Systematic Design of Active Constraint Switching Using Classical Advanced Control Structures

Abstract

An important task of the supervisory control layer is to maintain optimal operation. To achieve this, we need to change control objectives when constraints become active (or inactive) as a result of disturbances. In most process plants, the supervisory layer uses classical PID-based advanced control structures, but there is no systematic way of designing such structures. Here, we propose a systematic procedure to design the supervisory control layer using single-loop classical advanced control structures, such that the process achieves steady-state optimal operation when the active constraints change. The active constraints can be on the manipulated variable (MV, input) or on the controlled variable (CV, output). In this paper, we consider all three possible cases: CV−CV switching, which involves selectors; CV−MV switching, which does not need any special structure if we pair according to the input saturation pairing rule; and MV−MV switching, which uses split range control or some similar structure. We illustrate our methodology with two case studies.

Economic Control Structure Selection for Two-Layered Real-Time Optimization Systems

“…If for a particular application the gain is significant, one can design overriding controllers based on the knowledge of the possible active constraint regions, as proposed in Reyes-Luá and Skogestad. 30 The design of the active region identification block in Figure 3 is a complex research topic on its own and falls outside the scope of this paper. A short literature survey on approaches to change the CS when the active constraints change is provided in Section 4.3.4.…”

“…However, as recently pointed out in Reyes-Luá and Skogestad, 30 there is still little research on systematic design procedures of a supervisory control layer using advanced control for this purpose. Reyes-Luá and Skogestad 30 identified three types of switching cases: output CV to output CV switch; input MV to input MV switch; and input MV to output CV switch. For each case, the authors propose guidelines to construct the supervisory control layer based on override control and logical blocks.…”

“…In the general case, however, the set of active constraints of Problem may change for different values of the disturbances d . When this occurs, it is desirable for the control structure to switch to the new active set, which could be accomplished using overriding controllers . In the context of RTO, it has also been argued that the RTO layer should pass to the regulatory control layer not only the setpoint values but also the control structure itself, based on the detection of changes in the active set. ,, The detection of active constraint regions for the design of control structures with optimal operation in mind has been studied by Jacobsen and Skogestad …”

“…The nominal values of the disturbances are X 1 = 5, T 1 = 40, and T 200 = 25, and the disturbance ranges are X 1 ∈ [4.75, 5.25], T 1 ∈ [32, 48], and T 200 ∈[20,30]. The objective is to maximize the profit at the steady state, which is expressed as…”

Multiple-Input Single-Output Control for Extending the Steady-State Operating Range—Use of Controllers with Different Setpoints

“…When we need more than one input (u i , manipulated variable, MV) to cover the whole steady-state range for one output (y, controlled variable, CV), we can use three alternative classical control structures [5]:…”