Simulation of electricity markets using agent-based computational learning

De Oliveira, Fernando (2003) Simulation of electricity markets using agent-based computational learning. Doctoral thesis, University of London: London Business School. OPEN ACCESS


The purpose of this research is to conduct an analysis of how agent-based computational learning may contribute to a better understanding of pricing policies and strategic management of plant portfolio in electricity markets. The contributions of this thesis are methodological and theoretical with applications in policy analysis for electricity markets. At a policy level, this thesis applies agent-based simulation to the analysis of the impact of market design on the players' strategies and on the industry's performance as a whole. This represents the first detailed study of the New Electricity Trading Arrangements (NETA) in the England and Wales (E&W) electricity market, giving insights into the implications of NETA before its introduction. Further, this thesis addresses the issue of dominant position abuse by individual players in electricity markets. The context is a real application to the E&W electricity market as part of a Competition Commission Inquiry. The research contributions are in the areas of both market power and market design policy issues. At a methodological level, this thesis presents two contributions: the Finite Automata Dynamic Game (FADG) and the Plant Trading Game. The FADG models learning and adaptation in N-player extensive form games of incomplete information, where co-evolutionary automata learn and adapt together. The plant trading game is a large coordination game, simulating how players optimally learn and adapt in order to trade electricity plants. At a theoretical level, this thesis presents three contributions. First, it develops a stylised model for conduct-evaluation in electricity markets, which is applied to the analysis of market power abuse and regulatory policy. Second, it studies plant trading within the context of a Cournot game. Third, it shows that, in an FADG, best response is a necessary but not a sufficient condition for rational behaviour.

More Details

Item Type: Thesis (Doctoral)
Subject Areas: Management Science and Operations
Date Deposited: 25 Feb 2022 10:48
Date of first compliant deposit: 25 Feb 2022
Subjects: Electricity supply industry
Computer-based learning
Simulation models
Last Modified: 05 Mar 2022 22:21

Export and Share


Published Version - Text


View details on Dimensions' website

Downloads from LBS Research Online

View details

Actions (login required)

Edit Item Edit Item