Abstract: To address microgrid frequency fluctuations under complex source–load disturbances and the limited adaptability of parameter regulation in conventional virtual synchronous generators, a nonlinear model predictive control-based virtual synchronous generator frequency control strategy with dynamic comprehensive disturbance characterization is proposed. Frequency deviation, rate of change of frequency, and active power imbalance are integrated, and their weights are dynamically allocated according to the real-time contribution of each disturbance component to construct a dynamic comprehensive disturbance intensity. This disturbance information is further incorporated into nonlinear model predictive control, where virtual inertia and damping coefficients are directly selected as optimization variables to achieve coordinated rolling optimization of the control parameters. Simulation results under representative operating conditions show that the proposed dynamic comprehensive disturbance characterization improves frequency deviation suppression and dynamic recovery performance under complex source–load disturbances. In addition, nonlinear model predictive control further enhances the overall frequency response through rolling optimization of virtual inertia and damping, thereby improving the frequency regulation capability of low-inertia microgrids.