With the development of the low-altitude economy， Urban Air Mobility （UAM） confronts the challenges in risk management and efficiency due to dense flight operations. This study addresses the pre-flight phase of urban low-altitude Unmanned Aerial Vehicle （UAV） swarms， and introduces a two-stage optimal scheduling approach for flight plans utilizing complex network. Initially， considering the potential third-party risks of UAV flight， the four-dimensional trajectory for individual UAVs is pre-planned within a digital airspace grid environment to generate an initial four-dimensional flight plan for the UAV swarm. Subsequently， to address flight uncertainty， a conflict detection model is proposed to construct a complex network where flight plans are nodes and conflicts are edges. Important flight plans are identified by analyzing topological metrics of the complex network. Finally， an integrated flight plan optimal scheduling model is established， incorporating strategies such as ground holding， speed adjustment， and local re-routing. A two-stage optimization algorithm frame is developed to globally optimize key flight plans and then locally adjust remaining conflicts， and is solved based on the improved Fata morgana algorithm. Experimental results demonstrate that this method can significantly reduce or eliminate flight conflicts considering flight uncertainty， ensuring that delayed flight plans constitute no more than 3% of the total and increasing operational risk by no more than 1%. These findings offer technical support for urban low-altitude flight plan scheduling management， and are instrumental in fostering safe and orderly progression of urban air traffic.