A dynamic simulation method for a power system comprises the steps of performing low-order prediction according to a section initial value of a simulation equation to obtain an initial prediction value, and calculating a residual error through an error equation; Judging whether the residual error is converged under a preset condition or not, and if yes, performing low-order prediction on a next section; And otherwise, performing high-order iteration based on the residual until the residual converges to a preset condition, and outputting a simulation result until all integration steps are simulated. According to the technical scheme, a mode of combining high-order spectrum integration and a low-order integration algorithm is adopted, the high-order spectrum integration ensures the calculation precision, and the traditional low-order integration ensures the stability. High precision is achieved under the condition that the step length is not reduced, the influence of the local truncationerror of the low-order integral algorithm and the simulation cumulative error on the simulation result is greatly reduced, and the method is suitable for conventional transient stability simulation and medium and long-term dynamic simulation.