Page 45 - Distributed model predictive control for plant-wide systems

2

Model Predictive Control

2.1 Introduction
Predictive control appears to have been proposed independently by several people, more or less
simultaneously. The pioneers were mostly industrial practitioners who implemented predictive
control several years before the first publications appeared, so the publication dates do not tell
thewholestory.
The first description of MPC control applications was presented by Richalet et al.ofthe
French company Adersa in 1976’s conference [65] and later summarized in 1978’s Automat-
ica paper [4]. They described their approach as model predictive heuristic control (MPHC).
The solution software was referred to as IDCOM. The emphasis of MPHC was on a control
methodology which could be applied to problems too difficult to be handled by conventional
PID control, but which was based on intuitive concepts and offered ease of tuning [66].
In the early 1970s, with an initial application in 1973 [66], engineers at Shell Oil developed
their own independent MPC technology. Cutler and Ramaker presented details of an uncon-
strained multivariable control algorithm which they named dynamic matrix control (DMC) in
1979 [5]. Constraint handling, however, was still somewhat ad hoc at that time. Engineers at
Shell Oil addressed this weakness by posing the DMC algorithm as a quadratic program (QP)
in which input and output constraints appear explicitly, and Cutler et al. first described the
quadratic programming solution of dynamic matrix control (QDMC) algorithm in 1983 [67].
Several years later, a more comprehensive description was published in [68]. This method
emphasized optimal plant operation under constraints, and computed the control signal by
repeatedly solving a linear programming (LP) problem. DMC went on to become the most
well known of the commercial predictive control products [66].
All of these proposals shared the essential features of predictive control: an explicit internal
model, the receding horizon idea, and computation of the control signal by optimizing pre-
dicted plant behavior. Currently predictive control has become the most popular algorithm, and
the only advanced control methodology which has made a significant impact on industrial con-
trol engineering. The main reasons behind its success in these applications are as follows [26]:

• It handles multivariable control problems naturally.
• It can take account of actuator limitations.

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.