A self-adaptive hybrid approach for wind speed forecasting

Abstract

Wind power, as a promising renewable energy source, has environmental benefits, as well as economic and social ones. To evaluate wind energy properly and efficiently, this study proposes a hybrid forecasting approach that combines the Extreme Learning Machine (ELM), which rarely presents in literature on wind speed forecasting, the Ljung-Box Q-test (LBQ) and the Seasonal Auto-Regressive Integrated Moving Average (SARIMA) to enhance the accuracy of wind speed forecasting. The proposed hybrid method is examined by forecasting the mean daily and mean monthly wind speed of four wind speed observation sites located in northwestern China. The results confirm that, compared with other popular models (ARIMA，SARIMA，Back-Propagation neural network (BP) and ELM), the hybrid forecasting method improves the predictions of daily and monthly wind speed, which indicates that the developed hybrid method exhibits stronger forecasting ability. The forecasting results also suggest the hybrid approach has better generality and practicability in different wind farms.