Optimization of Methodology for Identifying Partial Discharges in Insulating Liquids Through Optical Radiation Analysis Using Machine Learning

The emission of optical radiation is associated with electrical discharges in high-voltage power systems. Detecting, measuring, and analyzing this radiation can enable the identification of the type of electrical discharge and the determination of its source. This article presents the results of measurements and analysis of optical spectra in the context of their application for recognizing different forms of electrical discharges in selected insulating liquids. Optical radiation in the ultraviolet, visible light, and near-infrared range, emitted by electrical discharges in insulating liquids commonly used in electrical engineering, such as mineral oil, natural ester, and synthetic ester, was analyzed. As part of the research, an innovative methodology for measuring and analyzing optical radiation emitted by electrical discharges was proposed, utilizing machine learning (ML) tools for precise analysis and interpretation of measurement data. The research results enrich current knowledge with new parameters and measurement indicators, representing another stage in the development of diagnostic methods that can be used to build expert systems and new diagnostic techniques for high-voltage insulation systems.