A novel wind speed prediction approach based on wind speed fluctuation characteristic extraction and dual decomposition

The accuracy of wind speed forecasting plays a crucial role in the operation of the power grid. However, the intermittency and randomness of wind present significant challenges to the accuracy of wind speed predictions. An improved hybrid model for wind speed forecasting is proposed in this paper. The model proposes a dual decomposition method that leverages the advantages of discrete wavelet transform and improved complete ensemble empirical mode decomposition with adaptive noise to extract wind speed features and effectively uncover latent feature information. An adaptive interval segment model is proposed to extract wind speed fluctuation features, combined with K-means to further segment the wind speed series. This model effectively overcomes the limitations of traditional wind speed feature extraction methods. Beluga whale optimization is employed for hyperparameter optimization in deep learning models, enabling both single-step and multi-step wind speed forecasting. Finally, to validate the effectiveness of the proposed model, wind speed data from three different wind farms in China were collected. Experimental analysis demonstrates that the proposed model outperforms other models in both single-step and multi-step wind speed forecasting. The results indicate that our proposed model could serve as a promising approach for short-term wind speed forecasting in the future.