Towards Real-Time Estimation of Li-ion Battery Characteristics for BMS with Storage-Limited Processors

In order to improve the efficiency and extend the cycle life of battery systems, accurate estimation algorithms implemented in battery management systems (BMS) are crucial to calculate critical battery performance indices. Real-time state of charge (SOC) and battery capacity determination are vital for battery diagnostics and prognostics and are the core function of these algorithms. However, the present typical algorithms are difficult to implement in BMS designed with computation- and storage-limited processors. Therefore, it is essential to investigate different real-time estimation algorithms' accuracy and computational complexity. After a summary and classification of various SOC and SOH estimation methods, a typical online adaptive model-based state observer framework for SOC and battery capacity estimation is proposed, and their algorithm accuracy and complexity are compared using the processor-in-the-loop (PIL) test. In particular, a co-estimation algorithm combining recursive least squares with the forgetting factor (FF-RLS) and dual-central differential Kalman filter (CDKF) is implemented, resulting in a mean absolute error of less than 0.62 % for SOC and a mean absolute relative error of less than 1.39 % for SOH.