With advances in low-altitude flight technologies, unmanned aerial vehicles (UAVs) enabled by three-dimensional maneuverability and minimal reliance on ground infrastructure—have become an effective means to improve response efficiency in disaster relief logistics. Focusing on urban emergency supply delivery under typhoon conditions, this study develops a three-dimensional path planning model that jointly incorporates wind-field disturbances, dynamically varying energy consumption, and communication constraints, and establishes a multi-objective optimization framework coupling energy use and timeliness. A 3D cuboid obstacle representation and axis-aligned bounding box (AABB) intersection detection are adopted to enforce collision-avoidance constraints, which are embedded into an NSGA-II multi-objective evolutionary algorithm. The algorithm is further enhanced via hybrid encoding and a penalty mechanism. Using constructed urban building-cluster scenarios and complex wind-field environments, comparative experiments against MOEA/D show that the improved NSGA-II achieves an average hypervolume (HV) increase of approximately 67.6%, with higher solution stability and a broader, more diverse Pareto front, demonstrating superior solving capability and robustness. To assess control robustness under dynamic disturbances, sensitivity analyses are conducted with respect to wind speed, communication range, and the spatial distribution of remote task points. The results reveal systematic effects of external disturbances on UAV swarm control and scheduling decisions. Accordingly, an adaptive dynamic scheduling mechanism is formulated for three operational phases: initial emergency response, post-disaster recovery, and normal operations. This work provides theoretical support and technical references for UAV-enabled emergency logistics scheduling under extreme weather, and offers key decision-making insights for integrating the low-altitude economy with resilient urban emergency management systems.