This study uses a dynamic 5-bus test case implemented via the AMES Wholesale Power Market Test Bed to investigate strategic capacity withholding by generation companies (GenCos) in restructured wholesale power markets under systematically varied demand conditions. The strategic behaviors of the GenCos are simulated by means of a stochastic reinforcement learning algorithm motivated by human-subject laboratory experiments. The learning GenCos attempt to improve their earnings over time by strategic selection of their reported supply offers. This strategic selection can involve both physical capacity withholding (reporting of lower-than-true maximum operating capacity) and economic capacity withholding (reporting of higher-than-true marginal costs). We explore the ability of demand conditions to mitigate incentives for capacity withholding by letting demand bids vary from 100% fixed demand to 100% price-sensitive demand.
Available at: http://works.bepress.com/leigh-tesfatsion/47/