Capacity estimation of lithium-ion cells by combining model-based and data-driven methods based on a sequential extended Kalman filter

Estimating the capacity of lithium-ion cells employed in electric vehicles is challenging because of the complex vehicle conditions and inconsistent cell decay. This paper proposes a novel capacity estimation method realized by combining model-based and data-driven methods based on a sequential extended Kalman filter (SEKF), to improve the accuracy, and reliability of capacity estimation. First, cycle aging tests are conducted on four cells under different aging stress. Second, the state-of-charge and capacity of the cells are co-estimated using a third-order extended Kalman filter (EKF) driven by dynamic data obtained under different aging stages. The advantages and disadvantages of this data-driven method are investigated. Third, a discrete Arrhenius aging model (DAAM) is developed to estimate the capacity, and its parameter mismatch problem is addressed. Finally, an SEKF estimator is proposed to integrate the capacities obtained using these methods. The SEKF comprises two EKFs connected in series: one to update the model parameters of the DAAM via the feedback provided by the third-order EKF and the other to combine the capacities from the third-order EKF and DAAM. The experimental results show that the proposed SEKF is suitable for capacity estimation with excellent accuracy and stability under different aging stress and dynamic conditions. (C) 2020 Elsevier Ltd. All rights reserved.

Automated summary

This paper proposes a novel capacity estimation method by combining model-based and data-driven methods based on SEKF to improve accuracy and reliability. Through cycle aging tests and dynamic data, it successfully estimates the state-of-charge and capacity of the batteries, while addressing a parameter mismatch problem.