An Improved Data‐Driven Spatio‐Temporal Downscaling Framework for Photovoltaic Power Prediction
- 26 May 2026
- journal article
- Published by Wiley in Advanced Theory and Simulations
- Vol. 9 (5)
Abstract
Photovoltaic (PV) power prediction methods often exhibit limited accuracy because they rely on low‐resolution numerical weather prediction data that fails to capture microclimate characteristics. To address it, this study proposes a hybrid data‐driven PV prediction framework built around a novel Meteorological Spatial Downscaling Module (MSDM). It first systematically extracts key data features through a Spatio‐Temporal Feature Engineering Module (ST‐FEM). Subsequently, within the MSDM, an Improved Spatial Averaging‐Based Meteorological Field Synthesis (I‐SAMS) method is employed to effectively fuse high‐quality, real‐time meteorological data from neighboring reference stations, generating a high‐resolution, localized meteorological field for the target station. The experimental results demonstrate that the proposed ST‐FEM+I‐SAMS framework achieves significant improvements in prediction performance, reducing the normalized Root Mean Square Error (nRMSE) and normalized Mean Absolute Error (nMAE) by 54.95% and 59.32%, respectively, compared to the baseline model without spatial information. Furthermore, it exhibits robustness under highly variable meteorological conditions. Therefore, these findings indicate that the I‐SAMS method is appropriate for integration with ST‐FEM, which effectively utilizes complex data from neighboring stations as virtual meteorological sensors, providing a cost‐effective, data‐driven solution to improve PV power prediction accuracy.Keywords
This publication has 47 references indexed in Scilit:
- A novel spatiotemporal relation fused network for solar photovoltaic power forecastingEnergy and AI, 2026
- A novel wind power prediction model based on PatchTST and temporal convolutional networkEnvironmental Progress & Sustainable Energy, 2025
- Research on Photovoltaic Power Prediction Method Based on Dynamic Similar Selection and Bidirectional Gated Recurrent UnitAdvanced Theory and Simulations, 2025
- Day-ahead hourly photovoltaic power forecasting using attention-based CNN-LSTM neural network embedded with multiple relevant and target variables prediction patternEnergy, 2021
- Deep learning based forecasting of photovoltaic power generation by incorporating domain knowledgeEnergy, 2021
- A multi-step ahead photovoltaic power prediction model based on similar day, enhanced colliding bodies optimization, variational mode decomposition, and deep extreme learning machineEnergy, 2021
- Extensive comparison of physical models for photovoltaic power forecastingApplied Energy, 2021
- Solar photovoltaic generation forecasting methods: A reviewEnergy Conversion and Management, 2018
- Exploring the potential of tree-based ensemble methods in solar radiation modelingApplied Energy, 2017
- Univariate and multivariate methods for very short-term solar photovoltaic power forecastingEnergy Conversion and Management, 2016