Joint contrastive learning with multi-level feature extraction for wind turbine operation condition clustering
Measurement Science and Technology · 19.06.2026 · 37 (25) · 256101
DOI: 10.1088/1361-6501/ae71f6 ↗Kurz gefasst
Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 17.06.2026. Verfasst von Lingwen Bao und 6 weiteren Autor:innen. Die Arbeit nennt 48 Literaturangaben. Gefördert von: National Key Research, the Development Program of China, Major Science and Technology Special Project of Sichuan Province, National Natural Science Foundation of China.
Abstract (Originalsprache)
Wind turbine condition monitoring requires robust clustering methods to address operational heterogeneity caused by environmental variations. This paper proposes a joint contrastive learning framework with multi-level feature extraction to enhance the generalization of clustering. The methodology implements: 1) Generation of multi-variant temporal representations through data augmentation, followed by hierarchical feature extraction projecting signals into distinct embedding spaces (subsequence, instance, and cluster levels); 2) A contrastive learning framework that systematically captures discriminative and complementary features through multi-level temporal comparisons; 3) Integration of domain knowledge via joint contrastive constraints to optimize cluster separability. Evaluations of two operational wind farms and the UEA benchmark demonstrate improvements of 9.88% and 8.64% in silhouette coefficient and Davies–Bouldin index for field data, respectively, with a 5.14% accuracy enhancement on public datasets. Experimental results validate the framework’s effectiveness in operational condition clustering under dynamic environments.
Förderung
- National Key Research, the Development Program of China
- Major Science and Technology Special Project of Sichuan Province
- National Natural Science Foundation of China
Zitieren
APA
Bao, L., Ding, K., Wang, S., Zhang, K., Zheng, Q., Luo, L., & Ding, G. (2026). Joint contrastive learning with multi-level feature extraction for wind turbine operation condition clustering. Measurement Science and Technology, 37(25), 256101. https://doi.org/10.1088/1361-6501/ae71f6
Vancouver
Bao L, Ding K, Wang S, Zhang K, Zheng Q, Luo L, et al. Joint contrastive learning with multi-level feature extraction for wind turbine operation condition clustering. Measurement Science and Technology. 2026;37(25):256101. doi:10.1088/1361-6501/ae71f6
BibTeX
@article{Bao2026,
author = {Bao, Lingwen and Ding, Kun and Wang, Shuying and Zhang, Kai and Zheng, Qing and Luo, Liangmei and Ding, Guofu},
title = {Joint contrastive learning with multi-level feature extraction for wind turbine operation condition clustering},
journal = {Measurement Science and Technology},
year = {2026},
volume = {37},
number = {25},
pages = {256101},
doi = {10.1088/1361-6501/ae71f6}
}
EUVIMED