Abstract: A VAV(variable air volume) air-conditioning internal model is established based on experimentation and the analysis of the dynamic behavior of the system.And the whole system is decomposed into seven subsystems.MPC(model predictive control) is used for each subsystem to carry on the local optimum control separately.While ensuring network connectivity and the full sharing of information between the various subsystems,the local goals of the various MPC will be combined into the system-level goal,thus the on-line optimization problem for large-scale VAV air-conditioning system is transformed into small scale distributed optimization problems.Simulation and experimentation study demonstrate the control effect of the system proposed.