Study On The Application Of Support Vector Machine And Relevance Vector Machine In The Dga Of Transformer Oil

Real-time state of power transformer can be known by transformer online monitoring and fault diagnosis analysis,which can timely reflect the potential faults of transformer,providing reliable guarantee for the safe and stable operation of power grid.Dissolved gas analysis(DGA)of transformer is one of the important methods for the diagnosis of transformer fault analysis.Through the type and content comparison and analysis of all kinds of characteristics gas dissolved in transformer oil,transformer internal hidden faults can be reflected to some extents.This article is based on DGA diagnosis method for transformer fault diagnosis.In this paper,mechanism of gases dissolution of transformer oil is first analyzed,and then the corresponding relationship between transformer fault type and gas content of transformer oil is discussed.Traditional diagnostic methods of DGA are briefly introduced,and the advantages and disadvantages of the three-ratio method and the improved three-ratio method of oil-immersed power transformer are analyzed.Then neural network diagnosis method is discussed as an example of new modern diagnostic methods.Then,in order to further increase the accuracy of diagnosis,support vector machine(SVM)is adopted in the DGA diagnosis.Diagnosis result of SVM is discussed,and simulation shows that SVM has higher accuracy,which proves the feasibility of the method based on SVM,and then the improved method is proposed to enhance the accuracy of diagnosis.At last,considering a series of unique advantages of the relevance vector machine(RVM)compared with other artificial intelligence methods,the relevance vector machine is applied to the oil-immersed power transformer fault diagnosis,and diagnosis effect is discussed.At the same time,a large number of measured data is used to analyze accuracy of different methods.The improved three ratio method,BP neural network,support vector machine and relevance vector machine method are finally adopted for comparative analysis,discusses.The results confirmed that the improved support vector machine and relevance vector machine method has higher accuracy and good feasibility,which is betterthan the traditional diagnosis methods.