Genetic programming method for satellite optimization design with quantification of multi-granularity model uncertainty

Utilizing digital tools for satellite optimization design is vital for supporting decision-making in the actual engineering of satellites. The higher the accuracy of the simulation model, the more precise the satellite performance evaluation, and the more valuable the optimization results. Traditional heuristic algorithms have been successful in optimizing satellite parameters but face challenges when dealing with component-level optimization of satellite composition and structure. To address this issue, this paper presents a genetic programming method for satellite optimization design with quantification of multi-granularity model uncertainty. It defines a multi-granularity simulation model for satellites and presents a method for quantifying model uncertainty. Building upon this foundation, it designs genetic programming tree structures and genetic operations, introducing granularity switching criteria to enable on-demand switching of model granularity. Furthermore, based on the correlation between satellite capabilities and subsystems, it defines an active crossover criterion at the subsystem level to expedite convergence speed further. Numerical simulation cases demonstrate the effectiveness of this method, which enables rapid optimization design of satellite component models, providing timely and efficient assistance for engineering applications.