Extended Gradient-Based Model for Real-Time Determination of Local Temperature-Dependent Currents Within Lithium-Ion Batteries

Abstract

Knowledge of local temperature-dependent current distributions helps battery management systems (BMS) to ensure an optimal operation. However, current measurements for all cells within a battery pack are technically not feasible and common model-based methods are too complex for a real-time application on simple BMS computing hardware. We already published a model to determine local cell currents based on the linearization of temperature-current dependencies. During evaluation with different cells, however, this model exhibited weaknesses for longer cycles with high discharge current. Therefore, we propose an extended version of this model that ensures reliable results also for such load profiles. For this purpose, subspace identification methods are used, which allow a purely data-based, user-friendly and robust model identification. We compare two different algorithms, which both will be shown to provide good results. The parameterization of this extended model is still based on only few measurement data, which can be easily determined. The memory requirement remains very low and the calculation of the model is simple enough to meet real-time requirements even on simple control units.