|
 |
| |
|
|
|
| |
Strategy Discovery Using Learning Agents,
A Power Market Example
|
| |
|
|
|
| |
Speaker:
|
Dr. Kong-wei LYE
Research Engineer
Planning & Operations Management Group
Singapore Institute of Manufacturing Technology (SIMTech), A*STAR
|
|
| |
Date:
|
19 December 2005 (Monday) |
|
| |
Time:
|
3:00 pm to 5:00 pm |
|
| |
Venue:
|
Meeting Room 4.4, Level 4
School of Information Systems
|
|
| |
In this talk, I shall present part of my PhD work in the use of agent learning in strategy discovery. The wholesale power market in the USA is a repeated single-sided auction market in which power generation companies (gencos) offer their supplies and an Independent System Operator buys on behalf of power consumers. In the late 1990's, prices in California's power market skyrocketed and blackouts came one after another. This prompted many to suspect that gencos colluded to manipulate the market. There was however, little evidence pointing to their collusion. By modeling gencos as learning agents, and through simulating the market, I show that simple learning algorithms are sufficient for gencos to discover independent strategies that allow them to act collectively like a monopoly, regardless of the elasticity of demand.
|
|
| |
Kong-wei Lye is a research engineer in the Planning and Operations Management group at Singapore Institute of Manufacturing Technology (SIMTech), A*STAR. His research interests include multi-agent systems, game theory, and machine learning. Currently, he is also interested in applying control theory to supply chain modeling and management. Prior to joining SIMTech, Lye was a systems analyst at Port of Singapore Authority where he was a member of the team that developed PortNet, PSA's Internet business portal.
Lye received his B.A.Sc(Hons) in Computer Engineering from Nanyang Technological University in 1996, and the M.S. and Ph.D. degrees in Electrical and Computer Engineering from Carnegie Mellon University in 1999 and 2004 respectively. |
|
We look forward to welcome you at this Research Talk. |
|
|
