@Article{ee.2026.074698,
AUTHOR = {Weijia Tang, Qiang Li, Ningyu Zhang},
TITLE = {Multi-Source Fusion with Patch-Guided Multi-Task Learning for Power Prediction of Offshore Wind Farm Clusters},
JOURNAL = {Energy Engineering},
VOLUME = {},
YEAR = {},
NUMBER = {},
PAGES = {{pages}},
URL = {http://www.techscience.com/energy/online/detail/25684},
ISSN = {1546-0118},
ABSTRACT = {Large-scale offshore wind farm clusters (OWFCs) have been increasingly connected to the power grid, and requires advanced forecasting models to enhance the prediction accuracy of OWFC’s power output. This paper proposes a multi-source fusion with patch-guided multi-task learning for power prediction of offshore wind farm clusters. Unlike traditional graph-based approaches that rely on predefined topological relationships, which are limited in capturing the highly similar but rapidly changing meteorological conditions among closely spaced offshore farms, the proposed model employs a parameter-sharing multi-task learning network to achieves both independence and correlation among offshore wind farm clusters, followed by utilizing a dynamically weighted multi-task loss function to gradually optimize the network parameters. Moreover, the proposed model applies the patch-guided feature learning module further enables natural alignment and fusion of multi-source data. To demonstrate the performance of the proposed model, experiments were conducted on offshore wind farm clusters in three different regions. The results show that the proposed model can obtain an average accuracy improvement of around 24.31% for MAE and 19.14% for RMSE, ensuring prediction accuracy and robustness.},
DOI = {10.32604/ee.2026.074698}
}