Abstract. Neutral un-effectual grounded system (NUGS) is widely used in Chinese medium voltage power distribution network. However, existing earthed fault location methods in NUGS are not satisfactory in practical application due to weak fault current and unstable arc. Based on analysis of spectrum characteristics of current signals in single-phase earthed fault feeder, this paper proposed an earthed fault location method based on support vector machine (SVM). Firstly, the zero sequence current signal was sampled, and its spectrum was gotten by discrete Fourier transform. Secondly, the three dimensional feature vector that constituted by fundament, harmonic, inter harmonic component, had been extracted as the input of SVM. Furthermore, the SVM was trained with the samples obtained by simulation under different fault distance, different fault resistance and different fault close angle. Finally, the earthed fault feeder is selected using the trained SVM. The simulation results indicate that the fault location by the presented method is more precise.