A hybrid model based on data preprocessing strategy and error correction system for wind speed forecasting

Abstract

Wind speed forecasting is crucial in exploiting wind energy and integrating power grid. This study presents a novel hybrid model, which includes decomposition module with real-time decomposition strategy, forecasting module and error correction module. In this model, the raw wind speed series is decomposed with empirical wavelet transform into several subseries. The Elman neural network is employed as predictor for each subseries. In addition, a new error correction system is proposed to capture the hidden information from wind speed and enhance the forecasting capability. In the error correction system, a quasi-real-time decomposition strategy is constructed to obtain errors of each subseries. The variational mode decomposition-autoregressive integrated moving average approach is built to predict the error series and complete the error correction task. Two experiments covering eight wind speed datasets and ten compared models are utilized to verify the effectiveness of the proposed model. The results show that: (a) the developed error correction system is an effective way to enhance forecasting performance of the decomposition based model; (b) the error series can be effectively repaired to increase the forecasting accuracy by the combination of the variational mode decomposition method and the autoregressive integrated moving average method; (c) the proposed model outperforms the compared conventional models in short-term wind speed forecasting.