Classification of the Power Transformers using Dissolved Gas Analysis

Abstract

Least squares support vector machines (LS-SVM), artificial neural networks (ANN) and the traditional methods based in the dissolved gas analysis were employed for the detection of incipient faults in power transformers. The diagnosis criteria’s commonly used for dissolved gas analysis (DGA) in transformer insulating oil are Doernenburg Ratio Method, Rogers Ratio Method, IEC 60599, IEEE C57.104-1991 and Duval Triangle. In this paper, a comparative study using ANN, LS-SVM and DGA for the incipient fault detection in power transformers is presented. The artificial neural network and least squares support vector machines approaches results are compared with the DGA results obtained from analysis of the traditional methods already consolidated in the technical literature.