Research On Transformer Fault Diagnosis Method Based On Grey System Theory

Power transformer is a very important power distribution and transmission equipment in power grid,and it is also one of the most core equipment in the whole power system.According to the concentration of dissolved gas in transformer oil to determine its operation state,timely and effectively find the fault of power transformer,is conducive to the staff to prepare early,avoid causing immeasurable economic losses to the society,so the transformer fault diagnosis has very important practical significance.However,the research of transformer fault can not be limited to fault diagnosis.Accurate prediction of dissolved gas concentration in transformer oil can find potential faults in the transformer and provide important reference for fault diagnosis of transformer.In order to solve the problems of low fault tolerance rate and code missing of dissolved gas fault data in different grades of transformer oil,the Euclidean distance in grey correlation coefficient is replaced by dynamic time warping distance,and the comprehensive weight of fault index is introduced to form grey correlation model based on dynamic time warping distance,Then,the standard fault reference sequence of transformer is obtained by using fuzzy c-means clustering algorithm,and the grey correlation degree model based on fuzzy c-means clustering algorithm is formed.Finally,the two models and three ratio method are combined to form a transformer fault combination diagnosis method based on improved grey correlation degree.The example analysis results show that the dynamic time warping distance and comprehensive weight can deal with the lack of code of dissolved gas data in transformer oil.At the same time,the method can diagnose the faults of different grades of transformers with considerable speed and accuracy.In order to accurately predict the concentration of dissolved gas in transformer oil,aiming at the problems of large long-term prediction error and instability of the basic grey prediction model,this paper first optimizes the basic grey prediction model from three aspects,combines the method with the best optimization effect,and then constantly updates the old information with new information,Finally,Markov chain is used to modify the long-term prediction results,so as to obtain the prediction method of dissolved gas concentration in transformer oil based on Gray Markov chain with small error and good stability.The result of example analysis shows that the average residual value is the smallest and the prediction effect is the best when the background value and initial value of the basic grey prediction model are optimized at the same time.This method can eliminate the error accumulation effect in the long-term prediction and control the prediction accuracy in a reasonable range.