Photovoltaic Power Prediction Based on Improved Long Short-Term Memory Network

As global efforts toward carbon reduction intensify, photovoltaic (PV) power generation has become a key component of the renewable energy sector. However, PV power output is highly influenced by factors such as weather conditions, temperature, and solar irradiance, exhibiting significant nonlinearity and non-stationarity, which pose challenges for accurate forecasting. This study develops a PV power prediction model based on the Long Short-Term Memory (LSTM) network, using data from the Changzhou PV power station as an example. Through data preprocessing, model construction and optimization, and performance evaluation, the results demonstrate the LSTM model's ability to effectively capture the nonlinear features and long-term dependencies of PV power output. The proposed model achieves high prediction accuracy, meeting practical requirements for PV power dispatching. This research provides technical insights that may enhance grid operation stability and PV station scheduling efficiency, offering valuable support for further studies on renewable energy utilization and energy structure optimization.