Frequency sensitivity analysis of battery states and parameters for data-agnostic online estimation

Recent research on battery state estimation typically focuses on battery modeling and estimation algorithms, while poor data quality can also lead to unsatisfactory estimation accuracy. It results from a lack of sufficient frequency components in battery current for state-parameter co-estimation. Conventional approaches using current injection are developed to increase the data quality, but such an intrusive approach degrades the battery operation and thus suffers from limited applicability. This article coordinates the frequency sensitivity analysis with the estimator design, aiming to propose a data-agnostic online estimator (DOE). The battery states and parameters can be accurately estimated using the proposed DOE without data adaptation. Specifically, this article first derives the limitations of existing estimation techniques. The DOE structure is then proposed and identified as robust regardless of the data frequency information. The scheme is experimentally verified with drive cycles at different temperatures, and results show that the DOE outperforms the conventional control groups.