100% found this document useful (1 vote)

381 views174 pages

Getting Started Guide: Model Predictive Control Toolbox™

porcoddio

Uploaded by

porco gesu

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

100% found this document useful (1 vote)

381 views174 pages

Getting Started Guide: Model Predictive Control Toolbox™

porcoddio

Uploaded by

porco gesu

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 174

Model Predictive Control Toolbox™

Getting Started Guide

Alberto Bemporad
N. Lawrence Ricker
Manfred Morari

R2019b
How to Contact MathWorks

Latest news: www.mathworks.com

Sales and services: www.mathworks.com/sales_and_services

User community: www.mathworks.com/matlabcentral

Technical support: www.mathworks.com/support/contact_us

Phone: 508-647-7000

The MathWorks, Inc.

1 Apple Hill Drive
Natick, MA 01760-2098
Model Predictive Control Toolbox™ Getting Started Guide
© COPYRIGHT 2005–2019 by The MathWorks, Inc.
The software described in this document is furnished under a license agreement. The software may be used
or copied only under the terms of the license agreement. No part of this manual may be photocopied or
reproduced in any form without prior written consent from The MathWorks, Inc.
FEDERAL ACQUISITION: This provision applies to all acquisitions of the Program and Documentation by,
for, or through the federal government of the United States. By accepting delivery of the Program or
Documentation, the government hereby agrees that this software or documentation qualifies as commercial
computer software or commercial computer software documentation as such terms are used or defined in
FAR 12.212, DFARS Part 227.72, and DFARS 252.227-7014. Accordingly, the terms and conditions of this
Agreement and only those rights specified in this Agreement, shall pertain to and govern the use,
modification, reproduction, release, performance, display, and disclosure of the Program and
Documentation by the federal government (or other entity acquiring for or through the federal government)
and shall supersede any conflicting contractual terms or conditions. If this License fails to meet the
government's needs or is inconsistent in any respect with federal procurement law, the government agrees
to return the Program and Documentation, unused, to The MathWorks, Inc.
Trademarks
MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See
www.mathworks.com/trademarks for a list of additional trademarks. Other product or brand
names may be trademarks or registered trademarks of their respective holders.
Patents
MathWorks products are protected by one or more U.S. patents. Please see
www.mathworks.com/patents for more information.
Revision History
October 2004 First printing New for Version 2.1 (Release 14SP1)
March 2005 Online only Revised for Version 2.2 (Release 14SP2)
September 2005 Online only Revised for Version 2.2.1 (Release 14SP3)
March 2006 Online only Revised for Version 2.2.2 (Release 2006a)
September 2006 Online only Revised for Version 2.2.3 (Release 2006b)
March 2007 Online only Revised for Version 2.2.4 (Release 2007a)
September 2007 Online only Revised for Version 2.3 (Release 2007b)
March 2008 Online only Revised for Version 2.3.1 (Release 2008a)
October 2008 Online only Revised for Version 3.0 (Release 2008b)
March 2009 Online only Revised for Version 3.1 (Release 2009a)
September 2009 Online only Revised for Version 3.1.1 (Release 2009b)
March 2010 Online only Revised for Version 3.2 (Release 2010a)
September 2010 Online only Revised for Version 3.2.1 (Release 2010b)
April 2011 Online only Revised for Version 3.3 (Release 2011a)
September 2011 Online only Revised for Version 4.0 (Release 2011b)
March 2012 Online only Revised for Version 4.1 (Release 2012a)
September 2012 Online only Revised for Version 4.1.1 (Release 2012b)
March 2013 Online only Revised for Version 4.1.2 (Release 2013a)
September 2013 Online only Revised for Version 4.1.3 (Release 2013b)
March 2014 Online only Revised for Version 4.2 (Release R2014a)
October 2014 Online only Revised for Version 5.0 (Release 2014b)
March 2015 Online only Revised for Version 5.0.1 (Release 2015a)
September 2015 Online only Revised for Version 5.1 (Release 2015b)
March 2016 Online only Revised for Version 5.2 (Release 2016a)
September 2016 Online only Revised for Version 5.2.1 (Release 2016b)
March 2017 Online only Revised for Version 5.2.2 (Release 2017a)
September 2017 Online only Revised for Version 6.0 (Release 2017b)
March 2018 Online only Revised for Version 6.1 (Release 2018a)
September 2018 Online only Revised for Version 6.2 (Release 2018b)
March 2019 Online only Revised for Version 6.3 (Release 2019a)
September 2019 Online only Revised for Version 6.3.1 (Release 2019b)
Contents

Introduction
1
Model Predictive Control Toolbox Product Description . . . . . . 1-2
Key Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-4

Building Models
2
MPC Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2
Plant Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2
Input Disturbance Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4
Output Disturbance Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5
Measurement Noise Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-6

Signal Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-8

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-8
Outputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-8

Construct Linear Time Invariant Models . . . . . . . . . . . . . . . . . 2-10

Transfer Function Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-10
Zero/Pole/Gain Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-10
State-Space Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-11
LTI Object Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-12
LTI Model Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-15

Specify Multi-Input Multi-Output Plants . . . . . . . . . . . . . . . . 2-17

v
CSTR Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-20

Linearize Simulink Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-22

Linearization Using MATLAB Code . . . . . . . . . . . . . . . . . . . . 2-22
Linearization Using Linear Analysis Tool in Simulink Control
Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-25

Linearize Simulink Models Using MPC Designer . . . . . . . . . . 2-32

Define MPC Structure By Linearization . . . . . . . . . . . . . . . . . 2-32
Linearize Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-37
Specifying Operating Points . . . . . . . . . . . . . . . . . . . . . . . . . 2-39
Connect Measured Disturbances for Linearization . . . . . . . . 2-50

Identify Plant from Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-53

Identify Plant from Data at the Command Line . . . . . . . . . . . 2-53
Working with Impulse-Response Models . . . . . . . . . . . . . . . . 2-56

Design MPC Controllers

3
Design Controller Using MPC Designer . . . . . . . . . . . . . . . . . . 3-2

Design MPC Controller at the Command Line . . . . . . . . . . . . 3-24

Design MPC Controller in Simulink . . . . . . . . . . . . . . . . . . . . . 3-38

Control of a Single-Input-Single-Output Plant . . . . . . . . . . . . 3-60

Control of a Multi-Input Single-Output Plant . . . . . . . . . . . . . 3-63

Control of a Multi-Input Multi-Output Nonlinear Plant . . . . 3-97

vi Contents
1

Introduction

• “Model Predictive Control Toolbox Product Description” on page 1-2

• “Acknowledgments” on page 1-3
• “Bibliography” on page 1-4
1 Introduction

Model Predictive Control Toolbox Product Description

Design and simulate model predictive controllers

Model Predictive Control Toolbox provides functions, an app, and Simulink® blocks for
designing and simulating model predictive controllers (MPCs). The toolbox lets you
specify plant and disturbance models, horizons, constraints, and weights. By running
closed-loop simulations, you can evaluate controller performance.

You can adjust the behavior of the controller by varying its weights and constraints at run
time. To control a nonlinear plant, you can implement adaptive and gain-scheduled MPCs.
For applications with fast sample rates, you can generate an explicit model predictive
controller from a regular controller or implement an approximate solution.

For rapid prototyping and embedded system implementation, the toolbox supports
automatic C-code and IEC 61131-3 Structured Text generation.

Key Features
• App for interactive design of MPC controllers
• Runtime adjustment of weights and constraints
• Adaptive and gain-scheduled MPCs for controlling systems with nonlinear dynamics
• Explicit MPC and approximate solution for guaranteed worst-case execution time
• Economic MPC with arbitrary nonlinear cost function and constraints
• Computationally efficient quadratic programming (QP) solver, and support for third-
party solvers
• Support for C-code generation (with Simulink Coder™) and IEC 61131-3 Structured
Text generation (with Simulink PLC Coder™)

1-2
Acknowledgments

Acknowledgments
MathWorks would like to acknowledge the following contributors to Model Predictive
Control Toolbox.

Alberto Bemporad
Professor of Control Systems, IMT Institute for Advanced Studies Lucca, Italy.
Research interests include model predictive control, hybrid systems, optimization
algorithms, and applications to automotive, aerospace, and energy systems. Fellow of
the IEEE®. Author of the Model Predictive Control Simulink library and commands.
N. Lawrence Ricker
Professor of Chemical Engineering, University of Washington, Seattle, USA. Research
interests include model predictive control and process optimization. Author of the
Model Predictive Control Simulink library and commands.
Manfred Morari
Professor at the Automatic Control Laboratory and former Head of Department of
Information Technology and Electrical Engineering, ETH Zurich, Switzerland.
Research interests include model predictive control, hybrid systems, and robust
control. Fellow of the IEEE, AIChE, and IFAC. Co-author of the first version of the
toolbox.

1-3
1 Introduction

Bibliography
[1] Allgower, F., and A. Zheng, Nonlinear Model Predictive Control, Springer-Verlag, 2000.

[2] Camacho, E. F., and C. Bordons, Model Predictive Control, Springer-Verlag, 1999.

[3] Kouvaritakis, B., and M. Cannon, Non-Linear Predictive Control: Theory & Practice,
IEE Publishing, 2001.

[4] Maciejowski, J. M., Predictive Control with Constraints, Pearson Education POD, 2002.

[5] Prett, D., and C. Garcia, Fundamental Process Control, Butterworths, 1988.

[6] Rossiter, J. A., Model-Based Predictive Control: A Practical Approach, CRC Press,
2003.

1-4
2

Building Models

• “MPC Modeling” on page 2-2

• “Signal Types” on page 2-8
• “Construct Linear Time Invariant Models” on page 2-10
• “Specify Multi-Input Multi-Output Plants” on page 2-17
• “CSTR Model” on page 2-20
• “Linearize Simulink Models” on page 2-22
• “Linearize Simulink Models Using MPC Designer” on page 2-32
• “Identify Plant from Data” on page 2-53
2 Building Models

MPC Modeling
Model predictive controllers use plant, disturbance, and noise models for prediction and
state estimation. The model structure used in an MPC controller appears in the following
illustration.

Plant Model
You can specify the plant model in one of the following linear-time-invariant (LTI) formats:

• Numeric LTI models — Transfer function (tf), state space (ss), zero-pole-gain (zpk)
• Identified models (requires System Identification Toolbox™) — idss, idtf, idproc,
and idpoly

The MPC controller performs all estimation and optimization calculations using a
discrete-time, delay-free, state-space system with dimensionless input and output
variables. Therefore, when you specify a plant model in the MPC controller, the software
performs the following, if needed:

2-2
MPC Modeling

1 Conversion to state space — The ss command converts the supplied model to an LTI
state-space model.
2 Discretization or resampling — If the model sample time differs from the MPC
controller sample time (defined in the Ts property), one of the following occurs:

• If the model is continuous time, the c2d command converts it to a discrete-time

LTI object using the controller sample time.
• If the model is discrete time, the d2d command resamples it to generate a
discrete-time LTI object using the controller sample time.
3 Delay removal — If the discrete-time model includes any input, output, or internal
delays, the absorbDelay command replaces them with the appropriate number of
poles at z = 0, increasing the total number of discrete states. The InputDelay,
OutputDelay, and InternalDelay properties of the resulting state-space model
are all zero.
4 Conversion to dimensionless input and output variables — The MPC controller
enables you to specify a scale factor for each plant input and output variable. If you
do not specify scale factors, they default to 1. The software converts the plant input
and output variables to dimensionless form as follows:

xp k + 1 = Apxp k + BSiup k
−1 −1
yp k = So Cxp k + So DSiup k .

where Ap, B, C, and D are the constant zero-delay state-space matrices from step 3,
and:

• Si is a diagonal matrix of input scale factors in engineering units.

• So is a diagonal matrix of output scale factors in engineering units.
• xp is the state vector from step 3 in engineering units (including any absorbed
delay states). No scaling is performed on state variables.
• up is a vector of dimensionless plant input variables, including manipulated
variables, measured disturbances, and unmeasured input disturbances.
• yp is a vector of dimensionless plant output variables.

The resulting plant model has the following equivalent form:

xp k + 1 = Apxp k + Bpuu k + Bpvv k + Bpdd k

yp k = Cpxp k + Dpuu k + Dpvv k + Dpdd k .

2-3
2 Building Models

−1
Here, Cp = So C, Bpu, Bpv, and Bpd are the corresponding columns of BSi. Also, Dpu,
−1
Dpv, and Dpd are the corresponding columns of So DSi. Finally, u(k), v(k), and d(k) are
the dimensionless manipulated variables, measured disturbances, and unmeasured
input disturbances, respectively.

The MPC controller enforces the restriction of Dpu = 0, which means that the
controller does not allow direct feedthrough from any manipulated variable to any
plant output.

Input Disturbance Model

If your plant model includes unmeasured input disturbances, d(k), the input disturbance
model specifies the signal type and characteristics of d(k). See “Controller State
Estimation” for more information about the model.

The getindist command provides access to the model in use.

The input disturbance model is a key factor that influences the following controller
performance attributes:

• Dynamic response to apparent disturbances — The character of the controller

response when the measured plant output deviates from its predicted trajectory, due
to an unknown disturbance or modeling error.

• Asymptotic rejection of sustained disturbances — If the disturbance model predicts a

sustained disturbance, controller adjustments continue until the plant output returns
to its desired trajectory, emulating a classical integral feedback controller.

You can provide the input disturbance model as an LTI state-space (ss), transfer function
(tf), or zero-pole-gain (zpk) object using setindist. The MPC controller converts the
input disturbance model to a discrete-time, delay-free, LTI state-space system using the
same steps used to convert the plant model on page 2-2. The result is:

xid k + 1 = Aidxid k + Bidwid k

d k = Cidxid k + Didwid k .

where Aid, Bid, Cid, and Did are constant state-space matrices, and:

• xid(k) is a vector of nxid ≥ 0 input disturbance model states.

2-4
MPC Modeling

• dk(k) is a vector of nd dimensionless unmeasured input disturbances.

• wid(k) is a vector of nid ≥ 1 dimensionless white noise inputs, assumed to have zero
mean and unit variance.

If you do not provide an input disturbance model, then the controller uses a default
model, which has integrators with dimensionless unity gain added to its outputs. An
integrator is added for each unmeasured input disturbance, unless doing so would cause
a violation of state observability. In this case, a static system with dimensionless unity
gain is used instead.

Output Disturbance Model

The output disturbance model is a special case of the more general input disturbance
model. Its output, yod(k), is directly added to the plant output rather than affecting the
plant states. The output disturbance model specifies the signal type and characteristics of
yod(k), and it is often used in practice. See “Controller State Estimation” for more details
about the model.

The getoutdist command provides access to the output disturbance model in use.

You can specify a custom output disturbance model as an LTI state-space (ss), transfer
function (tf), or zero-pole-gain (zpk) object using setoutdist. Using the same steps as
for the plant model on page 2-2, the MPC controller converts the specified output
disturbance model to a discrete-time, delay-free, LTI state-space system. The result is:

xod k + 1 = Aodxod k + Bodwod k

yod k = Codxod k + Dodwod k .

where Aod, Bod, Cod, and Dod are constant state-space matrices, and:

• xod(k) is a vector of nxod ≥ 1 output disturbance model states.

• yod(k) is a vector of ny dimensionless output disturbances to be added to the
dimensionless plant outputs.
• wod(k) is a vector of nod dimensionless white noise inputs, assumed to have zero mean
and unit variance.

If you do not specify an output disturbance model, then the controller uses a default
model, which has integrators with dimensionless unity gain added to some or all of its
outputs. These integrators are added according to the following rules:

2-5
2 Building Models

• No disturbances are estimated, that is no integrators are added, for unmeasured plant
outputs.
• An integrator is added for each measured output in order of decreasing output weight.

• For time-varying weights, the sum of the absolute values over time is considered
for each output channel.
• For equal output weights, the order within the output vector is followed.
• For each measured output, an integrator is not added if doing so would cause a
violation of state observability. Instead, a gain with a value of zero is used instead.

If there is an input disturbance model, then the controller adds any default integrators to
that model before constructing the default output disturbance model.

