Evolutionary turbine selectionfor wind power predictions

Abstract

Statistical methods have shown great success in short-term prediction of wind power in the recent past. A preselection of turbines is presented that is based on the segmentation of the area around the target turbine with a specific radius. Small problem instances allow a rigorous comparison of different input sets employing various regression techniques and motivate the application of evolutionary algorithms for finding adequate features. The optimization problem turns out to be difficult to solve, while strongly depending on the target turbine and the prediction technique.