Revisiting the dual extended Kalman filter for battery state-of-charge and state-of-health estimation: A use-case life cycle analysis

One of the most discussed topics in battery research is the state-of-charge (SOC) and state-of-health (SOH) determination of traction batteries. Unfortunately, neither is directly measurable and both must be derived from sensor signals using model-based algorithms. These signals can be noisy and erroneous, leading to an inaccurate estimate and, hence, to a limitation of usable battery capacity. A popular approach tackling these difficulties is the dual extended Kalman filter (DEKF). It consists of two extended Kalman filters (EKFs), that synchronously estimate both the battery states and parameters. An analysis of the reliability of the DEKF estimation against realistically fading battery parameters is still a widely discussed subject. This work investigates the DEKF performance from a high-level perspective, involving different load dynamics and SOH stages. A numerical optimization-based approach for the crucial filter parameterization is employed. We show that the DEKF partly improves the accuracy of the SOC estimation compared to the simple EKF over battery lifetime within the operational limits of an automotive application. However, capacity and internal resistance estimation becomes unreliable and partly diverges from the reference under constant and realistic load scenarios coupled with advanced degradation. As a consequence, a downstream use of both parameters in a SOC or SOH estimation is hampered over the battery lifetime. Extensions are needed to improve reliability and enable employment in real-world applications.