Wind speed prediction using measurements from neighboring locations and combining the extreme learning machine and the AdaBoost algorithm

Wind speed prediction plays an essential role in wind energy utilization. However, most existing studies of wind speed forecasting used data from one location to build models and forecasts, which limited the accuracy of wind speed forecasting. Therefore, to improve the prediction accuracy at a target location, this study proposes a multiple-point model based on data from multiple locations for short-term wind speed prediction. The model, which utilizes wind speed measurements from neighboring locations and combines the extreme learning machine (ELM) with the AdaBoost algorithm, is named the multiple-point-AdaBoost-ELM model. Data from seventeen automatic meteorological stations in the Heihe River Basin are used, four stations at different positions are taken as target stations for multi-time-scale wind speed prediction, and six models and several metrics are involved for comparative analysis and comprehensive evaluation. The results show that: (1) the prediction performance of the proposed multiple-point-AdaBoost-ELM model is significantly superior to that of the compared single-point models; (2) the prediction accuracy of the multiple-point-AdaBoost-ELM model is relatively less affected by the prediction time-scale than that of the corresponding single-point model; and (3) the stations located at the center of multiple stations can obtain more accurate prediction results than those located near the edges of the region. Therefore, the proposed multiple-point-AdaBoost-ELM model is a more promising method than traditional single-point modeling methods. The proposed method fully uses historical wind speed at surrounding locations to enhance the wind speed predictions at target locations, makes up for the deficiency of the wind speed forecasting using data from one location, and expands a new way for wind speed prediction.