Abstract: Distinguishing the types of single-phase grounding faults in distribution network can formulate fault maintenance strategies and improve fault handling efficiency. Distribution automation equipment, as an important carrier of rapid fault identification and treatment in distribution network, has great differences in the principle and effect of fault classification, and its accuracy rate cannot meet the working requirements of power system. Therefore, a tree structure fault type identification method based on decision tree and multi-core residual network (CART-MRN) is proposed. Firstly, the tree fault classification structure is established, and the multi-domain fault characteristics of voltage and current at fault points are extracted by Fourier transform and empirical mode decompsition (EMD) decomposition. Secondly, by combining feature analysis and information gain, a fusion algorithm model suitable for different small current grounding systems is established. Finally, through the verification and comparison experiment of the recorded wave data on the spot, it is proved that this method can be adopted to quickly and effectively complete the single-phase grounding fault classification and identification, and it is more adaptable.