Research on Peak Power Prediction of Lithium-Ion Battery Modules for Energy Storage Frequency Regulation Scenarios

Abstract Lithium-ion Battery Energy Storage Systems (BESS) are widely applied in grid frequency regulation due to fast response. Accurate maximum sustained power prediction is critical for safe operation. However, cell inconsistencies in internal resistance, temperature, and State of Energy (SOE) make module State of Power (SOP) prediction challenging. This paper proposes a peak power prediction method for lithium-ion battery modules in BESS frequency regulation. First, cell SOE is estimated using an electro-thermal coupled model and the adaptive unscented Kalman filter (AUKF), instead of the commonly used state of charge (SOC), considering frequency regulation applications. Further, a "reference-difference" model is employed to estimate SOE differences between cells in a module. The proposed method explicitly estimates SOE differences among cells, allowing module-level SOP prediction to account for cell inconsistency. Then, the module-level SOP prediction method is proposed, taking inconsistencies in SOE and internal resistance into consideration to predict the peak sustained maximum power demand over a future period during grid frequency regulation. Finally, an experimental platform is established to verify the proposed method with a battery module. The method is validated at 15, 25, and 35 °C over a 300 s prediction window, with power prediction errors of 6.5%, 3.2%, and 5.3%, respectively.