A Hybrid Machine Learning Framework for Day-Ahead PV Power Forecasting with NWP Integration and Spatial Downscaling

Zishuo Liu; Zezheng Han

doi:10.54691/wcdqjj94

Authors

Zishuo Liu
Zezheng Han

DOI:

https://doi.org/10.54691/wcdqjj94

Keywords:

Spatio-temporal Feature Integration; LightGBM; Random Forest; Multi-source Data Fusion.

Abstract

Accurate forecasting of photovoltaic (PV) power is essential for secure grid operation, but remains difficult due to meteorological uncertainty and the coarse spatial resolution of numerical weather prediction (NWP) data. To address these challenges, this study develops a hybrid framework that integrates physical modeling with machine learning. The PVWatts model is first applied as a physical baseline. Building on this, historical power records, temporal statistical features, and NWP data are incorporated into a LightGBM regression model for day-ahead forecasting. Comparative t-tests confirm that NWP features significantly improve predictive accuracy, while scenario partitioning using global horizontal irradiance (GHI) further enhances model adaptability. To resolve spatial mismatch, a Random Forest–based downscaling method is proposed, reconstructing high-resolution meteorological features from site-level data. These downscaled features are integrated into the forecasting model, yielding higher accuracy, robustness, and performance under diverse operating conditions. Overall, this study establishes a multi-level PV forecasting system that combines physical mechanisms and data-driven learning. The proposed approach achieves strong adaptability and reliable accuracy, providing effective support for renewable energy integration, smart grid scheduling, and PV plant operation.

Downloads

Download data is not yet available.

References

[1] Das U K, Tey K S, Seyedmahmoudian M, et al. Forecasting of photovoltaic power generation and model optimization: A review[J]. Renewable and Sustainable Energy Reviews, 2018, 81(1): 912-928..

[2] Gaboitaolelwe J, Zungeru A M, Yahya A, et al. Machine learning based solar photovoltaic power forecasting: A review and comparison[J]. IEEE Access, 2023, 11: 40820-40845.

[3] Markovics D, Mayer M J. Comparison of machine learning methods for photovoltaic power forecasting based on numerical weather prediction[J]. Renewable and Sustainable Energy Reviews, 2022, 161: 112364.

[4] Wang K, Qi X, Liu H. Photovoltaic power forecasting based LSTM-convolutional network[J]. Energy, 2019, 189: 116225.

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

[6] Li G, Xie S, Wang B, et al. Photovoltaic power forecasting with a hybrid deep learning approach[J]. IEEE Access, 2020, 8: 175871-175880.

[7] De Giorgi M G, Congedo P M, Malvoni M. Photovoltaic power forecasting using statistical methods: impact of weather data[J]. IET Science, Measurement & Technology, 2014, 8(3): 90-97.

[8] Li P, Zhou K, Lu X, et al. A hybrid deep learning model for short-term PV power forecasting[J]. Applied Energy, 2020, 259: 114216.

[9] Li G, Wang H, Zhang S, et al. Recurrent neural networks based photovoltaic power forecasting approach[J]. Energies, 2019, 12(13): 2538.

[10] Kaaya I, Ascencio-Vásquez J. Photovoltaic power forecasting[C]// Solar Radiation: Measurement, Modeling and Forecasting Techniques for Photovoltaic Solar Energy Applications. Cambridge: Academic Press, 2022: 195.

[11] Wang K, Qi X, Liu H. A comparison of day-ahead photovoltaic power forecasting models based on deep learning neural network[J]. Applied Energy, 2019, 251: 113315.