Power Generation Prediction of Photovoltaic Based on the XGBoost-Enhanced CNN-LSTM-Attention Model

Solar energy is currently one of the most mainstream forms of clean energy. Both the government and relevant companies have dedicated substantial human and material resources to the advancement of solar energy. The prediction of solar power generation provides strong data support for the government and enterprises in formulating energy policies and long-term plans. However, solar power generation is influenced by various factors such as irradiance and temperature, and the relationship between these factors and solar power generation is nonlinear and unclear, which leads to significant fluctuations in solar power output. Traditional prediction models struggle to handle such complex relationships. To address this issue, this paper proposes a novel hybrid model for predicting solar power generation and compares it with several traditional prediction models.