Abstract: The dispatching of the wind-thermal combined power generation system is significantly affected by the wind power uncertainty and wind-thermal synergistic characteristics. Therefore, this paper proposes a day-ahead and intra-day stages cooperative dispatching method for a wind-thermal combined power generation system based on the improved robust optimization of wind power uncertainty sets. First, a day-ahead and intra-day stages wind-thermal cooperative dispatching model is established, and the uncertainty set of wind power is constructed based on the "closed-loop feedback-type" data-driven robust optimization. Next, a wind-thermal cooperative robust dispatching method considering the improvement of the uncertainty set of wind power historical prediction error is proposed to achieve the dynamic random search of the day-ahead and intra-day stages wind-thermal cooperative dispatching plan. Then, the uncertain dispatching model is transformed using the robust duality theory, and the transformed mathematical model is solved by the branch and bound method, which is improved based on the feedback flow relaxation method and the effective uncertain domain space identification constraint method. Finally, cases validation is conducted on the IEEE-39 node system, the results show that, compared with the conventional wind-thermal bundled dispatching modes, the proposed method enhances the wind-thermal coordinated dispatching ability and economic efficiency, and significantly promotes wind power consumption.