Improved volumetric noise-adaptive H-infinity filtering for accurate state of power estimation of lithium-ion batteries with multi-parameter constraint considering low-temperature influence.

Wang, Shunli; Hu, Bohan; Zhou, Lei; Liu, Yuyang; Fernandez, Carlos; Blaabjerg, Frede

Abstract

Currently, the field of new energy is booming. Batteries containing lithium-ion have become an important component of new energy vehicles. The key parameters to accurately estimate the battery state depend on the State of Charge (SOC) and the State of Power (SOP). This paper presents a comprehensive investigation into the modeling of lithium-ion batteries. It employs equivalent modeling techniques to develop a fractional order model with enhanced accuracy and an online parameter identification algorithm. Furthermore, an optimized H-infinity filtering algorithm is implemented, and a multiple-parameter constraint mechanism based on the SOP prediction method is proposed for lithium-ion batteries. The traditional least square method is improved first to estimate lithium-ion batteries' state of charge and power. Given the phenomenon of "data redundancy", the Limited Memory Multi-Innovation Least Squares (LM-MILS) method is introduced, which can realize the online identification of the parameters to be identified. Then, to optimize the traditional H-infinity filtering algorithm for the poor applicability of nonlinear systems, a Volumetric Noise-Adaptive H-infinity Filtering (VN-AHF) is designed. It is an algorithm that compensates and corrects the generated errors and improves the estimation accuracy of the charged state. Finally, to improve the low accuracy of estimation strategy under single constraint conditions, a multi-parameter constrained estimation strategy for lithium-ion battery power state is developed, which can achieve high precision estimation of power state. The simulation results demonstrate that the optimized least-squares method exhibits enhanced accuracy, with a simulated voltage error of less than 0.045 V. Once the estimation has reached a stable point, the maximum error is reduced to 0.043 V, representing 1.004 % of the nominal voltage. The volumetric noise adaptive H-infinity filtering algorithm demonstrates robust and accurate performance, as evidenced by its ability to achieve minimum Error (1.002 %), minimum MAE (0.964 %), and minimum RMSE (1.539 %). Furthermore, the algorithm exhibits exceptional precision in power state estimation, with an estimated accuracy of 98 %. The multi-parameter constraint-based power state estimation method for Li-ion batteries has an estimation error of less than 68 W for the discharging state and 53 W for the charging state under the DST condition, and an estimation error of less than 86 W for the discharging state and 75 W for the charging state under the BBDST condition. Finally, through experiments conducted under various working conditions and based on these results, the improved algorithm is theoretically capable of operating the battery safely and efficiently with high accuracy.

Citation

WANG, S., HU, B., ZHOU, L., LIU, Y., FERNANDEZ, C. and BLAABJERG, F. 2025. Improved volumetric noise-adaptive H-infinity filtering for accurate state of power estimation of lithium-ion batteries with multi-parameter constraint considering low-temperature influence. Journal of energy storage [online], 115, article number 115999. Available from: https://doi.org/10.1016/j.est.2025.115999

• Journal Article Type: Article
• Acceptance Date: Feb 23, 2025
• Online Publication Date: Feb 27, 2025
• Publication Date: Apr 15, 2025
• Deposit Date: Mar 3, 2025
• Publicly Available Date: Feb 28, 2026
• Journal: Journal of energy storage
• Print ISSN: 2352-152X
• Electronic ISSN: 2352-1538
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 115
• Article Number: 115999
• DOI: https://doi.org/10.1016/j.est.2025.115999
• Keywords: Lithium-ion battery; State of charge; State of power; Limited memory multiple information least square; H-infinity filtering
• Public URL: https://rgu-repository.worktribe.com/output/2741672