Research On Algorithm Of Two-Layered Model Predictive Control: Steady-state Optimization And Dynamic Control

Model predictive control(MPC) is a class of advanced computer optimization control algorithms based on models, it many appears in two-layered structure in process industry, called two-layered model predictive control.. The two-layered structure, i.e., the steady state optimization layer and the dynamic control layer, is dominant in MPC technology. Compared with conventional predictive control technology, two-layered structure predictive control can deeply improve the potential economic benefits, while keep the process existing configuration and stable. However, the absorption of two-layered structure brings some new problems for predictive control. Firstly, in the layer of steady state optimization, the feasibility judgment and soft constraints adjustment is the primary problem to be addressed, secondly, the problem is how to track the target calculated by steady state optimization without static error.Aiming at the aforementioned problems, this paper studied on the steady state optimization and dynamic control of the two-layered structure prediction control, and the results are as follows:In the first part, the function of steady state optimization is that the set-points of automatic optimization and realize the real-time optimization(RTO) design fixed-point asymptotic tracking, also realize the optimization of MPC which corresponds to the process of the local economy, respectively, the corresponding is the steady-state target calculation target tracking method and the steady-state target calculation of self-optimization method. When optimization is not feasible, the need for soft constraints adjustment, which will determine the feasibility and soft constraints adjustment and combination of soft constraints adjustment priority descending method has higher practical value. For soft constraints adjustment and prioritization methods, a detailed comparison of calculation time difference between ascending and descending strategy.In the second part, the function of the dynamic control layer is to dynamically track the optimal set points obtained from the calculation of the steady state optimization layer, and to solve the problem of the transient characteristics of the control system. The dynamic matrix control algorithm is simulated and compared, and it is concluded that it has the advantages of convenient on-line calculation, strong real-time control and so on. Therefore, this paper focuses on the tracking algorithm based on dynamic matrix control, including the derivation and carding of the algorithm.In the third part, the steady state optimization and the tracking algorithm based on dynamic matrix are combined to form a two-layered structure predictive control method, which is applied to control the TRMS. The simulation results show the two-layered structure predictive control algorithm has a better tracking performance without static error.