Page 93 - Distributed model predictive control for plant-wide systems

5

Local Cost Optimization-based

Distributed Model Predictive

Control

5.1 Introduction
Consider a distributed system as shown in Figure 4.1, which is composed of many interacting
subsystems, each of which is controlled by a subsystem-based controller, which in turn is
able to exchange information with other subsystem-based controllers. The control objective
is to achieve a specific global performance of the entire system (or a common goal of all
subsystems).
For controlling such a system, the distributed (or decentralized) framework, where each
subsystem is controlled by an independent controller, is usually adopted despite the result-
ing global performance is not as good as a centralized solution. The reasons are as follows:
(1) the classical centralized control solution is often impractical for its lack of tolerance to
control faults and the large computational cost. The whole system is out of control when the
centralized controller fails, and the control integrity cannot be guaranteed when a control com-
ponent fails. (2) The distributed framework, in contrast, has the advantages of fault tolerance,
less computation, and being flexible to the system structure. (3) The development of com-
munication network technologies in process industries, which allows the distributed control
technologies and methodologies to exchange information for improving control, promotes the
development of distributed control solutions [27, 47, 57].
Among the distributed solutions, the distributed model predictive control (DMPC), which
controls each subsystem by a separate local model predictive control (MPC), has become more
and more prosperous [27, 47, 57] since it not only inherits the MPC’s ability to explicitly
accommodate constraints [22, 66, 67, 71, 88] but also possesses the advantages of the dis-
tributed framework as previously mentioned. However, as pointed out by the authors of [32, 41,
42], the performance of the DMPC is, in most cases, worse than that of the centralized MPC.
To improve the global performance of the DMPC, several coordination strategies have
appeared in the literature that accommodate different cost functions for the subsystem-based
MPC. The simplest and most adopted strategy is that each local controller minimizes its own

Distributed Model Predictive Control for Plant-Wide Systems, First Edition. Shaoyuan Li and Yi Zheng.
© 2015 John Wiley & Sons (Asia) Pte Ltd. Published 2015 by John Wiley & Sons (Asia) Pte Ltd.