Page 66 - Distributed model predictive control for plant-wide systems

40 Distributed Model Predictive Control for Plant-Wide Systems

MPC

u 1 * y 1 u 2 * y * 2 u 3 * y 3 u * m-1 y m-1 u * m y m

S *
S Na
S 1 S 2 S *
S S
Plant-wide system 3 Na-1

Figure 3.1 The centralized MPC control structure

and that are commonly encountered in theory and practice. Referring to [11], we will introduce
the centralized MPC control structure, the single-layer DMPC, and the hierarchical DMPC in
this chapter.

3.2 Centralized MPC

As illustrated in Figure 3.1, when there is only one MPC controller that has access to all
actuators and sensors of the process network and thus directly controls the physical network,
to control the entire system, then the control structure is referred to as a centralized MPC
control structure.
The advantage of a centralized MPC control structure in general is that it can obtain the
best performance, and it has been extensively studied in the literature. There are a large
number of reference materials available for designing the centralized MPC, in particular
for small-scale systems. In addition, there are many mature commercial centralized MPC
software packages available which make it more convenient for control application engineers
to apply centralized MPC to industrial processes. However, there are several issues that
obstacle the implementation of a centralized MPC control structure to large-scale plant-wide
systems, such as follows:

• Global communication which leads to a severe requirement on communication speed and
network safety.
• Heavy computational requirements which make it only suitable for very slow large-scale
systems.
• Commercial, legal, and political issues related to unavailability of information and restricted
control access.
• Undesirable properties with respect to reliability, scalability, error tolerance, and structure
flexibility.