Since test conditions are quite different from actual working conditions, it’s difficult for conventional test method to identify the exact value of synchronous machine dynamic parameters which are related to saturation and eddy currents. Thus, conventional test method cannot satisfy requirements of offline/online security analysis and control. Based on PMU measurements and taking transient processes of d-axis and q-axis stator windings into account, this paper proposes an online algorithm to estimate synchronous machine parameters in Park model. With stator voltage and field voltage as input signals, stator current and field current as output signals, the proposed algorithm chooses particle swarm algorithm as optimization algorithm. Practical examples show that, compared with the design parameters, the similarity between simulation data derived from identified parameters and measured data is much better. The proposed method can effectively improve the identification accuracy of synchronous machine parameters.