Abstract: It is difficult to solve the dynamic trajectory of transient voltage through differential algebraic equations in large-scale interconnected power system with low calculation speed, which cannot meet the requirements of online analysis. In this paper, an online method method for transient voltage trajectory based on Koopman theory is proposed. Firstly, the multi-mode coupling oscillation characteristics of voltage dynamic response in the transient process of the complex power system after faults are analyzed. Then, a transient voltage trajectory prediction method based on Hankel matrix enhanced extended dynamic mode decomposition (Hankel-EDMD) algorithm is proposed, where the measurement data are expanded by Hankel matrix to reduce the dependence on response data, enhancing the calculation efficiency of the proposed method. Finally, the accuracy of the proposed trajectory prediction method is further improved through dynamic tracking Koopman modes extraction by means of recursive sliding window. The simulation results of the Nordic32 test system and the standard system of the China Institute of Electrical Science and Technology show that the RMSE of transient voltage trajectory prediction using the proposed method is only 0.42% and 0.44%, which is much smaller than that of the current advanced algorithms, and the prediction time is at least one order of magnitude lower than that of current methods, which proves that the proposed method can realize the online prediction of transient voltage trajectory through a short measurement time window (0.5 s), achieving higher computational efficiency and accuracy. Meanwhile, the proposed method has good robustness of measurement noise, which provides a certain technical support for subsequent transient voltage stability analysis and control.