摘要：AbstractEconomic optimization is a key tool in ensuring competitive chemical plant operation. Traditional steady-state real-time optimization (RTO) is suboptimal in many applications where the plant exhibits frequent transitions or slow dynamics, thus requiring the use of dynamic RTO (DRTO). Additionally, DRTO algorithms exhibit faster response when able to account for the behavior of the underlying model predictive control (MPC) systems. This work seeks to combine closed-loop (CL) prediction of the plant response under the action of MPC with a scenario based robust modeling approach to account for plant uncertainty. The CL prediction is handled by directly modeling the MPC calculations and reformulating the resulting multilevel optimization problem as a single-level mathematical program with complementarity constraints (MPCC). The proposed robust CL DRTO formulation is compared against a single-scenario nominal CL DRTO in terms of maximizing economic performance in a case study involving a nonlinear CSTR. The robust DRTO is shown to outperform the nominal DRTO in this metric on average across the scenarios tested.
