Performance-propagating strategy framework for autonomous power trading

• Main Author: Chen, Yongwei
• Other Authors: Bo An
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Simulation and modeling
• Online Access: https://hdl.handle.net/10356/147894
• Institution: Nanyang Technological University

The Power Trading Agent Competition is a simulation platform for autonomous trading agents to operate between the wholesale market and the retail market, acting as the middleman who buys energy from the wholesale market and resells it to consumers in the retail market in the form of contracts. Many research have been conducted to understand what types of strategies allow autonomous agents to earn the most profits in the long run. Some of the proven strategies focus on optimizing individual performance in each market and not the overall strategy, even though the performance in markets might be interrelated. It is known that the performance of a trading agent in one market affects its performance in another market, yet the interrelationship between the two markets is not well-explored due to the complex nature of it. Therefore, the research question for this work is: ”Can the complexity of multiple markets be broken down into simpler machine learning problems to be used to assemble a holistic and coherent agent in the PowerTAC environment?” This work develops a performance-propagating strategy framework that can be used by many brokers to incorporate a holistic strategy. It involves analysing the relationships within each market and between each market, and then developing a strategy for each relationship in view of the propagating effect. Empirical results from this work have discovered out that while agents deploying a performance-propagating strategy is less effective in sustaining profitability when there are only a few competing brokers, agents with a performance-propagating strategy tends to outperform other competing brokers when more brokers are in the market. Therefore, the performance-propagating strategy framework is a viable framework that can be applied to any other agent’s current individual strategies to improve the total profitability of the agent.