Hybrid GM(1,1)-NARnet one hour ahead wind power prediction

Grey system theory deals with systems characterized by the uncertainty of the partially known/unknown information. The traditional Grey prediction model GM(1,1) has been widely used in different short-term prediction applications including wind power forecast. However, it is proved that it cannot provide sufficient prediction accuracy. In this paper a new approach for short-term wind power prediction is proposed. The suggested technique is a hybrid method comprised of the GM(1,1) forecasting model and the Nonlinear Auto Regressive neural network (NARnet) method. The forecasting precision of the proposed method is examined by applying it to an actual wind power data set. The experimental results confirm that the proposed technique outperforms the traditional GM(1,1), GM(1,1)-ARMA, and the persistence method.