Study On The Models Of Fault Diagnosis About Power Transformer Based On Dissolved Gases Analysis

ABSTRACTThe operation reliability of the power transformer, the key equipment in electrical power system, is closely related to the safety and stability of electrical power system. In order to overcome the technological difficulties encountered in the course of insulation fault diagnosis on the base of dissolved gases analysis (DGA), several kinds of mathematic models and actualized methods are brought forward to improve the reliability and veracity of fault diagnosis of transformers. The research works are shown mainly as followings:1) A modified fuzzy multi-criteria method is brought forward for insulation fault diagnosis of transformer. Furthermore, a new method for insulation fault diagnosis is proposed on the base of fuzzy multi-criteria together with rule reasoning. With fuzzy diagnosis in the method, fault reasons are filtrated and then the reasons in low probability are prohibited, Moreover, the left reasons are testified by using rule reasoning and then final concludes are drawn in much less misjudge probability and better results.2) On the base of geometry characteristics of C- partition to sample set of DGA data of transformer, a method is put forward to compute effective radius of neighbor field of a sample, number of clusters and values of initial centers. Moreover, the fuzzy C- means cluster model with adaptive weight is brought forward in the first time and then the fault classifier is designed for insulation fault diagnosis of transformer.3) According to the weakness of the degree of gray of gray incidence (DGI), a new formula to compute DGI is put forward and then incidence order criterion is ascertained. Moreover, a new DGI model for fault diagnosis is proposed according to the further analysis to relationship between fault reason and content of oil dissolved gases of transformer. Instance analysis results show that the method can be used for diagnosis to insulation fault and its location inside transformer with much higher veracity. 4) For the first time, gray cluster theory is introduced to insulation fault diagnosis of transformer. With research on whitenization weight function of gray class, a principle and method to ascertain it are put forward and then a gray cluster model is built for DGA to transformer faults. Instance analysis results present that the method is effective and break the new path for fault diagnosis of transformer.5) The gray prediction theory is introduced to fault diagnosis of transformer in this paper. It is testified that model GM(1,1) is effective for prediction of oil-dissolved gases in transformers. Meanwhile, a new method to compute the back-ground values is proposed according to weakness of traditional model GM(1,1). It is testified that prediction precision to oil-dissolved gases is improved and increasing trend of chromatogram values can be predicted effectively by instance analysis. Moreover, fault types and location can be predicted by the method combined with above fault diagnosis models, 6) Validity of above fault diagnosis models for transformers is analyzed and the analysis results show the higher correctness probability is achieved by the all above models proposed in this paper than that by traditional gas ratio methods. Furthermore, it is ensured which fault type is suitable to be diagnosed by use of each diagnosis models.