Learn to Bid as a Price-Maker Wind Power Producer

We consider the problem of a Wind power producer (WPP) participating in short-term power markets, that faces significant imbalance costs due to its non-dispatchable and uncertain production. Additionally, some WPPs have a large enough market share to influence market prices with their bidding decisions, thereby rendering price forecasts unreliable—commonly referred to as the price-maker setting. We model this problem as a contextual multi-armed bandit problem that leverages contextual information, such as market and generation forecasts, and accounts for the price-maker effect. We show that our algorithm achieves vanishing regret, compared to an omniscient oracle, ensuring convergence to optimal policy in the long run. The algorithm’s performance is evaluated against various benchmark strategies using a numerical simulation of the German day-ahead and real-time markets.