Application of S-Transform-based Artificial Neural Network to Wind Speed Forecasting

In this paper a new artificial neural network (ANN) method is proposed to forecast wind speed forecasting. The use of wind power generation (WPG) is expected to reduce CO2 as the framework of environmental preservation. However, output of WPG is affected by the meteorological conditions significantly. As the first stage of research, this paper focuses on wind speed that affects the output of WPG significantly. The proposed method is based on the intelligent system integration of GRBFN (Generalized Radial Based Function Network) of ANN and the S-Transform of the pre-filtering technique. GRBFN is used as the forecasting model to deal with nonlinear time series of wind speed. The S-Transform is employed to extract features of input variables as a pre-filtering technique in the forecasting model. Furthermore, the two-staged forecasting method is proposed to reduce the forecasting errors. EPSO of evolutionary computation is applied to optimizing the weights between neurons of GRBFN. The proposed method is successfully applied to real wind speed data of Miyakojima in Japan.