Steady State Optimization Inside Model Predictive Control

One of the major assets of Model Predictive Controllers (MPCs) lies in their ability to manipulate constraints, both on inputs and outputs. When control objectives for process outputs are determined in terms of a bounded range, or if there are more inputs than outputs, it may happen that, even if all control objectives are satisfied, some degrees of freedom remain available. We can then envisage to use them for a steady state optimization. This paper describes the principle of incorporating a steady state economic optimization into a MPC. It allows to: ¿ avoid offsets on controlled outputs because of inappropriate economic objectives given by an upper control level, ¿ optimize a linear objective function subject to certain hypothesis. This optimizer is currently incorporated into the Shell Multivariable Optimizing Controller (SMOC) and is running with success in several Shell refineries.