Measurement Noise Model

One controller design objective is to distinguish disturbances, which require a response,
from measurement noise, which should be ignored. The measurement noise model
specifies the expected noise type and characteristics. See “Controller State Estimation”
for more details about the model.

Using the same steps as for the plant model on page 2-2, the MPC controller converts the
measurement noise model to a discrete-time, delay-free, LTI state-space system. The
result is:

xn k + 1 = Anxn k + Bnwn k
yn k = Cnxn k + Dnwn k .

Here, An, Bn, Cn, and Dn are constant state space matrices, and:

• xn(k) is a vector of nxn ≥ 0 noise model states.

• yn(k) is a vector of nym dimensionless noise signals to be added to the dimensionless
measured plant outputs.
• wn(k) is a vector of nn ≥ 1 dimensionless white noise inputs, assumed to have zero
mean and unit variance.

If you do not supply a noise model, the default is a unity static gain: nxn = 0, Dn is an nym-
by-nym identity matrix, and An, Bn, and Cn are empty.

For an mpc controller object, MPCobj, the property MPCobj.Model.Noise provides

access to the measurement noise model.

2-6
See Also

Note If the minimum eigenvalue of DnDnT is less than 1x10–8, the MPC controller adds
1x10–4 to each diagonal element of Dn. This adjustment makes a successful default
Kalman gain calculation more likely.

The controller adjusts these in order to achieve its goals.

Unmeasured disturbances

These are independent inputs of which the controller has no direct knowledge, and for
which it must compensate.

Outputs
The plant outputs are the dependent variables (outcomes) you wish to control or monitor.
As shown in “MPC Modeling” on page 2-2, there are two types:

Measured outputs

The controller uses these to estimate unmeasured quantities and as feedback on the
success of its adjustments.

Unmeasured outputs

The controller estimates these based on available measurements and the plant model. The
controller can also hold unmeasured outputs at setpoints or within constraint boundaries.

You must specify the input and output types when designing the controller. See “Input
and Output Types” on page 2-14 for more details.

2-8
See Also

Construct Linear Time Invariant Models

Model Predictive Control Toolbox software supports the same LTI model formats as does
Control System Toolbox software. You can use whichever is most convenient for your
application and convert from one format to another. For more details, see “Basic Models”
(Control System Toolbox).

Transfer Function Models

A transfer function (TF) relates a particular input/output pair. For example, if u(t) is a
plant input and y(t) is an output, the transfer function relating them might be:

Y(s) s+2
= G(s) = 2 e−1.5s
U(s) s + s + 10

This TF consists of a numerator polynomial, s+2, a denominator polynomial, s2+s+10,

and a delay, which is 1.5 time units here. You can define G using Control System Toolbox
tf function:

Gtf1 = tf([1 2], [1 1 10],'OutputDelay',1.5)

Transfer function:
s + 2
exp(-1.5*s) * ------------
s^2 + s + 10

Zero/Pole/Gain Models
Like the TF format, the zero/pole/gain (ZPK) format relates an input/output pair. The
difference is that the ZPK numerator and denominator polynomials are factored, as in

s + 0.45
G(s) = 2.5
(s + 0.3)(s + 0.1 + 0.7i)(s + 0.1 − 0.7i)

(zeros and/or poles are complex numbers in general).

You define the ZPK model by specifying the zero(s), pole(s), and gain as in

poles = [-0.3, -0.1+0.7i, -0.1-0.7i];

Gzpk1 = zpk(-0.45,poles,2.5);

2-10
Construct Linear Time Invariant Models

State-Space Models
The state-space format is convenient if your model is a set of LTI differential and
algebraic equations. For example, consider the following linearized model of a continuous
stirred-tank reactor (CSTR) involving an exothermic (heat-generating) reaction [1].

dC′ A
= a11C′ A + a12T′ + b11T′c + b12C′ Ai
dt

dT′
= a21C′ A + a22T′ + b21T′c + b22C′ Ai
dt

where CA is the concentration of a key reactant, T is the temperature in the reactor, Tc is

the coolant temperature, CAi is the reactant concentration in the reactor feed, and aij and
bij are constants. See the process schematic in “CSTR Schematic” on page 2-11. The
primes (e.g., C′A) denote a deviation from the nominal steady-state condition at which the
model has been linearized.

CSTR Schematic

Measurement of reactant concentrations is often difficult, if not impossible. Let us assume

that T is a measured output, CA is an unmeasured output, Tc is a manipulated variable,
and CAi is an unmeasured disturbance.

The model fits the general state-space format

dx
= Ax + Bu
dt

y = Cx + Du

where

2-11
2 Building Models

C′ A T′c T′
x= , u = , y =
T′ C′ Ai C′ A

a11 a12 b11 b12 0 1 00

A= , B = , C = , D =
a21 a22 b21 b22 1 0 00

The following code shows how to define such a model for some specific values of the aij
and bij constants:
A = [-0.0285 -0.0014
-0.0371 -0.1476];
B = [-0.0850 0.0238
0.0802 0.4462];
C = [0 1
1 0];
D = zeros(2,2);
CSTR = ss(A,B,C,D);

This defines a continuous-time state-space model. If you do not specify a sampling period,
a default sampling value of zero applies. You can also specify discrete-time state-space
models. You can specify delays in both continuous-time and discrete-time models.

Note In the CSTR example, the D matrix is zero and the output does not instantly
respond to change in the input. The Model Predictive Control Toolbox software prohibits
direct (instantaneous) feedthrough from a manipulated variable to an output. For
example, the CSTR model could include direct feedthrough from the unmeasured
disturbance, CAi, to either CA or T but direct feedthrough from Tc to either output would
violate this restriction. If the model had direct feedthrough from Tc, you can add a small
delay at this input to circumvent the problem.

LTI Object Properties

The ss function in the last line of the above code creates a state-space model, CSTR,
which is an LTI object. The tf and zpk commands described in “Transfer Function
Models” on page 2-10 and “Zero/Pole/Gain Models” on page 2-10 also create LTI objects.
Such objects contain the model parameters as well as optional properties.

LTI Properties for the CSTR Example

The following code sets some of the CSTR model's optional properties:

2-12
Construct Linear Time Invariant Models

CSTR.InputName = {'T_c','C_A_i'};
CSTR.OutputName = {'T','C_A'};
CSTR.StateName = {'C_A','T'};
CSTR.InputGroup.MV = 1;
CSTR.InputGroup.UD = 2;
CSTR.OutputGroup.MO = 1;
CSTR.OutputGroup.UO = 2;
CSTR

The first three lines specify labels for the input, output and state variables. The next four
specify the signal type for each input and output. The designations MV, UD, MO, and UO
mean manipulated variable, unmeasured disturbance, measured output, and unmeasured
output. (See “Signal Types” on page 2-8 for definitions.) For example, the code specifies
that input 2 of model CSTR is an unmeasured disturbance. The last line causes the LTI
object to be displayed, generating the following lines in the MATLAB® Command Window:

A =
C_A T
C_A -0.0285 -0.0014
T -0.0371 -0.1476

B =
T_c C_Ai
C_A -0.085 0.0238
T 0.0802 0.4462

C =
C_A T
T 0 1
C_A 1 0

D =
T_c C_Ai
T 0 0
C_A 0 0

Input groups:
Name Channels
MV 1
UD 2

Output groups:
Name Channels

2-13
2 Building Models

MO 1
UO 2

Continuous-time model

Input and Output Names

The optional InputName and OutputName properties affect the model displays, as in the
above example. The software also uses the InputName and OutputName properties to
label plots and tables. In that context, the underscore character causes the next character
to be displayed as a subscript.

Input and Output Types

General Case

As mentioned in “Signal Types” on page 2-8, Model Predictive Control Toolbox software
supports three input types and two output types. In a Model Predictive Control Toolbox
design, designation of the input and output types determines the controller dimensions
and has other important consequences.

For example, suppose your plant structure were as follows:

Plant Inputs Plant Outputs

Two manipulated variables (MVs) Three measured outputs (MOs)
One measured disturbance (MD) Two unmeasured outputs (UOs)
Two unmeasured disturbances (UDs)

The resulting controller has four inputs (the three MOs and the MD) and two outputs (the
MVs). It includes feedforward compensation for the measured disturbance, and assumes
that you wanted to include the unmeasured disturbances and outputs as part of the
regulator design.

If you didn't want a particular signal to be treated as one of the above types, you could do
one of the following:

• Eliminate the signal before using the model in controller design.

• For an output, designate it as unmeasured, then set its weight to zero.
• For an input, designate it as an unmeasured disturbance, then define a custom state
estimator that ignores the input.

2-14
Construct Linear Time Invariant Models

Note By default, the software assumes that unspecified plant inputs are manipulated
variables, and unspecified outputs are measured. Thus, if you didn't specify signal
types in the above example, the controller would have four inputs (assuming all plant
outputs were measured) and five outputs (assuming all plant inputs were manipulated
variables).

For model CSTR, the default Model Predictive Control Toolbox assumptions are incorrect.
You must set its InputGroup and OutputGroup properties, as illustrated in the above
code, or modify the default settings when you load the model into MPC Designer.

Use setmpcsignals to make type definition. For example:

CSTR = setmpcsignals(CSTR,'UD',2,'UO',2);

sets InputGroup and OutputGroup to the same values as in the previous example. The
CSTR display would then include the following lines:

Input groups:
Name Channels
Unmeasured 2
Manipulated 1

Output groups:
Name Channels
Unmeasured 2
Measured 1

Notice that setmpcsignals sets unspecified inputs to Manipulated and unspecified

outputs to Measured.

LTI Model Characteristics

Control System Toolbox software provides functions for analyzing LTI models. Some of the
more commonly used are listed below. Type the example code at the MATLAB prompt to
see how they work for the CSTR example.

Example Intended Result

dcgain(CSTR) Calculate gain matrix for the CSTR model's
input/output pairs.

2-15
2 Building Models

Example Intended Result

impulse(CSTR) Graph CSTR model's unit-impulse response.
linearSystemAnalyzer(CSTR) Open the Linear System Analyzer with the
CSTR model loaded. You can then display
model characteristics by making menu
selections.
pole(CSTR) Calculate CSTR model's poles (to check
stability, etc.).
step(CSTR) Graph CSTR model's unit-step response.
zero(CSTR) Compute CSTR model's transmission zeros.

References
[1] Seborg, D. E., T. F. Edgar, and D. A. Mellichamp, Process Dynamics and Control, 2nd
Edition, Wiley, 2004, pp. 34–36 and 94–95.

Specify Multi-Input Multi-Output Plants

Most MPC applications involve plants with multiple inputs and outputs. You can use ss,
tf, and zpk to represent a MIMO plant model. For example, consider the following model
of a distillation column [1], which has been used in many advanced control studies:

12.8e−s −18.9e−3s 3.8e−8.1s u1

y1 16.7s + 1 21.0s + 1 14.9s + 1
= u2
y2 6.6e−7s −19.4e−3s 4.9e−3.4s
u3
10.9s + 1 14.4s + 1 13.2s + 1

Outputs y1 and y2 represent measured product purities. The controller manipulates the
inputs, u1 and u2, to hold each output at a specified setpoint. These inputs represent the
flow rates of reflux and reboiler steam, respectively. Input u3 is a measured feed flow rate
disturbance.

The model consists of six transfer functions, one for each input/output pair. Each transfer
function is the first-order-plus-delay form often used by process control engineers.

Specify the individual transfer functions for each input/output pair. For example, g12 is
the transfer function from input u1 to output y2.

g11 = tf( 12.8, [16.7 1], 'IOdelay', 1.0,'TimeUnit','minutes');

g12 = tf(-18.9, [21.0 1], 'IOdelay', 3.0,'TimeUnit','minutes');
g13 = tf( 3.8, [14.9 1], 'IOdelay', 8.1,'TimeUnit','minutes');
g21 = tf( 6.6, [10.9 1], 'IOdelay', 7.0,'TimeUnit','minutes');
g22 = tf(-19.4, [14.4 1], 'IOdelay', 3.0,'TimeUnit','minutes');
g23 = tf( 4.9, [13.2 1], 'IOdelay', 3.4,'TimeUnit','minutes');

Define a MIMO system by creating a matrix of transfer function models.

DC = [g11 g12 g13

g21 g22 g23];

Define the input and output signal names and specify the third input as a measured input
disturbance.

DC.InputName = {'Reflux Rate','Steam Rate','Feed Rate'};

DC.OutputName = {'Distillate Purity','Bottoms Purity'};
DC = setmpcsignals(DC,'MD',3);

-->Assuming unspecified input signals are manipulated variables.

2-17
2 Building Models

Review the resulting system.

DC =

From input "Reflux Rate" to output...

12.8
Distillate Purity: exp(-1*s) * ----------
16.7 s + 1

6.6
Bottoms Purity: exp(-7*s) * ----------
10.9 s + 1

From input "Steam Rate" to output...

-18.9
Distillate Purity: exp(-3*s) * --------
21 s + 1

-19.4
Bottoms Purity: exp(-3*s) * ----------
14.4 s + 1

From input "Feed Rate" to output...

3.8
Distillate Purity: exp(-8.1*s) * ----------
14.9 s + 1

4.9
Bottoms Purity: exp(-3.4*s) * ----------
13.2 s + 1

Input groups:
Name Channels
Measured 3
Manipulated 1,2

Output groups:
Name Channels
Measured 1,2

Continuous-time transfer function.

2-18
See Also

References
[1] Wood, R. K., and M. W. Berry, Chem. Eng. Sci., Vol. 28, pp. 1707, 1973.

where CA is the concentration of a key reactant, T is the temperature in the reactor, Tc is

the coolant temperature, CAi is the reactant concentration in the reactor feed, and aij and
bij are constants. The primes (e.g., C′A) denote a deviation from the nominal steady-state
condition at which the model has been linearized.

Measurement of reactant concentrations is often difficult, if not impossible. Let us assume

that T is a measured output, CA is an unmeasured output, Tc is a manipulated variable,
and CAi is an unmeasured disturbance.

The model fits the general state-space format

dx
= Ax + Bu
dt

y = Cx + Du

where

C′ A T′c T′
x= , u = , y =
T′ C′ Ai C′ A

2-20
CSTR Model

a11 a12 b11 b12 0 1 00

A= , B = , C = , D =
a21 a22 b21 b22 1 0 00

The following code shows how to define such a model for some specific values of the aij
and bij constants:

A = [-0.0285 -0.0014
-0.0371 -0.1476];
B = [-0.0850 0.0238
0.0802 0.4462];
C = [0 1
1 0];
D = zeros(2,2);
CSTR = ss(A,B,C,D);

The following code sets some of the CSTR model's optional properties:

CSTR.InputName = {'T_c', 'C_A_i'};

CSTR.OutputName = {'T', 'C_A'};
CSTR.StateName = {'C_A', 'T'};
CSTR.InputGroup.MV = 1;
CSTR.InputGroup.UD = 2;
CSTR.OutputGroup.MO = 1;
CSTR.OutputGroup.UO = 2;

To view the properties of CSTR, enter:

CSTR

2-21
2 Building Models

Linearize Simulink Models

Generally, real systems are nonlinear. To design an MPC controller for a nonlinear system,
you can model the plant in Simulink.

Although an MPC controller can regulate a nonlinear plant, the model used within the
controller must be linear. In other words, the controller employs a linear approximation of
the nonlinear plant. The accuracy of this approximation significantly affects controller
performance.

To obtain such a linear approximation, you linearize the nonlinear plant at a specified
operating point.

Note Simulink Control Design software must be installed to linearize nonlinear Simulink
models.

You can linearize a Simulink model:

• From the command line.

• Using the Linear Analysis Tool.
• Using MPC Designer. For an example, see “Linearize Simulink Models Using MPC
Designer” on page 2-32.

Linearization Using MATLAB Code

This example shows how to obtain a linear model of a plant using a MATLAB script.

For this example the CSTR model, CSTR_OpenLoop, is linearized. The model inputs are
the coolant temperature (manipulated variable of the MPC controller), limiting reactant
concentration in the feed stream, and feed temperature. The model states are the
temperature and concentration of the limiting reactant in the product stream. Both states
are measured and used for feedback control.

