Cross linked polyethylene(XLPE) cables are commolly used in the power grid of mining area, but the operation environment is relatively bad and partial discharge(PD) of cables often occurs.Aiming at the problem of PD identification in the cable, the propagation characteristics of PD signal in XLPE cable are studied.According to propagation characteristics, method combining empirical mode decomposition(EMD) with Teager energy operator is proposed to identify the initial wave head of PD signal at both ends of the cable.The method greatly improves the ability of anti-noise in wave head identification.The radial basis function(RBF) neural network is used to train the training samples.Combined with the time difference between the PD signal and the measurement point at both ends of the cable, the accurate location of PD in XLPE cable is realized.PSCAD/EMTDC is used to build the cable simulation circuit.Simulation results show that proposed method has high PD identification accuracy and small identification error.