Abstract: Due to the existence of various disturbances and uncertainties in complex flow environments, the design of flow control systems needs to consider these uncertainties and to adjust the control law automatically to adapt to random and abrupt disturbances. The closed-loop optimal control based on low-order linear models, which is widely used at present, can obtain the predetermined control effect with a small amount of control input, but the modelling accuracy and the unmodeled dynamics of the reduced-order models can greatly limit the adaptive ability of such control strategies. In this study, aiming at the shock buffet problem of airfoils in transonic flow, a model-free adaptive control (MFAC) based on the data-driven method was carried out to eliminate the fluctuating load caused by the shock buffet in different freestream states. The control uses the unsteady Reynolds Average Navier-Stokes (URANS) method for the flow field simulation, the trailing-edge flap as the actuator, and the lift coefficient as the feedback signal. When the flow state changes, the data-driven MFAC uses the input and output data to convert the flow system into a dynamic linearized data model online and minimizes the performance index to obtain the control law, to make the system automatically work in the optimal or close to the optimal state. The simulation results show that MFAC is better than existing open-loop control and proportional control strategies, and can completely suppress the buffet load even if the freestream state changes.