Obtain Steady-State Operating Point

The operating point defines the nominal conditions at which you linearize a model. It is
usually a steady-state condition.

2-22
Linearize Simulink Models

Suppose that you plan to operate the CSTR with the output concentration, C_A, at
2 kmol/m3. The nominal feed concentration is 10 kmol/m3, and the nominal feed
temperature is 300 K. Create an operating point specification object to define the steady-
state conditions.

opspec = operspec('CSTR_OpenLoop');
opspec = addoutputspec(opspec,'CSTR_OpenLoop/CSTR',2);
opspec.Outputs(1).Known = true;
opspec.Outputs(1).y = 2;

op1 = findop('CSTR_OpenLoop',opspec);

Operating point search report:

---------------------------------

Operating point search report for the Model CSTR_OpenLoop.

(Time-Varying Components Evaluated at time t=0)

Operating point specifications were successfully met.

States:
----------
(1.) CSTR_OpenLoop/CSTR/C_A
x: 2 dx: -4.6e-12 (0)
(2.) CSTR_OpenLoop/CSTR/T_K
x: 373 dx: 5.49e-11 (0)

Inputs:
----------
(1.) CSTR_OpenLoop/Coolant Temperature
u: 299 [-Inf Inf]

Outputs:
----------
(1.) CSTR_OpenLoop/CSTR
y: 2 (2)

The calculated operating point is C_A = 2 kmol/m3 and T_K = 373 K. Notice that the
steady-state coolant temperature is also given as 299 K, which is the nominal value of the
manipulated variable of the MPC controller.

To specify:

• Values of known inputs, use the Input.Known and Input.u fields of opspec

2-23
2 Building Models

• Initial guesses for state values, use the State.x field of opspec

For example, the following code specifies the coolant temperature as 305 K and initial
guess values of the C_A and T_K states before calculating the steady-state operating
point:

opspec = operspec('CSTR_OpenLoop');
opspec.States(1).x = 1;
opspec.States(2).x = 400;
opspec.Inputs(1).Known = true;
opspec.Inputs(1).u = 305;

op2 = findop('CSTR_OpenLoop',opspec);

Operating point search report:

---------------------------------

Operating point search report for the Model CSTR_OpenLoop.

(Time-Varying Components Evaluated at time t=0)

Operating point specifications were successfully met.

States:
----------
(1.) CSTR_OpenLoop/CSTR/C_A
x: 1.78 dx: -1.42e-14 (0)
(2.) CSTR_OpenLoop/CSTR/T_K
x: 377 dx: 5.68e-14 (0)

Inputs:
----------
(1.) CSTR_OpenLoop/Coolant Temperature
u: 305

Outputs: None
----------

Specify Linearization Inputs and Outputs

If the linearization input and output signals are already defined in the model, as in
CSTR_OpenLoop, then use the following to obtain the signal set.

io = getlinio('CSTR_OpenLoop');

Otherwise, specify the input and output signals as shown here.

2-24
Linearize Simulink Models

io(1) = linio('CSTR_OpenLoop/Coolant Temperature',1,'input');

io(2) = linio('CSTR_OpenLoop/Feed Concentration',1,'input');
io(3) = linio('CSTR_OpenLoop/Feed Temperature',1,'input');
io(4) = linio('CSTR_OpenLoop/CSTR',1,'output');
io(5) = linio('CSTR_OpenLoop/CSTR',2,'output');

Linearize Model

Linearize the model using the specified operating point, op1, and input/output signals,
io.

sys = linearize('CSTR_OpenLoop',op1,io)

sys =

A =
C_A T_K
C_A -5 -0.3427
T_K 47.68 2.785

B =
Coolant Temp Feed Concent Feed Tempera
C_A 0 1 0
T_K 0.3 0 1

C =
C_A T_K
CSTR/1 0 1
CSTR/2 1 0

D =
Coolant Temp Feed Concent Feed Tempera
CSTR/1 0 0 0
CSTR/2 0 0 0

Continuous-time state-space model.

Linearization Using Linear Analysis Tool in Simulink Control

Design
This example shows how to linearize a Simulink model using the Linear Analysis Tool,
provided by the Simulink Control Design software.

2-25
2 Building Models

Open Simulink Model

This example uses the CSTR model, CSTR_OpenLoop.

open_system('CSTR_OpenLoop')

Specify Linearization Inputs and Outputs

The linearization inputs and outputs are already specified for CSTR_OpenLoop. The input
signals correspond to the outputs from the Feed Concentration, Feed Temperature,
and Coolant Temperature blocks. The output signals are the inputs to the CSTR
Temperature and Residual Concentration blocks.

To specify a signal as a linearization input or output, first open the Linearization tab. To
do so, in the Simulink Apps gallery, click Linearization Manager. Then, in the Simulink
model window, click the signal.

To specify the signal as a:

• Linearization input, on the Linearization tab, in the Insert Analysis Points gallery,
click Input Perturbation.
• Linearization output, on the Linearization tab, in the Insert Analysis Points gallery,
click Output Measurement.

Open Linear Analysis Tool

To open the Linear Analysis Tool, in the Apps gallery, click Model Linearizer.

2-26
Linearize Simulink Models

Specify Residual Concentration as Known Trim Constraint

In the Simulink model window, click the CA output signal from the CSTR block. Then, on
the Linearization tab, in the Insert Analysis Points gallery, click Trim Output
Constraint.

2-27
2 Building Models

In the Linear Analysis Tool, on the Linear Analysis tab, select Operating Point >
Trim Model.

In the Trim the model dialog box, on the Outputs tab:

• Select the Known check box for Channel - 1 under CSTR_OpenLoop/CSTR.

• Set the corresponding Value to 2 kmol/m3.

Create and Verify Operating Point

In the Trim the model dialog box, click Start trimming.

The operating point op_trim1 displays in the Linear Analysis Workspace.

2-28
Linearize Simulink Models

Double click op_trim1 to view the resulting operating point.

In the Edit dialog box, select the Input tab.

The coolant temperature at steady state is 299 K, as desired.

Linearize Model

On the Linear Analysis tab, in the Operating Point drop-down list, select op_trim1.

2-29
2 Building Models

Click Step to linearize the model.

This option creates the linear model linsys1 in the Linear Analysis Workspace and
generates a step response for this model. linsys1 uses optrim1 as its operating point.

The step response from feed concentration to output CSTR/2 displays an interesting
inverse response. An examination of the linear model shows that CSTR/2 is the residual
CSTR concentration, C_A. When the feed concentration increases, C_A increases initially
because more reactant is entering, which increases the reaction rate. This rate increase
results in a higher reactor temperature (output CSTR/1), which further increases the
reaction rate and C_A decreases dramatically.

Export Linearization Result

If necessary, you can repeat any of these steps to improve your model performance. Once
you are satisfied with your linearization result, in the Linear Analysis Tool, drag and drop

2-30
See Also

it from the Linear Analysis Workspace to the MATLAB Workspace. You can now use
your linear model to design an MPC controller.

See Also
Linear Analysis Tool | linearize

Related Examples
• “Design MPC Controller in Simulink” on page 3-38
• “Design Controller Using MPC Designer” on page 3-2
• “Design MPC Controller at the Command Line” on page 3-24

2-31
2 Building Models

Linearize Simulink Models Using MPC Designer

This topic shows how to linearize Simulink models using MPC Designer. To do so, open
the app from a Simulink model that contains an MPC Controller block. For this example,
use the CSTR_ClosedLoop model.

sys = 'CSTR_ClosedLoop';
open_system(sys)

In the model window, double-click the MPC Controller block.

In the Block Parameters dialog box, ensure that the MPC Controller field is empty, and
click Design to open MPC Designer.

Using MPC Designer, you can define the MPC structure by linearizing the Simulink
model. After you define the initial MPC structure, you can also linearize the model at
different operating points and import the linearized plants.

Note If a controller from the MATLAB workspace is specified in the MPC Controller
field, the app imports the specified controller. In this case, the MPC structure is derived
from the imported controller. In this case, you can still linearize the Simulink model and
import the linearized plants.

Define MPC Structure By Linearization

This example shows how to define the plant input/output structure in MPC Designer by
linearizing a Simulink model.

On the MPC Designer tab, in the Structure section, click MPC Structure.

2-32
Linearize Simulink Models Using MPC Designer

2-33
2 Building Models

Specify Signal Dimensions

In the Define MPC Structure By Linearization dialog box, in the MPC Structure section,
if the displayed signal dimensions do not match your model, click Change I/O Sizes to
configure the dimensions. Any unmeasured disturbances or unmeasured outputs in your
model are not detected by the MPC Controller block. Specify the dimensions for these
signals.

Tip In the MPC Controller Block Parameters dialog box, in the Default Conditions tab,
you can define the controller sample time and signal dimensions before opening MPC
Designer.

2-34
Linearize Simulink Models Using MPC Designer

Select Plant Input/Output Signals

Before linearizing the model, assign Simulink signal lines to each MPC signal type in your
model. The app uses these signals as linearization inputs and outputs.

In the Simulink Signals for Plant Inputs and Simulink Signals for Plant Outputs
sections, the Block Path is automatically defined for manipulated variables, measured
outputs, and measured disturbances. MPC Designer detects these signals since they are
connected to the MPC Controller block. If your application has unmeasured disturbances
or unmeasured outputs, select their corresponding Simulink signal lines.

To choose a signal type, use the Selected option buttons.

2-35
2 Building Models

Click Select Signals.

In the Simulink model window, click the signal line corresponding to the selected signal
type.

The signal is highlighted, and its block path is added to the Select signals dialog box.

In the Select signals dialog box, click Add Signal(s).

In the Define MPC Structure By Linearization dialog box, the Block Path for the selected
signal type updates.

2-36
Linearize Simulink Models Using MPC Designer

Note If your model has measured disturbances, you must connect the corresponding
plant inputs to the signal line connected to the md port of the MPC Controller block. For
more information, see “Connect Measured Disturbances for Linearization” on page 2-50.

Specify Operating Point

In the Simulink Operating Point section, in the drop-down list, select an operating
point at which to linearize the model.

For information on the different operating point options, see “Specifying Operating
Points” on page 2-39.

Note If you select an option that generates multiple operating points for linearization,
MPC Designer uses only the first operating point to define the plant structure and
linearize the model.

Define Structure and Linearize Model

Click Define and Linearize.

The app linearizes the Simulink model at the specified operating point using the specified
input/output signals, and adds the linearized plant to the Data Browser.

Also, a default controller, which uses the linearized plant as its internal model, and a
default simulation scenario are created.

MPC Designer uses the input/output signal values at the selected operating point as
nominal values.

Linearize Model
After you define the initial MPC structure, you can linearize the Simulink model at
different operating points and import the linearized plants. Doing so is useful for
validating controller performance against modeling errors.

On the MPC Designer tab, in the Import section, click Linearize Model.

2-37
2 Building Models

Select Plant Input/Output Signals

In the Simulink Signals for Plant Inputs and Simulink Signals for Plant Outputs
sections, the input/output signal configuration is the same as you specified when initially
defining the MPC structure.

You cannot change the signal types and dimensions once the structure has been defined.
However, for each signal type, you can select different signal lines from your Simulink
model. The selected lines must have the same dimensions as defined in the current MPC
structure.

Specify Operating Point

In the Simulink Operating Point section, in the drop-down list, select the operating
points at which to linearize the model.

2-38
Linearize Simulink Models Using MPC Designer

For information on the different operating point options, see “Specifying Operating
Points” on page 2-39.

Linearize Model and Import Plant

Click Linearize and Import.

MPC Designer linearizes the Simulink model at the defined operating point using the
specified input/output signals, and adds the linearized plant to the Data Browser.

If you select the Use selected operating point to update nominal values as well
option, the app updates the controller nominal values using the operating point signal
values.

If you select an option that generates multiple operating points for linearization, the app
linearizes the model at all the specified operating points. The linearized plants are added
to the Data Browser in the same order in which their corresponding operating points are
defined. If you choose to update the nominal values, the app uses the signal values from
the first operating point.

Specifying Operating Points

In the Simulink Operating Point section, in the drop-down list, you can select or create
operating points for model linearization. For more information on finding steady-state
operating points, see “About Operating Points” (Simulink Control Design) and “Compute
Steady-State Operating Points from Specifications” (Simulink Control Design).

Select Model Initial Condition

To linearize the model using the initial conditions specified in the Simulink model as the
operating point, select Model Initial Condition.

2-39
2 Building Models

The model initial condition is the default operating point for linearization in MPC
Designer.

Linearize at Simulation Snapshot Times

To linearize the model at specified simulation snapshot times, select Linearize At.
Linearizing at snapshot times is useful when you know that your model reaches an
equilibrium state after a certain simulation time.

2-40
Linearize Simulink Models Using MPC Designer

In the Enter snapshot times to linearize dialog box, in the Simulation snapshot times
field, enter one or more simulation snapshot times. Enter multiple snapshot times as a
vector.

Click OK.

If you enter multiple snapshot times, and you selected Linearize At from the:

• Define MPC Structure By Linearization dialog box, MPC Designer linearizes the
model using only the first snapshot time. The nominal values of the MPC controller are
defined using the input/output signal values for this snapshot.
• Linearize Simulink Model dialog box, MPC Designer linearizes the model at all the
specified snapshot times. The linearized plant models are added to the Data Browser
in the order specified in the snapshot time array. If you selected the Use selected
operating point to update nominal values as well option, the nominal values are
set using the input/output signal values from the first snapshot.

Compute Steady-State Operating Point

To compute a steady-state operating point using numerical optimization methods to meet

your specifications, select Trim Model.

2-41
2 Building Models

In the Trim the model dialog box, enter the specifications for the steady-state values at
which you want to find an operating point. You can specify values for states, input signals,
and output signals.

2-42
Linearize Simulink Models Using MPC Designer

Click Start Trimming.

MPC Designer creates an operating point for the given specifications. The computed
operating point is added to the Simulink Operating Point drop-down list and is
selected.

For examples showing how to specify the conditions for a steady-state operating point
search, see “Compute Steady-State Operating Points from Specifications” (Simulink
Control Design).

Compute Operating Point at Simulation Snapshot Time

To compute operating points using simulation snapshots, select Take Simulation

Snapshot. Linearizing the model using operating points computed from simulation
snapshots is useful when you know that your model reaches an equilibrium state after a
certain simulation time.

2-43
2 Building Models

In the Enter snapshot times to linearize dialog box, in the Simulation snapshot times
field, enter one or more simulation snapshot times. Enter multiple snapshot times as a
vector.

Click Take Snapshots.

MPC Designer simulates the Simulink model. At each snapshot time, the current state of
the model is used to create an operating point, which is added to the drop-down list and
selected.

If you entered multiple snapshot times, the operating points are stored together as an
array. If you selected Take Simulation Snapshot from the:

2-44
Linearize Simulink Models Using MPC Designer

• Define MPC Structure By Linearization dialog box, MPC Designer linearizes the
model using only the first operating point in the array. The nominal values of the MPC
controller are defined using the input/output signal values for this operating point.
• Linearize Simulink Model dialog box, MPC Designer linearizes the model at all the
operating points in the array. The linearized plant models are added to the Data
Browser in the same order as the operating point array.

In MPC Designer, the Linearize At and Take Simulation Snapshot options generally
produce the same linearized plant and nominal signal values. However, since the Take
Simulation Snapshot option first computes an operating point from the snapshot before
linearization, the results can differ.

Select Existing Operating Point

Under Existing Operating Points, select a previously defined operating point at which
to linearize the Simulink model. This option is available if one or more previously created
operating points are available in the drop-down list.

2-45
2 Building Models

If the selected operating point represents an operating point array created using multiple
snapshot times, and you selected an operating point from the:

Select Multiple Operating Points

To linearize the Simulink model at multiple existing operating points, select Linearize at
Multiple Points. This option is available if there are more than one previously created
operating points in the drop-down list.

