Reinforcement Learning-Driven Decision-Making in Deregulated Electricity Markets Involving Greedy Agent-Based Participants

The transition from conventional power systems to smart grids has established the foundation for the integration of advanced artificial intelligence algorithms aimed at addressing optimization challenges within deregulated electricity markets. This paper explores the dynamic intersection of deregulated energy markets, economic dispatch principles, and the integration of sophisticated Reinforcement Learning algorithms. Specifically, the role of the Q-Learning algorithm in shaping the decisionmaking processes of a generation company, with the objective of maximizing revenue in a competitive environment, is studied. Additionally, a novel greedy reward system has been embedded to enhance the learning capabilities of an agent, guiding it towards the formulation of optimal power generation decisions. The numerical results of our paper highlight the importance of implementing reinforcement learning approaches in power systems and emphasize the significance of reward engineering in Reinforcement Learning research related to revenue maximization within the evolving landscape of deregulated energy markets.