Research On Transformer Winding Deformation Diagnosis Method Based On Frequency Response Curve Characteristics

The power transformer is the key device for the connection and power transfer of different voltage levels in the power system.Its health status plays a decisive role in the normal operation of the power system.Transformer winding deformation is the important precursor of transformer fault.The timely diagnosis of winding deformation is helpful to discover the hidden danger of transformer fault.The characteristics of frequency response curve to determine the transformer winding deformation is an effective method to diagnose winding deformation.Aiming at the problems of missed diagnosis and misjudgment of frequency response curve method in the diagnosis of transformer winding deformation faults,the related research is studied and three corresponding solutions are proposed in this paper.(1)Aiming at the problem that the existing correlation coefficient method cannot determine the deformation whether the measured frequency response curve of the transformer winding and the reference frequency response curve have the same shape but different amplitudes,the diagnosis method of transformer winding deformation based on signal distance algorithm is proposed in this paper.PSpice circuit simulation software is used to simulate three faults of axial deformation,radial deformation and axis offset of the transformer winding,so as to obtain frequency response data under different winding deformation degrees.Compared with the fourteen methods including correlation coefficient method and Euclidean distance,and combined with actual transformer frequency measurement data to verify,the results of simulation experiments and measured experiments show that as the degree of deformation of the winding increases,the mutual distance value continuously increases between[0,1],and the slight deformation of the winding can be diagnosed.It proves that this method effectively solves the problems of the correlation coefficient method,and the method has high sensitivity and reliability.(2)Aiming at the problem that "manual empirical method" for diagnosing the fault type of winding frequency response curve easily leads to misjudgment,the identification method of transformer winding deformation fault type based on signal distance and Support Vector Machine(SVM)is proposed in this paper.This method introduces the mutual distance applied in the field of power grid fault line selection as the fault identification feature of transformer winding.The improved grid search method and cross-validation are used to optimize the SVM kernel function parameter g and penalty factor C.Compared with the recognition result based on the sub-band correlation coefficient and the SVM,as well as based on the amplitude and frequency change rate of the resonance point and the SVM,the simulation experiments show that the accuracy of winding deformation fault recognition based on sub-band mutual distance and SVM is improved by 7.14%and 3.58%,respectively.(3)Aiming at the problem that the transformer winding deformation monitor on the market is little bulky,the transformer winding deformation monitoring software based on mobile phone APP and QT software system are designed in this paper.The "constant frequency mode","sweep mode" signal command transmission,frequency response data reception,data analysis,data processing and winding amplitude frequency characteristic curve drawing are realized.Furthermore,the real-time monitoring function of transformer winding deformation based on the signal distance algorithm is implemented on the mobile phone APP.