Forecasting the wind power generation using Box–Jenkins and hybrid artificial intelligence

Abstract

Purpose The purpose of this paper is to forecast wind power generation in an area through different methods, and then, recommend the most suitable one using some performance criteria. Design/methodology/approach The Box–Jenkins modeling and the Neural network modeling approaches are applied to perform forecasting for the last 12 months. Findings The results indicated that among the tested artificial neural network (ANN) model and its improved model, artificial neural network-genetic algorithm (ANN-GA) with RMSE of 0.4213 and R2 of 0.9212 gains the best performance in prediction of wind power generation values. Finally, a comparison between ANN-GA and ARIMA method confirmed a far superior power generation prediction performance for ARIMA with RMSE of 0.3443 and R2 of 0.9480. Originality/value Performance of the ARIMA method is evaluated in comparison to several types of ANN models including ANN, and its improved model using GA as ANN-GA and particle swarm optimization (PSO) as ANN-PSO.