The partial dischange of GIS equipment has potential safety hazard

which can not be ignored. The UHFmethod is used to study the typical partial discharge process in GIS

and some partial discharge defect models are established to conduct a large number of experiments. The characteristic parameters of partial discharge are obtained via the experimental data

and the characteristics of partial discharge are studied by using the characteristic spectrum of partial discharge. The characteristic quantities are selected and classified

and different types of characteristic quantities are chosen to be substituted into the auto-regressive and moving average（ARMA）model to analyze the development trend prediction of partial discharge in GIS. The results show that: 1)for the linear parameters of partial discharge

the ARMA model can accurately predict partial discharge development trend

and the overall prediction error is less than 10%;2)for the step parameters of partial discharge

the ARMA model can predict partial discharge development trend

but is difficult to give accurate prediction;and 3)for the non-linear parameters of partial discharge

the ARMA model cannot predict partial discharge development trend

but can give extremes and peaks of the parameters for reference.