Skip to Main Content
We consider an agent-based model of forward trading in electricity. Electricity suppliers and generators buy and sell power in three stylized markets: the forward market, the intra-day or prompt market and the real-time spot or ancillary services market. Using computational learning, we develop a model whereby agents' strategies are determined by evolved neural networks of arbitrary size and topology. In a high carbon system similar to today's conventional fossil-fuel based supply stacks, simple strategies for both agents emerge. When substantial wind generation is included, however, these strategies are seen to be no longer appropriate. New insights relating to the impact of wind on fossil generator market power have substantial implications for price formation and the investment signals regarding peaking capacity.