Abstract: The supply air volume fluctuates in a large range or even interrupts during the switching process of mine main ventilators, which is easy to cause the gas concentration to exceed the limit. The commonly used model-based main ventilator switchover control method is difficult to handle the constraints well. The intelligent algorithm relies on expert knowledge and experience, which is subjective and arbitrary. In order to solve the above problems, taking the switchover process of main ventilators in the No.2 Mine of China Pingmei Energy and Chemical Group Co., Ltd. as the research background, the optimal control of supply air volume in the switchover process of main ventilators is studied. Based on the hydrodynamics equation and the concept of graph theory, the dynamic model of the main ventilator switchover process is established. The Taylor expansion is used to linearize it to reduce the computational complexity. The switchover process of mine main ventilator is strongly nonlinear, and the system state is constrained. It is a discrete multi-input and multi-output system. The model predictive control (MPC) algorithm is used to study the air volume control problem. The MPC system of the main ventilator switchover process is designed, which converts the air volume optimization into a quadratic programming problem. The primal-dual neural network is used to solve the optimization problem online to realize the real-time optimal control of the supply air volume during the main ventilator switchover process. The test results show that the MPC system can effectively switch the main ventilator, and the air volume of four air doors can well track the reference value during the switchover process. The operation time of each sampling time is 0.027 s, which meets the real-time requirements of the switchover process. The maximum fluctuation of supply air volume is only 0.9%, which is significantly better than the traditional PID control effect.