Chaotic time series prediction model of wind power based on ensemble empirical mode decomposition-approximate entropy and reservoir

Abstract

According to the chaotic feature of wind power time series, a combined short-term wind power forecasting approach based on ensemble empirical mode decomposition (EEMD)-approximate entropy and echo state network (ESN) is proposed. Firstly, in order to reduce the calculation scale of partial analysis for wind power and improve the wind power prediction accuracy, the wind power time series is decomposed into a series of wind power subsequences with obvious differences in complex degree by using EEMD-approximate entropy. Then, the forecasting model of each subsequence is created with least squares support vector machine (LSSVM), ESN and EEMD-ESN improved with the regularized high frequency parts. Finally, the simulation is performed by using the real data collected from a certain wind farm, the results show that the EEMD-ESN model is better in the training speed and forecasting accuracy, than those obtained from the least square support vector machine (LSSVM) model, which provides a new useful reference for the short-term forecasting of wind power in online engineering application.