Improving real-world execution of optimized trading schedules for large-scale battery storage systems through data-driven component parametrization

Large-scale battery storage systems (BESS) play a key role in ancillary services and are set to contribute significantly to short-term energy trading. However, linear BESS optimization models for energy trading are often based on simplified assumptions, such as fixed component efficiencies. These simplifications fail to capture crucial operational constraints and result in discrepancies between scheduled and actual state of charge (SOC), leading to unfulfilled power delivery and financial penalties. This study addresses this gap by using field data from a real BESS to parametrize linear load-dependent efficiency models for inverters and transformers. Furthermore, the models are validated in the field by assessing their accuracy in calculating power delivery. This includes accounting for component efficiencies and SOC dynamics and comparing the results to a reference test. The data-driven parametrization presented in this study achieved a reduction of 78.2% in unfulfilled energy delivery and a 71.7% reduction in balancing energy costs caused by deviations compared to the reference test. It also significantly decreased the BESS round-trip efficiency deviation between modeled and measured values, with a 4.2 percentage point improvement over the reference test. The linear inverter model achieved a deviation from the actual measured round-trip efficiency of only 0.55 percentage points. These findings highlight the importance of accurate efficiency modeling in minimizing SOC deviations and fulfilling planned schedules in energy trading applications. Finally, this work proposes a methodology that is broadly applicable not only for energy trading with BESS, but also for ancillary services and multi-use operation.