Page 65 - Distributed model predictive control for plant-wide systems

3

Control Structure of Distributed

MPC

3.1 Introduction

As introduced in Chapter 2, MPC is a control strategy typically used in a discrete-time control
context, i.e., control actions are determined in discrete control cycles of a particular duration.
From the beginning of one control interval until the beginning of the next control interval, the
manipulated variables (input actions) stay fixed, i.e., a zero-order hold strategy is employed.
The objective of the predictive control is to determine those actions that optimize the behavior
of the system and minimize costs as specified through the performance index. In order to
find the optimal inputs that could be able to lead to the best performance of the closed-loop
system, the predictive controller uses the prediction model to predict the future behavior of
the system under various inputs and measurable disturbances over a certain prediction horizon.
Once the predictive controller has determined the inputs that minimize the system performance
index, it implements the resulting optimal inputs until the beginning of the next time instant.
And in the next time instant, the predictive controller determines new inputs using updated
information. Hence, the predictive control operates in a receding horizon or rolling horizon
fashion to determine its actions, and sometimes is called receding horizon control.
The control structure of MPC is a very general concept. It includes how to schedule the
controllers, the inputs/outputs of each controller, and the number of controllers that constitute
the MPC control system, e.g., the control system can consist of a single MPC controller or
multiple MPC controllers. Some properties in which MPC control structure can differ are

• the number of controllers,
• the sensors and actuators that each MPC controller is able to access to,
• the information feeding to each MPC,
• the authority relations between the MPCs,
• the communication protocol that the controllers have among one another, serial or parallel.

It is difficult to define all types of structure due to the large amount of properties that they
can have. Fortunately, some mostly accepted types of MPC control structure can be identified,

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.