Consistency assessment for Lithium-ion battery energy storage systems based on low-rank matrix approximation and non-parametric Kullback-Leibler divergence

Abstract

Consistency is a critical factor that influences the energy efficiency and operational safety of lithium-ion battery energy storage systems (ESSs). Conducting consistency assessment is essential for ESSs. Specifically, evaluating the consistency levels of ESSs is crucial for their health management. Furthermore, it is necessary to locate inconsistent battery packs for subsequent equalization maintenance and replacement as consistency deteriorates. This paper proposes a general assessment framework for implementing consistency evaluation and inconsistency identification in ESSs. First, considering the high-dimensionality and sparsity of charge-discharge voltage cycle data, a consistency evaluation indicator for battery groups is developed based on the rank of the approximated low-rank matrix, which accounts for inherent heterogeneity and uncertainties. Then, inconsistency identification is performed by calculating the non-parametric Kullback-Leibler divergences between the voltage distributions of each monitored battery pack and the normal packs. Finally, an industrial application to a real-world energy storage power station is conducted to show the effectiveness and robustness of the proposed method. The obtained results demonstrate that the consistency levels of battery groups can be effectively evaluated and inconsistent battery packs can be accurately identified. The proposed consistency assessment framework meets the online application requirements of real-world battery management systems, thereby holding great potential for practical health management in ESSs.