A comprehensive arc model is presented. On account of the complex characteristics of arc grounding, a new method of the fault line selection based on the characteristics of transient zero sequence current and the difference of two-phase current with support vector machine (SVM) is proposed. Besides, the relation between transient zero sequence current and the D-value of fault phase and non-fault phase is studied. Then the wavelet analysis is used to transform the signal into the feature band (625~1250Hz) within the correlation analysis. The correlation coefficients of each feeder can be used as the inputs of SVM classification algorithm, and the line selection process of single phase arc-grounding fault is established. Simulation are performed in the EMTP, the data is processed by Matlab. The results show that this method can correctly select the fault line for different neutral point grounding mode, different distance and different fault time.