Combining Physical Mechanisms with Data-Driven Approaches for Short-Term Photovoltaic Power Forecasting

Accurate photovoltaic (PV) power forecasting is critical for the reliable and stable operation of the power grid. Most existing studies directly use weather conditions as data-driven inputs without incorporating PV physics models to improve the quality of input weather condition features. Therefore, a novel PV power forecasting method that combines physical models and data-driven approaches is proposed. First, a physical model of the PV system is established to convert solar irradiance data into the instantaneous observed irradiance received by the PV system. Then, this newly obtained irradiance data, along with other weather features, is used to reconstruct the data-driven inputs, achieving short-term power forecasting by combining physical models and data-driven approaches. Finally, simulation analysis is conducted using a real PV system dataset. The results show a significant improvement in forecasting accuracy, validating the effectiveness of the proposed method. The results indicate that this article combines physical mechanisms with data-driven approaches, fully considering the advantages and disadvantages of two commonly used models, greatly improving prediction accuracy.