A novel wind turbine health condition monitoring method based on composite variational mode entropy and weighted distribution adaptation. (May 2021)

• Record Type: Journal Article
• Authors: Ren, He; Liu, Wenyi; Shan, Mengchen; Wang, Xin; Wang, Zhengfeng
• Is Part Of: Renewable energy. Volume 168(2021)
• Page Start: 972
• Page End: 980
• Publication Date: 2021-05
• Subjects: Wind turbine -- Fault diagnosis -- Composite variational mode entropy (CVME) -- Weighted distribution adaptation (WDA) -- Health condition monitoring (HCM); Renewable energy sources -- Periodicals; Power resources -- Periodicals; Énergies renouvelables -- Périodiques; Ressources énergétiques -- Périodiques; 333.794
• DOI: 10.1016/j.renene.2020.12.111
• Languages: English
• ISSNs: 0960-1481

Abstract

Abstract: Aimed at the problem that the complicated working condition of wind turbine and the lack of sufficient target samples, which makes it difficult to conduct effective health condition monitoring (HCM), a novel method based on composite variational mode entropy (CVME) and weighted distribution adaptation (WDA) is proposed in this paper. A series of mode components are first obtained by performing variational mode decomposition (VMD) on the signals under various working conditions. The mode components are analyzed on multi-scale, and then the fuzzy entropy is extracted at different scales. The extracted CVME is input as a feature vector into WDA. The WDA method can effectively reduce the discrepancy of data distribution between the source and target domains by adjusting the weight of the marginal distribution and the conditional distribution, and solve the problem of class imbalance in domains by a weight matrix. The transferability evaluation is used to select the feature sets under auxiliary working conditions with high similarity to the target feature set as the source samples. Finally, the source and target samples are input into the classifier for training and testing. Compared with traditional fault diagnosis methods, experiment shows that the proposed method has higher accuracy in wind turbine fault diagnosis under variable working conditions. Graphical abstract: Image 1 Highlights: CVME comprehensively evaluates the complexity of vibration signals. WDA effectively reduces the discrepancy of data distribution between domains. The transferability evaluation improves the effect of transfer learning. CVME-WDA is effective for wind turbine fault diagnosis under variable working conditions.