In response to spacecraft control system design challenges, a novel intelligent optimal control method based on value functions and Back Propagation (BP) neural network modeling was proposed. By establishing a flight dynamics simulation model, a large number of characteristic points were generated. Based on time-frequency domain indices, a multi-dimensional performance evaluation system was established to assess control performance. Using the value function, the optimal parameter set was selected to establish a neural network mapping relationship, with flight state as input and control parameters as output. This formed an intelligent control algorithm with optimal performance. The feasibility of the proposed method was verified through simulations.