A new dynamic integrated approach for wind speed forecasting

Abstract

Wind energy is considered as one of the most promising and economical renewable energy. In order to insure maximum yield of wind energy, it is vital to evaluate wind energy potential of the wind farms. Since wind energy is proportional to the cube of wind speed, the evaluation of wind energy potential assessment comes down to the wind speed forecasting. In this paper, the wind speed is predicted by utilizing a new dynamic integrated approach. The novelties of this method mainly include: firstly, the Phase Space Reconstruction (PSR) is employed to dynamically choose the input vectors of the forecasting model; secondly, the data preprocessing approach, named the Kernel Principal Component Analysis (KPCA), is proposed to efficiently extract the nonlinear characteristics of the high-dimensional feature space reconstructed by the PSR; thirdly, Core Vector Regression (CVR) model, whose parameters are determined by the Competition Over Resource (COR) heuristic algorithm, is adopted to the model for quick computational speed; finally, the Grey Relational Analysis, Diebold-Mariano and Pesaran-Timmermann statistic are treated as evaluation tools to assess the forecasting effectiveness of this approach. The empirical results show that this integrated approach can significantly improve forecasting effectiveness and statistically outperform some other benchmark methods in terms of the directional forecasting and level forecasting.