Dimensionally constrained adversarial attack and defense in wind power forecasting
PLOS One · 27.03.2026 · 21 (3) · e0345284
DOI: 10.1371/journal.pone.0345284 ↗Kurz gefasst
Fachartikel in der Zeitschrift PLOS One, online erschienen am 27.03.2026. Verfasst von Yangming Min und 5 weiteren Autor:innen. Beteiligte Einrichtung: College of electrical engineering, Guizhou University, GuiYang, Guizhou, China. Hauptthemen laut MEDLINE: Wind, Neural Networks, Computer. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Science and Technology Project of Guizhou Province, Innovation Fund of Guizhou University Survey and Design Institute.
Abstract (Originalsprache)
Deep neural networks (DNNs) have achieved remarkable success in wind power forecasting, but DNNs are vulnerable to adversarial attacks that can severely degrade forecast accuracy. Existing studies primarily emphasize attack effectiveness and pay limited attention to attack stealthiness. In this paper, a dimension-constrained momentum iterative fast gradient sign method (DC-MI-FGSM) is proposed for wind power forecasting, which generates highly stealthy perturbations by applying the momentum update mechanism during attack optimization and limiting the perturbation dimensions of input samples. To defend against this attack, a denoising autoencoder (DAE)-based preprocessing defense strategy is developed for wind power forecasting, which resists adversarial attacks by mapping adversarial samples back to their corresponding clean forms. The effectiveness of the proposed attack and defense methods is validated on the public SDWPF dataset under both white-box and black-box scenarios. Compared with existing attacks, DC-MI-FGSM achieves a lower average perturbation percentage (APP), indicating superior attack stealthiness. Meanwhile, it causes more severe degradation in forecasting accuracy, as measured by MAPE, RMSE, and MAE, demonstrating stronger attack effectiveness. For defense, the proposed DAE-based preprocessing strategy effectively mitigates adversarial perturbations, significantly reducing forecasting errors while preserving the original accuracy on clean data. Moreover, it consistently outperforms adversarial training in terms of robustness and usability.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- College of electrical engineering, Guizhou University, GuiYang, Guizhou, China.
- Guizhou University Survey and Design Institute Co., Ltd., Guiyang, Guizhou, China.
- Electric Power Science Research Institute, Guizhou Power Grid Co., Ltd., Guiyang, Guizhou, China.
- Zunyi Power Supply Bureau, Guizhou Power Grid Co., Ltd., Zunyi, Guizhou, China.
ORCID-Profile: Yangming Min, Congmei Jiang
Förderung
- Science and Technology Project of Guizhou Province
- Innovation Fund of Guizhou University Survey and Design Institute
Zitieren
APA
Min, Y., Jiang, C., Zhang, L., Wen, X., Tu, J., & Zhang, J. (2026). Dimensionally constrained adversarial attack and defense in wind power forecasting. PLOS One, 21(3), e0345284. https://doi.org/10.1371/journal.pone.0345284
Vancouver
Min Y, Jiang C, Zhang L, Wen X, Tu J, Zhang J. Dimensionally constrained adversarial attack and defense in wind power forecasting. PLOS One. 2026;21(3):e0345284. doi:10.1371/journal.pone.0345284
BibTeX
@article{Min2026,
author = {Min, Yangming and Jiang, Congmei and Zhang, Liangheng and Wen, Xiankui and Tu, Junling and Zhang, Jing},
title = {Dimensionally constrained adversarial attack and defense in wind power forecasting},
journal = {PLOS One},
year = {2026},
volume = {21},
number = {3},
pages = {e0345284},
doi = {10.1371/journal.pone.0345284}
}
EUVIMED