Performance analysis of four modified approaches for wind speed forecasting

Abstract

Providing accurate wind speed prediction algorithms has become increasingly significant because of the important technological and economic impacts of wind speed on wind power generation. In this study, two combined strategies for wind speed forecasting are proposed and followed. Four approaches derived from these strategies are employed. The first two approaches employ Particle Swarm Optimization (PSO) to optimize the parameters in the first-order and second-order adaptive coefficient (FAC and SAC) methods. The remaining two approaches employ the decomposition of wind speed data into seasonal and trend components. The seasonal component is represented by the seasonal exponential adjustment, and the trend component is predicted by the hybrid models of PSO and FAC or SAC. By employing these four approaches, the daily mean wind speed is forecasted for four observation sites in Gansu, China. Multiple evaluation methods are used to assess forecasting quality: the mean square error, the mean absolute percentage error and the relative error. It is found that all four approaches better predict wind speed than the original FAC and SAC models.