2-46
Linearize Simulink Models Using MPC Designer

In the Specify multiple operating points dialog box, select the operating points at which to
linearize the model.

To change the operating point order, click an operating point in the list and click Up or
Down to move the highlighted operating point within the list.

Click OK.

If you selected Linearize at Multiple Points from the:

• Define MPC Structure By Linearization dialog box, MPC Designer linearizes the
model using only the first specified operating point. The nominal values of the MPC
controller are defined using the input/output signal values for this operating point.
• Linearize Simulink Model dialog box, MPC Designer linearizes the model at all the
specified operating points. The linearized plant models are added to the Data
Browser in the order specified in the Specify multiple operating points dialog box.

View/Edit Operating Point

To view or edit the selected operating point, under View/Edit, click the Edit option.

2-47
2 Building Models

In the Edit dialog box, if you created the selected operating point from a simulation
snapshot, you can edit the operating point values.

2-48
Linearize Simulink Models Using MPC Designer

If the selected operating point represents an operating point array, in the Select
Operating Point drop-down list, select an operating point to view.

If you obtained the operating point by trimming the model, you can only view the
operating point values.

2-49
2 Building Models

To set the Simulink model initial conditions to the states in the operating point, click
Initialize model. You can then simulate the model at the specified operating point.

When setting the model initial conditions, MPC Designer exports the operating point to
the MATLAB workspace. Also, in the Simulink Configuration Parameters dialog box, in the
Data Import/Export section, it selects the Input and Initial state parameters and
configures them to use the states and inputs in the exported operating point.

To reset the model initial conditions, for example if you delete the exported operating
point, clear the Input and Initial state parameters.

Connect Measured Disturbances for Linearization

If your Simulink model has measured disturbance signals, connect them to the
corresponding plant input ports and to the md port of the MPC Controller block. If you
have multiple measured disturbances, connect them to the MPC Controller using a vector
signal. As discussed in “Define MPC Structure By Linearization” on page 2-32, MPC

2-50
Linearize Simulink Models Using MPC Designer

Designer automatically detects the measured disturbances connected to the MPC

Controller block and sets them as plant inputs for linearization.

Since the measured disturbances connected to the md port are selected as linearization
inputs, you must connect the plant measured disturbance input ports to the selected
signal line, as shown in the following:

Correct MD Connection

If you connect the plant measured disturbance input ports to the corresponding signals
before the Mux block, as shown in the following, there is no linearization path from the
signals at the md port to the plant. As a result, when you linearize the plant using MPC
Designer, the measured disturbance channels linearize to zero.

2-51
2 Building Models

Incorrect MD Connection

Identify Plant from Data

When designing a model predictive controller, you can specify the internal predictive
plant model using a linear identified model. You use System Identification Toolbox
software to estimate a linear plant model in one of these forms:

• State-space model — idss

• Transfer function model — idtf
• Polynomial model — idpoly
• Process model — idproc
• Grey-box model — idgrey

You can estimate the plant model programmatically at the command line or interactively
using the System Identification app.

Identify Plant from Data at the Command Line

This example shows how to identify a plant model at the command line. For information
on identifying models using the System Identification app, see “Identify Linear Models
Using System Identification App” (System Identification Toolbox).

Load the measured input/output data.

load plantIO

This command imports the plant input signal, u, plant output signal, y, and sample time,
Ts to the MATLAB® workspace.

Create an iddata object from the input and output data.

mydata = iddata(y,u,Ts);

You can optionally assign channel names and units for the input and output signals.
mydata.InputName = 'Voltage';
mydata.InputUnit = 'V';
mydata.OutputName = 'Position';
mydata.OutputUnit = 'cm';

Typically, you must preprocess identification I/O data before estimating a model. For this
example, remove the offsets from the input and output signals by detrending the data.

2-53
2 Building Models

mydatad = detrend(mydata);

You can also remove offsets by creating an ssestOptions object and specifying the
InputOffset and OutputOffset options.

For this example, estimate a second-order, linear state-space model using the detrended
data. To estimate a discrete-time model, specify the sample time as Ts.
ss1 = ssest(mydatad,2,'Ts',Ts)

ss1 =
Discrete-time identified state-space model:
x(t+Ts) = A x(t) + B u(t) + K e(t)
y(t) = C x(t) + D u(t) + e(t)

A =
x1 x2
x1 0.8942 -0.1575
x2 0.1961 0.7616

B =
Voltage
x1 6.008e-05
x2 -0.01219

C =
x1 x2
Position 38.24 -0.3835

D =
Voltage
Position 0

K =
Position
x1 0.03572
x2 0.0223

Sample time: 0.1 seconds

Parameterization:
FREE form (all coefficients in A, B, C free).
Feedthrough: none
Disturbance component: estimate
Number of free coefficients: 10

2-54
Identify Plant from Data

Use "idssdata", "getpvec", "getcov" for parameters and their uncertainties.

Status:
Estimated using SSEST on time domain data "mydatad".
Fit to estimation data: 89.85% (prediction focus)
FPE: 0.0156, MSE: 0.01541

You can use this identified plant as the internal prediction model for your MPC controller.
When you do so, the controller converts the identified model to a discrete-time, state-
space model.

By default, the MPC controller discards any unmeasured noise components from your
identified model. To configure noise channels as unmeasured disturbances, you must first
create an augmented state-space model from your identified model. For example:

ss2 = ss(ss1,'augmented')

ss2 =

A =
x1 x2
x1 0.8942 -0.1575
x2 0.1961 0.7616

B =
Voltage v@Position
x1 6.008e-05 0.004448
x2 -0.01219 0.002777

C =
x1 x2
Position 38.24 -0.3835

D =
Voltage v@Position
Position 0 0.1245

Input groups:
Name Channels
Measured 1
Noise 2

Sample time: 0.1 seconds

Discrete-time state-space model.

2-55
2 Building Models

This command creates a state-space model, ss2, with two input groups, Measured and
Noise, for the measured and noise inputs respectively. When you import the augmented
model into your MPC controller, channels in the Noise input group are defined as
unmeasured disturbances.

Working with Impulse-Response Models

You can use System Identification Toolbox software to estimate finite step-response or
finite impulse-response (FIR) plant models using measured data. Such models, also known
as nonparametric models, are easy to determine from plant data ([1] and [2]) and have
intuitive appeal.

Use the impulseest function to estimate an FIR model from measured data. This
function generates the FIR coefficients encapsulated as an idtf object; that is, a transfer
function model with only numerator coefficients. impulseest is especially effective in
situations where the input signal used for identification has low excitation levels. To
design a model predictive controller for this plant, you can convert the identified FIR
plant model to a numeric LTI model. However, this conversion usually yields a high-order
plant, which can degrade the controller design. For example, the numerical precision
issues with high-order plants can affect estimator design. This result is particularly an
issue for MIMO systems.

Model predictive controllers work best with low-order parametric models. Therefore, to
design a model predictive controller using measured plant data, you can:

• Estimate a low-order parametric model using a parametric estimator, such as ssest.

• Initially identify a nonparametric model using impulseest, and then estimate a low-
order parametric model from the response of the nonparametric model. For an
example, see [3].
• Initially identify a nonparametric model using impulseest, and then convert the FIR
model to a state-space model using idss. You can then reduce the order of the state-
space model using balred. This approach is similar to the method used by ssregest.

References
[1] Cutler, C., and F. Yocum, "Experience with the DMC inverse for identification,"
Chemical Process Control — CPC IV (Y. Arkun and W. H. Ray, eds.), CACHE, 1991.

2-56
See Also

[2] Ricker, N. L., "The use of bias least-squares estimators for parameters in discrete-time
pulse response models," Ind. Eng. Chem. Res., Vol. 27, pp. 343, 1988.

[3] Wang, L., P. Gawthrop, C. Chessari, T. Podsiadly, and A. Giles, "Indirect approach to
continuous time system identification of food extruder," J. Process Control, Vol. 14,
Number 6, pp. 603–615, 2004.

Design MPC Controllers

• “Design Controller Using MPC Designer” on page 3-2

• “Design MPC Controller at the Command Line” on page 3-24
• “Design MPC Controller in Simulink” on page 3-38
• “Control of a Single-Input-Single-Output Plant” on page 3-60
• “Control of a Multi-Input Single-Output Plant” on page 3-63
• “Control of a Multi-Input Multi-Output Nonlinear Plant” on page 3-97
3 Design MPC Controllers

Design Controller Using MPC Designer

This example shows how to design a model predictive controller for a continuous stirred-
tank reactor (CSTR) using MPC Designer.

CSTR Model

The following differential equations represent the linearized model of a continuous

stirred-tank reactor (CSTR) involving an exothermic reaction:

dC′ A
= a11C′ A + a12T′ + b11T′c + b12C′ Ai
dt

dT′
= a21C′ A + a22T′ + b21T′c + b22C′ Ai
dt

where the inputs are:

• CAi — Concentration of reagent A in the feed stream (kgmol/m3)

• Tc — Reactor coolant temperature (degrees C)

and the outputs are:

• T — Reactor temperature (degrees C)

• CA — Residual concentration of reagent A in the product stream (kgmol/m3)

The prime terms, such as C′A, denote a deviation from the nominal steady-state condition
at which the model has been linearized.

Measurement of reagent concentrations is often difficult. For this example, assume that:

• Tc is a manipulated variable.

3-2
Design Controller Using MPC Designer

• CAi is an unmeasured disturbance.

• T is a measured output.
• CA is an unmeasured output.

The model can be described in state-space format:

dx
= Ax + Bu
dt

y = Cx + Du

where,

C′ A T′c T′
x= , u = , y =
T′ C′ Ai C′ A

a11 a12 b11 b12 0 1 00

A= , B = , C = , D =
a21 a22 b21 b22 1 0 00

For this example, the coolant temperature has a limited range of ±10 degrees from its
nominal value and a limited rate of change of ±4 degrees per sample period.

Create a state-space model of a CSTR system.

A = [-0.0285 -0.0014; -0.0371 -0.1476];

B = [-0.0850 0.0238; 0.0802 0.4462];
C = [0 1; 1 0];
D = zeros(2,2);
CSTR = ss(A,B,C,D);

Import Plant and Define MPC Structure

mpcDesigner

3-3
3 Design MPC Controllers

On the MPC Designer tab, in the Structure section, click MPC Structure.

In the Define MPC Structure By Importing dialog box, in the Select a plant model or an
MPC controller table, select the CSTR model.

Since CSTR is a stable, continuous-time LTI system, MPC Designer sets the controller
sample time to 0.1 Tr, where Tr is the average rise time of CSTR. For this example, in the
Specify MPC controller sample time field, enter a sample time of 1.

By default, all plant inputs are defined as manipulated variables and all plant outputs as
measured outputs. In the Assign plant i/o channels section, assign the input and output
channel indices such that:

• The first input, coolant temperature, is a manipulated variable.

3-4
Design Controller Using MPC Designer

• The second input, feed concentration, is an unmeasured disturbance.

• The first output, reactor temperature, is a measured output.
• The second output, reactant concentration, is an unmeasured output.

3-5
3 Design MPC Controllers

3-6
Design Controller Using MPC Designer

Click Define and Import.

The app imports the CSTR plant to the Data Browser. The following are also added to the
Data Browser:

• mpc1 — Default MPC controller created using sys as its internal model.
• scenario1 — Default simulation scenario.

The app runs the default simulation scenario and updates the Input Response and
Output Response plots.

Tip To view the response plots side-by-side, on the View tab, in the Tiles section, click
Left/Right.

3-7
3 Design MPC Controllers

Once you define the MPC structure, you cannot change it within the current MPC
Designer session. To use a different channel configuration, start a new session of the
app.

Define Input and Output Channel Attributes

On the MPC Designer tab, select I/O Attributes.

In the Input and Output Channel Specifications dialog box, in the Name column, specify a
meaningful name for each input and output channel.

In the Unit column, optionally specify the units for each channel.

Since the state-space model is defined using deviations from the nominal operating point,
set the Nominal Value for each input and output channel to 0.

Keep the Scale Factor for each channel at the default value of 1.

3-8
Design Controller Using MPC Designer

Click OK.

The Input Response and Output Response plot labels update to reflect the new signal
names and units.

Configure Simulation Scenario

On the MPC Designer tab, in the Scenario section, click Edit Scenario > scenario1.

In the Simulation Scenario dialog box, increase the Simulation duration to 30 seconds.

In the Reference Signals table, in the first row, specify a step Size of 2 and a Time of 5.

In the Signal column, in the second row, select a Constant reference to hold the
concentration setpoint at its nominal value.

3-9
3 Design MPC Controllers

The default scenario is configured to simulate a step change of 2 degrees in the reactor
temperature, T, at a time of 5 seconds.

Click OK.

The response plots update to reflect the new simulation scenario configuration.

In the Data Browser, in the Scenarios section, click scenario1. Click scenario1 a
second time, and rename the scenario to stepT.

3-10
Design Controller Using MPC Designer

Configure Controller Horizons

On the Tuning tab, in the Horizons section, specify a Prediction horizon of 15 and a
Control horizon of 3.

3-11
3 Design MPC Controllers

The response plots update to reflect the new horizons. The Input Response plot shows
that the control actions for the manipulated variable violate the required coolant
temperature constraints.

Define Input Constraints

In the Design section, click Constraints.

In the Constraints dialog box, in the Input Constraints section, enter the coolant
temperature upper and lower bounds in the Min and Max columns respectively.

Specify the rate of change limits in the RateMin and RateMax columns.

3-12
Design Controller Using MPC Designer

Click OK.

3-13
3 Design MPC Controllers

The Input Response plot shows the constrained manipulated variable control actions.
Even with the constrained rate of change, the coolant temperature rises quickly to its
maximum limit within three control intervals.

Specify Controller Tuning Weights

On the Tuning tab, in the Design section, click Weights.

In the Input Weights table, increase the manipulated variable (MV) Rate Weight to
0.3. Increasing the MV rate weight penalizes large MV changes in the controller
optimization cost function.

In the Output Weights table, keep the default Weight values. By default, all unmeasured
outputs have zero weights.

3-14
Design Controller Using MPC Designer

Since there is only one manipulated variable, if the controller tries to hold both outputs at
specific setpoints, one or both outputs will exhibit steady-state error in their responses.
Since the controller ignores setpoints for outputs with zero weight, setting the
concentration output weight to zero allows reactor temperature setpoint tracking with
zero steady-state error.

Click OK.

3-15
3 Design MPC Controllers

The Input Response plot shows the more conservative control actions, which result in a
slower Output Response.

Eliminate Output Overshoot

Suppose the application demands zero overshoot in the output response. On the
Performance Tuning tab, drag the Closed-Loop Performance slider to the left until
the Output Response has no overshoot. Moving this slider to the left simultaneously
increases the manipulated variable rate weight of the controller and decreases the output
variable weight, producing a more robust controller.

3-16
Design Controller Using MPC Designer

When you adjust the controller tuning weights using the Closed-Loop Performance
slider, MPC Designer does not change the weights you specified in the Weights dialog
box. Instead, the slider controls an adjustment factor, which is used with the user-
specified weights to define the actual controller weights.

This factor is 1 when the slider is centered; its value decreases as the slider moves left
and increases as the slider moves right. The weighting factor multiplies the manipulated
variable and output variable weights and divides the manipulated variable rate weights
from the Weights dialog box.

To view the actual controller weights, export the controller to the MATLAB workspace,
and view the Weights property of the exported controller object.

3-17
3 Design MPC Controllers

Test Controller Disturbance Rejection

In a process control application, disturbance rejection is often more important than

setpoint tracking. Simulate the controller response to a step change in the feed
concentration unmeasured disturbance.

On the MPC Designer tab, in the Scenario section, click Plot Scenario > New
Scenario.

In the Simulation Scenario dialog box, set the Simulation duration to 30 seconds.

In the Unmeasured Disturbances table, in the Signal drop-down list, select Step.

In the Time column, specify a step time of 5 seconds.

3-18
Design Controller Using MPC Designer

Click OK.

The app adds new scenario to the Data Browser and creates new corresponding Input
Response and Output Response plots.

