Short-Term Wind Power Prediction Based on Multiple Decompositions and WOA-LSSVM Model

In response to the significant uncertainty of wind power and the limitations of a single model in capturing the nonlinear characteristics of signals, this paper introduces a wind power prediction model based on multiple decompositions and the Least Squares Support Vector Machines (LSSVM) optimized by the Whale Optimization Algorithm (WOA), referred to as the WOA-LSSVM model. In this model, the raw wind power data is initially decomposed using the Complete Ensemble Mode Decomposition with Adaptive Noise (CEEMDAN) to obtain several Intrinsic Mode Functions. Subsequently, the sample entropy of each component is calculated, and these Intrinsic Mode Functions are classified into three components: high, medium, and low frequency, using K-means clustering. The Variational Mode Decomposition (VMD) is then employed to further process the high-frequency components. Finally, the processed data is fed into the LSSVM, and the hyperparameters of the LSSVM are fine-tuned using the WOA algorithm. Experimental results based on measured data from a wind farm in Turkey reveal that, in comparison to other models, the proposed model exhibits the highest determination coefficient, the smallest root mean square error, and the lowest mean absolute error, demonstrating its suitability for short-term wind power prediction.