Abstract: Accurate and rapid prediction of frequency trajectories following system disturbances is beneficial for stability assessment and emergency control. Proposes a method for predicting frequency trajectory curves of power systems after disturbances based on Physics-Informed Neural Networks (PINNs). This method integrates the rotor motion equation and generator governor characteristics to establish differential equations that describe the system disturbance process, guiding the training of the neural network model. The paper uses the New England 10-machine 39-bus system as a case study for simulation research. By comparing the results with simulations from DIgSILENT and other purely data-driven neural network models, it is demonstrated that physics-informed neural networks significantly reduce the required training data and achieve high accuracy with a simpler network structure.