Ageing condition assessment of oil-paper insulation using near infrared spectroscopy detection and analytical technique

• Main Authors: Feng Tang, Yin Zhang, Bin Yuan, Yuan Li, Wen-Bo Zhang, Guan-Jun Zhang
• Format: Article
• Language: English
• Published: Wiley 2019-01-01
• Series: The Journal of Engineering
• Subjects: transformer oil; regression analysis; least squares approximations; infrared spectroscopy; infrared spectra; power transformer insulation; paper; data analysis; ageing; power transformers; S; industrial applications; NIRS analysis; quantitative analysis model; effectively spectral information; competitive adaptive reweighted sampling; spectral data quality; evaluation precision; data analytic algorithms; high-voltage power transformer; main insulation type; ageing condition; chemical tests; physical tests; fast response; analytical technique; infrared spectroscopy detection; oil-paper insulation; condition assessment
• Online Access: https://digital-library.theiet.org/content/journals/10.1049/joe.2018.8516

Near Infrared Spectroscopy (NIRS), as one of analysis technologies, has shown promisingly industrial applications for significant properties for recent decades such as fast response, preciseness, non-intrusion etc. Here, the authors employed NIRS coupled with a series of physical and chemical tests to assess the ageing condition of oil-paper insulation, which is responsible for the main insulation type of high-voltage power transformer. Among these procedures, the data analytic algorithms are of utmost importance to determine the evaluation precision. After plenty of trial-and-error, Savitzky-Golay (S-G) convolution was finally utilised to de-noise samples and improve the spectral data quality. The competitive adaptive reweighted sampling (CARS) was used to select the optimal wavelength combination of NIRS, which is found able to fully extract the effectively spectral information and reduce dimensions of spectral data. Based on the above-mentioned techniques, the quantitative analysis model of NIRS was established by partial least squares (PLS), which could synthetically process the spectral data and the degree of polymerisation (DP) of paper samples. The results indicated that compared with the traditional detection methods, the NIRS analysis is a powerful and informative tool to characterise the condition of oil-paper insulation without intrusion or damage to transformers.