Radial basis function neural network based short-term wind power forecasting with Grubbs test

Abstract

Accurate prediction on wind power generation plays an important role in power system dispatching and wind farm operation. The Radial Basis Function (RBF) neural network, owing to its superior performance of linear/nonlinear algorithm with respect to fast convergence and accurate prediction, is very suitable for wind power forecasting. Based on the historical data from a wind farm composed of wind speed, environmental temperature, and power generation, the authors develop a short-term wind power prediction model for one-hour-ahead forecasting using a RBF neural network. Due to the existence of incorrect values in the original data, the Grubbs test is conducted to preprocess the samples. In the case study, the forecasting results are compared with the actual wind power outputs. The simulation shows that the presented method could provide accurate and stable forecasting.