Learning outcome

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.

Knowledge

The candidate will understand how:

• system identification can be used to build models for dynamic systems from known input and output data
• prediction error based methods for system identification works
• direct subspace based methods for system identification works, both for open and closed loop system identification, and how these methods are based on numerical methods from linear algebra
• various methods for system identification may be analysed for consistency and efficiency

Skills

The candidate will be able to:

• use observed data to build dynamic models for systems
• use prediction error methods for system identification
• use subspace based methods for system identification to build dynamic models from both open and closed loop system experiments
• analyse and validate the identified models

General competence

The candidate will:

• understand how concepts from system identification have a wider use
• understand scientific precision, and the role of theorems and proofs
• get experience in presenting scientific theory and applications

Course Description

• Linear algebra, QR and SVD decompositions; projection theory, e.g., orthogonal and oblique projections
• Regularisation and regression methods like Ordinary Least Square (OLS), Principal Component Analysis (PCA) and Regression (PCR) and Partial Least Squares (PLS)
• System and realization theory for dynamic and linear stochastic systems
• Methods for subspace based system identification; identification of system order; methods as CVA, N4SID, MOESP and DSR
• Subspace identification of closed loop systems
• Prediction Error Methods (PEM) for system identification
• Recursive methods for system identification
• Model validation

Teaching and Learning Methods

The course content is to be presented by lecturer(s) and PhD students in a seminar form with variations, depending on the number of students. The students will be expected to work on case studies which are to be solved using computational tools.

Assessment Methods

The final test will count 100% in the course grade.

Minor adjustments may occur during the academic year, subject to the decision of the Dean