The invention provides a multi-level optimization control system and method based on a Stackelberg-Nash differential game, and the system comprises a multi-decision-maker robust control module, a Stackelberg-Nash differential game module, an event-driven generation module, an evaluation network learning module and a system construction module, the multi-decision-maker robust control module guarantees the robustness of the system through an ADP-based robust control method, and the multi-decision-maker robust control module guarantees the robustness of the system through an ADP-based robust control method. The Stackelberg-Nash differential game module describes a multi-level non-cooperative game behavior of a complex system by using a Stackelberg game model, designs cost functions of a leader and a follower, and uses the cost functions to update a control strategy, and the event-driven generation module designs event-driven conditions suitable for multiple decision makers, updates the control strategy when the driving conditions are violated, and generates a multi-level non-cooperative game. The evaluation network learning module trains an evaluation network and solves a solution of a coupling HJ equation under event driving to obtain an optimal control strategy, the system construction module applies the obtained optimal control strategy to a multi-level complex system containing multiple decision makers, and the multi-level optimization control system and method based on the Stackelberg-Nash differential game are formed.