Global predictive control: A unified control structure for decoupling setpoint tracking, feedforward compensation and disturbance rejection dynamics

Abstract

Predictive control has now reached a mature level. However, a unification of long-range predictive controllers, such as unified predictive control (UPC), generalised predictive control (GPC) and partial state model reference control (PSMRC) is still lacking. The paper presents global predictive control (GlobPC), which unifies, for stable processes, these controllers. GlobPC is based on an internal model framework where three independent control criteria are minimised: one for tracking, one for regulation and one for feedforward. Using three criteria allows for the desired tracking to be obtained, and for regulation and feedforward to behaviours in an optimal way, while keeping them perfectly separated. The GlobPC structure also exhibits a clean separation of the stochastic and deterministic parts of the control algorithm. The state–space model representation is used for a more uniform and easy programming of GlobPC for both single-input single-output and multiple-input multiple-output processes. An application to a pilot plant illustrates the benefits of the method.