Graduate
school in Systems, Optimization, Control and Networks (SOCN) |
Graduate school courses scheduled in the academic year
2009-2010
Fall
2009
"MODEL
PREDICTIVE CONTROL"
Lecturer : Jan
MACIEJOWSKI
Spring 2010
"COMPLEX
NETWORKS : STRUCTURE AND DYNAMICS "
Lecturer : Santo FORTUNATO (ISI Foundation, Torino,
Italy)
"MODELLING,
ESTIMATION AND CONTROL OF BIOSYSTEMS "
Lecturers : Philippe BOGAERTS (ULB), Denis DOCHAIN (UCL)
"LMI
OPTIMIZATION WITH APPLICATIONS IN CONTROL"
Lecturer : Didier HENRION (LAAS-CNRS, Toulouse, France)
Detailed
contents
Lecturer : Jan
MACIEJOWSKI (University of Cambridge, UK)
Description
Model Predictive Control (MPC) is the only `advanced’ control methodology (ie more advanced than PID) which has found wide application in the process industries. It offers advantages which make it very attractive for other industries too, such as automotive and aerospace, and its use in such industries is being actively explored at present.
Contents
The course will start with the basic ideas of MPC, together with some specific examples of its advantages over “classical” control. It will then discuss the structure of MPC controllers, present possible variations (such as non-quadratic cost functions and stabilised predictions), and deal with important practicalities, especially disturbance feedforward and disturbance modelling. A state-space framework will be used, but the connection with the well-known GPC framework will be made. The course will then survey the state of recent MPC-related research, covering efficient computation, stability and robustness, prioritisation of objectives, the use of nonlinear models, the application of MPC to hybrid systems (which contain logic or mode switches as well as continuous dynamics), and distributed MPC. The course will be illustrated throughout with examples from various applications.
The course will be presented over 6 3-hour sessions, each session consisting of 2 lectures.
Session 1:
1. Introduction: Motivation, Basic ideas. Motivation: Industrial success of MPC; reasons for this success; examples of applications; future applications. Basic ideas: Receding horizon, reference trajectory, coincidence points, offset-free tracking, use with unstable plant. A little history.
2. Basic formulation and solution of MPC. Formulation: Linear model, quadratic cost, linear constraints. Prediction, disturbance models and observers. Examples. Solution: QP. Brief presentation of active set and interior point methods. Efficient ordering of variables. Constraint softening. Examples. Structure of MPC controller: Piecewise-affine.
Session 2:
3. The GPC formulation. Prediction using transfer functions. Prediction with a disturbance model. The GPC model. State-space interpretation.
4. Introduction to MPC Toolbox and hands-on experience.
Session 3:
5. Other formulations. Feedforward from measured disturbances. Stabilised predictions. Non-quadratic cost (1-norm or infinity-norm) - motivation and solution. Zones, funnels,coincidence points. PFC.
6. Stability. Strategies for ensuring stability: terminal constraints; infinite horizons; contraction conditions. Feasibility implies stability. Detailed presentation of Rawlings-Muske approach to infinite horizons in presence of constraints.
Session 4:
7. Tuning. Special cases: `Mean-level control', `Deadbeat control', `Perfect control'. Tuning parameters: Horizons, weights, disturbance model, observer.
8. Robust MPC; Uncertainty models: Norm-bounded; polytopes. Lee-Yu tuning; LQG-LTR tuning; relations to industrial practice. Outline of LMI approach to robust constrained MPC. Output admissible sets, positive-invariant sets, etc. Optimising over feedback policies instead of open-loop actions.
Session 5:
9. Two case studies. Shell heavy oil fractionator. Newell and Lee evaporator.
10. Hands-on experience with case-studies.
Session 6:
11. Nonlinear MPC. Using nonlinear internal models: pros and cons. Approximate solution using repeated relinearisation. Strategies for handling nonlinear internal models. Cost reduction rather than minimization.
12. Perspectives. Exploiting spare degrees of freedom - ideal resting values, fault-tolerance; Constraint management; Mixed Logic Dynamic (MLD) systems; links to hybrid control.
Supporting
material
The course will be based on Prof. Maciejowski’s book “Predictive Control with Constraints” (Prentice-Hall, 2001, ISBN 0 201 39823 0) but will also contain some more recent material.
Dates :
September 30 and october 01, 02,
19, 20, 21, 2009
Schedule : 09h15-12h15
Local : Auditorium Arenbergkasteel KAST01.07
!!! ATTENTION 01 october 15h00 - 18h00 Room 00.24 Dept of Electrical Engineering
This
course will take place at the Katholieke
Universiteit Leuven,
Department Elektrotechniek
ESAT, Kasteelpark Arenberg 10, 3001 Heverlee
2.COMPLEX NETWORKS : STRUCTURE AND DYNAMICS
Description :
The modern science of networks is perhaps the most popular and
promising research area within
complex systems.
Many systems can be represented as networks, but the latter are
characterized by a set of
common statistical
regularities, which are crucial to understand their structure and
function.
Contents :
The course will be divided into 6 sessions of 2 and a half hours each.
