Abstract: It is important to identify partial discharge(PD) type in distribution cable for monitoring cable running state and extending cable life. The fingerprint library used for pattern recognition is generally constructed based on a single defect. If we detect a PD signal based on the traditional pulse amplitude-time series broadband detection system when there is random interference source or multi-station discharge source, misjudgment will occur in the process of pattern recognition. Aiming at pd defects of distribution cables, a method of pd classification and recognition is proposed based on equivalent time-frequency analysis, fuzzy C-means clustering and PRPD spectrum. The pulse waveform is firstly classified according to its characteristic value, and then is identified by PRPD spectrum, which can effectively solve the problem that the traditional direct identification method cannot correctly judge the multi-local discharge source. Firstly, a partial discharge testing system is established, and different types of partial discharge waveforms are obtained. Then, the time-domain waveform of the original discharge is transformed into T-F mode by equivalent time-frequency analysis, and the data in T-F mode is classified by fuzzy C-means clustering analysis, and the PRPD spectrum of each type of discharge pulse is extracted. Finally, the discharge type is identified according to the PRPD spectrum, and the accurate classification and identification of the local discharge signals of the multi-discharge power supply in distribution cables is realized. Experimental results show that this method can classify and identify discharge types effectively in the case of multiple single local discharge sources and two mixed local discharge sources.