In the Data Browser, in the Scenarios section, rename NewScenario to distReject.

3-19
3 Design MPC Controllers

In the Output Response plots, the controller returns the reactor temperature, T, to a
value near its setpoint as expected. However, the required control actions cause an
increase in the output concentration, CA to 6 kgmol/m3.

Specify Concentration Output Constraint

Previously, you defined the controller tuning weights to achieve the primary control
objective of tracking the reactor temperature setpoint with zero steady-state error. Doing
so enables the unmeasured reactor concentration to vary freely. Suppose that unwanted
reactions occur once the reactor concentration exceeds a 3 kgmol/m3. To limit the reactor
concentration, specify an output constraint.

On the Tuning tab, in the Design section, click Constraints.

3-20
Design Controller Using MPC Designer

In the Constraints dialog box, in the Output Constraints section, the second row of the
table, specify a Max unmeasured output (UO) value of 3.

In the Output Constraints section, click Constraint Softening Settings.

By default, all output constraints are soft, meaning that their MinECR and MaxECR
values are greater than zero. To soften the unmeasured output (UO) constraint further,
increase its MaxECR value.

Click OK.

3-21
3 Design MPC Controllers

In the Output Response plots, once the reactor concentration, CA, approaches 3
kgmol/m3, the reactor temperature, T, starts to increase. Since there is only one
manipulated variable, the controller makes a compromise between the two competing
control objectives: Temperature control and constraint satisfaction. A softer output
constraint enables the controller to sacrifice the constraint requirement more to achieve
improved temperature tracking.

Since the output constraint is soft, the controller maintain adequate temperature control
by allowing a small concentration constraint violation. In general, depending on your
application requirements, you can experiment with different constraint settings to achieve
an acceptable control objective compromise.

3-22
See Also

Export Controller

In the Tuning tab, in the Analysis section, click Export Controller to save the
tuned controller, mpc1, to the MATLAB workspace.

Delete Plants, Controllers, and Scenarios

To delete a plant, controller, or scenario, in the Data Browser, right-click the item you
want to delete, and select Delete. You can also click the item and hit Delete on the
keyboard.

You cannot delete the current controller. Also, you cannot delete a plant or scenario if it is
the only listed plant or scenario.

If a plant is used by any controller or scenario, you cannot delete the plant.

To delete multiple plants, controllers, or scenarios, hold Shift and click each item that
you want to delete.

References
[1] Seborg, D. E., T. F. Edgar, and D. A. Mellichamp, Process Dynamics and Control, 2nd
Edition, Wiley, 2004, pp. 34–36 and 94–95.

Design MPC Controller at the Command Line

This example shows how to create and test a model predictive controller from the
command line.

Define Plant Model

This example uses the plant model described in “Design Controller Using MPC Designer”
on page 3-2. Create a state-space model of the plant and set some of the optional model
properties.

A = [-0.0285 -0.0014; -0.0371 -0.1476];

B = [-0.0850 0.0238; 0.0802 0.4462];
C = [0 1; 1 0];
D = zeros(2,2);
CSTR = ss(A,B,C,D);

CSTR.InputName = {'T_c','C_A_i'};
CSTR.OutputName = {'T','C_A'};
CSTR.StateName = {'C_A','T'};
CSTR.InputGroup.MV = 1;
CSTR.InputGroup.UD = 2;
CSTR.OutputGroup.MO = 1;
CSTR.OutputGroup.UO = 2;

Create Controller

To improve the clarity of the example, suppress Command Window messages from the
MPC controller.

old_status = mpcverbosity('off');

Create a model predictive controller with a control interval, or sample time, of 1 second,
and with all other properties at their default values.

Ts = 1;
MPCobj = mpc(CSTR,Ts)

MPC object (created on 26-Aug-2019 22:28:46):

---------------------------------------------
Sampling time: 1 (seconds)
Prediction Horizon: 10
Control Horizon: 2

3-24
Design MPC Controller at the Command Line

Plant Model:
--------------
1 manipulated variable(s) -->| 2 states |
| |--> 1 measured output(s)
0 measured disturbance(s) -->| 2 inputs |
| |--> 1 unmeasured output(s)
1 unmeasured disturbance(s) -->| 2 outputs |
--------------
Indices:
(input vector) Manipulated variables: [1 ]
Unmeasured disturbances: [2 ]
(output vector) Measured outputs: [1 ]
Unmeasured outputs: [2 ]

Disturbance and Noise Models:

Output disturbance model: default (type "getoutdist(MPCobj)" for details)
Input disturbance model: default (type "getindist(MPCobj)" for details)
Measurement noise model: default (unity gain after scaling)

Weights:
ManipulatedVariables: 0
ManipulatedVariablesRate: 0.1000
OutputVariables: [1 0]
ECR: 100000

State Estimation: Default Kalman Filter (type "getEstimator(MPCobj)" for details)

Unconstrained

View and Modify Controller Properties

Display a list of the controller properties and their current values.

get(MPCobj)

Ts: 1
PredictionHorizon (P): 10
ControlHorizon (C): 2
Model: [1x1 struct]
ManipulatedVariables (MV): [1x1 struct]
OutputVariables (OV): [1x2 struct]
DisturbanceVariables (DV): [1x1 struct]
Weights (W): [1x1 struct]
Optimizer: [1x1 struct]

3-25
3 Design MPC Controllers

Notes: {}
UserData: []
History: 26-Aug-2020 22:28:46

The displayed History value will be different for your controller, since it depends on
when the controller was created. For a description of the editable properties of an MPC
controller, enter mpcprops at the command line.

Use dot notation to modify these properties. For example, change the prediction horizon
to 15.

MPCobj.PredictionHorizon = 15;

You can abbreviate property names provided that the abbreviation is unambiguous.

Many of the controller properties are structures containing additional fields. Use dot
notation to view and modify these field values. For example, you can set the measurement
units for the controller output variables. The OutputUnit property is for display
purposes only and is optional.

MPCobj.Model.Plant.OutputUnit = {'Deg C','kmol/m^3'};

By default, the controller has no constraints on manipulated variables and output

variables. You can view and modify these constraints using dot notation. For example, set
constraints for the controller manipulated variable.

MPCobj.MV.Min = -10;
MPCobj.MV.Max = 10;
MPCobj.MV.RateMin = -3;
MPCobj.MV.RateMax = 3;

You can also view and modify the controller tuning weights. For example, modify the
weights for the manipulated variable rate and the output variables.

MPCobj.W.ManipulatedVariablesRate = 0.3;
MPCobj.W.OutputVariables = [1 0];

You can also define time-varying constraints and weights over the prediction horizon,
which shifts at each time step. Time-varying constraints have a nonlinear effect when they
are active. For example, to force the manipulated variable to change more slowly towards
the end of the prediction horizon, enter:

MPCobj.MV.RateMin = [-4; -3.5; -3; -2.5];

3-26
Design MPC Controller at the Command Line

MPCobj.MV.RateMax = [4; 3.5; 3; 2.5];

The -2.5 and 2.5 values are used for the fourth step and beyond.

Similarly, you can specify different output variable weights for each step of the prediction
horizon. For example, enter:

MPCobj.W.OutputVariables = [0.1 0; 0.2 0; 0.5 0; 1 0];

You can also modify the disturbance rejection characteristics of the controller. See
setEstimator, setindist, and setoutdist for more information.

Review Controller Design

Generate a report on potential run-time stability and performance issues.

review(MPCobj)

3-27
3 Design MPC Controllers

3-28
Design MPC Controller at the Command Line

In this example, the review command found two potential issues with the design. The
first warning asks whether the user intends to have a weight of zero on the C_A output.
The second warning advises the user to avoid having hard constraints on both MV and
MVRate.

Perform Linear Simulations

Use the sim function to run a linear simulation of the system. For example, simulate the
closed-loop response of MPCobj for 26 control intervals. Specify setpoints of 2 and 0 for
the reactor temperature and the residual concentration respectively. The setpoint for the
residual concentration is ignored because the tuning weight for the second output is zero.

T = 26;
r = [0 0; 2 0];
sim(MPCobj,T,r)

3-29
3 Design MPC Controllers

3-30
Design MPC Controller at the Command Line

You can modify the simulation options using mpcsimopt. For example, run a simulation
with the manipulated variable constraints turned off.

MPCopts = mpcsimopt;
MPCopts.Constraints = 'off';
sim(MPCobj,T,r,MPCopts)

3-31
3 Design MPC Controllers

3-32
Design MPC Controller at the Command Line

The first move of the manipulated variable now exceeds the specified 3-unit rate
constraint.

You can also perform a simulation with a plant/model mismatch. For example, define a
plant with 50% larger gains than those in the model used by the controller.

Plant = 1.5*CSTR;
MPCopts.Model = Plant;
sim(MPCobj,T,r,MPCopts)

3-33
3 Design MPC Controllers

3-34
Design MPC Controller at the Command Line

The plant/model mismatch degrades controller performance slightly. Degradation can be

severe and must be tested on a case-by-case basis.

Other options include the addition of a specified noise sequence to the manipulated
variables or measured outputs, open-loop simulations, and a look-ahead option for better
setpoint tracking or measured disturbance rejection.

Store Simulation Results

Store the simulation results in the MATLAB Workspace.

[y,t,u] = sim(MPCobj,T,r);

3-35
3 Design MPC Controllers

The syntax suppresses automatic plotting and returns the simulation results. You can use
the results for other tasks, including custom plotting. For example, plot the manipulated
variable and both output variables in the same figure.

figure
subplot(2,1,1)
plot(t,u)
title('Inputs')
legend('T_c')
subplot(2,1,2)
plot(t,y)
title('Outputs')
legend('T','C_A')
xlabel('Time')

3-36
See Also

Restore the mpcverbosity setting.

mpcverbosity(old_status);

Design MPC Controller in Simulink

This example shows how to design a model predictive controller for a continuous stirred-
tank reactor (CSTR) in Simulink using MPC Designer.

This example requires Simulink Control Design software to define the MPC structure by
linearizing a nonlinear Simulink model.

If you do not have Simulink Control Design software, you must first create an mpc object
in the MATLAB workspace. For more information, see “Design Controller Using MPC
Designer” on page 3-2 and “Design MPC Controller at the Command Line” on page 3-24.

CSTR Model

A CSTR is a jacketed nonadiabatic tank reactor commonly used in the process industry.

3-38
Design MPC Controller in Simulink

An inlet stream of reagent A feeds into the tank at a constant rate. A first-order,
irreversible, exothermic reaction takes place to produce the product stream, which exits
the reactor at the same rate as the input stream.

The CSTR model has three inputs:

• Feed Concentration (CAi) — The concentration of reagent A in the feed stream

(kgmol/m3)
• Feed Temperature (Ti) — Feed stream temperature (K)
• Coolant Temperature (Tc) — Reactor coolant temperature (K)

The two model outputs are:

• CSTR Temperature (T) — Reactor temperature (K)

• Concentration (CA) — Concentration of reagent A in the product stream, also referred
to as the residual concentration (kgmol/m3)

The control objective is to maintain the residual concentration, CA, at its nominal setpoint
by adjusting the coolant temperature, Tc. Changes in the feed concentration, CAi, and
feed temperature, Ti, cause disturbances in the CSTR reaction.

The reactor temperature, T, is usually controlled. However, for this example, ignore the
reactor temperature, and assume that the residual concentration is measured directly.

Open Simulink Model

open_system('CSTR_ClosedLoop')

3-39
3 Design MPC Controllers

Connect Measured Disturbance To MPC Controller Block

In the Simulink model window, double-click the MPC Controller block.

In the Block Parameters dialog box, on the General tab, in the Additional Inports
section, check the Measured disturbance (md) option.

Click Apply to add the md inport to the controller block.

3-40
Design MPC Controller in Simulink

In the Simulink model window, connect the Feed Temperature block output to the md
inport.

Open MPC Designer App

In the MPC Controller Block Parameters dialog box, click Design to open MPC
Designer.

Note This step requires Simulink Control Design software to linearize the Simulink
model. For more information, see “Linearize Simulink Models Using MPC Designer” on
page 2-32.

If you do not have Simulink Control Design software, you must first create an mpc object
in the MATLAB workspace.

3-41
3 Design MPC Controllers

Define MPC Structure

In MPC Designer, on the MPC Designer tab, in the Structure section, click MPC
Structure.

In the Define MPC Structure By Linearization dialog box, in the Controller Sample
Time section, specify a sample time of 0.1.

In the MPC Structure section, click Change I/O Sizes to add the unmeasured
disturbance and measured disturbance signal dimensions.

In the MPC Block Signal Sizes dialog box, specify the number of input/output channels of
each type:

Click OK.

3-42
Design MPC Controller in Simulink

In the Define MPC Structure By Linearization dialog box, in the Simulink Signals for
Plant Inputs section, the app adds a row for Unmeasured Disturbances (UD).

3-43
3 Design MPC Controllers

3-44
Design MPC Controller in Simulink

The manipulated variable, measured disturbance, and measured output are already
assigned to their respective Simulink signal lines, which are connected to the MPC
Controller block.

In the Simulink Signals for Plant Inputs section, select the Unmeasured
Disturbances (UD) row, and click Select Signals.

In the Simulink model window, click the output signal from the Feed Concentration block.

The signal is highlighted and its block path is added to the Select Signal dialog box.

In the Select Signals dialog box, click Add Signal(s).

In the Define MPC Structure By Linearization dialog box, in the Simulink Signals for
Plant Inputs table, the Block Path for the unmeasured disturbance signal is updated.

Linearize Simulink Model

Linearize the Simulink model at a steady-state equilibrium operating point where the
residual concentration is 2 kgmol/m3. To compute such an operating point, add the CA
signal as a trim output constraint, and specify its target constraint value.

In the Simulink model window, right-click the signal line connected to CA outport of the
CSTR block, and select Linear Analysis Points > Trim Output Constraint.

3-45
3 Design MPC Controllers

The CA signal can now be used to define output specifications for calculating a model
steady-state operating point.

In the Define MPC Structure By Linearization dialog box, in the Simulink Operating
Point section, in the drop-down list, under Create New Operating Point, click Trim
Model.

In the Trim the model dialog box, in the Outputs tab, check the box in the Known
column for Channel-1 and specify a Value of 2.

3-46
Design MPC Controller in Simulink

This setting constrains the value of the output signal during the operating point search to
a known value.

Click Start Trimming.

In the Define MPC Structure By Linearization dialog box, in the Simulink Operating
Point section, the computed operating point, op_trim1, is added to the drop-down list
and selected.

In the drop-down list, under View/Edit, click Edit op_trim1.

3-47
3 Design MPC Controllers

In the Edit dialog box, on the State tab, in the Actual dx column, the near-zero derivative
values indicate that the computed operating point is at steady-state.

Click Initialize model to set the initial states of the Simulink model to the operating
point values in the Actual Values column. Doing so enables you to later simulate the
Simulink model at the computed operating point rather than at the default model initial
conditions.

In the Initialize Model dialog box, click OK.

To reset the model initial conditions, for example if you delete the exported operating
point, clear the Input and Initial state parameters.

Close the Edit dialog box.

3-48
Design MPC Controller in Simulink

In the Define MPC Structure By Linearization dialog box, click Define and Linearize to
linearize the model.

The linearized plant model is added to the Data Browser. Also, the following are added
to the Data Browser:

• A default MPC controller created using the linearized plant as an internal prediction
model
• A default simulation scenario

Define Input/Output Channel Attributes

On the MPC Designer tab, in the Structure section, click I/O Attributes.

In the Input and Output Channel Specifications dialog box, in the Name column, specify
meaningful names for each input and output channel.

In the Unit column, specify appropriate units for each signal.

3-49
3 Design MPC Controllers

The Nominal Value for each signal is the corresponding steady-state value at the
computed operating point.

Click OK.

Define Disturbance Rejection Simulation Scenarios

The primary objective of the controller is to hold the residual concentration, CA, at the
nominal value of 2 kgmol/m3. To do so, the controller must reject both measured and
unmeasured disturbances.

In the Scenario section, click Edit Scenario > scenario1.

