Multi-layer fusion model based on decomposition denoising and intelligent algorithms for wind speed prediction

The proportion of wind energy within the renewable energy market has steadily increased, making it a key driver in advancing global clean energy development. Owing to the complexity of atmospheric perturbations and nonlinear features of wind speed data, this paper presents a multi-layer fusion model combining atmospheric perturbations and intelligent algorithms. Firstly, establishing a new data processing method, Savitzky-Golay optimized Variational Mode Decomposition (S-VMD), which uses the VMD to decompose the wind speed sequence into multiple IMFs and denoises the high-frequency components using a Savitzky-Golay filter based on the composite multiscale entropy (CMSE) of each modal component. Then, the Slime Mould Algorithm-based Optimizer (SMABO) is proposed to refine the CNN-BIGRU hybrid model for wind speed prediction. Given the characteristics of wind, this research combines the Weibull distribution with the Lorenz equation to derive the sequence of atmospheric disturbances, which are used to quantify atmospheric uncertainty. Additionally, the prediction error is corrected through the optimization of the Extreme Gradient Boosting (XGBOOST) algorithm for the model's error sequence. Compared with other models, the proposed hybrid model achieves an overall improvement of 16.6 %, 13.1 % and 12.9 % in prediction accuracy on three different wind farm datasets, respectively, while maintaining the MAE below 0.2.