Research On Intelligent Diagnosis Method Of Dransformer Fault Based On Dissolved Gas Analysis

With the rapid development of world economy and science and technology,the global electricity consumption is increasing year by year,and the scale of power grid is expanding year by year.People put forward higher requirements for the quality and reliability of power supply.Transformer is the core equipment of the power grid,once there is a fault,it is easy to cause a large area of power grid fault,resulting in greater economic losses.Improving the technical level of transformer fault diagnosis and optimizing the fault diagnosis method can effectively improve the stability and reliability of the power grid,which plays a very important role in ensuring the quality of people’s production and life.Transformer fault diagnosis is a complex technology,which not only needs to master the basic knowledge and test methods of transformer,but also needs to combine experience for comprehensive diagnosis.The application of intelligent algorithm provides more choices for transformer fault diagnosis.The traditional diagnosis method takes dissolved gas analysis in oil as a typical representative.It can judge the type and severity of fault by detecting the composition and concentration of fault gas in oil.The accuracy is insufficient and the diagnosis efficiency is low.In view of the limitations of traditional diagnosis methods,this paper establishes a transformer fault diagnosis model based on CART algorithm and DGA technology,and uses BO algorithm to optimize the selection process of super parameters of cart model.Simulation results show that BO-CART model has better combination of parameters,simpler model structure,higher diagnostic accuracy than traditional CART algorithm,and can solve the problem of difficult selection of super parameters in traditional decision tree.In order to improve the accuracy of transformer fault diagnosis,based on the LightGBM algorithm and DGA technology,a LightGBM transformer fault diagnosis method based on grid search optimization is proposed.In the fault diagnosis model,five kinds of fault characteristic gases in DGA diagnosis method are taken as the original data of the model,and the grid search method and "k-fold cross validation method" are used to automatically find the optimal parameter combination of LightGBM model.The support vector machine,decision tree and random forest method are compared and analyzed,and the performance of the model is evaluated by confusion matrix.The simulation results show that the LightGBM model has higher diagnostic accuracy,can effectively diagnose transformer faults,and has feasibility and certain advantages.