Optimal Economic Dispatch Scheduling in Competitive Energy Market Utilizing a Greedy Q-Learning Algorithm

In the realm of deregulated electricity markets, selection of power plant units for electricity generation is considered crucial for their optimal operation. In this study, an agent-involved model simulating a competitive energy market consisted of two generation companies aiming to cover load demand and maximize revenue is examined. The agent is trained through the utilization of the Q-Learning algorithm embedded with a greedy reward system. The results clearly illustrate that the agent driven supplier attains the optimal economic dispatch schedule, underscoring the effectiveness of a single training session, which enables the agent to learn the most cost-effective approach to cover maximum capacity while simultaneously minimizing testing errors. Finally, the findings of this study emphasize the potential for improved decision-making for more autonomous economic dispatch processes in the dynamic landscape of competitive energy markets.