Short-Term Wind Power Prediction based on Adaptive Hybrid Swarm Intelligence Optimization

Abstract

Wind power prediction is crucial for ensuring the safe and stable operation of high-penetration renewable energy power systems. Addressing the nonlinear and non-stationary characteristics of wind power sequences and the challenges of hyperparameter optimization in deep learning models, this paper proposes a short-term wind power prediction method based on adaptive hybrid swarm intelligence optimization. First, an adaptive elite hybrid algorithm (SAE-PSO-GWO) integrating Grey Wolf Optimizer and Particle Swarm Optimization is constructed, incorporating elite opposition-based learning and parameter adaptive mechanisms to balance global exploration and local exploitation capabilities. Subsequently, a prediction model combining multi-scale CNN and LSTM is designed to extract multi-scale temporal features and model long-term dependencies. Finally, SAE-PSO-GWO is employed to automatically optimize the model hyperparameters. Experimental results demonstrate that the proposed algorithm exhibits superior convergence accuracy and stability on the CEC2010 test functions. On wind power datasets, the optimized model achieves MAE, RMSE, SMAPE, and R² values of 57.89, 116.87, 14.17%, and 0.9680, respectively, representing improvements of approximately 14.38%, 19.48%, 4.90%, and 1.83% compared to baseline models. This verifies the effectiveness of the proposed method in enhancing prediction accuracy and robustness.