An experimental investigation of three new hybrid wind speed forecasting models using multi-decomposing strategy and ELM algorithm

Abstract

The wind speed forecasting is important for the wind power industry to achieve the intelligent management. Three new hybrid methods using the WPD (Wavelet Packet Decomposition), the EMD (Empirical Mode Decomposition) and the ELM (Extreme Learning Machine) are presented for wind speed multi-step predictions. In the proposed architectures, the WPD is chosen to decompose the actual wind speed series into several sub-layers, while the EMD is adopted to further decompose the obtained LF (Low Frequency) sub-layers, the obtained HF (High Frequency) sub-layers and all the obtained sub-layers into a number of IMFs (Intrinsic Mode Functions), respectively. Finally, the ELM is used to complete the wind speed predicting computation for these decomposed wind speed sub-layers. To investigate the performance of the proposed hybrid models in the wind speed multi-step forecasting and find which kind of signal decomposing approach is the most suitable for the ELM based wind speed forecasting, the PM (Persistent Model), the ARIMA model, the SVM model, the ELM model, the WPD-ELM model, the WPD-EMD (LF)-ELM model, the WPD-EMD (HF)-ELM model and the WPD-EMD-ELM model are all included in the forecasting performance comparisons. The results of the two real experiments indicate that among all the involved models, the WPD-EMD (LF)-ELM model has the best predicting performance.