To address the problem that the predictive control parameters of the model are fixed and set according to experience during satellite rendezvous and docking trajectory tracking control, a DQN (Deep Q Network) optimized MPC (Model Predictive Control) method is proposed. A relative motion model of the satellite based on the ROE (relative orbit element) is established, considering the thrust constraints. Then DQN is used to improve the MPC so that the satellite can independently select the appropriate control parameters according to the current environment, which has stronger robustness. Through the theoretical study and simulation analysis of the satellite orbit motion model, the feasibility of the DQN-optimized MPC control scheme is verified. The simulation analysis compares the trajectory tracking control effects of using only MPC and DQN-optimized MPC algorithm, which validates the efficiency and excellence of the DQN-optimized MPC, and the simulation results show that the control accuracy has been improved by 9.8%, and the steady-state time has been shortened by 30.6%.