Study On Transformer Faults Diagnosis Based On The Ensemble Learning With Several Bayesian Classifiers

The power transformer is the key equipment of power system,which is related to the safe and economic operation of power grid.To ensure the reliability of transformer,it is necessary to identify the transformer faults in advance.Due to the complexity of transformer conditions in the case of faults,the existing diagnostic methods can not completely determine the types of transformer fault.At the same time,with the increase of voltage level and the expansion of power grid scale in our country,the number of transformers increases sharply.It is imperative to explore new diagnostic methods.In this paper,the major faults of the transformer is analyzed and the gas production mechanism that produced by insulation materials is summarized.Then artificial intelligence algorithm,such as Bayesian network classifier algorithm is introduced.Specific work is as follows:Firstly,due to the factors of human operation and other disturbances in the field,it is inevitable that gas data will coantiannoise and outlier data during the process of collection and transmission.Therefore,the outlier data mining algorithm based on local outlier factor is used to filter and eliminate some bad data.Secondly,to meet the requirements of the Bayesian classifier algorithm,the ChiMerge discretization algorithm is used to discrete the continuous data.Finally,this paper proposes a combined fault diagnosis method based on multiple Bayesian network classifiers.Based on the previous work,the fault diagnosis of the transformer is carried out by using the combination fault diagnosis method.The experimental results show that the accuracy of the proposed method is higher than that of the single Bayesian classifier,and thusimproving the reliability of the transformer fault diagnosis.