Research On Transformer Fault Diagnosis And Prediction Based On Improved Support Vector Machine

Power transformer is a complex and extremely important electrical equipment in power grid.It plays a role in changing voltage and transmitting electric energy.Its operation state is directly related to the stability and safety of the whole power system.Once the power transformer fails,it is likely to cause significant economic losses and social impact.Therefore,it is very important to judge the operation state and fault type of power transformer.In addition,based on the monitoring data and artificial intelligence technology,predicting the potential fault of transformer in advance is of great significance to avoid or reduce the economic loss caused by transformer fault shutdown.The traditional dissolved gas analysis(DGA)technology based on dissolved gas analysis in oil can carry out the preliminary fault diagnosis of transformer according to the composition of dissolved gas in transformer oil.However,with the increasing requirements of modern power system for fault diagnosis accuracy,DGA analysis method can not meet the requirements.To solve this problem,a transformer fault diagnosis method based on the combination of DGA and support vector machine(SVM)is proposed to improve the accuracy of fault diagnosis.At the same time,in order to solve the problem of insufficient classification accuracy of support vector machine,K-means clustering algorithm is introduced to optimize the transformer fault diagnosis model of SVM parameters,and then a diagnosis method with high accuracy and fast convergence speed is established.In order to make the fault diagnosis more targeted,a power transformer fault location identification model based on the fusion of three evidence bodies: the relative content and ratio of dissolved gas in oil and electrical test data through D-S evidence theory is constructed.In the example analysis,through the analysis of 174 groups of transformer operation state sample information obtained in advance,the influence of different preprocessing methods on the classification accuracy is studied,and the accuracy of different diagnosis methods is compared and verified by examples.The research results can provide strong theoretical support for operators.At present,the prediction method of dissolved gas content in oil does not consider the interaction factors among oil temperature,load and characteristic gas,so that the transformer fault can not be found in time.In view of this problem,a transformer fault prediction method based on support vector machine algorithm and cross validation idea is proposed.By analyzing the correlation between various factors,the factors with strong correlation with the gas to be predicted are extracted,and the information with weak correlation is eliminated.An improved support vector regression prediction(SVR)model is established,which combines the prediction results with the fault diagnosis method,It is used to predict the fault trend of the transformer in advance,and then take a reasonable maintenance scheme for the transformer.In the example analysis,through the selection of kernel function and the training of five characteristic gases,the optimal parameters are obtained,and the prediction curve and relative error of gases are analyzed.Using the same data samples to test different prediction models,the research results show that the improved support vector machine regression prediction model has higher prediction accuracy than the previous prediction models,and can provide a valuable reference for the maintenance arrangement of power transformer.