Abstract: The inconsistence of DRTO and MPC models in the two layer structure of integrating DRTO and MPC will deteriorate the effect of optimization. Hence, this paper proposes a multi model generalized predictive controller based on the two layer structure of DRTO. The upper structure is DRTO, which mainly solves the optimization problem of the economic objective function. Dynamic process model is adopted to get a real time updating optimal trajectory. In the lower one, a multi model generalized predictive controller is adopted to replace the previous single model controller in order to attenuate the influence of disturbances. Multiple fixed models and adaptive model are used to identify system dynamic characteristics so as to decrease the errors caused by the inconsistence of two levels’ models. Meanwhile, the transient performance of the system and the ability to regulate the jumping parameter may be improved. Finally, a case study illustrates the feasibility and efficiency of the proposed method．