Data Driven Analysis of Aged Insulating Oils by UV-Vis Spectroscopy and Principal Component Analysis (PCA)

Abstract

The ageing of the insulating oil in a power transformer is a major concern for the power utilities. The effective monitoring of the insulation medium is very critical to prevent failures in the power system network. In this work, a natural ester based oil (NEO) is considered for the study of its ageing characteristics using ultraviolet visible (UV-Vis) spectroscopy analysis combined with principal component analysis (PCA) technique. The NEO considered gives an added advantage of being a biodegradable substitute to the conventional mineral oil (MO). The oil is subjected to accelerated thermal stress in a sealed vessel ageing test setup at 150°C for up to 2000 hours. The samples are taken out at intervals of 500 hours for carrying out the required tests. The UV-Vis spectroscopy is performed for both fresh and aged oil samples to observe the variation in the absorbance spectrum in all the intervals over a wide wavelength range of 350 nm to 800 nm. PCA is a statistical technique which is employed to conduct the data driven analysis of the oil samples. This multivariate method is an important tool for dimensionality reduction and helps in understanding which variables are more relevant in ageing assessment by various plots related to PCA. The findings showed that the UV-Vis analysis combined with PCA helps in evaluating the ageing of a sample by classifying and analyzing the data.