The Application Research Of Support Vector Machines In Transformer Fault Diagnosis

Power transformer is one of the most important electrical equipment in the power system, it is very important to find incipient faults as early as possible, and it is also an important problem for power department. In this paper, we quest for using artificial intelligent methods to transformer fault diagnosis, major study of Bayesian network classification, support vector machine classification and Euclidean distance classification. According to the Euclidean distances between the transformer's state sorts, we build the multi-classification model of Support Vector Machines; Considering Naive Bayesian prone to misjudge when the samples is less, we construct a new classifier combine na?ve Bayesian and SVM; Considering the deficiencies of one-against-one SVM classifier, such as having inseparable regions and the tardy classified speed, we present a new method to improve one-against-one SVM classifier using Euclidean distance classifier, and construct a new classification model. The experiment results show that the improved classifier has excellence in both classification accuracy and computation efficiency.