Main power cable incipient faults are single phase short circuits with short duration and little effect on system operation, so the detection of cable incipient faults is needed to avoid them develop into permanent faults. The phase current characteristics of cable current caused by different disturbances are compared and the incipient faults current characteristic quantities are extracted by multi-scale wavelet transform, and fault current high frequency detail coefficients, low frequency approximation coefficients and the maximum module are taken as compound criterions to detect the transient process of cable over-currents. The time-frequency eigenvectors of cable incipient faults are created and the gray correlation analysis is used to obtain the relevancy of fault currents between fault currents of under-test and standard cable incipient faults in order to identify the incipient faults. At last, the cable line model is built in PSCAD/EMTDC, and the correction of the proposed method is verified by simulation analysis.