Research On Method Of On-line Monitoring Of Transformer Winding Deformation Based On Magnetic Field Measurement

As an important equipment for transmitting and transforming electric energy,transformer is affected by electric power resultting from many times external shortcircuits.The impact of electric power may cause irreversible deformation of the winding,which has become the primary cause of transformer winding inter-turn short circuit fault.Therefore,it is of great practical significance to carry out on-line monitoring of transformer winding deformation timely and accurately.In this paper,based on the fact that deformation of the transformer windings will directly affect the magnetic field distribution near the winding,the indirect method of transformer winding deformation on-line monitoring by on-line measuring the transformer magnetic field distribution is put forward.Firstly,the correspondence between transformer winding deformation and its magnetic field distribution is studied.The normal and common deformation models of transformer windings are established by ANSYS,and the corresponding relationship between transformer winding deformation and magnetic field distribution is analyzed qualitatively and quantitatively.The results show that the distribution of the magnetic field of the transformer has different characteristics with the different types of winding deformation.Secondly,the transformer winding deformation classification method based on the distribution characteristics of transformer magnetic field is proposed.Based on the quantitative analysis of transformer magnetic field before and after winding deformation,five kinds of coefficients for determining transformer winding deformation types are defined,and the transformer winding deformation is classified by these five coefficients.The results show that the method can classify the transformer winding deformation intuitively and accurately.Then,based on support vector machine,the transformer winding deformation classification method is proposed.Support vector machine(SVM)was trained and tested by sample of ANSYS simulation.The parameters of support vector machine(SVM)were optimized by cross validation and grid search.According to the influence of the sensor installation scheme on the accuracy of winding deformation classification,an improved particle swarm optimization algorithm is proposed to optimize the magnetic sensor installation.The results show that the proposed method can be used to accurately classify the transformer winding deformation with fewer sensors.At last,a method to determine the degree of transformer winding deformation is put forward.The degree of transformer winding deformation is measured by Euclidean distance.According to the optimal sensor installation scheme obtained by PSO algorithm,by analyzing the relationship between the severity of different types of transformer windings and the Euclidean distance,a method of determining the severity of various types of transformer deformation is proposed,and the determination threshold of transformer winding deformation is given.