In order to improve the ability of suppressing disturbance in the pulverizing system, an improved multi-model predictive control algorithm is designed on the basis of considering the historical output and predicted output of each controlled variable in pulverizing system. The algorithm introduces the proportion, integral and differential effect on the deviation of the controlled variables in the objective function, thus improving the dynamic regulation process of the system and reducing the fluctuation of the primary air flow and the outlet temperature, which is important to the precise distribution of the primary air. Simulation results show that the proposed algorithm can improve the ability of the system to resist the disturbance of coal moisture and the stability of a wide range of operating conditions and can realize the safe, stable and economic operation of the pulverizing system.