A short term wind speed forecasting method using signal decomposition and extreme learning machine

Abstract

In this study, a novel hybrid model using signal decomposition technique and extreme learning machine (ELM) is developed for wind speed forecasting. In the proposed model, signal decomposition technique, namely wavelet packet decomposition (WPD), is utilized to decompose the raw non-stationary wind speed data into relatively stable sub-series; then, ELMs are employed to predict wind speed using these stable sub-series, eventually, the final wind speed forecasting results are calculated through combination of each sub-subseries prediction. To evaluate the forecasting performance, real historical wind speed data from a wind farm in China are employed to make short term wind speed forecasting. Compared with other forecasting method mentioned in the paper, the proposed hybrid model WPD-ELM can improve the wind speed forecasting accuracy.