Fault Diagnosis Of Electronic Transformer Based On Wavelet Energy Entropy

In recent years,electronic transformers have been widely used with the large-scale construction of intelligent substations,because of their significant advantages in insulation,cost and performance compared with traditional electromagnetic transformers.Electronic transformers have become a research hotspot.However,in the long-term operation of electronic transformers,it is prone to occur deviations such as variation ratio,drift,and even overall failure,which result in distortion of the measurement results.in more serious cases,protection errors may occur which directly affects the normal operation and function realization of the twice system.In addition,the practical time of electronic transformers is not too long,the accumulated experience and fault samples are few,which result in the lack of effective diagnostic means.Therefore,the research on the fault diagnosis of electronic transformers is of great significance.This thesis focuses on the diagnosis of fault patterns such as fixed deviation,drift deviation and variable ratio deviation which are common to electronic transformers.On the basis of introducing the structure,working principle and fault mathematical model of electronic transformers,aiming at the characteristics of time domain abrupt change and large energy distribution in different frequency bands,this thesis proposes an electronic transformer fault diagnosis method based on wavelet energy entropy.This method uses the SVM(Support Vector Machine)method to classify fault patterns that are distinguished by the energy entropy feature,which can effectively diagnose a single fault category.For the problem of spectral entropy is easily overlapped and different fault types cannot be effectively separated when composite faults occur.A new diagnostic method based on the combination of stratified energy entropy and interval estimation is proposed.The energy entropy expectation of each layer under the decomposition scale is defined,the wavelet entropy bands of each scale of the sample are calculated under the premise of confidence,the fault identification problem of overlapping regions of energy entropy is solved by the matching of wavelet entropy bands,which can effectively distinguish composite faults.The dynamic model experiment of grid faults and the above-mentioned transformer fault modes are carried out on the Nanrui Jibao PCS931 line protection device,the obtained experimental data is analyzed and processed by MATLAB platform and the effectiveness of the above methods are verified.