Model Predictive Control – Basic MPC

John Bagterp Jorgensen
Niels Kjolstad Poulsen
Sten Bay Jorgensen

February 19-23, 2007 and March 19-21, 2007
IMM & Department of Chemical Engineering
Technical University of Denmark

Course description:
This course gives an introduction to model predictive control (MPC)
technology with an emphasis on industrial applications. Model predictive
control is by far the most industrially successful advanced process
control (APC) technology. MPC can be applied for multivariate processes
with constraints, strong interactions, time delays and otherwise difficult
dynamics. Operational benefits of MPC include increased production
capacity, decreased material and energy consumption, and reduced product
variability. Typical payback times for MPC projects are less than 1 year.

This course is tailored for industrial participants. The purposeis to
enable participants to structure, design and tune model predictive
controllers for industrial processes. This involves identification and
selectionof processes that can be better controlled by MPC than by
classical PID technology, selection of control structure (MVs, CVs, DVs),
identification of a predictive model from process data as well as design
and tuning of a model based estimator, predictor, and constrained
regulator. These elements will be used to construct and implement a model
predictive control system that will be tested by simulation. The
presentation of the topics will be with an emphasis on practical and
industrially relevant issues.

Course requirements:
The participants must have a working knowledge of Matlab and be familiar
with basic principles of process control.

Course Schedule and Content:

February 19-23: Lectures & Exercises
Day 1: Introduction to simulation technology and implementation of a
simple MPC
Day 2: Optimization principles and unconstrained MPC
Day 3: Constrained MPC and models for linear MPC
Day 4: State estimation and system identification
Day 5: Tuning and implementation examples.

February 24–March 18: Self study
The participants have the opportunity to do an MPC project on a process
that they select themselves. The instructors will provide feedback tothis
voluntary MPC project.

March 19-21:
Day 6: Presentation of case studies and automatic tuning procedures.
Day 7: Closed loop identification for MPC
Day 8: Input design for identification of models for MPC & case study

Course Location:
Technical University of Denmark, DK-2800 Kgs Lyngby, Denmark
(10 km North of Copenhagen, Denmark)

Industrial participants: 15.000 DKR (approx. USD 2.500)
Academic participants: University agreement. 5 ECTS points

Please register with John Bagterp Jorgensen, [log in to unmask], no later than
february 8th.

John Bagterp Jørgensen, PhD
Assistant Professor of Scientific Computing

Informatic and Mathematical Modelling
Scientific Computing
Building 305, Office 109
Technical University of Denmark
DK-2800 Kgs. Lyngby

Direct Phone: +45 45253088