Application of rough set to build transformer fault diagnosis model based on multivariate decision tree
Heming Li · Advanced Technology of Electrical Engineering and Energy · 2004
Through decision tree to build the model of transformer fault diagnosis, not only fault samples with nonnumerical values can be processed, but also rules can be learned and knowledge can be simplified with better adaptive capability. However, there exists certain limitation to describe complex cause and reason relationships through univariate decision tree, in which any symptom may be detected repeatedly along one path. In this paper, transformer fault diagnosis approach based on multivariate decision tree is presented. To improve analytical efficiency, several symptoms can be detected synchronously in one node of the multivariate decision tree. Based on advantages of rough set, such as knowledge reduction and classification, through discernibility matrix of rough set to select symptoms and construct multivariate decision tree for transformer fault diagnosis, the diagnosis knowledge can be reduced. And the diagnosis model is visual and easy to understand. Results of comparison test verify the effectiveness of the approach.