Research On Transformer Fault Diagnosis Based On Vector Machine

Intelligent fault diagnosis of transformer is the main link to promote the development of the smart grid.With the development of sensors and computers,the potential failure of transformer is found under the guidance of the new thinking of “Internet”.Therefoer,research on transformer fault diagnosis based on vector machine is proposed,which based on DGA technology and DAG-SVM algorithm to detect transformer state quickly and effectively.Firstly,the single DAG-SVM transformer fault diagnosis model is used to diagnose the transformer,through simulation analysis,the accuracy of single DAG-SVM transformer fault diagnosis model based on linear kernel function,radial basis kernel function,polynomial kernel function and S kernel function is 62.98%,73.68%,70.18% and 70.18% respectively.In conclusion,there are lower accuracy rate for single DAG-SVM transformer fault diagnosis model based on different kernel function,which because of two main reasons include the poor quality of the training sample and the defect of the model itself.Therefore,the dissertation presents two improved DAG-SVM fault diagnosis models of transformer based on improved GA sample selection and hybrid kernel function respectively.The simulation analysis shows that the accuracy rate of two improved models more than 77%,but there are a slight improvement compared with the unimproved single DAG-SVM transformer fault diagnosis model.The improved single DAG-SVM transformer fault diagnosis model cannot fully meet the actual requirements,so that the DAG-SVM fault diagnosis model of transformer based on improved integration is proposed in this paper.The samples are selected as training samples,and the hybrid kernel function DAG-SVM transformer fault diagnosis model is used as weak classifier for Bagging integration to obtain the corresponding strong classifier.Because the strong classifier will refuse to classifying some samples,which will affect the accuracy of the result,so that the BP neural network is used to compensate and improve it.Then the DAG-SVM-BP fault diagnosis model of transformer based on Bagging is obtained.The simulation results show that the accuracy of this model is 92.98%.Finally,based on MATLAB GUI,an expert system for transformer fault diagnosis is established,and the representative model is embedded in it to achieve friendly man-machine interaction.It is convenient for practical engineering application.