Research On The Online Fault Diagnosis Of Power Transformer Based On A Variety Of Theoretical Integration

A safe and reliable system of the power generation,transmission,substation,power supply is the guarantee of normal operation of the whole power system.The transformer is the cornerstone of a series of links.No matter which link is inseparable from the transformer.How to make the transformer can be reliable and stable operation is crucial,because the whole electric power system safe and stable operation will largely depend on its healthy operation.The key to the problem is How can timely and effective to find out if the transformer normal work.The operating state of transformer on-line monitoring will provide a good way.Based on the widely on the basis of relative literatures.I fully researched power transformer internal fault characteristics of the gas produced and Mastered the physical and chemical mechanism of them.Detailed understanding of the different gas characteristics can corresponding internal fault types.The traditional detection method.Finally we understand traditional detection methods and the latest intelligent fault diagnosis methods.In combination with my meactual work,the main content of my paper is as follows:(1)Illuminating the purpose and significance of transformer on-line fault diagnosis,Present research situation and application prospect of the fault diagnosis technology at home and abroad.Summarizing the methods currently used in the field of fault diagnosis as well as the method of artificial intelligence and the transformer fault diagnosis on the latest trends in development in the future.(2)Studying the characteristics of power transformer insulationstructure,the reliability assessment and inner common faults of transformer and dissolved gas in transformer oil.(3)Based on the connection oil dissolved gases concentration with different fault,there is a certain fuzziness in the boundary of three ratio method.Using fuzzy mathematics technology processing modified the improved three ratio method in the fuzzy boundary point.Establishing a fuzzy comprehensive evaluation method based on the improved three ratio method,and Through the fault samples demonstrate the effectiveness of the diagnosis of fault model.(4)Based on the connection oil dissolved gases concentration with different fault,Using the characteristics of the gas in the oil content regard as the characteristics of the artificial neural network training vector matrix.Make full use of artificial neural network has many characteristics,including parallel processing,learning and memory,adaptive ability and robustness.Puts forward two by far the most common neural network(BP neural network and RBF neural network)and constructing two sets of power transformer fault diagnosis model based on their neural network,applicating the MATLAB software to simulated tests The two models.Finally,we analysis and compare the transformer diagnosis model validity based on The BP neural network and RBF neural network.(5)Introduced the basic concept of D-S evidence theory and deep analysising the principle of D-S evidence theory and D-S evidence theory synthesis rules.Then put forward the transformer on-line fault diagnosis method based on evidence theory.By analyzing a large number of fault data and known in the history of the failure and the experience accumulated for a long time and using the improved method of three ratio method and the characteristics of the gas as a source of evidence to Structure diagnosis model framework.Verifying the model through the actual fault samples,and the final result shows that the established model can achieve satisfactory diagnosis.(6)We brief analyze the deficiency of single fault mathematical theory model.We based on this,established the various theoretical integration oftransformer fault diagnosis model.This model is based on the three methods of intelligent integrated products,including fuzzy comprehensive evaluation,RBF artificial neural network,D-S evidence theory.Allotting their uncertainty before the preliminary diagnosis results of the fuzzy comprehensive evaluation and RBF artificial neural network are normalized respectively.The initial MASS function is determined and the objective of probability distribution is realized.The structure of the complexity of the distribution function is avoided.Through the laws of the synthesis of D-S evidence theory to get the final MASS function and diagnosis results are obtained.The effectiveness of this model are verified through the fault samples.(7)Analyzed a variety of intelligent transformer fault diagnosis model in this paper is effective.Through a large number of fault samples in the diagnosis of data analysis show that the performances of the fusion of many theories in this paper,the diagnostic accuracy of transformer fault diagnosis model is higher than the single fault model of mathematical theory.It can for a single failure model of mathematical theory is weak places cannot even diagnosis has high diagnostic accuracy can be achieved.At the power transformer in the on-line fault diagnosis,this diagnosis is one of the most promising diagnostic method In the future.(8)In this paper,at the end of the chapter on the contents of the previous chapters are summarized,and made a overview the fusion range of theoretical research achievements of transformer fault diagnosis model.Finally we put forward the deficiency of the diagnosis model and how to further improve the research direction in the future.