Online evaluation of the ability of power system to resist power disturbance is of great significance for power dispatcher to understand the power grid frequency stability situation in real time. In this paper, a low-order simulation model for frequency security analysis based on dynamic characteristic node clustering is proposed for large-scale interconnected synchronous systems with obvious time-space distribution characteristics. Based on the electromechanical transient time-domain simulation results of contingency fault set, the coherency identification and clustering of node dynamic frequency characteristics are carried out. Based on the single machine equivalent model and frequency lead-lag characteristics of each coherency node set, a low-order simulation model for online analysis of frequency security is established, which can be used for online fast and refined evaluation of frequency security after contingency fault set. The accuracy and effectiveness of the proposed method are verified by an actual power grid.