A gradient-based strategy for integrating Real Time Optimizer (RTO) with Model Predictive Control (MPC)

In the process industries it is often desirable that advanced controllers, such as model predictive controllers (MPC), control the plant ensuring stability and constraints satisfaction, while an economic criterion is minimized. Usually the economic objective is optimized by an upper level Real Time Optimizer (RTO) that passes steady state targets to a lower dynamic control level. The drawback of this structure is that the RTO employs complex stationary nonlinear models to perform the optimization and has a sampling time larger than the controller one. As a consequence, the economic setpoints calculated by the RTO may be inconsistent for the dynamic layer. In this paper an MPC that explicitly integrates the RTO structure into the dynamic control layer is presented. To overcome the complexity of this one-layer formulation a first order approximation of the RTO cost function is proposed, which provides a low-computational-cost suboptimal solution. It is shown that the proposed strategy ensures convergence and recursive feasibility under any change of the economic function. The strategy is tested in a simulation on a subsystem of a fluid catalytic cracking (FCC) unit.