Imitation Learning-Based Strategic Bidding in Electricity Market Considering Price-Quantity Pairs

In the context of deregulated electricity markets, market participants, particularly producers, often engage in strategic bidding to maximize their own profits. However, the growing dynamics of the market environment and the increasing complexity of bidding price-quantity pairs pose significant challenges to the optimization of bidding strategies. To address these challenges, this paper proposes a deep reinforcement learning (DRL) approach to optimize bidding strategies in spot electricity markets considering the bidding actions on price-quantity pairs. First, a bilevel optimization model is proposed, which considers both price and quantity, providing a more realistic strategy space for producers. Then, to improve learning performance in high-dimensional action spaces, a framework combining imitation learning with DRL is developed to leverage expert experiences in bidding strategies from historical data. Finally, numerical simulations show that our model considering price-quantity pairs increases producers’ profits, while the imitation learning-based framework enhances the learning convergence efficiently.