A gradient-based Economic MPC suitable for industrial applications

Model predictive control (MPC) is typically implemented as a lowerstage of a hierarchical control structure. The upper level stages - known as RealTime Optimizer (RTO) - are devoted to compute, by means of a stationaryoptimization, the targets that the dynamic control stage (MPC) should reach toeconomically optimize the operation of the process. A different alternative consist inincorporating the economic optimization performed by the RTO stage directly in thedynamic optimization that solves the MPC stage. In this way, a single stage controlstructure could be implemented, avoiding the frequent inconsistencies that shows thecommunication between the two stages. In this work a new MPC formulation thatexplicitly integrates the RTO structure into the dynamic control layer is presented.The main properties of the proposed strategy are the simplicity - provided that ituses a gradient-based approximation for the economic cost, the guarantee of stabilityand recursive feasibility and an extended domain of attraction. Several simulationsof a subsystem of a fluid catalytic cracking (FCC) unit were performed to test thecontroller.