Aiming at the problems that fault recognition rate is low and the reliability is difficult to make quantitative analysis of fault line identification method for power system, a fault line identification approach based on improved Hilbert-Huang Transform and identification confidence degree is proposed. Zero sequence current of each line can be decomposed by applying the improved HHT and the instantaneous phase angle and amplitude information of fault feature point are obtained, and fault line selection voting mechanisms are got by comparing phase and angle relationship. Feature points instantaneous amplitude is used to judge the credibility of the vote and define the identification confidence degree, and then the result of fault line identification is obtained. ATP/EMTP simulator is adopted to build five-feeder distribution network simulation model, with which different fault position, initial fault phase, earth resistance are simulated and the proposed method is validated. The simulation results show the proposed approach has high accuracy, shorten calculation time, and is not affected by the fault position, initial fault phase, and earth resistance.