Abstract: Photovoltaic (PV) power forecasting is challenged by its inherent variability. Pure data-driven models struggle with generalization under data scarcity and complex weather conditions. In this paper, we introduce a physics-informed deep learning hybrid model (PIDL-HM) that systematically generates physically grounded input features (e.g., plane-of-array irradiance and module temperature) for a convolutional neural network–long short-term memory (CNN–LSTM), establishing a principled integration framework beyond simple ensemble methods. Rigorously validated across multiple PV power plants in China and Australia for 15-min, 4-h, and 24-h forecasting, our approach demonstrates superior performance, with a reduction of up to 8.93% in root mean square error compared to a purely data-driven baseline. Crucially, the model shows remarkable data efficiency, maintaining high accuracy with only three months of training data, and exceptional robustness, providing a 6.87% improvement in performance under strong cross-seasonal data distribution shifts. This work provides a reliable and data-efficient forecasting solution, establishing the PIDL-HM as a foundational element for next-generation forecasting systems.

基于物理信息深度学习实现数据高效且稳健的光伏发电功率预测

[1]Agga A, Abbou A, Labbadi M, et al., 2021. Short-term self consumption PV plant power production forecasts based on hybrid CNN–LSTM, ConvLSTM models. Renewable Energy, 177:101-112.

[2]Das UK, Tey KS, Seyedmahmoudian M, et al., 2017. SVR-based model to forecast PV power generation under different weather conditions. Energies, 10(7):876.

[3]Hay JE, 1979. Calculation of monthly mean solar radiation for horizontal and inclined surfaces. Solar Energy, 23(4):301-307.

[4]IEA (International Energy Agency), 2024. Renewables 2024: Analysis and Forecasts to 2030. https://www.iea.org/reports/renewables-2024

[5]Liu LY, Liu DR, Sun Q, et al., 2017. Forecasting power output of photovoltaic system using a BP network method. Energy Procedia, 142:780-786.

[6]Mattei M, Notton G, Cristofari C, et al., 2006. Calculation of the polycrystalline PV module temperature using a simple method of energy balance. Renewable Energy, 31(4):553-567.

[7]Mayer MJ, 2022. Benefits of physical and machine learning hybridization for photovoltaic power forecasting. Renewable and Sustainable Energy Reviews, 168:112772.

[8]Mayer MJ, Gróf G, 2021. Extensive comparison of physical models for photovoltaic power forecasting. Applied Energy, 283:116239.

[9]Niccolai A, Dolara A, Ogliari E, 2021. Hybrid PV power forecasting methods: a comparison of different approaches. Energies, 14(2):451.

[10]Ogliari E, Dolara A, Manzolini G, et al., 2017. Physical and hybrid methods comparison for the day ahead PV output power forecast. Renewable Energy, 113:11-21.

[11]Rao Z, Yang ZM, Yang XP, et al., 2024. TCN-GRU based on attention mechanism for solar irradiance prediction. Energies, 17(22):5767.

[12]SAMR (State Administration for Market Regulation), 2021. Technical Requirements for Dispatching Side Forecasting System of Wind or Photovoltaic Power, GB/T 40607-2021. National Standards of the People’s Republic of the China(in Chinese).

[13]Sun FP, Li LH, Bian DX, et al., 2025. Photovoltaic power prediction based on multi-scale photovoltaic power fluctuation characteristics and multi-channel LSTM prediction models. Renewable Energy, 246:122866.

[14]Zhu CY, Tu YB, Wei QY, et al., 2025. Short-term power prediction of photovoltaic power station based on LSTM-XGBoost model. Solar Energy, 300:113819.

[15]Zhu JB, Li MR, Luo L, et al., 2023. Short-term PV power forecast methodology based on multi-scale fluctuation characteristics extraction. Renewable Energy, 208:141-151.