lEarn: A Reinforcement Learning Based Bidding Strategy for Generators in Single sided Energy Markets

We aim to increase the profit of a given generator participating in a single-sided wholesale energy market. We model the market clearing mechanism and the behavior of other generators competing in the market. We utilize interesting structures in the data to classify generators and then build novel supervised competition models for each class of generators. We leverage these models to build an interactive system through which we discover better bidding strategies for the given generator using reinforcement learning (RL). We relax several assumptions made in existing works in order to make the problem more relevant to real life. Our MDP formulation enables us to tackle action space explosion in an efficient way. Further, our state formulation enables us to compute optimal actions across all time-steps of the day in parallel. We compare the performance of the proposed RL based bidding agent with the historical real world performance of a generator in a wholesale energy market.