F7-1: System Identification

Aalborg University Esbjerg

 

Instructor

Youmin Zhang

Department of Computer Science and Engineering

Aalborg University Esbjerg

Niels Bohrs Vej 8

6700 Esbjerg, Denmark

Phone: (+45) 7912 7741                    Fax:   (+45) 7912 7710

Email:  ymzhang@cs.aue.auc.dk      Homepage: http://www.cs.aue.auc.dk/~ymzhang/

Time

Wednesdays (Onsdag), 8:15 a.m. - 12:00 noon

Location

B201

Recommended Textbook
  • Lennart Ljung, System Identification: Theory for the User, 2nd Edition, Prentice-Hall, 1999 (ISBN: 0-13-656695-2)

References
  • L. Ljung and T. Glad, Modeling of Dynamic Systems, Prentice Hall, 1994 (ISBN 0-13-597097-0)
  • T. Soderstrom and P. Stoica, System Identification, Prentice Hall International, 1994 (0-13-127606-9).

 

 

Class Schedule and Downloads

 

Lecture #

Contents

Lecture 1

 

8 Sept.

Introduction - General introduction to modelling and system identification

1)     Theory and experiment based modelling methods;

2)     Parametric and non-parametric models and identification methods;

3)     Procedure of system identification.

Reading: 

1)     Textbook: Chapter 1, Sections 4.1-4.3

2)     L. Ljung, From Data to Model: A Guided Tour of System Identification, Linköping University, Sweden, Report No. LiTH-ISY-R-1652, 1994.

Lecture Notes: Lecture 1

Lecture 2

 

29 Sept.

Non-recursive (off-line) methods

1)     Least-Squares (LS) method and its variants;

2)     Instrumental variable methods;

3)     Prediction error methods.

Reading: Textbook: Sections 7.1-7.3, 7.5-7.

Exercise: 7G.1, 7E.1

Lecture Notes: Lecture 2

Lecture 3

 

6 Oct.

Recursive (on-line) methods (I)

1)     Recursive Least-Squares (RLS) methods;

2)     Tacking and forgetting factor techniques.

Reading: Textbook: Chapter 11.

Exercise: 1) 11E.1, 2) 11T.1, 3) Try to derive the weighted RLS from the weighted LS.

Lecture Notes: Lecture 3

Lecture 4

 

13 Oct.

Recursive (on-line) methods (II)

1)     Kalman filter for parameter estimation;

2)     Recursive instrumental variable methods;

3)     Recursive prediction error methods;

4)     Recursive pseudolinear regressions;

5)     Comparison of different methods;

6)     Newly developed sliding window blockwise least-squares algorithms.

Reading:

1)     Textbook: Chapter 11;

2)     J. Jiang and Y. M. Zhang (2004), A Novel Variable-Length Sliding Window Blockwise Least-Squares Algorithm for On-Line Estimation of Time-Varying Parameters, International Journal of Adaptive Control and Signal Processing, 18(6): 505-521.

3)     J. Jiang and Y. M. Zhang (2004), A Revisit to Block and Recursive Least Squares for Parameter Estimation, International Journal of Computers and Electrical Engineering, 30(5): 403-416.

Lecture Notes: Lecture 4

Lecture 5

 

20 Oct.

On-line identification methods (III), summary of the course, and practical aspects and applications of system identification

1)     Input signals and persistent excitation;

2)      Model structure selection;

3)      Model validation;

4)     Practical aspects and applications of system identification.

Reading:  Textbook: Chapters 13-17
Lecture Notes: Lecture 5

Review

 

4 Jan. 2005

Slides: Review

Examination

Time:  6 Jan. 2005,  kl. 9.00 – 11.00

 

Exercises for examination:  Problem, Solution

 

 

Course-related Projects

Proposals of the course related projects are listed as following:

    Development of New Sliding-Window Blockwise Least Squares Identification Algorithms with Applications to Fault Diagnosis

·         Extension to a recently developed sliding-window batch/blockwise Least-Squares (LS) identification algorithms

·         Applications of the developed new identification algorithm for a fault diagnosis application

·         Matlab/Simulink simulation and implementation with application to a physical system selected

 

    References:

1.      J. Jiang and Y. M. Zhang (2004), A Novel Variable-Length Sliding Window Blockwise Least-Squares Algorithm for On-Line Estimation of Time-Varying Parameters, International Journal of Adaptive Control and Signal Processing, 18(6): 505-521.

2.      J. Jiang and Y. M. Zhang (2004), A Revisit to Block and Recursive Least Squares for Parameter Estimation, International Journal of Computers and Electrical Engineering, 30(5): 403-416.

 

    Subspace Identification Algorithms with Applications to Parameters Estimation in State-space Models

 

·         Review on the subspace identification algorithms

·         Development of subspace-based identification algorithms for parameter estimation of control effectiveness, i.e., parameters in B matrix of the system matrix set {A, B, C, D}

·         Development of subspace-based identification algorithms for parameter estimation in system matrix set {A, B, C, D}

·         Matlab/Simulink simulation and implementation with application to a physical system selected

 

    System Identification and Model Predictive Control (MPC) for Process Control

 

·         Development of on-line identification algorithms for process control, for example, subspace identification approach

·         Development of MPC control with identified process model

·         Integration and interaction of identification and MPC control

·         Simulation and verification using Matlab/Simulink

·        Implementation in closed-loop for control of certain engineering systems, for example (petro-)chemical, oil & gas or other industrial processes

 

This homepage is created and maintained by Youmin Zhang.