Image processing based analysis of transformer oil

Abstract

Transformer oil is a mixture of hydrocarbons which can tolerate high temperature and is a excellent insulator. This not only serves as insulator but also as a coolant. Besides this suppresses sparking, arching and corona. Variation in transformer performance will ultimately affect on power supplying company’s revenue and consumers. Hence, to prevent the transformer failure, oil analysis is very essential and there are several methods to diagnose the health of the oil. Various tests like Dissolved Gas Analysis (DGA), Electrical Diagnostics etc. can evaluate the transformer health but this method is very vast, dangerous during oil transaction and lengthy. Image processing based analysis is a quick software based analysis which inputs images of transformer oil, processes it based on suitable image processing techniques and determines the prominent characteristic features of transformer oil i.e. acidity, power factor, NN (Neutralization Number), dissipation factor etc. due to ageing color of transformer oil changes and also these factors also deteriorated. These factors and transformer oil color are very closely interrelated. Also the presence of different gases like H2,C2H6,C2H4,C2H2,CH4, CO2, CO, O2, N2 etc. are very important to be detected in order to detect the kind of fault occurred and about transformer history. Thus in this project the presence of oil in ppm level is detected using linear regression method. Also by using the quantity of these gases the kind of fault in transformer is found out using two methods: key gas method and iec method and GUI models are built for easy graphical interface.