Research On Fault Diagnosis Method Of Power Transformer Based On NRS And VNWOA-LSSVM

Power transformer is the key electrical equipment in the power system,once the failure,will have a serious impact on the entire power system and its users,so to ensure its safe and stable operation,and timely deal with potential transformer failures,reduce the occurrence of major accidents,for the stable operation of the power system and the steady development of China’s power grid is of great significance.In view of the current problems of low transformer fault diagnosis accuracy and complex diagnosis methods,this paper conducts the following research:Firstly,the traditional transformer fault diagnosis technology and the diagnosis technology after integrating intelligent algorithm are further studied and the dissolved gas analysis(DGA)method in oil is determined to be used as the basic technology for model construction.Aiming at the problem of complex transformer fault data and ambiguous boundaries,this paper proposes to use neighborhood rough sets(NRS)to reduce the fault data.Since the transformer fault sample data is nonlinear and indivisible,LSSVM is selected as the transformer fault classifier by using the Least Square Support Vector Machine(LSSVM)to transform the quadratic optimization problem into a linear equation system solution.The IEC three-ratio data,commonly used DGA data and NRS reduced data were used as input of LSSVM for fault diagnosis experiments,and the feasibility of NRS in handling transformer fault data and the superiority of LSSVM were verified by experiments,and the NRS-LSSVM diagnostic model was established.Secondly,aiming at the problems of low optimization accuracy and slow convergence speed of the current Whale Optimization Algorithm(WOA),the von Neumann topology(VN)is introduced into the whale algorithm,and the Von Neumann Whale Optimization Algorithm(VNWOA)is proposed by constructing a VN topology for each whale individual.The local search capability of the algorithm is enhanced and the convergence speed is accelerated.By selecting the test function for verification,the global optimization ability of the improved WOA algorithm is significantly improved,which is suitable for the optimization solution of LSSVM hyperparameters.Finally,a fault diagnosis method for VNWOA-LSSVM transformer based on NRS reduction model is proposed.The fault data after NRS reduction is used as the input of each model to diagnose transformer faults,and the basic LSSVM,GA-LSSVM,PSO-LSSVM and WOA-LSSVM diagnostic models are compared with the experiments,and the results show that the best adaptation of the proposed model is the highest,the speed of achieving the global optimal solution is the fastest,and the diagnostic accuracy reaches 98.6%,which proves that the proposed model has high accuracy for transformer fault diagnosis.And in the face of complex situations,it still has strong stability,with good application prospects.