A Decision Tree Approach for Power Transformer Insulation Fault Diagnosis

A novel transformer insulation fault diagnosis method is proposed based on a decision tree in this paper. In terms of history samples library of transformer faults, the method applies entropy based information gain as heuristic information to select test attributes, and uses ID3 algorithm to generate the decision tree. Then, pruning in the tree to eliminate noises, and distilling classification rules are handled. The research shows the method not only possesses rapid induction learning ability and classification speed, but also can effectively compress data and save memory, and is an effect transformer fault diagnosis method. In the end, a practical application indicates the validities of the method.