Accurate and interpretable state of health estimation of lithium-ion batteries based on multimodal laser-induced plasma spectroscopic sensing

Abstract

Lithium-ion batteries are a core component of next-generation power systems and electric vehicles. As the key parameter of the battery management system (BMS), state of health (SOH) significantly affects battery operational performance, making its accurate estimation crucial. Existing methods struggle to simultaneously balance efficiency, generalization, and interpretability. A significant quantitative correlation exists between the lithium content in the anode and the SOH. By capturing laser spectroscopic signals and monitoring the lithium content, accurate estimation and reliable prediction of SOH can be achieved. In this study, a novel multimodal laser spectroscopy sensing method is proposed, which enables the verification of battery aging mechanisms, accurate estimation of SOH, and reliable prediction of service life. Specifically, the content of the lithium element is accurately monitored by collecting spectral signals from the battery anode, while ultrasonic and image information supplement additional detection insights into batteries like density and hardness. This integration significantly enhances the accuracy and precision of battery state estimation. A dedicated multimodal laser spectroscopic sensing network (MLSS-Net) is developed to fuse these multimodal features, enabling precise SOH estimation and reliable prediction of battery remaining lifespan. Validation experiments demonstrate the efficacy of the proposed method. This work offers a dependable technical solution for Li-ion battery SOH evaluation.