Continous battery optimization on the Intra-day market

Abstract

This research aims to optimize the charging schedules of a Battery Energy Storage Sys- tem (BESS) on the Dutch intraday electricity market using reinforcement learning (RL). The study explores the effectiveness of threshold models developed by Bertrand and Papavasiliou within the Dutch market context, examining their performance with and without allowing to trade outside of BESS capacities (which causes imbalance). Initially, the RL model with imbal- ance demonstrated significant potential, showing higher profitability over traditional strategies. However, its high volatility and occasional extreme losses indicated substantial risk, making the more stable RL model without imbalance a safer alternative for conservative trading strategies. The analysis highlighted the importance of considering battery size and constraints in BESS optimization, as larger capacities showed higher profitability but required careful management to avoid significant imbalances and associated costs. This thesis contributes to the under- standing of BESS optimization in the Dutch intraday market, providing valuable insights into the application of RL and the potential benefits of considering market imbalances and battery constraints in trading strategies.