The Advanced Bidding Strategy for Power Generators Based on Reinforcement Learning

Abstract

Facing the competition in the maturing electricity markets, the power producers have to devise innovative bidding strategies to maximize profits while balancing the costs. In this article a new method for supply bid design based on reinforcement learning of agents is proposed for a thermal power producer. Through their bidding strategies, the producers are following the composite objective function of maximizing profits while maintaining the utilization factor. The performance indicators, which include profit of each unit and its utilization factor, are presented in learning and non-learning scenarios for the Slovenian power system. The results show that the agents that use the reinforcement learning bidding strategy consistently outperform those that use simple non-learning bidding on the electricity market.

Keywords:

• agent-based modeling
• bidding strategy
• power market simulation
• power system economics
• reinforcement learning