Abstract: To thoroughly analyze the impact of wingtip connection on the aerodynamic characteristics of composite drones, this study establishes a high-fidelity numerical simulation method based on polyhedral meshes. The accuracy of the computational approach was validated using a three-dimensional low-Reynolds-number FX63-137 straight wing case. Within the Reynolds number range of 1.5×10⁵ to 2.0×10⁵, a combined strategy of theoretical analysis and numerical simulation was employed to systematically investigate variations in the lift coefficient, drag coefficient, and lift-to-drag ratio of both the composite as a whole and its individual units. Based on the findings, a predictive model was developed to characterize how the aerodynamic performance of the formation varies with the number of units. The results indicate that wingtip connection can significantly enhance the maximum lift-to-drag ratio of composite drones. For instance, a two-unit composite achieves an improvement of approximately 29.85% compared to a single unit. This enhancement is attributed to a notable increase in the aspect ratio of the composite, leading to a lift distribution closer to the ideal pattern. Additionally, wingtip connection eliminates tip vortex interference from central units and reduces induced drag. However, as the number of units increases, this beneficial effect gradually diminishes. For example, when the number of units increases from 9 to 10, the improvement in the maximum lift-to-drag ratio at the same angle of attack decreases to about 1.84%. This attenuation occurs because the incremental aerodynamic benefit from adding more units becomes limited and is increasingly diluted relative to the growing reference area of the formation. Moreover, the aerodynamic characteristics of an individual aircraft vary significantly depending on its position within the composite drone. Specifically, the enhancement in aerodynamic performance for the aircraft positioned on both sides is only approximately 50% of that achieved by the central aircraft. Further validation confirms that the proposed predictive model demonstrates high accuracy and applicability in estimating the aerodynamic characteristics of formation drones across various configurations, with errors in both lift and drag coefficients within 2%.