Systematic design of advanced control structures

Abstract

PID-based advanced control structures are commonly used in process industries. However, their design mostly relies on the experience of the control and process engineers, and not always considers optimal operation. This thesis introduces a framework for the design of PID-based control structures, also considering steady-state optimal operation. In the proposed procedure, we use a priority list of constraints to make operation feasible. We also identify the relevant active constraint switches in the supervisory layer and how to handle each case. The active constraints can be on manipulated variables (MV, input) or on controlled variables (CV, output), and the switching cases are: •CV to CV constraint switching: use selectors. •MV to MV constraint switching: use split range control or alternatively controllers with different setpoints or input (valve) position control. •MV to CV constraint switching: use nothing if the input saturation pairing rule is followed; otherwise, use an MV to MV scheme and a selector to take over control when the main MV saturates.

Control structures that extend the operating range for the controlled variable by using more than one manipulated variable are of interest because they can handle MV to MV constraint switching and MV to CV constraint switching. This thesis gives a closer look to standard split range control and proposes a systematic design procedure for this control structure, considering the different dynamic effects of each MV on the CV. Standard split range control has intrinsic limitations in terms of tuning because it only has one common controller and one design parameter for each MV. To overcome this, we propose a generalized split range controller that uses a baton strategy to allow using independent controllers for each MV and the same setpoint. However, sometimes it is optimal to have different setpoints, and in this thesis we identify in which cases it is optimal to use different controllers with different setpoints and describe a procedure to find the optimal setpoint deviations. The thesis contains several cases demonstrating various classical control structures for supervisory control and it also gives some brief guidelines when to use classical control schemes and when to use Model Predictive Control (MPC).