Multi-step wind speed forecasting based on a hybrid decomposition technique and an improved back-propagation neural network

Accurate wind speed forecasting (WSF) not only ensures stable power system operation but also contributes to enhancing the competitiveness of wind power companies in the market. In this paper, a hybrid prediction model based on secondary decomposition algorithm (SDA) is proposed for WSF. First, wavelet transform (WT) is used to decompose the wind speed sequence into approximate and detailed components. Second, the obtained detailed components are further decomposed by symplectic geometry mode decomposition (SGMD). Then, the marine predators algorithm-optimized back-propagation neural network (BPNN) is used to predict the new subsequences. The case study was implemented on 4 datasets. The experimental results show that, first, the proposed hybrid model has the highest prediction accuracy and the best robustness among all the compared models in 1-4-step prediction. Second, the proposed hybrid decomposition strategy has significant utility in reducing the difficulty of WSF. After adding SDA, the average improvement levels of MAPE in 1-4-step prediction were 85.64%, 84.93%, 81.08% and 80.67%, respectively. Third, the re-decomposition of the details obtained by WT can improve the prediction accuracy. After the re-decomposition of the details obtained by WT, the proposed WT-SGMD-MPA-BP model leads to the average improvement percentages of 44.44%, 61.69%, 50.56% and 49.28% in RMSE compared with WT-MPA-BP model in various horizons. The proposed model provides valuable reference for WSF. In future work, the performance of the model for other nonlinear sequences is worth exploring.