Transformer Fault Diagnosis Model Based on Discrete Hopfield Neural Network

Abstract

To overcome the low accuracy of traditional transformer fault diagnosis methods, a new method is proposed combining analysis of dissolved gas (DGA) in oil with discrete Hopfield neural network (DHNN). The eigenvectors of transformer fault diagnosis were obtained via improved Rogers three-ratio method, then a fault diagnosis model of the discrete Hopfield neural network was established. To further test the generalization performance of the established model, a simulation test was done by taking examples of 4 major types of fault diagnosis of the main transformer in a certain substation. The simulation result shows that the fault diagnosis model based on discrete Hopfield neural network has high accuracy, fast speed and good generalization performance, and the practicability and validity of the proposed method are also verified.