Selective initial state criteria to enhance convergance rate of Q-Learning algotithm in power system stability application

Author

Hadidi, Ramtin ; Jeyasurya, Benjamin

Author_Institution

Fac. of Eng. & Appl. Sci., Memorial Univ., St. John´´s, NL

fYear

2009

fDate

3-6 May 2009

Firstpage

486

Lastpage

489

Abstract

In this paper, a modified Q-Learning algorithm is proposed to enhance the convergence speed of the conventional algorithm to reach a near optimal policy. This is achieved by using selective initial state criteria (SISC) instead of choosing initial state randomly in each episode. The proposed method is implemented to control power system stabilizers to enhance power system stability. The validity of modified Q-Learning has been tested on a 2 area, 4 machines power system.

Keywords

learning (artificial intelligence); power system stability; Q-learning algorithm; convergence rate; power system stability; reinforcement learning; selective initial state criteria; Control systems; Convergence; Dynamic programming; Electronic mail; Learning; Power engineering and energy; Power system control; Power system modeling; Power system stability; Power systems; Convergence; Q-Learning; power system stability; reinforcement learning;

fLanguage

English

Publisher

ieee

Conference_Titel

Electrical and Computer Engineering, 2009. CCECE '09. Canadian Conference on

Conference_Location

St. John´s, NL

ISSN

0840-7789

Print_ISBN

978-1-4244-3509-8

Electronic_ISBN

0840-7789

Type

conf

DOI

10.1109/CCECE.2009.5090182

Filename

5090182