Study On Fault Diagnosis Methods For The Oil-Immersed Power Transformers

As one of the most widely used transformers in the Power System,oil-immersed power transformers not only have become an indispensable part in power transmission and distribution,but also guarantee the safety and stability of the whole system.Thus,it becomes essential to put more emphasis on the research of fault diagnosis so as to ensure the safety and efficient operation of the Power System and even the whole society.The approaches of fault diagnosis about transformers could be improved because of its complicated structure and large calculation.The transformers’ fault classification and its corresponding reasons were elaborated based on the analysis of fault mechanism firstly.And then,there is an intensive study of change regulations of characteristic gas in some kinds of fault types through excavating historical data.On the basis of this research,a further explore on the application of multiple radio of gas in fault diagnosis was made.And after that,approaches of electrical tests and the assisted methods of fault diagnosis on transformers were concluded.At last,the three methods of transformers fault diagnosis would be further studied.Firstly,combined the technology of oil dissolved gas analysis with the fault criterion of multiple gas radio,the method of fault diagnosis that is based on Variable Precision Rough Set was raised,which matches the fault category with fault characteristic gas so as to do the diagnosis of fault types.Specifically,Establish a decision table about fault diagnosis of transformers at first,and then simplify this table by means of the attribute reduction algorithm with the weight of significance and information,which followed by constructing the decision table with the minimal attributes for the fault diagnosis of transformers.Finally,diagnostic rules which are the standard and orderly could be extracted in accordance with the new decision table.Secondly,the combination of the theory of VPRS and the RBF neural network was introduced into the fault diagnosis of transformers.In other words,the training samples of RBF neural network was established,which was contained with minimal input vectors on the basis of the fault category of transformers without redundant information.After that,the acquired samples were sent into the RBF neural network for better performances of fault diagnosis which could simplify the structure of RBF neural network.Thirdly,the method about fault diagnosis of transformers which is based on a VPRS-RBF network optimized by Particle Swarm Optimization algorithm was proposed so as to optimize the parameters of RBF neural network.And this method could efficiently reduce the training time of RBF neural network.Finally,the feasibility and effectiveness of these three kinds of transformers fault diagnosis methods was verified by a certain test samples.The results showed that the accuracy of the three fault diagnosis methods is about 90% and its efficiency is better than traditional methods.And,the method about fault diagnosis of transformers which is based on a VPRS-RBF network optimized by Particle Swarm Optimization algorithm is better in accuracy and efficiency than the other two methods.