To address the problem of degraded tracking performance in small UAV swarms operating in near-ground environments due to wind speed disturbances, a hybrid controller scheduling strategy based on wind speed grading is proposed to achieve robustness control for the formation system across all operating conditions. First, a wind disturbance model is established, decomposing the composite disturbance into a slowly varying mean wind, turbulence components, and unknown bounded disturbances, and the wind intensity is classified based on the Beaufort Scale. On this basis, a three-level progressive control architecture is designed. For low wind speeds (Levels 1-2), an optimal control law is developed within a reinforcement learning framework to balance energy consumption and tracking accuracy; for moderate wind speeds (Levels 3), a distributed cooperative extended state observer is introduced to improve disturbance estimation accuracy through network collaboration; for high wind speeds (Levels 4-5), adaptive integral sliding mode compensation is integrated to effectively suppress severe disturbances. Based on the leader-follower consensus principle, a unified analytical framework is constructed using Lyapunov theory and input-to-state stability theory, with a hysteresis-based switching strategy designed to avoid controller chattering at wind level boundaries. Theoretical analysis proves that the formation states exhibit global asymptotic stability and bounded convergence. Simulation experiments demonstrate that the proposed method effectively suppresses disturbances under varying wind speeds and communication topologies, while exhibiting superior performance in energy efficiency, disturbance rejection capability, and dynamic response speed. The approach provides an efficient and reliable solution for autonomous navigation and cooperative operations of near-ground UAV swarms, offering strong practical engineering value.