Dempster Shafer Evidence Theory Application for Fault Diagnosis of Power Transformers

Abstract

In this paper, advance diagnosis in power transformers, which is one of the most equipment of power systems. Real gas data from Dissolve Gas Analysis has been used for fault diagnosis. Multi-Layer Perceptron Neural Network, Support Vector Machine and Naive Bayes classifiers are used for fault diagnosis. The data set is included in a preprocessing step for the operation of statistical learning algorithms and also has been used as a training and test data set for classification algorithms. The results from the classifiers are compared. Then, the classifier results are combined with Dempster Shafer Evidence Theory, one of the most effective Data Fusion techniques. For this, mass functions for Data Fusion are obtained from the outputs of the classifiers, and the fusion process is performed using the Dempster Shafer Combination Rule. It is seen that the fusion method has better diagnostic accuracy compared to individual classifiers.