Study On Insulation Fault Diagnosis Of Transformer Based On Dissolved Gas-in-oil Analysis

Dissolved Gas-in-oil Analysis (DGA) is a very effective method of transformer fault diagnosis. The principle of gas generation in transformer is analysed, the traditional methods and artificial intelligence methods for transformer fault diagnosis are introduced. Levenberg-Marquardt algorithm is used to train ANN, the convergence of ANN is accelerated and the propbability of plunging into local minima is decreased, at the same time multiple ANNs method is used, the degree of belief of ANN diagnosis result is given by the variance of multiple ANNs diagnosis result, and the error of diagnosis result is decreased. For representing the fuzzy characteristic of transformer fault, a fuzzy mathematics method integrating key gases method and modified IEC tri-ratio method is presented. A method integrating ANN and fuzzy mathematics by D-S evidence theory is presented, and an expert system for transformer fault diagnosis based on this method is realized.