The predictive model is a main topic in the modeling predictive control(MPC) of the HVAC control field. According to Shenzhen Guoying building, the water storage air conditioning system was studied and modeled. This paper proposes a modeling method that combines the TRNSYS and MATLAB to model the water storage air conditioning system. The MPC method demonstrates good performance on the control of the water storage air conditioning system. And the MPC method requires a predictive model of the controlled object. So this paper focuses on the modeling of the water storage air conditioning system. A joint simulation platform of TRNSYS and MATLAB is built. Using rapid development function of TRNSYS build the models of the water tank and plate heat exchangers. The model of the water storage air conditioning system is identified by using artificial neural networks and the collected operation data through the joint simulation platform. Simulation results show that the prediction model match well the test data. The model identified has high accuracy and strong generalization ability.