Research On Transformer Fault Diagnosis Method Based On Neighborhood Rough Set And Grey Wolf Algorithm Optimized Support Vector Machine

Transformers play an important role in power systems.Its operation status is inseparable from the stability of the power grid.Once the transformer is abnormal or faulty,it will adversely affect the power supply and distribution of the power grid.Dissolved Gas Analysis(DGA)in oil is a relatively accurate method for transformer fault diagnosis.This paper studies the relationship between transformer fault data and transformer fault types based on DGA,and adopts the idea of combining fault data processing with diagnostic model optimization to improve the accuracy of transformer fault diagnosis.Through the development of the transformer fault diagnosis system,the two are integrated to realize the process of online monitoring.The main research contents of this paper are as follows:Firstly,the detection technology of dissolved gas in transformer oil is discussed.Several detection methods of transformer faults are systematically analyzed,and then the method of dissolved gas analysis in oil used in this paper is derived.The causes of transformer fault gases are introduced and the fault gases of transformers are discussed.The different fault types of the transformer and the different gas types generated when the transformer fails are identified,and the relationship between the fault types and fault gases is mapped.The relevant content of online detection and determine the detection process of dissolved gas in oil are introduced,and the process is brought into the online detection method as well.Secondly,a Neighborhood Rough Sets(NRS)reduction model is constructed.The relationship between the ratio of transformer fault data and faults is studied.The ratio type data different from the single gas form is used as the fault diagnosis sample in this paper.Seventeen groups of alternative ratio data are generated and reduced by NRS,and the reduced eight ratio data is obtained.The diagnostic accuracy of three ratio data,seventeen ratio data,basic DGA data and reduced eight ratio data were compared by using the basic support vector machine diagnostic model.Through the experimental analysis,using NRS to reduce the transformer fault ratio data can effectively remove the redundant information in the transformer fault data.Thirdly,a GWO-SVM transformer fault diagnosis method based on NRS reduction model is proposed.The eight-ratio data reduced by NRS is used as the diagnostic sample,and the Grey Wolf Optimizer(GWO)is used to optimize the parameters of the Support Vector Machine(SVM)to construct the transformer fault diagnosis model.And the basic SVM diagnostic model,Genetic Algorithms(GA)optimized SVM diagnostic model,Particle Swarm Optimization(PSO)optimized SVM diagnostic model and GWO-SVM diagnostic model are used to diagnose transformer faults to compare and verify the accuracy of the diagnostic model in this paper.Through experimental analysis,the GWO-SVM model established in this paper has a high accuracy rate.Finally,a transformer fault diagnosis system based on Lab VIEW is developed.The system includes user login module,data input module,data reduction module,fault diagnosis module and fault early warning module.By calling MATLAB script,the NRS program and GWO-SVM diagnostic model are embedded in the transformer fault diagnosis system and the fault diagnosis is visualized,which reflects the integrity of the research content in this paper.