Research On Transformer Fault Diagnosis Method Based On Imprecise Probability

With the large-scale construction of UHV projects,China has entered the era of UHV AC-DC hybrid power grids.The mutual influence of AC and DC systems is further aggravated.The problem of“strong and weak cross-cutting" is outstanding,and it is facing new challenges to ensure the safe and stable operation of large power grids.The premise of safe and stable operation of large power grids is the safe and stable operation of power equipment.As one of the most important power transmission and transformation equipment in power systems,power transformers evaluate the operation status of transformers through advanced and reasonable methods,and timely grasp the operating status?The fault situation is of great significance for improving the safety and economy of the operation of the large power grid.With a large number of new equipment and new technologies connected to the power grid,especially the extensive operation of UHV substations,the lack of relevant transformer fault records and test data has become a difficult problem that plagues the evaluation of transformer operating conditions.Under these conditions,many potential faults cannot be accurately predicted and there is no convincing transformer diagnostic result.In the case of insufficient statistical data,the traditional transformer fault diagnosis method based on the large number theorem is no longer applicable.The non-precise probability can replace the original accurate single-valued probability in the form of interval probability in the absence of the fault sample,which becomes an effective method for transformer fault diagnosis.The brief probability theory in this paper includes both accurate probability and inexact probability theory.The inaccurate probability focuses on describing the effective methods of small sample events.Then the probability estimation method is introduced on the basis of the Imprecise Dirichlet Model(IDM),and the inaccurate probability of IDM estimation of transformer fault operation is analyzed.There are many types of transformer faults and various ways to collect fault information.This paper selects the grounding current,partial discharge,oil chromatographic test data and insulation of transformer core clamp which are directly related to transformer fault diagnosis by Pearson correlation coefficient method.Five types of data parameters,such as oil characteristic test data and coil insulation related test data,are used as fault diagnosis basis.Based on this,a new method of transformer diagnosis based on inexact Dirichlet model and reliability network classifier is established.The new fault diagnosis method can output reliable transformer fault diagnosis results in the case of insufficient sample data and test data,and can find latent transformer problems,solve the defects of traditional transformer fault diagnosis means,and significantly enhance transformer fault diagnosis level.