Study On Statistical Parameters And Cluster Diagnosis For Aged Oil-Paper Classification

As the main insulation used in oil-immersed transformer, oil-paper suffers from thermal and electrical aging stress in the long-term running, which deteriorate the insulating performance and threaten the reliability of transformer directly. Aging research work is of vital importance in insulation subject, which ensures the secure and economy are reliable operation of power system.The oil tank is designed for the oil-paper under electrical and thermal thermal stresses. This paper presents a new oil-paper cavity modle to simulate the turn-to-turn insulation defects in transformers. A total of impregnated specimens are placed in the oil tank for the accelerated aging experiment.Aging experiment of this modle under electrical and thermal stresses are taken and partial discharges and degree of polymerization of specimens are measured during the experiment. There are 27 statistical operators are extracted from the partial discharge. The change law of statistical operators is analysed. There are principal parameters are extracted from the 27 statistical operators by factor analysis and compose the parameter matrix. The change law of principal parameters is analysed.The PD parameter vectors composed by either the statistical operators or the principal parameters are clustered by four types of clustering analysis methods, the c-mean clustering, the fuzzy c-mean clustering, the possibilistic c-means clustering and the kernel-based possibilistic c-mean clustering. The clustering results show that the robust clustering results can be obtained by using clustering algorithms with both the PD statistical operators and principal parameters. The possibilistic c-means clustering and the kernel-based possibilistic c-mean clustering show advantages to the c-mean clustering and the fuzzy c-mean clustering method for oil-paper aging diagnosis.