Partial discharge signal denoising method based on frequency spectrum clustering and local mean decomposition. Issue 10 (19th February 2021)

• Record Type: Journal Article
• Authors: Li, Shanjun; Sun, Sashuang; Shu, Qin; Chen, Minwei; Zhang, Dakun; Zhou, Dianbo
• Is Part Of: IET science, measurement & technology. Volume 14:Issue 10(2020)
• Page Start: 853
• Page End: 861
• Publication Date: 2021-02-19
• Subjects: signal denoising -- interference suppression -- pattern clustering -- partial discharge measurement -- white noise -- power apparatus -- insulation testing -- time‐frequency analysis -- Hilbert transforms -- statistical analysis; electrical equipment insulation -- frequency spectrum clustering -- local mean value decomposition -- narrow‐band interference frequencies -- white noise -- local mean decomposition -- intrinsic mode functions -- empirical mode decomposition -- denoised PD signal -- denoised components -- PD denoising methods -- background noise -- partial discharge signal denoising method -- K‐means clustering; Measurement -- Periodicals; Electrical engineering -- Periodicals; Electronics -- Periodicals; Nanotechnology -- Periodicals; Electromagnetism -- Periodicals; Medical instruments and apparatus -- Periodicals; 621.3
• DOI: 10.1049/iet-smt.2020.0061
• Languages: English
• ISSNs: 1751-8822

Abstract

Abstract : Suppressing the background noise of partial discharge (PD) is one of the key issues for accurately diagnosing the state of electrical equipment insulation. To solve this problem, this study proposes a new denoising method based on frequency spectrum clustering and local mean value decomposition. First, the K ‐means clustering is employed on the frequency spectrum to pick out narrow‐band interference frequencies. Next, the PD signal with white noise is decomposed by local mean decomposition into different product function components, and the components contain more information about time–frequency than the intrinsic mode functions originated from empirical mode decomposition. Besides, the adaptive threshold is utilised to eliminate white noise in the components. Finally, the denoised PD signal is synthesised by these denoised components. The proposed method and three traditional methods are applied to simulated and field‐detected noisy PD signals, respectively. The results of the evaluation indicators confirm that the proposed method is better than the existing PD denoising methods.