Researchers have developed an online framework for estimating the state of health of lithium-ion batteries using only a one-minute segment of relaxation voltage. Described in a study published in the journal Ionics, the approach addresses longstanding operational challenges in battery management systems, where conventional online monitoring algorithms frequently suffer from sensitivity to varied charging and discharging protocols alongside limited architectural interpretability.

By exclusively utilizing relaxation voltage data, the proposed method mitigates the influence of varying operational strategies on state of health estimation. Specifically, a one-dimensional convolutional neural network is employed to automatically extract intrinsic features from the relaxation voltage sequences.

To process these extracted voltage features efficiently while preserving physical meaning, the research team designed a physics-informed lightweight transformer. The neural architecture incorporates electrochemical principles directly into parameter-free, physics-parameterized attention heads, establishing global mathematical dependencies between relaxation sequence features and the overall state of health while maintaining computational efficiency.

Validation across both public benchmarks and real-world energy storage datasets demonstrated that the framework achieved a mean absolute percentage error of 0.45 percent. The authors reported that these predictive outcomes confirm the stability and generalizability of the model across differing battery chemistries and operational environments, offering a viable method to optimize in-operando performance and battery service life.

The research paper, published in the journal Ionics, details the online estimation architecture conceived by lead investigator Weiguang Hu along with Xiaozhuang Zhang, Lv He, Kailong Chu, Junxi Yang, Zicong Chen, and Ziqi Huang.

Subject of Research: Lithium-ion battery state of health estimation using relaxation voltage and physics-informed transformers

Article Title: An online SOH estimation method using only 1-minute relaxation voltage

Article References: Hu, W., Zhang, X., He, L., Chu, K., Yang, J., Chen, Z., & Huang, Z. (2026). An online SOH estimation method using only 1-minute relaxation voltage. Ionics. https://doi.org/10.1007/s11581-026-07615-8

Image Credits: AI Generated

DOI: 10.1007/s11581-026-07615-8

Keywords: battery health, state of health, lithium-ion batteries, relaxation voltage, physics-informed transformer, convolutional neural network, battery management systems, energy storage, electrochemical principles, deep learning, computational efficiency, mean absolute percentage error, online estimation, battery degradation, transformer architecture, attention mechanism