This paper addresses the need for simulating real combat aircrafts of different types in target simulation through Target drone. Initially, a maneuver library containing various basic tactical actions is established, where each maneuver corresponds to a specific control command. Subsequently, in order to enable intelligent maneuver decision-making, the study employs the Deep Reinforcement Learning (DRL) method to train a model that takes air combat situations as input parameters within defined boundary constraints, which based on the Soft Actor-Critic (SAC) algorithm framework specifically, while the model select action commands from the maneuver library. Through analysis of the simulation results, while the rewards stabilize at a relatively substantial value, the effectiveness and feasibility of this approach are validated. The proposed algorithm contributes to simulating advanced fighter aircraft with autonomous decision-making capabilities as realistic targets to a certain extent.