Session 1:
Networks: definitions, characteristics, basic concepts in
graph theory, centrality measures,
weighted graphs
Session 2:
Real World Networks: examples, small-world properties,
clustering coefficient,
fat-tailed degree distributions
Session 3:
Network models: Erdoes-Renyi graphs, models based on
preferential
attachment, other models
Session 4:
Community structure I: elements of community structure, basic
problems and
classical methods
Session 5:
Community structure II: new methods, modularity, testing and
significance of
clustering
Session 6:
Dynamics on networks: resilience/percolation, epidemic
spreading, social dynamics,
navigation, synchronization
Dates : February 02, 03, 05, 16, 17, 19, 2010.
Schedule : 09h15-12h15
Support : paper 1
Lecture I, Lecture II, Lecture III, Lecture IV, Lecture V, Lecture VI.
This course will take place at CESAME, Bâtiment Euler, 4, av. G. Lemaître, 1348 Louvain-la-Neuve
3.MODELLING, ESTIMATION AND CONTROL OF BIOSYSTEMS
Course description :
Aim : The objective of this course is to give an introduction and cover recent aspects of dynamical modeling, monitoring and control of biochemical processes. The course will cover the following topics :
Dynamical modeling of biochemical processes : the notion of reaction networks and mass balance modeling will be introduced and used to build a general dynamical model for bioprocesses, both stirred tank reactors (described by ODE’s (ordinary differential equations)) and incompletely mixed reactors, such as fixed bed or fluidised bed reactors as well as population balance models (described by PDE’s (partial differential equations)). Mathematical concepts of the general dynamical model, including model reduction and stability, as well as microbial ecology concepts like the competitive exclusion principle, will be studied. The link with metabolic engineering will also be explicated. The course will also cover the identification of bioprocess models (including the structural and pratical model identifiability and the design of optimal experiments for parameter estimation). It will also address simulation issues related to PDE models and the use of reduction methods for this type of models.
Monitoring : this part of the course will be dedicated to the design applications of state observers (Luenberger observers, Kalman filters, asymptotic observers, robust observers, …) and parameter estimation algorithms (in particular to estimate reaction rates and yield coefficients).
Control : the course will emphasize optimal control and (adaptive) linearizing control (including adaptive extremum seeking). The choice of these control approaches will be motivated in the context of bioprocess applications.
Several practical applications will be used to illustrate the techniques and principles covered in this course. Examples will include problems from the food industry and the pharmaceutical industry to the environment and the (waste) water treatment.
Contents :
Module |
Themes |
duration |
S1 |
Biosystem Modeling : introduction, model classes, property analysis |
1.5h |
S1 |
Parameter identification (I) |
1.5h |
S2 |
Parameter identification (II) |
2h |
S2 |
Optimal experiment design |
1h |
S3 |
State estimation |
3h |
S4 |
An overview of applications (food engineering, population models, microbial ecology, environmental processes, biotech industries) |
2h |
S4 |
Numerical simulation (with introduction to PDEs) |
1h |
S5 |
process optimization (optimal control and adaptive extremum seeking) |
3h |
S6
|
Control in practice: successful and not so succesful approaches with emphasis on batch and fed-batch applications |
3h |
Dates : March 15, 16, 17, 2010.
Schedule : 09h15 - 12h15 and 14h - 17h
ATTENTION !! On March 15 the course will begin at 9h00 until 12h00 and 14h00 until 17h00.
This course will take place at CESAME, Bâtiment Euler, 4, av. G. Lemaître, 1348 Louvain-la-Neuve
Support :
D. Dochain :
A. Vande Wouwer, Ph. Bogaerts :
4. LMI OPTIMIZATION WITH APPLICATIONS IN CONTROL
Lecturer : Didier HENRION (Laas CNRS-Toulouse, FRANCE)
Description
This is a course for graduate students or researchers with a background in linear control systems, linear algebra and convex optimization.
The focus is on semidefinite programming (SDP), or optimization over linear matrix inequalities (LMIs), an extension of linear programming to the cone of positive semidefinite matrices. Since the 1990s, LMI methods have found numerous applications mostly in combinatorial optimization, systems control and signal processing.
Contents
Slides in PDF format will be posted here soon. See here
for the last versions of the slides.
The course starts with fundamental mathematical features of
linear matrix inequalities:
Then we cover latest achievements in semidefinite
programming and real algebraic geometry:
For the labs, we use the YALMIP interface, SeDuMi and
PENBMI to
define and solve LMI and BMI problems under the Matlab environment. We
also briefly survey recent developments in semidefinite solvers and
software packages:
The end of the course focuses on the application of LMI techniques to solve several control problems traditionally deemed as difficult, such as robustness analysis of linear and nonlinear systems, or design of fixed-order robust controllers with H-infinity specifications. The originality of the approach is in the simultaneous use of algebraic or polynomial techniques (as opposed to classical state-space methods) and modern convex optimization techniques:
Homework :
Homeworks are handed out during the course. Some of this material is
used during the labs. Full written solutions to the homeworks and labs
(with Matlab scripts) are available on request.
