The selection of suitable control structures has an important influence on the economic performance of process systems in the presence of disturbances. Economics has been incorporated in the control structure selection problem using different formulations based on different criteria. The back-off approach is based on the idea of minimizing the economic loss that results from the need to back off from the active constraints to avoid violating them in the presence of disturbances. On the other hand, self-optimizing control schemes aim at selecting controlled variables and constant setpoint values, such that the economic loss with respect to optimal operation is minimized in the presence of disturbances. This paper presents a comprehensive study of different formulations of the back-off approach that pays attention to steady-state feasibility in the presence of disturbances. We argue that the back-off approach that selects controlled variables and optimal setpoint values by minimizing the average cost in the presence of disturbances is a global self-optimizing control approach. The performance of the different formulations is compared by means of three different case studies.

In industrial chemical plants, the selection of controlled variables, manipulated variables, and setpoint values directly impacts the economics of plant operation. The economic operation of a controlled plant can be improved using measurements to optimize the setpoint values and/or change the control structure online. These corrective actions are typically implemented by a supervisory setpoint control layer, such as a real-time optimization system. In a recent work, the authors studied the steady-state back-off approach for control structure selection that selects the optimal control structure and setpoint values by minimizing the average economic cost while guaranteeing feasibility in the presence of disturbances. In the present paper, we extend the aforementioned back-off approach by considering the presence of an upper real-time optimization layer. We present formulations for control structure selection that consider changes in the setpoint values as a function of measured (or estimated) disturbances and changes to the control structure when the set of active constraints changes. The usefulness of the proposed formulations is demonstrated in simulation on a linear example, an evaporator process, and a reaction-distillation plant with recycle.