Study On Transformer Fault Diagnosis Based On MGWO-SVM And PCA Feature Reconstruction

Power transformer is one of the most important equipments in power system.Its operation state is directly related to the security and stability of the whole power grid.With the development of the power industry,the installed capacity of the power system is increasing day by day.Undoubtedly,the transformer is facing the increasingly serious problems of material deterioration and insulation aging,even leading to major power failure accidents,and causing serious economic losses.Therefore,it is of great theoretical and practical significance to conduct in-depth and effective fault diagnosis research on the power transformer and guide the maintenance and repair of the transformer.The fault diagnosis method of power transformer based on dissolved gas data(DGA data)in oil can find the latent fault of transformer accurately and timely and improve the operation and maintenance level of transformer.On this basis,this paper attempts to combine an intelligent algorithm and a data processing method with support vector machine(SVM)which is good at solving the problem of non-linear small sample classification to research a new method of power transformer fault diagnosis.The main contents of this paper are as follows:(1)In this paper,SVM is applied to power transformer fault diagnosis,and its parameters are optimized by grey wolf optimization algorithm(GWO).Aiming at the disadvantages of slow convergence speed and easy to fall into local optimum in the later iteration of GWO algorithm,the renewal mechanism of α,β,δ wolf is improved,differential evolution mechanism is introduced into it,and an improved grey wolf optimization algorithm(MGWO)is proposed.The experimental results show that MGWO algorithm has rich population diversity and excellent global search ability,which can be used as an effective tool for SVM parameter optimization.(2)The evaluation index and reference type of power transformer state are determined,and the feature reconstruction of DGA data set is carried out by principal component analysis(PCA),so as to eliminate the feature confusion caused by data modeling.Min-max normalization is performed on DGA data set and PCA feature reconstructed DGA data set to build SVM model.(3)Training MGWO-SVM transformer fault diagnosis model by using the DGA data set and PCA feature reconstructed DGA data set,and test set samples are used to verify the diagnosis ability of each model.Comparing diagnosis results of grey wolf optimization algorithm,particle swarm optimization(PSO),genetic algorithm(GA)and IEC three ratio method with MGWO-SVM model.The experimental results show that the MGWO-SVM model based on the PCA feature reconstructed DGA data set has the highest model accuracy and fault diagnosis accuracy,has excellent generalization and prediction ability,which canprovide solid technical support for transformer fault diagnosis and has practical application value.