Wind-Solar Power Cooperative Prediction Method Based on Feature Fusion and Improved Optimized Deep Learning Algorithm

With the continuous expansion of renewable energy integration into power grids, high-accuracy power prediction has become a critical technology for ensuring the safe and stable operation of power systems. This paper proposes a wind-solar power cooperative prediction method based on feature fusion and an improved optimized deep learning algorithm. First, Spearman correlation analysis is employed to screen key meteorological features. Second, variational mode decomposition (VMD) and complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) are applied to perform secondary decomposition on power sequences, reducing nonstationarity. Furthermore, a multi-path bidirectional long short-term memory network (MPBiLSTM) prediction model is constructed, incorporating the Luong attention mechanism to dynamically allocate feature weights. Finally, an improved sparrow search algorithm (ISSA) is used to optimize the model's hyperparameters, and a deep error feedback mechanism is introduced to enhance prediction robustness. Experiments demonstrate that the proposed model achieves an average reduction of 36.9% in root mean square error (RMSE) compared to traditional methods on wind-solar datasets, with particularly significant improvements in prediction accuracy under cloudy and strong wind scenarios.