In the Simulation Scenario dialog box, in the Reference Signals table, in the Signal
drop-down list select Constant to hold the output setpoint at its nominal value.

In the Measured Disturbances table, in the Signal drop-down list, select Step.

Specify a step Size of 10 and a step Time of 0.

3-50
Design MPC Controller in Simulink

Click OK.

In the Data Browser, under Scenarios, click scenario1. Click scenario1 a second
time, and rename it MD_reject.

In the Scenario section, click Plot Scenario > New Scenario.

In the Simulation Scenario dialog box, in the Unmeasured Disturbances table, in the
Signal drop-down list, select Step.

Specify a step Size of 1 and a step Time of 0.

3-51
3 Design MPC Controllers

Click OK.

In the Data Browser, under Scenarios, rename NewScenario to UD_reject.

Arrange Output Response Plots

To make viewing the tuning results easier, arrange the plot area to display the Output
Response plots for both scenarios at the same time.

On the View tab, in the Tiles section, click Top/Bottom.

The plot display area changes to display the Input Response plots above the Output
Response plots.

Drag the MD_reject: Output tab up to the top plot.

3-52
Design MPC Controller in Simulink

Tune Controller Performance

In the Tuning tab, in the Horizon section, specify a Prediction horizon of 20 and a
Control horizon of 5.

The Output Response plots update based on the new horizon values.

Use the default controller constraint and weight configurations.

In the Performance Tuning section, drag the Closed-Loop Performance slider to the
right, which leads to tighter control of outputs and more aggressive control moves. Drag
the slider until the MD_reject: Output response reaches steady state in less than 2
seconds.

3-53
3 Design MPC Controllers

Drag the State Estimation slider to the right, which leads to more aggressive
unmeasured disturbance rejection. Drag the slider until the UD_reject: Output response
reaches steady state in less than 3 seconds.

3-54
Design MPC Controller in Simulink

Update Simulink Model with Tuned Controller

In the Analysis section, click the Update and Simulate arrow .

Under Update and Simulate, click Update Block Only. The app exports the tuned
controller, mpc1, to the MATLAB workspace. In the Simulink model, the MPC Controller
block is updated to use the exported controller.

Simulate Unmeasured Disturbance Rejection

In the Simulink model window, change the simulation duration to 5 seconds.

3-55
3 Design MPC Controllers

The model initial conditions are set to the nominal operating point used for linearization.

Double-click the Feed Concentration block.

In the Block Parameters dialog box, enter a Constant Value of 11 to simulate a unit step
at time zero.

3-56
Design MPC Controller in Simulink

Click Apply.

In the Simulink model window, click Run.

The Concentration output response is similar to the UD_reject response, however the
settling time is around 1 second later. The different result is due to the mismatch between
the linear plant used in the MPC Designer simulation and the nonlinear plant in the
Simulink model.

Simulate Measured Disturbance Rejection

In the Block Parameters: Feed Concentration dialog box, enter a Constant Value of 10 to
return the feed concentration to its nominal value.

Click OK.

In the Simulink model window, double-click the Feed Temperature block.

In the Block Parameters: Feed Temperature dialog box, enter a Constant Value of 310 to
simulate a step change of size 10 at time zero.

Click OK.

3-57
3 Design MPC Controllers

In the Simulink model window, click Run.

The Concentration output response is similar to the MD_reject response from the MPC
Designer simulation.

References
[1] Seborg, D. E., T. F. Edgar, and D. A. Mellichamp, Process Dynamics and Control, 2nd
Edition, Wiley, 2004, pp. 34–36 and 94–95.

• “Design Controller Using MPC Designer” on page 3-2

3-59
3 Design MPC Controllers

Control of a Single-Input-Single-Output Plant

This example shows how to control a double integrator plant under input saturation in
Simulink®.

Define Plant Model

The linear open-loop dynamic model is a double integrator.

plant = tf(1,[1 0 0]);

Design MPC Controller

Create the controller object with sampling period, prediction and control horizons.
Ts = 0.1;
p = 10;
m = 3;
mpcobj = mpc(plant, Ts, p, m);

-->The "Weights.ManipulatedVariables" property of "mpc" object is empty. Assuming defau

-->The "Weights.ManipulatedVariablesRate" property of "mpc" object is empty. Assuming d
-->The "Weights.OutputVariables" property of "mpc" object is empty. Assuming default 1.

Specify actuator saturation limits as MV constraints.

mpcobj.MV = struct('Min',-1,'Max',1);

Simulate Using Simulink

To run this example, Simulink is required.

if ~mpcchecktoolboxinstalled('simulink')
disp('Simulink is required to run this example.')
return
end

Simulate closed-loop control of the linear plant model in Simulink. The MPC Controller
block is configured to use the mpcobj object as its controller.
mdl = 'mpc_doubleint';
open_system(mdl)
sim(mdl)

-->Converting the "Model.Plant" property of "mpc" object to state-space.

-->Converting model to discrete time.

3-60
Control of a Single-Input-Single-Output Plant

Assuming no disturbance added to measured output channel #1.

-->The "Model.Noise" property of the "mpc" object is empty. Assuming white noise on eac

3-61
3 Design MPC Controllers

The closed-loop response shows good setpoint tracking performance.

bdclose(mdl)

See Also
MPC Controller | MPC Designer | mpc

More About
• “Control of a Multi-Input Single-Output Plant” on page 3-63
• “Control of a Multi-Input Multi-Output Nonlinear Plant” on page 3-97

3-62
Control of a Multi-Input Single-Output Plant

Control of a Multi-Input Single-Output Plant

This example shows how to design model predictive controller with one measured output,
one manipulated variable, one measured disturbance, and one unmeasured disturbance in
a typical workflow.

Define Plant Model

The discrete-time linear open-loop dynamic model is defined below with sample time Ts.

sys = ss(tf({1,1,1},{[1 .5 1],[1 1],[.7 .5 1]}));

Ts = 0.2;
model = c2d(sys,Ts);

Design MPC Controller

Define type of input signals: the first signal is a manipulated variable, the second signal is
a measured disturbance, the third one is an unmeasured disturbance.

model = setmpcsignals(model,'MV',1,'MD',2,'UD',3);

Create the controller object with sampling period, prediction and control horizons.

mpcobj = mpc(model,Ts,10,3);

-->The "Weights.ManipulatedVariables" property of "mpc" object is empty. Assuming defau

-->The "Weights.ManipulatedVariablesRate" property of "mpc" object is empty. Assuming d
-->The "Weights.OutputVariables" property of "mpc" object is empty. Assuming default 1.

Define constraints on the manipulated variable.

mpcobj.MV = struct('Min',0,'Max',1,'RateMin',-10,'RateMax',10);

For unmeasured input disturbances, its model is an integrator driven by white noise with
variance = 1000.

mpcobj.Model.Disturbance = tf(sqrt(1000),[1 0]);

Simulate Closed-Loop Response Using the sim Command

Specify simulation parameters.

Tstop = 30; % simulation time

Tf = round(Tstop/Ts); % number of simulation steps

3-63
3 Design MPC Controllers

r = ones(Tf,1); % reference signal

v = [zeros(Tf/3,1);ones(2*Tf/3,1)]; % measured disturbance signal

Run the closed-loop simulation and plot results.

sim(mpcobj,Tf,r,v)

Assuming no disturbance added to measured output channel #1.

-->The "Model.Noise" property of the "mpc" object is empty. Assuming white noise on eac

3-64
Control of a Multi-Input Single-Output Plant

Specify disturbance and noise signals in simulation option object.

SimOptions = mpcsimopt(mpcobj);
d = [zeros(2*Tf/3,1);-0.5*ones(Tf/3,1)];
SimOptions.Unmeas = d; % unmeasured input disturbance signal
SimOptions.OutputNoise=.001*(rand(Tf,1)-.5); % output measurement noise
SimOptions.InputNoise=.05*(rand(Tf,1)-.5); % noise on manipulated variables

Run the closed-loop simulation and save the results to the workspace.

[y,t,u,xp] = sim(mpcobj,Tf,r,v,SimOptions);

Plot results.

3-65
3 Design MPC Controllers

figure
subplot(2,1,1)
plot(0:Tf-1,y,0:Tf-1,r)
title('Output')
grid
subplot(2,1,2)
plot(0:Tf-1,u)
title('Input')
grid

Simulate Closed-Loop Response with Model Mismatch

Test the robustness of the MPC controller against a model mismatch. Specify the true
plant simModel, with a nominal output value of 0.1.

3-66
Control of a Multi-Input Single-Output Plant

simModel = ss(tf({1,1,1},{[1 .8 1],[1 2],[.6 .6 1]}));

simModel = setmpcsignals(simModel,'MV',1,'MD',2,'UD',3);
simModel = struct('Plant',simModel);
simModel.Nominal.Y = 0.1;
simModel.Nominal.X = -.1*[1 1 1 1 1];

Create option object.

SimOptions.Model = simModel;
SimOptions.plantinit = [0.1 0 -0.1 0 .05]; % Initial state of the true plant
SimOptions.OutputNoise = []; % remove output measurement noise
SimOptions.InputNoise = []; % remove noise on manipulated variables

% Run the closed-loop simulation and plot results.

sim(mpcobj,Tf,r,v,SimOptions)

-->Converting model to discrete time.

3-67
3 Design MPC Controllers

3-68
Control of a Multi-Input Single-Output Plant

Soften Constraints

Relax the constraints on manipulated variables from hard to soft.

mpcobj.MV.MinECR = 1;
mpcobj.MV.MaxECR = 1;

Define an output constraint.

mpcobj.OV.Max = 1.1;

Run a new closed-loop simulation.

sim(mpcobj,Tf,r,v)

3-69
3 Design MPC Controllers

Assuming no disturbance added to measured output channel #1.

-->The "Model.Noise" property of the "mpc" object is empty. Assuming white noise on eac

3-70
Control of a Multi-Input Single-Output Plant

MV constraint has been slightly violated while MO constraint has been violated more.
Penalize more output constraint and rerun the simulation.

mpcobj.OV.MaxECR = 0.001; % The closer to zero, the harder the constraint

sim(mpcobj,Tf,r,v)

Assuming no disturbance added to measured output channel #1.

-->The "Model.Noise" property of the "mpc" object is empty. Assuming white noise on eac

3-71
3 Design MPC Controllers

3-72
Control of a Multi-Input Single-Output Plant

Now MO constraint has been slightly violated while MV constraint has been violated more
as expected.

Change Kalman Gains Used in the Built-In State Estimator

Model Predictive Control Toolbox™ software provides a default Kalman filter to estimate
the state of plant, disturbance, and noise models. You can change the Kalman gains.

First, retrieve the default Kalman gains and state-space matrices.

[L,M,A1,Cm1] = getEstimator(mpcobj);

The default observer poles are:

e = eig(A1-A1*M*Cm1)

3-73
3 Design MPC Controllers

e =

0.5708 + 0.4144i
0.5708 - 0.4144i
0.4967 + 0.0000i
0.9334 + 0.1831i
0.9334 - 0.1831i
0.8189 + 0.0000i

Design a new state estimator by pole-placement.

poles = [.8 .75 .7 .85 .6 .81];

L = place(A1',Cm1',poles)';
M = A1\L;
setEstimator(mpcobj,L,M);

Simulate Open-Loop Response

Test the behavior of plant in open-loop using the sim command. Set the OpenLoop flag to
on, and provide the sequence of manipulated variables that excite the system.

SimOptions.OpenLoop = 'on';
SimOptions.MVSignal = sin((0:Tf-1)'/10);

As the reference signal will be ignored, we can avoid specifying it.

sim(mpcobj,Tf,[],v,SimOptions)

Assuming no disturbance added to measured output channel #1.

-->The "Model.Noise" property of the "mpc" object is empty. Assuming white noise on eac
-->Converting model to discrete time.

3-74
Control of a Multi-Input Single-Output Plant

3-75
3 Design MPC Controllers

Examine Steady-State Offset

To examine whether the MPC controller will be able to reject constant output
disturbances and track constant setpoint with zero offsets in steady-state, compute the
DC gain from output disturbances to controlled outputs using the cloffset command.

DC = cloffset(mpcobj);
fprintf('DC gain from output disturbance to output = %5.8f (=%g) \n',DC,DC);

DC gain from output disturbance to output = 0.00000000 (=3.55271e-15)

A zero gain means that the output will track the desired setpoint.

3-76
Control of a Multi-Input Single-Output Plant

Simulate Controller Using mpcmove

First, obtain the discrete-time state-space matrices of the plant.

[A,B,C,D] = ssdata(model);
Tstop = 5; % Simulation time
x = [0 0 0 0 0]'; % Initial state of the plant
xmpc = mpcstate(mpcobj); % Initial state of the MPC controller
r = 1; % Output reference signal

Store the closed-loop MPC trajectories in arrays YY, UU, and XX.

YY=[];
UU=[];
XX=[];

Run the simulation loop

for t=0:round(Tstop/Ts)-1
% Store states
XX = [XX,x]; %#ok<*AGROW>
% Define measured disturbance signal
v = 0;
if t*Ts>=10
v = 1;
end
% Define unmeasured disturbance signal
d = 0;
if t*Ts>=20
d = -0.5;
end
% Plant equations: output update (no feedthrough from MV to Y)
y = C*x + D(:,2)*v + D(:,3)*d;
YY = [YY,y];
% Compute MPC action
u = mpcmove(mpcobj,xmpc,y,r,v);
% Plant equations: state update
x = A*x + B(:,1)*u + B(:,2)*v + B(:,3)*d;
UU = [UU,u];
end

Plot the results.

figure
subplot(2,1,1)

3-77
3 Design MPC Controllers

plot(0:Ts:Tstop-Ts,YY)
grid
title('Output')
subplot(2,1,2)
plot(0:Ts:Tstop-Ts,UU)
grid
title('Input')

If at any time during the simulation you want to check the optimal predicted trajectories,
it can be returned by mpcmove as well. Assume you want to start from the current state
and have a set-point change to 0.5, and assume the measured disturbance is gone.

3-78
Control of a Multi-Input Single-Output Plant

r = 0.5;
v = 0;
[~,Info] = mpcmove(mpcobj,xmpc,y,r,v);

Extract the optimal predicted trajectories.

topt = Info.Topt;
yopt = Info.Yopt;
uopt = Info.Uopt;
figure
subplot(2,1,1)
stairs(topt,yopt)
title('Optimal sequence of predicted outputs')
grid
subplot(2,1,2)
stairs(topt,uopt)
title('Optimal sequence of manipulated variables')
grid

3-79
3 Design MPC Controllers

Linear Representation of MPC Controller

When the constraints are not active, the MPC controller behaves like a linear controller.
You can get the state-space form of the MPC controller.

LTI = ss(mpcobj,'rv');

Get the state-space matrices for the linearized controller.

[AL,BL,CL,DL] = ssdata(LTI);

Simulate linear MPC closed-loop system and compare the linearized MPC controller with
the original MPC controller with constraints turned off.

3-80
Control of a Multi-Input Single-Output Plant

mpcobj.MV = []; % No input constraints

mpcobj.OV = []; % No output constraints
Tstop = 5; %Simulation time
xL = zeros(size(BL,1),1); % Initial state of linearized MPC controller
x = [0 0 0 0 0]'; % Initial state of plant
y = 0; % Initial measured output
r = 1; % Output reference set-point
u = 0; % Previous input command
YY = [];
XX = [];
xmpc = mpcstate(mpcobj);
for t = 0:round(Tstop/Ts)-1
YY = [YY,y];
XX = [XX,x];
v = 0;
if t*Ts>=10
v = 1;
end
d = 0;
if t*Ts>=20
d = -0.5;
end
uold = u;
% Compute the linear MPC control action
u = CL*xL+DL*[y;r;v];
% Compare the input move with the one provided by MPCMOVE
uMPC = mpcmove(mpcobj,xmpc,y,r,v);
dispStr(t+1) = {sprintf('t=%5.2f, u=%7.4f (provided by LTI), u=%7.4f (provided by M
% Update plant equations
x = A*x + B(:,1)*u + B(:,2)*v + B(:,3)*d;
% Update controller equations
xL = AL*xL + BL*[y;r;v];
% Update output equations
y = C*x + D(:,1)*u + D(:,2)*v + D(:,3)*d;
end

Assuming no disturbance added to measured output channel #1.

