Læringsutbytte

A candidate who has passed the course will have a learning outcome in the form of acquired knowledge, skills, and general competence, as described below[1].

Knowledge

The candidate will:

1. Understand how different discrete time process models may be used to describe Multiple Input and Multiple Output (MIMO) dynamic systems, e.g. FIR models, step response models and state space models
2. Understand how discrete time process models may be used to construct prediction models, i.e., models used to predict the future behaviour of the process outputs
3. Understand how unconstrained Model Predictive Control (MPC) algorithms may be used to construct stabilizing and offset-free control methods.
4. Understand how process constraints may be included in the MPC algorithm in order to efficiently deal with constraints, both for linear and non-linear system models
5. Understand the similarities and the differences between predictive control methods and traditional state feedback based Linear Quadratic (LQ) based optimal control

Skills

The candidate will :

1. Be able to build models for describing dynamic MIMO systems
2. Be able to construct prediction models from different dynamic system models
3. Be able to use optimization theory in order to find the minimum of LQ and non-linear cost functions, both for unconstrained and constrained problems
4. Be able to formulate MPC methods for systems, based on both linearized and non-linear dynamic models

General competence

The candidate will:

1. Understand how MPC methods may be used to efficiently control complicated MIMO dynamic systems
2. Understand how MPC methods may be used as process support tools
3. Understand and provide the necessary foundation for further studies
4. Get experience in presenting scientific theory and applications

[1] Codes given in parentheses correspond to the codes given in the programme description.

Innhold

Mathematical programming/optimization methods. Optimization of dynamical systems (overview): model fitting, state estimation, data reconciliation. Discrete and continuous time systems, and parametrization of high order problems. Predictive control in LQ systems: DMC, GPC, state space models. Offset-free predictive control, and robust predictive control. Predictive control in nonlinear systems; efficient solution methods. Predictive control and controller efficiency. Overview of strategies for implementing control algorithms. Overview of commercially available predictive controllers.

Arbeids- og læringsformer

Lectures and an assignment are used.

Lectures are used to highlight the main topics of the course, facilitating knowledge.

During the semester the students will work with an assignment. This assignment will be based on problem based learning giving the students better understanding of specific subtopics in the course and general competence in writing scientific reports with citing and referencing sources.

An on-line, part-time version of the SCE study programme will start in the Fall 2015. The present course will be taught on-line from the fall/spring semester year 2017. However, the course will continue to be taught also as a traditional campus-based course. The course contents and learning material used in the course will be the same in both programmes, except that in the on-line programme, more emphasis will be put on video presentations, and the compulsory project task will be adjusted to suit the on-line programme.

Vurderingsformer

The final test counts 60% of the final course grade.

A mandatory assignment counts 40% of the final course grade.

It is necessary to pass the final test in order to achieve a passing grade.

Det tas forbehold om mindre justeringer i planen.