An On-line Monitoring and Diagnosis Method of Power Fault Based on Unsupervised Learning

Abstract

In order to obtain higher diagnosis accuracy, a power transformer fault diagnosis method based
on unsupervised learning improved deep belief network (IDBN) is proposed. Through this
method, the training process can be regarded as the initialization process of the weight
parameters of a deep BP network. This method makes IDBN overcome the problems of
traditional BP network falling into local optimum due to improper parameters of random
initialization weights, which leads to long training time and so on. Experiment results show that
the accuracy of the proposed method is better than that of support vector machine and artificial
neural network based on the same data set.