Improved gray wolf particle filtering and high-fidelity second-order autoregressive equivalent modeling for intelligent state of charge prediction of lithium-ion batteries.

Xie, Yanxin; Wang, Shunli; Fernandez, Carlos; Yu, Chunmei; Fan, Yongcun; Cao, Wen; Chen, Xianpei

Abstract

Abstract: The rapid development of new energy vehicles puts forward higher requirements for the lithium-ion batteries model construction, high-efficiency condition monitoring and collaborative estimation. An improved high-fidelity second-order autoregressive model is proposed and constructed, and the autoregressive model is integrated with the second-order equivalent circuit model, which can achieve an accurate and reliable description of the batteries internal dynamic change process. To achieve the accurate expression of the battery's external characteristics and internal state, the forgetting factor is combined with the recursive least square algorithm to improve the parameter identification accuracy and optimality while reducing the space complexity of the algorithm. A novel gray wolf particle filtering algorithm is proposed, which eliminates the particles severe degradation in traditional algorithms and enhances the ability of particles to resist degradation. The algorithm superiority and generalization are verified under complex working conditions. The experimental results show that the accuracy of the high-fidelity second-order autoregressive model can reach 99%, which can well simulate the complex chemical reaction process inside the lithium-ion battery. Experimental simulation is performed under constant current conditions. Compared with the extended Kalman filter, unscented Kalman filter, and particle filter algorithms, the gray wolf particle filter algorithm has reduced the root mean square error by 3.39%, 0.90% and 2.84%, and the mean absolute error has reduced by 1.95%, 0.51% and 2.22%. Under dynamic stress test conditions, the root mean square error is reduced by 1.54%, 0.33%, and 0.78%, and the average absolute error is reduced by 1.4%, 0.22%, and 0.76%. In addition, when tested under different environmental conditions, although the improved algorithm has a relatively long running time, the estimation accuracy of the algorithm is greatly improved and the execution efficiency is high. The improved algorithm provides a theoretical basis for the reliability and stability of the onboard operation of lithium-ion batteries.

Citation

XIE, Y., WANG, S., FERNANDEZ, C., YU, C., FAN, Y., CAO, W. and CHEN, X. 2021. Improved gray wolf particle filtering and high-fidelity second-order autoregressive equivalent modeling for intelligent state of charge prediction of lithium-ion batteries. International journal of energy research [online], 45(13), pages 19203-19214. Available from: https://doi.org/10.1002/er.7014

• Journal Article Type: Article
• Acceptance Date: Jun 23, 2021
• Online Publication Date: Jul 12, 2021
• Publication Date: Oct 25, 2021
• Deposit Date: Jul 1, 2021
• Publicly Available Date: Jul 13, 2022
• Journal: International journal of energy research
• Print ISSN: 0363-907X
• Electronic ISSN: 1099-114X
• Publisher: Wiley
• Peer Reviewed: Peer Reviewed
• Volume: 45
• Issue: 13
• Pages: 19203-19214
• DOI: https://doi.org/10.1002/er.7014
• Keywords: State of charge; Lithium-ion batteries; High-fidelity second-order autoregressive model; Gray wolf particle filtering
• Public URL: https://rgu-repository.worktribe.com/output/1375603