Research On Rural Power Network Transformer Fault Diagnosis Technology Based On EW Pentagon And AACM-ANN Information Fusion

As the hub equipment of rural power grid,the safe and stable operation of transformer is the necessary basis to ensure the normal operation of rural residents’ life and production.Therefore,further research on transformer fault diagnosis technology and improving the reliability of transformer operation is of great significance to ensure rural economic and social development.Dissolved gas analysis(DGA),as one of the most widely used transformer fault diagnosis methods,can effectively identify the transformer fault mode and fault severity.However,the traditional DGA diagnosis method ignores the energy difference between dissolved gas components in different oils when distinguishing transformer fault modes,and has the limitation of too absolute boundary,which directly affects the diagnosis performance of the model.In addition,due to the complex fault mechanism of transformer,it is difficult to obtain accurate and reliable diagnosis conclusions only by a single diagnosis method.In order to solve the above problems,based on the existing transformer fault diagnosis methods,this paper studies the transformer fault diagnosis technology from the perspectives of energy characterization and information fusion.The specific research results are as follows:(1)The cracking gas production law of mineral insulating oil is analyzed based on thermodynamic theory and thermodynamic enthalpy change theory.The generation energy of dissolved gas in oil is calculated according to the gas production model of mineral oil cracking and enthalpy analysis,which is used to quantitatively characterize the energy difference between dissolved gas components in different oils.On this basis,the correlation analysis method is used to evaluate the correlation between energy weighted characteristic parameters and transformer fault mode,and the effectiveness of energy weighted characteristic parameters is verified.The results of correlation analysis show that compared with the characteristic parameters of traditional DGA,there is a stronger correlation between energy weighted characteristic parameters and transformer fault mode.(2)Based on thermodynamic theory and thermodynamic enthalpy change theory,an energy weighted Pentagon(EW Pentagon)model is constructed by using spatial analysis theory,which is used to realize the discrimination of transformer fault mode.The effectiveness of the model is verified by example analysis.The results of case analysis show that compared with the Duval triangle1,IEC ratio,Duval Pentagon1 recommended by the current standard and Mansour Pentagon,the diagnostic accuracy of the EW Pentagon method was improved by 26.08%,27.25%,14.51%,and 16.47%,respectively.(3)Based on the basic idea of information fusion,combining asymmetric convolution,attention mechanism and artificial neural network technology,AACM-ANN information fusion model is constructed to realize transformer multi-source decision information fusion.Several simulation experiments are carried out to verify the feasibility and effectiveness of the model.The simulation experiment results show that compared with EW Pentagon,SVM and ELM primary diagnosis models,the average diagnostic accuracy of AACM-ANN is improved by 10.02%,7.95% and 8.78% respectively;Compared with DS and DSmT information fusion models,the average diagnostic accuracy of AACM-ANN is improved by 6.57% and 5.95% respectively.