Abstract:

With the widespread application of lithium-ion batteries, State of Power (SOP) prediction has become increasingly crucial as a key technology to ensure efficient and safe battery operation. This paper provides a comprehensive review of SOP prediction methods, analyzing four main approaches: map-based methods, mechanism model methods, equivalent circuit model (ECM) methods and data-driven methods. Additionally, the prediction of module-level SOP is discussed. The map-based method is straightforward. However, it requires multiple charge and discharge experiments, which are time-consuming and limit its applicability to single operating conditions. The mechanism model method, based on porous electrode theory and concentrated solution theory, can accurately describe internal battery reactions through partial differential equations. This method can account for the impact of internal battery parameters on power performance, but it comes with a relatively high computational complexity. The equivalent circuit model method employs circuit components to simulate the dynamic responses of batteries. This method can integrate multiple constraints, including voltage, current, and state of charge (SOC), to optimize performance while balancing accuracy and computational efficiency. The data-driven method leverages machine learning techniques, such as neural networks and support vector machines, to construct SOP prediction models directly from running data. It can also integrate with traditional mechanistic models to form hybrid architectures, where the prediction performance is contingent upon both data quality and quantity. For module-level SOP prediction, this paper highlights the impact of cell inconsistency on module power and discusses potential solutions. Finally, existing challenges and future development directions are summarized. Current SOP prediction technologies still face four major challenges: (1) Existing methods struggle to meet the specific requirements of energy storage scenarios. (2) The prediction accuracy and computational efficiency fail to meet the requirements of practical applications. (3) Battery aging introduces time-varying parameters, leading to model mismatches. (4) inconsistencies among battery modules exacerbate prediction difficulties. To address these challenges, future SOP prediction technologies will advance in several key areas, including high-precision modeling and optimization of solution strategies, dynamic updating of model parameters and constraint boundaries, and the development of approaches such as "weak cell identification—characteristic cell modeling—dynamic model parameter updates." These advancements will provide safer and more efficient battery management solutions for energy storage systems.

Key words: Lithium-ion battery, battery management system, battery model, state of power prediction