Short-term prediction of PV power by coupling QPSO-LSTM and improved mechanism modeling

The intermittency and stochasticity of solar energy leads to the strong uncertainty of its output. In the context of carbon neutrality, the photovoltaic (PV) power forecast has become the key technical bottleneck restricting its rapid deployment and connection with the power grid. This paper proposes a combination forecasting model for short-term PV power by coupling data-driven and mechanism modeling methods to improve the accuracy and reliability of PV power prediction. First, based on historical data, QPSO-LSTM and improved mechanism prediction model are established, respectively. Secondly, the minimum error sum method and entropy weight method are used to determine the weight of a single prediction model, and the combined prediction model is constructed. Thirdly, the cumulative prediction error and root mean square error are used as indicators to compare the accuracy and reliability of different prediction models. Finally, a photovoltaic power station in Zhangjiakou is taken as a case study. The results show that the combination forecasting model can effectively improve the accuracy.