Short-term wind speed forecasting based on improved empirical wavelet transform and least squares support vector machines

A new improved empirical wavelet transform (IEWT) method is proposed to treat with the nonlinearity and nonstationarity of original wind speed signal. This method decomposes wind speed signal into a set of band-limited sub-sequences to decrease instability. On this basis, combined with least squares support vector machine (LSSVM), a short-term wind speed forecasting model based on IEWT-LSSVM is proposed. The phase space reconstruction parameters and the hyper parameters of LSSVM model are optimized by simulated annealing particle swarm optimization (SAPSO). Finally, taking the wind speed data of a certain wind farm in North China as an example, the simulation results illustrate that the forecasting model based on IEWT-LSSVM can effectively track the change of wind speed signal, has high prediction accuracy in single-step prediction and multi-step prediction. © 2021, Solar Energy Periodical Office Co., Ltd. All right reserved.