To achieve the online early fault recognition of the power cable, a r ecognition method using the wavelet energy functions and the Selforganizing ne u ral network was presented. The difference of the zeroorder current between the early fault cable and the normal cable was extracted , and then the wavelet ener gy function value of the difference was calculated as the input vector of the se lforganizing neural network which recognized the fault pattern . The early cab l e faults comprised three type faults. The first type was the one phase fault cau sed by the decreasing insulation between one phase and the ground. The second is the two phases fault caused by the decreasing insulations between two phases or between two phases and the ground. The third is the three phases fault caused b y the decreasing insulations between three phases or between three phases and th e ground. Moreover, the recognition stability of the selforganizing neural net w ork was compared with the BP neural network by using the Euclidian distance. Fin ally, the simulation experiment shows that the proposed method is correct and re liable and the system is stable.