Transformer fault diagnosis based on bayesian network and rough set reduction theory
Qijia Xie, Hui-xiong Zeng, Ling Ruan, Xiao-ming Chen, Hailong Zhang · 2013
Bayesian network's capability of dealing with uncertain problems could be a proper solution to the unreliable conclusion drawn by transformer fault diagnosis due to incomplete data. This paper combined the Bayesian network classifier and rough set reduction theory together, set up the Bayesian network classification model based on expert knowledge and statistical data, integrated the data of DGA and electrical tests as the input set of diagnosis, actualized the probabilistic reasoning and sequencing of potential fault types, and improved the reliability of the diagnosis. Meanwhile, rough set reduction theory was used for minimum reduction of Bayesian network classification model, which effectively reduced the complexity of network structure, reduced the input of the model and better suited practical diagnosis. Experiment proved that this method is capable of dealing with missing information, embodies fault-tolerant feature and can achieve high accuracy. It's a kind of effective method for transformer fault diagnosis.