Transformers fault diagnosis based on Support Vector Machines and dissolved gas analysis

Transformers are one of the critical equipments of electrical system. This paper analyses the merits and flaws of the existing transformer-fault-diagnosis methods, and then utilizes SVC (Support Vector Classification), one of branches of SVM (Support Vector Machine) which has many advantages like great generalization, effectively avoiding "over-learning" and being able to learn from small scale sample, to analyze transformers dissolved gas. Because of the nonlinear classification relationship between DGA and fault diagnosis, this paper unifies the DGA data, selects proper kernel function, solution, and parameters, formulates the SVC with best effect, sets up a two-floor fault diagnosis tree, and ultimately builds the transformer-fault-diagnosis model. Through the self-test and comparison with Three Ratio Method as well as BP neural network, it is improved that the model built in this paper remarkably enhances the accuracy and efficiency of fault diagnosis.