To study the decomposition characteristics of SF

6

under different types of negative DC partial discharge（PD）

then to establish the correlation between the SF

6

decomposition characteristics and the PD type

SF

6

decompo-sition experiments were conducted on the SF

6

decomposition platform under four different types of PD. The resultsshow that the SF

6

gas will decompose under the negative DC PD caused by the four de

fects and generate five stabledecomposed components

namely

CF

4

CO

2

SO

2

F

2

SOF

2

and SO

2

. And the concentrations of the SF

6

decomposedcomponents and the characteristic component ratio have different change rules versus discharging time

which couldbe used for PD type characterization. In order to choose characteristic parameters that can effectively characterize thePD types

the SF

6

decomposed component concentration and the characteristic component ratio were treated by thespectral clustering algorithm to recognize the PD types. It is concluded that these two characteristic parameters both can effectively reflect PD type

and the characteristic component ratio behaves better.