Multi-step wind power forecast based on similar segments extracted by mathematical morphology

This paper proposes a pre-processing method to enhance the accuracy of wind power forecast. Instead of using the whole dataset indifferently for training, the proposed method only uses the segments that share the same pattern. In order to search for such segments in the historical data, a k-OCCO filter and a weighted multi-resolution morphological gradient (MMG) are employed. Afterwards, the forecast is conducted by the least square support vector machine (LS-SVM) model, using these segments for training. Simulation studies are carried out on wind power data to demonstrate the advantage of the proposed method, and the results have shown that both the accuracy and the stability of the LS-SVM model have been improved by introducing the proposed method.