Partial discharge is an important factor to cause insulation aging. Ultrasonic signal and high-frequency pulse current are selected as partial discharge monitoring parameters

a system based on ultrasonic and highfrequency current sensor network systems are adopted to on-line monitor changes in switchgear partial discharge

and support vector regression（ SVR） and back propagation（ BP） neural networks is respectively used for modeling analysis on amount of switchgear partial discharge. Experimental result data show that coefficient of determination of partial discharge estimation model based on SVR is 0. 867; coefficient of determination of partial discharge estimation model based on BP neural networks is 0. 375. The value of partial discharge estimation based on SVR is consistent with actual value

so SVR model is selected to estimate magnitude of switchgear partial discharge.