Optimal design of electricity auctions from deep learning:

Optimal auction design in electricity markets is important to ensure affordable, sustainable and clean energy. This paper, motivated by recent advances in finding optimal mechanisms using Deep Learning, introduces a method for discovering optimal designs for electricity auctions that minimize expected generation costs and incentivize generators to disclose their true unit costs. The results highlight the effectiveness of the method for recovering analytical solutions in capacity-constrained reverse auctions, and its flexibility to find optimal designs in scenarios where there is uncertainty in capacities and energy demand, costs are correlated, and unit costs are not identical between different time intervals during the day. Finally, experiments are conducted with real data from the Wholesale Electricity Market in Colombia during 2022, to evaluate the effect that the integration of solar and wind energy has on generation costs in the optimal auction.