At present, the control accuracy of wind tunnel flow field is difficult to meet the large aircraft requirement. In this paper, a multivariable model prediction control is introduced to design the flow field controller. To improve the ability to resist attack angle disturbance in a wind tunnel test, dynamic compensation of the attack angle is proposed to improve the predictive value of static pressure. Considering more new test conditions, a multi-model fusion method is used to solve the problem of establishing the prediction model of the new working conditions. To ensure the implementation of the controller, a tuning method is given based on the first order plus dead model of flow field and the disturbance of attack angle. The tests result in 2.4m wind tunnel verifies that the proposed wind tunnel predictive controller can effectively regulate wind tunnel flow field to achieve high precision, which shows the effectiveness of the propoed controller.