Investigation On Winding Deformation And Dissolved Gas Analysis For Transformer Fault Detection

As one of the key equipment connecting power grids of different voltage levels in power systems,power transformers are responsible for voltage conversion and power transmission between power grids of different voltage levels.The safe and stable operation of power transformers is essential for maintaining and ensuring the normal operation of power systems.Transformer fault diagnosis based on dissolved gases in oil is an effective way to find potential faults inside the transformer.However,there are many types of intelligent diagnostic methods currently used for transformer fault diagnosis,and each has its own advantages and disadvantages.Therefore,the study of integrated learning has practical research significance for the development of transformer fault diagnosis.At the same time,winding deformation is one of the common types of faults in power transformers.Although the power transformer windings will not seriously affect the operation of the transformer when slight deformation occurs,if the transformer is not repaired and maintained in time,the degree of winding deformation will be further aggravated.Even the dielectric breakdown has a short circuit between the turns,which seriously threatens the safety and stability of the power transmission system.Therefore,it is of great significance to carry out research on the detection of winding deformation.In summary,this paper mainly does the following work:(1)By analyzing the direct relationship between power transformer faults and dissolved gases in oil,the three representative algorithms of Adaboost,Bagging and Random Forest in integrated learning are studied.Different transformer fault diagnosis models are constructed by using decision tree as weak learner.The validity and accuracy of transformer fault diagnosis based on integrated learning are verified by test set and out-of-package estimation.(2)The wavelet time-frequency transform method is used to replace the traditional fast Fourier transform,and the test is carried out on the winding of a three-phase dry test transformer.The results show that the winding pulse frequency response curve based on Morlet wavelet transform reduces the noise influence and retains important extreme points.In addition,in the repeated test,the winding pulse frequency response curve based on the Morlet wavelet transform maintains good consistency.Under the winding deformation test,as the degree of failure increases,the change of the frequency response curve is more obvious and shows regularity.