Based on existing research, this paper analyzes and compares the phase information characters between partial discharge signal (PD) and its disturbance, and also extracts the feature of Partial Discharge with complex wavelets in noise environment. The research demonstrates that complex wavelets are more superior to original real wavelets in identifying Partial Discharge signal, and those simulated applications and the datum from laboratory can guarantee the effectiveness of the extraction of the feature of Partial Discharge and the restraint of the strong electromagnetic disturbance.