Chapter 20
Model Predictive Control (MPC)
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 1
Introduction on MPC
Scope
The combination of detailed process understanding with the
intelligent use of modern control systems (hardware, software and
technology) to achieve improved profitability.
Intelligent Use
Detailed Process of
Understanding Modern Control
Systems
Improved
Profitability
$
Introduction to Process Control Romagnoli & Palazoglu
Introduction on MPC
Introduced since 1980
It works in discrete time framework
Chapter 20
It partly overcomes the need of a feedback architecture
It integrates dynamic modeling and optimization
It is inherently multi-variable
It does more than set point tracking and
disturbance rejection …
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 3
History
• First practical application:
– DMC – Dynamic Matrix Control, early 1970s at Shell Oil
Chapter 20
– Cutler later started Dynamic Matrix Control Corp.
• Many successful industrial applications
• Theory (stability proofs etc) lagging behind 10-
20 years.
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 4
Current MPC Use
• Used in a majority of existing multivariable control
applications
Chapter 20
• Technology of choice for many new advanced
multivariable control application.
• Success rides on the computing power increase
• Has many important practical advantages
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 5
MPC Advantages
• Straightforward formulation, based on well
understood concepts
Chapter 20
• Explicitly handles constraints
• Explicit use of a model
• Well understood tuning parameters
– Prediction horizon
– Optimization problem setup
• Development time much shorter than for competing
advanced control methods
• Easier to maintain: changing model or specs does
not require complete redesign, sometimes can be
done on the fly
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 6
Side Objectives of
Model Predictive Control
1. Prevent violations of input and output constraints.
Chapter 20
2. Prevent excessive movement of the input (manipulated)
variables.
3. Drive some output (controlled) variables to their optimal
set points, while maintaining other outputs within
specified ranges.
4. If a sensor or actuator becomes not available anymore,
still control as much of the process as possible.
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 7
Model Predictive Control:
Conceptual schematics
Chapter 20
from ControlWiki
8
Block diagram
Chapter 20
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 9
Model Predictive Control: Basic Concepts
1. Future values of output variables are predicted using a dynamic
model of the process and current measurements at the k-th
sampling instant :
ŷ(k+1)=f [ŷ(k), u(k)]
Chapter 20
• Unlike time delay compensation methods, the predictions are made
for more than one time delay ahead.
2. The control calculations are based on both current measurements
and future predictions.
3. The manipulated variables, u(k), at the k-th sampling instant are
calculated so that they minimize an objective function J, e.g.:
y (k) ŷ(k)
k P 1 2
min J *
u(k)
k 1
• The reference trajectory y*(k) is based on set points calculated
using Real Time Optimization (RTO).
4. Inequality & equality constraints, and measured disturbances are
included in the control calculations.
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 10
MPC time trajectories
y*
Chapter 20
y* reference trajectory
Manipulated
variable
Figure 20.2 Basic concept for Model Predictive Control
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 11
Model Predictive Control: Calculations
1. At the k-th sampling instant, the values of the manipulated
variables, u, at the next M sampling instants, {u(k), u(k+1), …,
u(k+M -1)} are calculated.
• This set of M “control moves” is calculated so as to minimize
the predicted deviations from the reference trajectory over the
Chapter 20
next P sampling instants while satisfying the constraints.
• Typically, an LP or QP problem is solved at each sampling
instant.
• Terminology: M = control horizon, P = prediction horizon
2. Then the first “control move”, u(k), is implemented.
3. At the next sampling instant, k+1, the M-step control policy is
re-calculated for the next M sampling instants, k+1 to k+M, and
implement the first control move, u(k+1).
4. Then Steps 1 and 2 are repeated for subsequent sampling
instants.
Note: This approach is an example of a receding horizon
approach.
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 12
Model Predictive Control
MPC is like …
Chapter 20
… playing chess and planning moves ahead
13
Real-Time Optimization (RTO)
• The on-line calculation of optimal set-points, also called real-
time optimization (RTO), allows the profits from the process to
be maximized while satisfying operating constrains.
20
Chapter 19
• In real-time optimization (RTO), the optimum values of the set
points are re-calculated on a regular basis (e.g., every hour or
Chapter
every day).
• These repetitive calculations involve solving a constrained,
economic optimization problem, based on:
1. A steady-state model of the process, traditionally a linear one
2. Economic information (e.g., prices, costs)
3. A performance Index to be maximized (e.g., profit) or
minimized (e.g., cost).
Note: Items # 2 and 3 are sometimes referred to as an
economic model.
Model Predictive Control:
Conceptual schematics
Chapter 20
from Romagnoli & Palazoglu (2005),
“Introduction to Process Control” 15
When Should Predictive Control be Used?
1. Processes are difficult to control with standard PID
algorithm (e.g., large time constants, substantial time
delays, inverse response, etc.)
Chapter 20
2. There is significant process interactions between u and y.
• i.e., more than one manipulated variable has a significant effect on an
important process variable.
3. Constraints (limits) on process variables and manipulated
variables are important for normal control.
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 16
Model Predictive Control
• MPC displays its main strength when applied
to problems with
Chapter 20
a large number of manipulated and controlled
variables
constraints imposed on both the manipulated and
the controlled variables.
time delays
interaction between variables
multiple disturbances; if can be measured,
exploits the built-in feedforward capabilities of MPC
Introduction to Process Control Romagnoli & Palazoglu
Terminology:
y ↔ CV
u ↔ MV
ŷ(k +1) ... ŷ(k + P) d ↔ DV
Chapter 20
Figure 20.9 Flow chart for MPC calculations.
y*(k+1) … y*(k+M)
u(k+1) … u(k+M)
u(k+1)
from Seborg, Edgar, Mellichamp, Process 18
Dynamics and Control, 2nd Ed
Emerging MPC applications
Chapter 20
Introduction to Process Control Romagnoli & Palazoglu
Emerging MPC applications
Chapter 20
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 20
Chapter 20
from Seborg, Edgar, Mellichamp, Process
Dynamics and Control, 2nd Ed 21
