Quantitative prediction of aging state of oil-paper insulation based on Raman spectroscopy

• Published in: AIP advances Vol. 11; no. 3; pp. 035001 - 035001-8
• Main Authors: Chen, Xingang, Chen, Shuting, Yang, Dingkun, Luo, Hao, Yang, Ping, Cui, Weikang
• Format: Journal Article
• Language: English
• Published: Melville American Institute of Physics 01.03.2021; AIP Publishing LLC
• Subjects: Aging; Correlation coefficients; Feature extraction; Insulation; Least squares method; Raman spectroscopy; Regression analysis; Spectrum analysis; Support vector machines
• ISSN: 2158-3226; 2158-3226
• DOI: 10.1063/5.0035682

Accurate monitoring of insulation aging of oil-paper insulation power equipment such as oil-immersed transformers is a key and difficult point in the field of high voltage research. In this paper, a method based on Raman spectroscopy to diagnose the aging degree of oil-paper insulation is discussed. Raman detections of the samples were carried out on a self-built Raman detection platform. The partial least squares method was used to extract and analyze the spectral features. The aging time of the sample was used to supervise the feature extraction of oil-paper insulation Raman data, and the intrinsic mathematical relationship between the Raman features of oil-paper insulation and the aging was excavated. Finally, a quantitative aging diagnostic model based on Raman spectral features of oil-paper insulation to predict its aging state was built with the assistance of the support vector regression method. The results of aging time prediction for 30 test samples show that the mean square error is 0.0123 and the square of correlation coefficient is 0.987. The proposed method provides a new idea for Raman aging diagnosis of oil-paper insulation.