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

Xin Lai,Wei Yi,Yifan Cui,Chao Qin,Xuebing Han,Tao Sun,Long Zhou,Yuejiu Zheng

DOI: https://doi.org/10.1016/j.energy.2020.119233

IF: 9

2021-02-01

Energy

Abstract:<p>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.</p>

energy & fuels,thermodynamics

What problem does this paper attempt to address?

This paper proposes a novel lithium-ion battery capacity estimation method that combines model-driven and data-driven approaches, based on the Sequential Extended Kalman Filter (SEKF). The main objective of the study is to address the accuracy and reliability issues in lithium-ion battery capacity estimation for electric vehicles, especially under complex dynamic conditions and battery aging.

In the experimental part, the researchers conducted cycle aging tests on four batteries under different aging stresses, including accelerated aging experiments conducted at alternating temperatures of 45°C, -25°C, and -10°C. They first obtained the basic performance data of the batteries through standard capacity tests, mixed pulse power characterization tests, and dynamic condition tests. Then, the third-order Extended Kalman Filter (EKF) was used to jointly estimate the state-of-charge (SOC) and capacity of the batteries using dynamic data. In addition, a Discrete Aharoni Aging Model (DAAM) was developed to update the model parameters in real-time for capacity estimation, and these two methods were combined using the sequence extended Kalman filter (SEKF) to improve the accuracy and stability of capacity estimation.

The innovations of this paper are:
1. A novel battery capacity estimation framework and method that combines data-driven and model-driven approaches.
2. The application of the third-order EKF for simultaneous estimation of SOC and capacity in different aging stages of the battery.
3. The construction of a DAAM model with updateable parameters for real-time capacity estimation, with EKF used for parameter calibration.
4. The proposed SEKF algorithm, which combines the results of the third-order EKF and improved DAAM for capacity estimation, to enhance estimation accuracy and stability under different aging pressures and dynamic conditions.

Experimental verification shows that this method exhibits high accuracy and stability in capacity estimation under complex aging conditions.