Photovoltaic power prediction based on ResNet-LSTM model

Solar photovoltaic power generation is subject to randomness and fluctuations due to variations in solar radiation intensity, making accurate prediction of photovoltaic power generation both important and challenging. This paper employs an LSTM model based on residual networks (ResNet) using two meteorological variables as inputs. The residual connections add the inputs and outputs of the LSTM layers for prediction calculations. The model's accuracy is validated using actual measured data from the Alice Springs photovoltaic power station in Australia. The results show that the ResNet-LSTM model achieves lower RMSE, MSE, MAE, and R² error metrics compared to the LSTM model, with overall better performance. This prediction method is significant for ensuring grid stability and improving photovoltaic power utilization, with important implications for widespread application.