Deep reinforcement learning for simulating the strategic bidding behaviour of distributed flexibilities in smart markets

Abstract

Local flexibility markets could support the energy transition by providing distribution system operators with a new tool to manage grid congestions caused by the massive expansion of renewable generation. Fast implementation of such smart markets could meet the need for distributed local flexibility and support the integration of new sustainable energy sources. However, the risk of gaming and the difficulty of identifying such strategies hinder the development and implementation of these markets. Therefore, this study proposes a model approach, which leverages the substantial progress in research on reinforcement learning to simulate the strategic bidding of flexibilities in consideration of local grid characteristics.