Steady-state target real-time optimization for adaptive constrained generalized predictive control

Model predictive control plays an important role in hierarchical control. It receives the set-point from real time optimization layer hourly and gives a dynamic control signal to basic control loops or the system in minutes. To solve the problem brought by the frequency difference in hierarchy design, the model predictive control layer is divided into two parts: steady-state target optimization and dynamic predictive control. The steady-state target optimization receives the set-points and recalculates the targets every moment before the dynamic predictive control executes. However, for the case that system property varies, the steady-state optimization will lose its feasibility in fixed model. In this paper, an steady-state model updating mechanism is proposed along with the adaptive predictive model updating mechanism. Simulation results on a two tank model show that the steady state target is re-optimized in real time, and good dynamic performance is achieved.