The invention provides an unmanned vessel data driving reinforcement learning control method with specified performance for an unmanned surface vessel system. The method comprises the steps of S1, building an unmanned surface vessel mathematic model, S2, introducing a specified performance function, S3, designing an optimal controller of the unmanned vessel, and S4, designing weight updating ratesof the evaluator and the actuator. According to the method, simultaneous updating of the actuator and the evaluator can be realized, and the error can be within a specified range, so that the optimalcontrol strategy is obtained. Meanwhile, the method accelerates the convergence speed of the control system, and obviously improves the adaptability and reliability of the operation of the unmanned vessel system in an unknown environment.