Some features extracted from partial discharge(PD) can not directly reflect the physical characteristics of discharge. To solve this problem

the distribution characteristics of discharge amplitude and discharge frequency according to frequency phase are proposed. In this paper

a two-dimensional statistical matrix of the partial discharge

named by partial discharge phase matrix(PPM)

is obtained by counting the number of the different discharge amplitudes on different frequency phases in the multiple frequency cycles. The singular value decomposition of the two-dimensional statistical matrix is used to get the distribution feature of discharge amplitude and discharge frequency according to phase

and support vector machine is introduced to realize the partial discharge identification. The results show that

this feature can not only reveal the distribution characteristics of discharge amplitude and frequency in the frequency phase

but also distinguish the different discharge patterns with higher recognition precision

and detect the same discharge type with different sizes.