Page 151 - Distributed model predictive control for plant-wide systems

7

Networked Distributed Predictive

Control with Information Structure

Constraints

7.1 Introduction
The majority advantage of distributed model predictive control (DMPC) is that it has the char-
acteristics of good flexibility and error tolerance. This characteristic is based on the fact that the
controllers are relevantly independent from each other. It means that the number of systems
that each subsystem-based MPC communicates with will decrease, and then the flexibility
and the ability of the error-tolerance of the whole closed-loop control system will improve.
In addition, in some fields or processes, the global information are unavailable to controllers
(e.g., in a multi-intelligent vehicle system) for the management or the system scale reasons.
Thus, designing a DMPC which could significantly improve the global performance of the
closed-loop system with limited information structure constraints is valuable.
In the previous chapters, we introduced the basic DMPC and the Nash optimality-based
DMPC where each subsystem-based controller pursues the performance of a local subsystem.
Chapter 6 presents a method to improve global optimality based on global information.
This chapter will propose a coordination strategy which could improve the global per-
formance using appropriate network resources, where the optimization objective of each
subsystem-based MPC considers not only the performance of the corresponding local
subsystem but also those it has a direct impact on. In the optimization, each local controller
takes into account not only the impacts coming from its neighbors but also the impacts
applied to its neighbors for improving global performance. Both the algorithms, where each
subsystem-based controller communicates with each other once a control period and the
iterative algorithm are designed in this chapter.
For the noniterative algorithm, the closed-loop stability analysis is also provided for guiding
local MPC’s tuning. Moreover, the performance of the closed-loop system using the proposed
distributed MPC is analyzed and the application to the accelerated cooling and controlled
(ACC) process is presented to validate the efficiency of this method. For the iterative algorithm
where each subsystem-based MPC exchanges information several times during it solves its

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.