Schedule :
Monday : 14:00 - 17:30
Tuesday and Wednesday : 9:00 - 12:30
This
course will take place at the Katholieke
Universiteit Leuven,
Thermotechnisch Instituut, Kasteelpark Arenberg 41, room 01.02 (Aula
van de tweede hoofdwet), 3001 Heverlee
Support :
Part 1.0 - Part 1.1 - Part 1.2 - Part 1.3 - Part 1.4 - Part 1.5 - Part 1.6
Other courses scheduled in the academic year 2009-2010
5. NUMERICAL OPTIMAL CONTROL ALGORITHMS, AND APPLICATIONS IN RENEWABLE
ENERGY SYSTEMS
Lecturers : Moritz DIEHL (KUL, Leuven, Belgium) and B. HOUSKA(KUL, Leuven, Belgium)
Professor Responsible : Prof. Dr. Moritz Diehl
Telephone : 003216321884 Fax : Email :moritz.diehl@esat.kuleuven.be
Objectives :
Aim of the very interactive course is to provide the participants with a
strong working knowledge about the methods and applications of dynamic
optimization in engineering applications.
Programme to be followed :
The course will consist of lectures, interactive sessions and guided computer exercises.
Applications from several fields are treated in self-chosen tutorial
projects by the participants in the last two days of the course.
Particular emphasis is put on renewable
energy systems like wind power, seasonal heat pumps, or solar thermal
power plants.
A tentative list of treated topics is: Dynamic system modelling for
optimization, theory of nonlinear programming and optimal control,
dynamic programming, indirect versus direct approaches, simultaneous vs.
sequential approaches, parameter estimation and nonlinear least squares
problems, model
predictive control, application in chemical and mechanical engineering.
The software tool to be used is the open source tool ACADO - a toolkit for
automatic control and dynamic optimization.
Towards the end of the course every participant will be working on
formulating and solving a dynamic optimization problem of her/his own
choice, so it is encouraged to think about interesting applications of
dynamic optimization even before the course. The lectures and exercises
will be given by the organizers.
Minimum year of study: 4 th and 5 th year
Minimum level of English : high
Key words : numerical mathematics, optimization, direct optimal control methods, object oriented programming, interest in real world applications, dynamic system modeling,
Language (in which the Course will be taught) : English
Minimum : (Number of students required for the course to take place) : 10
Maximum : ( Total number of places, Home & non Home students, not to be exceeded) 30
Reserved for local Home students : ( This total is included in the Maximum number of places) 5
Prerequisites :
This course is aimed at 4th or 5th year master students with very strong
skills in mathematics and a working knowledge of programming in C and
MATLAB. Strong knowledge of analysis and linear algebra (2 years) is
Requested and knowledge of numerical mathematics is very helpful.
Course exam :
A short written exam for self-assessment and rehearsal will be held on Friday morning
and the remaining time is devoted to individual computer projects performed by the participants.
Schedule :
09:00 - 12:30 & 14:00 - 18:00 except for friday 14:00 - 16:00
6. NONLINEAR OBSERVER DESIGN : A DISSIPATIVE APPROACH
Lecturer : Jaime MORENO (UNAM, MEXICO)
Objectives :
The general objective of this minicourse (10 hours) is to give, first, a brief overview of different nonlinear observer design problems and methods. Then a recent approach for observer design, based on the dissipativity theory, will be presented and discussed. It will be shown that this dissipative approach unifies and generalizes several current design strategies. The general ideas will be illustrated with examples and possible applications for control.
Contents :
The course is divided in 4 sessions:
Dates : December 1 and 3, 2009
Schedule : 10:00 - 12:00 and 14:00 - 17:00.
Schedule : Auditorium of Service d'Automatique, 31 Boulevard Dolez, 7000 Mons (1st floor left).
7.BELGIAN FRANCQUI CHAIR 2009-2010 - PAUL VAN DOOREN
Prof.dr.
Friday, April
30, 2010: Model reduction of dynamical systems.
Friday, May 07,
2010: Dominant feature extraction and structured
matrices.
Friday, May 21, 2010:
Networks and graphs.
You
can register
for this occasion before April 1 by email to Martine Vermeiren (martine.vermeiren@ua.ac.be),
mentioning the number of participants.
Registration
You can register electronically by filling in the following form via the web :
http://www.uclouvain.be/sites/socn/graduate_registration.html
If you have problems with this, please contact Nathalie PONET
The admission is free for doctoral students and participants
from Belgian
academic institutions. Other participants are requested to pay a
registration
fee of EURO 500,- per
course but a waiver can be
obtained under special conditions (contact the secretariat).
Payment can be made by bank transfer to the account
n°310-0959001-48 with the
mention "AUTO2866 ACTIVITES DIDACTIQUES"
Secretariat
Nathalie PONET
CESAME - Bât. Euler
4, av. G. Lemaître
1348 Louvain-la-Neuve (Belgium)
Tél : 010/47 80 36
fax : 010/ 47 21 80
e-mail : nathalie.ponet@uclouvain.be
web site : http://www.uclouvain.be/sites/socn