부산대학교 도서관

To improve the effectiveness of ultra high frequency (UHF) method applied to partial discharge(PD) detection caused by insulators in gas insulated switchgear(GIS), and to achieve the development trend prediction, the developing process of PD is studied based on step test method. Phase resolved partial discharge (PRPD) patterns, phase resolved pulse sequence (PRPS) patterns and characteristic fingerprint information are extracted, then an information fingerprint library for the diagnosis of PD severity is built based on the effective information. Clustering algorithm and least square method are employed to build the division rule of development stages, PD's evolution mechanism is then studied considering with the experimental phenomenon. The prediction in the early stage of PD are studied based on Auto-Regressive and Moving Average (ARMA) model theory, the development trend of PD with step changed and nonlinear changed patterns are predicted.