The invention discloses a system for recognizing a partial discharge mode of power equipment, and the system comprises a signal preprocessing module which carries out the filtering and amplifying of asensor coupling signal. The signal processed by the signal preprocessing module is collected, and the data is transmitted to a data analysis module for analysis. A partial discharge PRPD graph and statistical features are calculated, and the statistical features of the partial discharge graph are inputted into a BP network neural algorithm input layer in a grouped manner, thereby obtaining different group of discharge results, and finally obtaining a discharge mode according to the different distributed weights. The system improves the recognition difficulty of the partial discharge mode. Thesystem can improve the accuracy and sensitivity through employing a high sampling rate and a high pulse capturing repetition rate.