-->The "Model.Noise" property of the "mpc" object is empty. Assuming white noise on eac

Display the results.

for t=0:round(Tstop/Ts)-1
disp(dispStr{t+1});
end

3-81
3 Design MPC Controllers

t= 0.00, u= 5.2478 (provided by LTI), u= 5.2478 (provided by MPCOBJ)

t= 0.20, u= 3.0134 (provided by LTI), u= 3.0134 (provided by MPCOBJ)
t= 0.40, u= 0.2281 (provided by LTI), u= 0.2281 (provided by MPCOBJ)
t= 0.60, u=-0.9952 (provided by LTI), u=-0.9952 (provided by MPCOBJ)
t= 0.80, u=-0.8749 (provided by LTI), u=-0.8749 (provided by MPCOBJ)
t= 1.00, u=-0.2022 (provided by LTI), u=-0.2022 (provided by MPCOBJ)
t= 1.20, u= 0.4459 (provided by LTI), u= 0.4459 (provided by MPCOBJ)
t= 1.40, u= 0.8489 (provided by LTI), u= 0.8489 (provided by MPCOBJ)
t= 1.60, u= 1.0192 (provided by LTI), u= 1.0192 (provided by MPCOBJ)
t= 1.80, u= 1.0511 (provided by LTI), u= 1.0511 (provided by MPCOBJ)
t= 2.00, u= 1.0304 (provided by LTI), u= 1.0304 (provided by MPCOBJ)
t= 2.20, u= 1.0053 (provided by LTI), u= 1.0053 (provided by MPCOBJ)
t= 2.40, u= 0.9920 (provided by LTI), u= 0.9920 (provided by MPCOBJ)
t= 2.60, u= 0.9896 (provided by LTI), u= 0.9896 (provided by MPCOBJ)
t= 2.80, u= 0.9925 (provided by LTI), u= 0.9925 (provided by MPCOBJ)
t= 3.00, u= 0.9964 (provided by LTI), u= 0.9964 (provided by MPCOBJ)
t= 3.20, u= 0.9990 (provided by LTI), u= 0.9990 (provided by MPCOBJ)
t= 3.40, u= 1.0002 (provided by LTI), u= 1.0002 (provided by MPCOBJ)
t= 3.60, u= 1.0004 (provided by LTI), u= 1.0004 (provided by MPCOBJ)
t= 3.80, u= 1.0003 (provided by LTI), u= 1.0003 (provided by MPCOBJ)
t= 4.00, u= 1.0001 (provided by LTI), u= 1.0001 (provided by MPCOBJ)
t= 4.20, u= 1.0000 (provided by LTI), u= 1.0000 (provided by MPCOBJ)
t= 4.40, u= 0.9999 (provided by LTI), u= 0.9999 (provided by MPCOBJ)
t= 4.60, u= 1.0000 (provided by LTI), u= 1.0000 (provided by MPCOBJ)
t= 4.80, u= 1.0000 (provided by LTI), u= 1.0000 (provided by MPCOBJ)

Plot the results.

figure
plot(0:Ts:Tstop-Ts,YY)
grid

3-82
Control of a Multi-Input Single-Output Plant

Running a closed-loop where all constraints are turned off is easy using sim. We just
specify an option in the SimOptions structure:

SimOptions = mpcsimopt(mpcobj);
SimOptions.Constr = 'off'; % Remove all MPC constraints
SimOptions.Unmeas = d; % unmeasured input disturbance
sim(mpcobj,Tf,r,v,SimOptions);

3-83
3 Design MPC Controllers

3-84
Control of a Multi-Input Single-Output Plant

Simulate Using Simulink®

To run this example, Simulink® is required.

if ~mpcchecktoolboxinstalled('simulink')
disp('Simulink(R) is required to run this part of the example.')
return
end

Recreate the MPC controller.

mpcobj = mpc(model,Ts,10,3);
mpcobj.MV = struct('Min',0,'Max',1,'RateMin',-10,'RateMax',10);
mpcobj.Model.Disturbance = tf(sqrt(1000),[1 0]);

3-85
3 Design MPC Controllers

-->The "Weights.ManipulatedVariables" property of "mpc" object is empty. Assuming defau

-->The "Weights.ManipulatedVariablesRate" property of "mpc" object is empty. Assuming d
-->The "Weights.OutputVariables" property of "mpc" object is empty. Assuming default 1.

The continuous-time plant to be controlled has the following state-space realization:

[A,B,C,D] = ssdata(sys);

Simulate closed-loop MPC in Simulink.

mdl1 = 'mpc_miso';
open_system(mdl1)
sim(mdl1)

Assuming no disturbance added to measured output channel #1.

-->The "Model.Noise" property of the "mpc" object is empty. Assuming white noise on eac

3-86
Control of a Multi-Input Single-Output Plant

3-87
3 Design MPC Controllers

3-88
Control of a Multi-Input Single-Output Plant

3-89
3 Design MPC Controllers

3-90
Control of a Multi-Input Single-Output Plant

Next, run a simulation with sinusoidal output noise. Assume output measurements are
affected by a sinusoidal measurement noise of frequency 0.1 Hz. You want to inform the
MPC object about this so that state estimates can be improved.

omega = 2*pi/10;
mpcobj.Model.Noise = 0.5*tf(omega^2,[1 0 omega^2]);

Revise the MPC design and specify a white Gaussian noise unmeasured disturbance with
zero mean and variance 0.1.

mpcobj.Model.Disturbance = tf(0.1);
mpcobj.weights = struct('MV',0,'MVRate',0.1,'OV',0.005);
mpcobj.predictionhorizon = 40;
mpcobj.controlhorizon = 3;

Run the simulation.

Tstop = 150;
mdl2 = 'mpc_misonoise';
open_system(mdl2)
sim(mdl2,Tstop)

3-91
3 Design MPC Controllers

-->Assuming output disturbance added to measured output channel #1 is integrated white

-->A feedthrough channel in NoiseModel was inserted to prevent problems with estimator

3-92
Control of a Multi-Input Single-Output Plant

3-93
3 Design MPC Controllers

3-94
Control of a Multi-Input Single-Output Plant

3-95
3 Design MPC Controllers

bdclose(mdl1)
bdclose(mdl2)

See Also
MPC Controller | MPC Designer | mpc

More About
• “Control of a Single-Input-Single-Output Plant” on page 3-60
• “Control of a Multi-Input Multi-Output Nonlinear Plant” on page 3-97

3-96
Control of a Multi-Input Multi-Output Nonlinear Plant

Control of a Multi-Input Multi-Output Nonlinear Plant

This example shows how to design a model predictive controller for a multi-input multi-
output nonlinear plant. The plant has 3 manipulated variables and 2 measured outputs.

Linearize Nonlinear Plant

To run this example, Simulink® and Simulink Control Design™ are required.
if ~mpcchecktoolboxinstalled('simulink')
disp('Simulink(R) is required to run this example.')
return
end
if ~mpcchecktoolboxinstalled('slcontrol')
disp('Simulink Control Design(R) is required to run this example.')
return
end

The nonlinear plant is implemented in Simulink model mpc_nonlinmodel and linearized

at the default operating condition using the linearize command from Simulink Control
Design.
plant = linearize('mpc_nonlinmodel');

Assign names to I/O variables.

plant.InputName = {'Mass Flow';'Heat Flow';'Pressure'};
plant.OutputName = {'Temperature';'Level'};
plant.InputUnit = {'kg/s' 'J/s' 'Pa'};
plant.OutputUnit = {'K' 'm'};

Design MPC Controller

Create the controller object with sampling period, prediction and control horizons:
Ts = 0.2;
p = 5;
m = 2;
mpcobj = mpc(plant,Ts,p,m);

-->The "Weights.ManipulatedVariables" property of "mpc" object is empty. Assuming defau

-->The "Weights.ManipulatedVariablesRate" property of "mpc" object is empty. Assuming d
-->The "Weights.OutputVariables" property of "mpc" object is empty. Assuming default 1.

Specify MV constraints.

3-97
3 Design MPC Controllers

mpcobj.MV = struct('Min',{-3;-2;-2},'Max',{3;2;2},'RateMin',{-1000;-1000;-1000});

Define weights on manipulated and controlled variables.

mpcobj.Weights = struct('MV',[0 0 0],'MVRate',[.1 .1 .1],'OV',[1 1]);

Simulate Using Simulink

Run simulation.

mdl1 = 'mpc_nonlinear';
open_system(mdl1)
sim(mdl1)

-->Converting model to discrete time.

-->Assuming output disturbance added to measured output channel #1 is integrated white
-->Assuming output disturbance added to measured output channel #2 is integrated white
-->The "Model.Noise" property of the "mpc" object is empty. Assuming white noise on eac

3-98
Control of a Multi-Input Multi-Output Nonlinear Plant

3-99
3 Design MPC Controllers

3-100
Control of a Multi-Input Multi-Output Nonlinear Plant

Modify MPC Design to Track Ramp Signals

In order to track a ramp, a triple integrator is defined as an output disturbance model on

both outputs.

outdistmodel = tf({1 0;0 1},{[1 0 0 0],1;1,[1 0 0 0]});

setoutdist(mpcobj,'model',outdistmodel);

Run simulation.

mdl2 = 'mpc_nonlinear_setoutdist';
open_system(mdl2)
sim(mdl2)

-->Converting model to discrete time.

-->The "Model.Noise" property of the "mpc" object is empty. Assuming white noise on eac

3-101
3 Design MPC Controllers

3-102
Control of a Multi-Input Multi-Output Nonlinear Plant

Simulate without Constraints

When the constraints are not active, the MPC controller behaves like a linear controller.

mpcobj.MV = [];

Reset output disturbance model to default

setoutdist(mpcobj,'integrators');

The input to the linear controller LTI is the vector [ym;r], where ym is the vector of
measured outputs, and r is the vector of output references.

LTI = ss(mpcobj,'r');

-->Converting model to discrete time.

Run simulation.

refs = [1;1]; % output references are step signals

mdl3 = 'mpc_nonlinear_ss';
open_system(mdl3)
sim(mdl3)

3-103
3 Design MPC Controllers

3-104
Control of a Multi-Input Multi-Output Nonlinear Plant

3-105
3 Design MPC Controllers

Compare Simulation Results

fprintf('Compare output trajectories: ||ympc-ylin|| = %g\n',norm(ympc-ylin));

disp('The MPC controller and the linear controller produce the same closed-loop traject

Compare output trajectories: ||ympc-ylin|| = 1.50163e-14

The MPC controller and the linear controller produce the same closed-loop trajectories.

bdclose(mdl1)
bdclose(mdl2)
bdclose(mdl3)

See Also
MPC Controller | MPC Designer | mpc

More About
• “Control of a Single-Input-Single-Output Plant” on page 3-60
• “Control of a Multi-Input Single-Output Plant” on page 3-63

3-106

Common questions

To linearize a Simulink model using MPC Designer, first open the Simulink model containing the MPC Controller block and open MPC Designer from its Block Parameters dialog . Define the plant input/output structure by assigning signal lines corresponding to each MPC signal type, including manipulated variables and measured/unmeasured disturbances . Specify operating points for linearization in the Simulink Operating Point section using steady-state conditions or simulation snapshots . Click 'Linearize Model' to linearize the model at the selected operating point, which uses these inputs/outputs as nominal values . The linearized plant is then used within the MPC Designer for controller design, integrating both modeling and controller design steps .

Specifying a sample time in an MPC controller ensures that the controller operates using a consistent time interval for calculations, enabling the conversion of continuous-time models to discrete-time for state-space representation, and aligning the model's time domain with the controller's . Defining input/output signal dimensions allows for correct integration and configuration of signal channels, such as manipulated variables and disturbances, ensuring that all signals match the expected structure and facilitating model linearization and control design in Simulink . This setup is crucial for accurate model predictive control and simulation results, as it influences delay handling, scaling of signal variables, and the detection of signal paths in the linearized model ."}

Simulating closed-loop MPC responses under noise conditions involves defining and incorporating disturbance and noise models into the control system and analyzing the output performance. Different noise types, such as white Gaussian noise or sinusoidal noise, provide insight into how random fluctuations and periodic disturbances affect the system, respectively. Sinusoidal measurement noise is particularly insightful for assessing periodic interference impacts on state estimation and control actions. In refining an MPC design, sinusoidal noise helps adjust noise models like `mpcobj.Model.Noise` in the MPC object, ensuring robust filters and estimators are set to handle such explicit noise forms . Incorporating these elements enhances the controller's ability to maintain performance and stability in noisy operational settings.

In Simulink models, when connecting measured disturbances for linearization, it is necessary to link the plants' input ports with the md port of the MPC Controller block. Measured disturbances must be represented as input signals so that MPC Designer can detect and utilize them as plant inputs during linearization . For multiple disturbances, a vector signal representing all disturbances is used . This connection is crucial for accurately modeling disturbance impacts and for validating control strategies within MPC, as disturbances can significantly alter system dynamics and affect the performance and robustness of the controller.

Setting constraints on manipulated variables within a Model Predictive Control (MPC) framework is crucial for ensuring that the control actions remain within feasible operational limits, thus safeguarding the system against dangerous or inefficient states. Constraints can include limits on the range of manipulated variables and their rates of change, which help in maintaining system stability and performance . In Simulink, such constraints are configured by accessing the Constraints dialog within the MPC Designer. Here, you can specify the minimum and maximum bounds as well as the rate of change limits for the manipulated variables, ensuring that their behavior aligns with the operational requirements of the process . Soft constraints can also be used, allowing for controlled violation of these limits if needed to maintain other control objectives, thereby offering a balance between different control goals . These constraints and settings can be dynamically adjusted during runtime to reflect changes in system dynamics or control priorities ."}

When handling multiple operating points for model linearization, MPC Designer provides options to linearize the model at all specified operating points or just the first one, depending on the dialog box used to initiate the process. If multiple operating points are selected from the Define MPC Structure By Linearization dialog box, only the first operating point is used, whereas selecting from the Linearize Simulink Model dialog box linearizes the model at all specified points in order . The linearized plant models are sequentially added to the Data Browser, affecting simulation results by capturing different system dynamics or conditions that can be analyzed for their impact on controller performance . This allows for robust testing and validation of the controller under various scenarios.

Specifying operating point-based nominal values enhances the design of an MPC controller in Simulink by allowing the controller to use the input/output signal values at a defined operating point as baseline conditions. This can improve the accuracy of the linearization process, ensuring that the linear model closely approximates the nonlinear plant behavior under specific conditions. This alignment facilitates better performance by using the appropriate steady-state values as reference points, thereby reducing modeling errors and improving the controller's response to disturbances. When multiple operating points are used, the nominal values are updated based on the first operating point, ensuring consistency with the initial system conditions . By starting with accurate nominal conditions, the controller can better accommodate dynamic changes and maintain system stability .

The integrator model of unmeasured disturbances plays a critical role in Model Predictive Control (MPC) by allowing the controller to compensate for disturbances affecting the plant outputs that are not directly measured. It helps the controller predict the plant output trajectory in the presence of these disturbances, ensuring robust performance over time. In Simulink, to implement unmeasured disturbances, you augment the plant model by creating a state-space model where the unmeasured disturbances are modeled as persistent inputs. This augmented model includes an integrator or a static gain for each disturbance input and is imported into the MPC controller, allowing it to consider these disturbances when predicting future outputs . The default approach is adding integrators with unity gain unless doing so would invalidate state observability in the system .

The robustness of an MPC controller against model mismatch can be tested in Simulink by simulating a closed-loop system with a plant that has different characteristics than the model used by the controller. For instance, you can define a plant with 50% larger gains than those in the model used by the controller to evaluate the impact of the mismatch on controller performance . During testing, key parameters to adjust include manipulated variable constraints, prediction horizon, and control horizons, as well as the plant/model mismatch which can significantly degrade controller performance . Additionally, you can incorporate noise into the measured outputs or manipulated variables and examine how the controller handles these disturbances .

