Bid filtering for congestion management in European balancing markets – A reinforcement learning approach
Résumé
Innovations for near real-time common European balancing markets are underway to meet the flexibility needs induced by the deployment of renewables and new market agents. Never have markets and real-time network operations been run so closely on a continental scale. Our paper investigates a filtering method for integrating congestion management and near real-time markets. Reinforcement Learning is applied to add the cost of physical delivery to bid prices to advantage/disadvantage bids that reduce/create congestion. We assess the impact of this new method on market welfare and congestion management costs and show that it brings significant efficiency gains compared to no filtering or a baseline filtering methodology.