Fault diagnosis model of power transformer based on an improved binary tree and the choice of the optimum parameters of multi-class SVM

An improved binary tree algorithm is proposed for the practical problem of the relativity position of the data sets for oil-immersed transformer in the pattern feature space. And a fault diagnosis model of dissolved gas analysis (DGA) based on an improved binary tree multi-class support vector machine (SVM) is constructed. This method overcomes the disadvantage that the traditional binary tree, which doesn´t consider the distributing situation of the data sets, constructs directly the SVM classifier. At the same time, the two-divided method presented by the paper is applied in the choice of the optimal parameters of SVM. The experiment is performed and this method acquires a better performance.