Lithium-Ion Battery State Estimation Based on Anode Strain Field Reconstitution Utilizing Optical Frequency Domain Reflectometry

Abstract

The state of charge (SOC) and state of health (SOH) in battery systems are crucial indicators for evaluating battery performance, playing a vital role in ensuring the normal operation of battery systems. In this study, a phase-sensitive optical frequency domain reflectometer was employed for real-time monitoring of strain fields in lithium battery anodes. Distributed strain and strain rate data were used as inputs to a feedforward neural network for predicting battery SOC. The results showed that the predictive accuracy of distributed strain data (98.3%) significantly outperformed single-point predictions (88.8%), demonstrating comparable accuracy (98.5%) to predictions based on electrical parameters (current, voltage). Additionally, features such as maximum strain in a single cycle and cumulative residual strain during cycling were utilized. A long short-term memory recurrent neural network was employed to predict battery SOH, achieving a prediction accuracy of 96.3%. The use of purely strain data enabled high-precision prediction of SOC and SOH without requiring any electrical information during battery operation. Moreover, the principle of distributed measurement allows simultaneous measurement of individual or multiple battery packs, thereby offering robust support for future battery system management.