A Short-Term Wind Power Forecasting Approach With Adjustment of Numerical Weather Prediction Input by Data Mining

This paper proposes a novel short-term wind power forecasting approach by mining the bad data of numerical weather prediction (NWP). Today's short-term wind power forecast (WPF) highly depends on the NWP, which contributes the most in the WPF error. This paper first introduces a bad data analyzer to fully study the relationship between the WPF error with several new extracted features from the raw NWP. Second, a hierarchical structure is proposed, which is composed of a K-means clustering-based bad data detection module and a neural network (NN)-based forecasting module. In the NN module, the WPF is fully adjusted based on the output of the bad data analyzer. Simulations are performed comparing with two other different methods. It proves that the proposed approach can improve the short-term wind power forecasting by effectively identifying and adjusting the errors from NWP.