Power Cable Faults Diagnosis Based on the Convex Hull Binary Tree SVM

Aiming at the early fault diagnosis problem with the limited number of training samples for the online power cable, the convex hull binary tree SVM is introduced after the new concepts of the convex hull distance and the partition distance are defined. In the optimal parameter condition, the multiclass classifications among the normal state and 3 kinds of early fault states are accomplished by using the 2-dimensional entropy feature vector of the amplitude and the frequency of the zero-order current. Comparing with 3 different methods, the classification accuracy and the classification speed are improved.