time horizon up to 48 hours. It is based on the large scale wind field prediction which is generated operationally by the German weather service. In this paper we focus on the reduction of the forecast error for the aggregated power output of wind farms in a spatially extended region. Due to spatial smoothing effects the error decreases considerably compared to a single site. This reduction strongly depends on the size of the region rather than on the number of wind farms it contains. We investigate the spatial smoothing effect using measured data from 30 sites in Germany. To generalise the result we consider several model ensembles of wind farms and the current distribution of wind turbines in Germany based on a statistical approach.