Power system topology automatic identification is the important foundation of automatic visual model generation of electromagnetic transient simulation for large-scale power system. Most traditional algorithms for power system topology identification are limited to single-tier description, which is hard to meet the demands of hierarchical clear layout and modular modeling in large-scale automatic electromagnetic transient modeling. Therefore, a new method of power system topology hierarchical identification based on the compression of incidence matrix and the update of branch pointer vector is proposed in this paper. This method realizes automatic search of substation nodes and hierarchical identification of ‘inter-station topology + in-station topology’ based on large-scale data source, without massive graph search and logical operation, and lays a topology foundation for subsequent work of automatic substation layout and element parameter mapping. The method is tested on a simplified system case and the actual large-scale system case of Southwest Grid of China, with its validity and adaptability verified. The method proposed can be used directly in automatic topology hierarchical identification and visual modeling of complex power system in provincial or regional scale. This work is supported by National Key Research and Development Program of China (No. 2018YFB0905200).