Robust Diagnosis of Capacity and SOC Consistency in Battery Pack Based on OCV Reconstruction in Real-Time Battery Management System

Abstract

Accurate consistency diagnosis of series-connected battery packs is crucial for the safety management of lithium-ion batteries. However, traditional methods for extracting and analyzing consistency indicators often require significant memory and computing resources, posing a challenge for embedded battery management system (BMS). In this regard, this work proposes a practical method for real-time diagnosis of state of charge (SOC) and capacity consistency. First, a low-complexity online identification method is employed to obtain the open-circuit voltage (OCV) of individual cells. These OCVs serve as raw indicators which are then reconstructed to extract consistency parameters. During OCV reconstruction, the extended Kalman filter (EKF) is employed to iteratively find optimal solutions with less memory and computational consumption. Additionally, the mean shift algorithm (MS) is adapted to enhance robustness and reliability under practical conditions, such as partially available data and inaccurate estimations from EKF. The proposed method is validated on a real BMS at varying temperatures of 5 °C, 25 °C, and 45 °C. The diagnosis errors for capacity and SOC are within 3.2% and 1.6%, respectively, at all temperatures, even with partially available data. Resource consumption analysis demonstrates that the proposed method maintains appropriate complexity and reduced storage requirements, making it ideal for embedded BMS applications.