Research On Fault Diagnose Of Power Transformer Based On Information Entropy And Grey Relational Analysis

It is significant to diagnose the transformer fault timely and improving the technical levelof transformer maintenance highly because the safe and stable operation of power transformeris the foundation for ensuring electricity supply, as well as for guaranteeing normalproduction and social life. The method based on dissolved gas analysis is a monitoringapproach widely used for power transformer fault diagnosis. Currently, the furtherimprovement of its accuracy and stability becomes a hot research topic.Firstly, the thesis summarizes the mechanism of power transformer oil dissolved gas,analyzes the content of dissolved gas when transformer is in normal operation and thecomposition and content change trend of oil dissolved gas when transformer is in failure. Thespecific composition of dissolved gas is given when transformer is in different fault. Secondly,the method of grey relational analysis is introduced into the fault diagnosis of powertransformer to overcome the defects of three ratio method. By collecting a large number of oildissolved gas data when power transformer in fault condition, the standard spectra for powertransformer fault diagnosis is constructed by using mean generation method. Thirdly, thebasic idea of grey relational analysis is expounded and the calculation principle of quantizedmodels for grey relational analysis is presented. Additionally, the rough correlation model isoptimized to overcome its drawback through introducing the concept of information entropyand assigning weight values objectively, the value of resolution coefficient is optimized. Theweighted correlation model for fault diagnosis is built and the diagnostic procedure is given.Meanwhile, the concept of combination forecast in divination is introduced, and on the basisof establishment of four separate correlation models, the weight values of the combinedmodel are derived by using minimum variance method, and the drawbacks of singlecorrelation model such as insufficient information utilization are overcome to some degree, sothe combined model embodies the obvious mathematical and physical values. Finally, it isproved that the stability and accuracy of fault diagnosis can be improved with fault diagnosisexamples using the weighted correlation model and the combined model.The theoretical basis is provided for the power transformer fault diagnosis technology,through the analysis of composition and content of oil dissolved gas in the faulty powertransformer and the optimization of the general grey correlation model. On the basis ofresearch of the defects in the traditional fault diagnosis methods, the study on powertransformer fault diagnosis is carried out so tentatively and exploratoring that it is practicalsignificance for improvement of the security, stability, and economic operation of transformer.