Varying the prediction and control horizons in a Model Predictive Control (MPC) controller affects the controller's ability to anticipate future disturbances and make corrective actions. A longer prediction horizon allows the controller to foresee potential future issues over a more extended period, potentially increasing control accuracy but also increasing computational demand. Conversely, a shorter horizon reduces the computational load but may not account for long-term effects as effectively . In Simulink, configuring these horizons is part of the controller setup process. You can specify prediction and control horizons using the MPC Designer by navigating to the Tuning tab and entering the desired values under the Horizon section. This adjustment can lead to changes in the output response plots, reflecting the controller's performance with different horizon settings .

Model Predictive Control Toolbox, User's Guide
No ratings yet
Model Predictive Control Toolbox, User's Guide
346 pages
MPC
100% (1)
MPC
250 pages
Simulink® PLC Coder™ User's Guide
No ratings yet
Simulink® PLC Coder™ User's Guide
203 pages
Inverted Pendulum State-Space Methods For Controller Design
No ratings yet
Inverted Pendulum State-Space Methods For Controller Design
16 pages
Overfitting and Underfitting
No ratings yet
Overfitting and Underfitting
3 pages
Model Predictive Control
100% (2)
Model Predictive Control
42 pages
Open Science and MATLAB Integration
No ratings yet
Open Science and MATLAB Integration
41 pages
Pattern Recognition and AI Using Matlab Textbook PDF
No ratings yet
Pattern Recognition and AI Using Matlab Textbook PDF
263 pages
Model Predictive Control Overview
100% (4)
Model Predictive Control Overview
55 pages
GOOD Notes For System Identification and Parameter Estimation
100% (1)
GOOD Notes For System Identification and Parameter Estimation
103 pages
Dq420ma English PDF
No ratings yet
Dq420ma English PDF
6 pages
Nur Aqilah Binti Mohamad Amin
100% (1)
Nur Aqilah Binti Mohamad Amin
35 pages
Matlab For Control Engineers Katsuhiko Ogata PDF
0% (1)
Matlab For Control Engineers Katsuhiko Ogata PDF
16 pages
Model Predictive Control Basics
100% (1)
Model Predictive Control Basics
50 pages
Alexander K. Belyaev, Hans Irschik, Michael Krommer (Eds.) - Mechanics and Model-Based Control of Advanced Engineering Systems (2014, Springer-Verlag Wien)
No ratings yet
Alexander K. Belyaev, Hans Irschik, Michael Krommer (Eds.) - Mechanics and Model-Based Control of Advanced Engineering Systems (2014, Springer-Verlag Wien)
317 pages
Neural Approaches Im Engeineering Matlab Supplement
No ratings yet
Neural Approaches Im Engeineering Matlab Supplement
218 pages
2019 - On The Control of Multi-Agent Systems - A Survey
No ratings yet
2019 - On The Control of Multi-Agent Systems - A Survey
164 pages
Holtz - Complex Vectors PDF
No ratings yet
Holtz - Complex Vectors PDF
9 pages
Fault Detection and Classification Using Machine Learning in MATLAB
100% (1)
Fault Detection and Classification Using Machine Learning in MATLAB
6 pages
Ljung L System Identification Theory For User
No ratings yet
Ljung L System Identification Theory For User
255 pages
Optimal Design of Control Systems Stochastic and Deterministic Problems
100% (2)
Optimal Design of Control Systems Stochastic and Deterministic Problems
423 pages
Analysis of Thyristor Power-Conditioned Motors by S.K. Pillai
No ratings yet
Analysis of Thyristor Power-Conditioned Motors by S.K. Pillai
19 pages
Neural Networks For Optimization and Signal Processing
No ratings yet
Neural Networks For Optimization and Signal Processing
549 pages
Recursive Methods in System Identification
100% (1)
Recursive Methods in System Identification
30 pages
Simulink Design Optimization - Getting Started Guide
No ratings yet
Simulink Design Optimization - Getting Started Guide
225 pages
T7 - State Feedback Analysis and Design - 2021
No ratings yet
T7 - State Feedback Analysis and Design - 2021
35 pages
Advanced Control Using Matlab PDF
100% (1)
Advanced Control Using Matlab PDF
564 pages
NGSPICE Guide for Circuit Designers
No ratings yet
NGSPICE Guide for Circuit Designers
5 pages
Simscape Ug
No ratings yet
Simscape Ug
886 pages
Digital Control Systems Overview
No ratings yet
Digital Control Systems Overview
20 pages
EE328 Power Electronics: Course Books
No ratings yet
EE328 Power Electronics: Course Books
36 pages
The IGBT Device: Physics, Design and Applications of The Insulated Gate Bipolar Transistor 2nd Edition B. Jayant Baliga Download
100% (1)
The IGBT Device: Physics, Design and Applications of The Insulated Gate Bipolar Transistor 2nd Edition B. Jayant Baliga Download
97 pages
Control System Development Using Dynamic Signal Analyzers: Application Note 243-2
No ratings yet
Control System Development Using Dynamic Signal Analyzers: Application Note 243-2
60 pages
Control of An Industrial Process Using PID Control Blocks PDF
No ratings yet
Control of An Industrial Process Using PID Control Blocks PDF
9 pages
System Design Through Matlab Control Toolbox and Simulink
No ratings yet
System Design Through Matlab Control Toolbox and Simulink
499 pages
Introduction To Electrical Engineering The Oxford Series in Electrical and Computer Engineering
No ratings yet
Introduction To Electrical Engineering The Oxford Series in Electrical and Computer Engineering
896 pages
Machine Learning in MATLAB - (Z-Library)
No ratings yet
Machine Learning in MATLAB - (Z-Library)
5 pages
(Synthesis Lectures On Engineering, Science, and Technology) Farzin Asadi - Signals and Systems With MATLAB and Simulink-Springer (2024)
No ratings yet
(Synthesis Lectures On Engineering, Science, and Technology) Farzin Asadi - Signals and Systems With MATLAB and Simulink-Springer (2024)
481 pages
1 Transient Response
No ratings yet
1 Transient Response
17 pages
Advances and Aplication in Sliding Model Control System (AhmadTaherAzar QuanminZhu) PDF
No ratings yet
Advances and Aplication in Sliding Model Control System (AhmadTaherAzar QuanminZhu) PDF
592 pages
Model Predictive Control Toolbox™ User's Guide
100% (3)
Model Predictive Control Toolbox™ User's Guide
256 pages
Model Predictive Control Toolbox PDF
No ratings yet
Model Predictive Control Toolbox PDF
384 pages
Matlab Control System Toolbox User's Guide (PDFDrive)
No ratings yet
Matlab Control System Toolbox User's Guide (PDFDrive)
1,816 pages
Usingcontrol Matlab
No ratings yet
Usingcontrol Matlab
1,626 pages
Control Ug
No ratings yet
Control Ug
1,958 pages
Model Predictive Control Toolbox™ Reference
100% (2)
Model Predictive Control Toolbox™ Reference
158 pages
MATLAB Control Systems
100% (1)
MATLAB Control Systems
182 pages
Get Start CS
No ratings yet
Get Start CS
278 pages
Libro de Matlab
No ratings yet
Libro de Matlab
132 pages
Model Predictive Control Toolbox: User's Guide
100% (1)
Model Predictive Control Toolbox: User's Guide
246 pages
MATLAB Control Sytem Toolbox User S Guide The Mathworks Full
No ratings yet
MATLAB Control Sytem Toolbox User S Guide The Mathworks Full
88 pages
Getting Started Guide: Model Predictive Control Toolbox™ 3
100% (1)
Getting Started Guide: Model Predictive Control Toolbox™ 3
142 pages
Release Notes For Model Predictive Control Toolbox™
No ratings yet
Release Notes For Model Predictive Control Toolbox™
55 pages
Getting Started Guide: Control System Toolbox™ 9
No ratings yet
Getting Started Guide: Control System Toolbox™ 9
280 pages
System Identification Toolbox™ User's Guide: Lennart Ljung
100% (1)
System Identification Toolbox™ User's Guide: Lennart Ljung
886 pages
Eg Index - Exam Index Copyright
No ratings yet
Eg Index - Exam Index Copyright
38 pages
(Metasol MMS) EN C01822-03-202312
No ratings yet
(Metasol MMS) EN C01822-03-202312
56 pages
The Current MasterFormat Divisions
100% (3)
The Current MasterFormat Divisions
3 pages
Personality PDF
No ratings yet
Personality PDF
10 pages
Action Research Proposal (Pupil Incentive Program)
No ratings yet
Action Research Proposal (Pupil Incentive Program)
6 pages
Valu-Based Pricing - FAQ
No ratings yet
Valu-Based Pricing - FAQ
2 pages
The Healthy Company Pitch Deck July20
No ratings yet
The Healthy Company Pitch Deck July20
20 pages
Chinese Vocabs of Resturant Dining
100% (1)
Chinese Vocabs of Resturant Dining
3 pages
Manage Meetings Assessment Guide
No ratings yet
Manage Meetings Assessment Guide
11 pages
Exclusive Model Management Contract
No ratings yet
Exclusive Model Management Contract
9 pages
2020 Scala Mens Summer Santana Only-Min
No ratings yet
2020 Scala Mens Summer Santana Only-Min
17 pages
Efficient EFF-WB Dust Filter System
No ratings yet
Efficient EFF-WB Dust Filter System
2 pages
Boyer-Roxas Vs Court of Appeals
No ratings yet
Boyer-Roxas Vs Court of Appeals
2 pages
Financial Accounting Lecture Notes
100% (1)
Financial Accounting Lecture Notes
3 pages
Nutritional Problems in India & World: Ways To Combat: DR Kalpana Kulshrestha
No ratings yet
Nutritional Problems in India & World: Ways To Combat: DR Kalpana Kulshrestha
39 pages
SRYLED P4.81 Outdoor 1R1G1B 3in1 SMD1921 Full Color LED Display (500x500mm Cabinet)
No ratings yet
SRYLED P4.81 Outdoor 1R1G1B 3in1 SMD1921 Full Color LED Display (500x500mm Cabinet)
4 pages
Introduction
No ratings yet
Introduction
16 pages
Microsoft Word - Gosavi - PRATIK-Orgnal
No ratings yet
Microsoft Word - Gosavi - PRATIK-Orgnal
7 pages
The Effects of Online Social in Uencers On Purchasing Behavior of Generation Z: An Empirical Study in Vietnam
No ratings yet
The Effects of Online Social in Uencers On Purchasing Behavior of Generation Z: An Empirical Study in Vietnam
13 pages
The Mitochondria Is The Powerhouse of The Cell
No ratings yet
The Mitochondria Is The Powerhouse of The Cell
4 pages
ESCR PDF
No ratings yet
ESCR PDF
3 pages
Multi-Objective Optimization of Injection Molding Process Parameters For Short Cycle Time and Warpage Reduction Using Conformal Cooling Channel
No ratings yet
Multi-Objective Optimization of Injection Molding Process Parameters For Short Cycle Time and Warpage Reduction Using Conformal Cooling Channel
10 pages
11kV VCB Penal - PDF - Relay - Fuse (Electrical)
No ratings yet
11kV VCB Penal - PDF - Relay - Fuse (Electrical)
17 pages
Storytelling in Advertising From Narrative
No ratings yet
Storytelling in Advertising From Narrative
12 pages
Android Final
No ratings yet
Android Final
1,185 pages
Dental Alloy Properties Guide
No ratings yet
Dental Alloy Properties Guide
4 pages
MVS Notes Unit-I
No ratings yet
MVS Notes Unit-I
16 pages
Financial Analytics
No ratings yet
Financial Analytics
26 pages
Khab
0% (1)
Khab
237 pages
Creating A Culture of Continuous Learning and Developmen
100% (2)
Creating A Culture of Continuous Learning and Developmen
15 pages

Getting Started Guide: Model Predictive Control Toolbox™

Uploaded by

Getting Started Guide: Model Predictive Control Toolbox™

Uploaded by

Model Predictive Control Toolbox™

Getting Started Guide

Latest news: www.mathworks.com

Sales and services: www.mathworks.com/sales_and_services

User community: www.mathworks.com/matlabcentral

Technical support: www.mathworks.com/support/contact_us

The MathWorks, Inc.

Signal Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-8

Construct Linear Time Invariant Models . . . . . . . . . . . . . . . . . 2-10

Specify Multi-Input Multi-Output Plants . . . . . . . . . . . . . . . . 2-17

Linearize Simulink Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-22

Linearize Simulink Models Using MPC Designer . . . . . . . . . . 2-32

Identify Plant from Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-53

Design MPC Controllers

Design MPC Controller at the Command Line . . . . . . . . . . . . 3-24

Design MPC Controller in Simulink . . . . . . . . . . . . . . . . . . . . . 3-38

Control of a Single-Input-Single-Output Plant . . . . . . . . . . . . 3-60

Control of a Multi-Input Single-Output Plant . . . . . . . . . . . . . 3-63

Control of a Multi-Input Multi-Output Nonlinear Plant . . . . 3-97

• “Model Predictive Control Toolbox Product Description” on page 1-2

Model Predictive Control Toolbox Product Description

• “MPC Modeling” on page 2-2

• If the model is continuous time, the c2d command converts it to a discrete-time

• Si is a diagonal matrix of input scale factors in engineering units.

The resulting plant model has the following equivalent form:

xp k + 1 = Apxp k + Bpuu k + Bpvv k + Bpdd k

Input Disturbance Model

The getindist command provides access to the model in use.

• Dynamic response to apparent disturbances — The character of the controller

• Asymptotic rejection of sustained disturbances — If the disturbance model predicts a

xid k + 1 = Aidxid k + Bidwid k

• xid(k) is a vector of nxid ≥ 0 input disturbance model states.

• dk(k) is a vector of nd dimensionless unmeasured input disturbances.

Output Disturbance Model

xod k + 1 = Aodxod k + Bodwod k

• xod(k) is a vector of nxod ≥ 1 output disturbance model states.

Measurement Noise Model

• xn(k) is a vector of nxn ≥ 0 noise model states.

For an mpc controller object, MPCobj, the property MPCobj.Model.Noise provides

The controller adjusts these in order to achieve its goals.

Construct Linear Time Invariant Models

Transfer Function Models

This TF consists of a numerator polynomial, s+2, a denominator polynomial, s2+s+10,

Gtf1 = tf([1 2], [1 1 10],'OutputDelay',1.5)

(zeros and/or poles are complex numbers in general).

poles = [-0.3, -0.1+0.7*i, -0.1-0.7*i];

where CA is the concentration of a key reactant, T is the temperature in the reactor, Tc is

Measurement of reactant concentrations is often difficult, if not impossible. Let us assume

The model fits the general state-space format

a11 a12 b11 b12 0 1 00

LTI Object Properties

LTI Properties for the CSTR Example

Input and Output Names

Input and Output Types

For example, suppose your plant structure were as follows:

Plant Inputs Plant Outputs

• Eliminate the signal before using the model in controller design.

Use setmpcsignals to make type definition. For example:

Notice that setmpcsignals sets unspecified inputs to Manipulated and unspecified

LTI Model Characteristics

Example Intended Result

Example Intended Result

Specify Multi-Input Multi-Output Plants

12.8e−s −18.9e−3s 3.8e−8.1s u1

g11 = tf( 12.8, [16.7 1], 'IOdelay', 1.0,'TimeUnit','minutes');

Define a MIMO system by creating a matrix of transfer function models.

DC = [g11 g12 g13

DC.InputName = {'Reflux Rate','Steam Rate','Feed Rate'};

-->Assuming unspecified input signals are manipulated variables.

Review the resulting system.

From input "Reflux Rate" to output...

From input "Steam Rate" to output...

From input "Feed Rate" to output...

Continuous-time transfer function.

where CA is the concentration of a key reactant, T is the temperature in the reactor, Tc is

Measurement of reactant concentrations is often difficult, if not impossible. Let us assume

poles = [-0.3, -0.1+0.7i, -0.1-0.7i];