Multi-classification Model And Fault Diagnosis Of Power Transformer Based On DGA

Transformer is the key equipment for power system to realize the power transmission and substation engineering.Its operation state directly affects the safety,reliability and stability of the whole power system.Oil-immersed transformers will inevitably fail during prolonged operation.How to diagnose and predict the transformer failure in the internal mechanism of the oil-immersed transformer and the dissolved gas content in the oil is the key to the safe operation of the transformer.In this paper,based on the fault diagnosis of oil immersed transformer,a method of feature extraction and machine learning is used to diagnose and predict transformer faults.The main contents of this paper are as follows:1.Establish the oil immersed transformer fault diagnosis model,combined with artificial intelligence of oil immersed transformer fault diagnosis method.Firstly,study gas source and dissolved gas in transformer oil,and analyzed gas production mechanism of gas in transformer oil,discussed the mapping relationship between the characteristic of transformer fault and gas in insulating oil,get different fault types and the characteristics of the relationship between gas.Based on the special correspondence between the characteristic gas and the transformer failure,through the comparison of the various diagnostic methods in the first chapter,it is necessary to find a more accurate and comprehensive method,and combine the artificial intelligence method to carry on the transformer fault diagnosis.2.Improved KPCA and LS-SVM fault diagnosis method.In this paper,an improved KPCA and LS-SVM fault diagnosis methods are proposed to improve the accuracy of KPCA feature extraction in the dissolved gas content.The method is used to determine the feature space obtained by KPCA nonlinear mapping,eliminating outliers,reducing the effect of outliers on KPCA feature extraction.In the fault diagnosis of oil immersed transformer,the output of LS-SVM is processed by KPCA,and the results prove the feasibility and validity of the method.3.A fault diagnosis method based on ant colony optimization least squares support vector machine is proposed.Aiming at the problem that the kernel function parameter and the penalty coefficient are difficult to be determined in LS-SVM,a transformer fault diagnosis method based on ant colony optimization least squares support vector machine is proposed.This method uses the ant colony algorithm to optimize the selection of parameters in LS-SVM,and improve the classification accuracy and precision of LS-SVM.fault type.To validate the LS-SVM method based on Ant Colony Optimization in the sample data set,the results show that the ant colony optimization after least squares support vector machine model classification accuracy rate reached 92.57%,compared with the LS-SVM algorithm,the classification accuracy rate 9.43 percentage points higher,and get the normal state and 6 kinds of typical fault types.