Photovoltaic Power Prediction Based on Multi-Model Integration

Photovoltaic (PV) power generation has been widely applied in recent years due to its renewable and environmentally friendly nature. However, the power output of PV generation is highly influenced by meteorological factors, resulting in significant randomness and volatility, which poses challenges to grid scheduling and energy management. Existing prediction methods, such as traditional statistical models and individual deep learning models, have limitations in handling complex nonlinear relationships and long-term dependencies. To address these issues, this paper proposes a photovoltaic power prediction method based on feature extraction and multi-model optimization integration. First, the importance of meteorological factors is evaluated using the XGBoost model to extract key features, and the feature set is further optimized using a weighted recursive feature elimination method. Then, an improved Long Short-Term Memory (LSTM) model is used to capture long-term dependencies in the time series, while TimesNet is employed for efficient time feature extraction and modeling. Finally, a self-adaptive weighted fusion strategy is used to integrate the prediction results to maximize the performance of each model. Experimental results show that the proposed method outperforms existing individual models in terms of both accuracy and stability, demonstrating significant advantages.