Wind Power Forecasting Based on Ensemble Empirical Mode Decomposition with Generalized Regression Neural Network Based on Cross-Validated Method

Abstract

The growth of wind power connected to the power grid has increased the importance of accurate wind power prediction that exhibits non-linearity and non-stationarity. The goal of this study is to forecast wind power by using the generalized regression neural network (GRNN) coupled with ensemble empirical mode decomposition (EEMD) and assessment of prediction accuracy. EEMD technologies are used to perform decomposition, and each intrinsic mode function is predicted and forecasted by using a GRNN based on cross-validated parameters. The forecasting results of the sub-series are superimposed as the results of wind power prediction. Results show that the proposed method has high prediction accuracy and is highly effective in forecasting wind power.