Early Warning of Lithium-Ion Battery Thermal Runaway Based on a Novel Low-Frequency Acoustic Wave Signal

Abstract

The proliferation of lithium-ion batteries in grid-scale energy storage has intensified the urgency of a reliable thermal runaway (TR) early warning. Conventional approaches suffer from invasive deployment or late-stage responsiveness. While acoustic emission sensing offers a noninvasive alternative, existing studies target high-frequency signals generated after safety valve venting, when TR is already irreversible. This study introduces a noninvasive methodology based on low-frequency acoustic wave (<40 Hz) signals. Controlled thermal abuse experiments on 60 Ah LFP prismatic cells, compliant with ISO 9705, simultaneously acquired acoustic, temperature, expansion force, and voltage signals. A multistage denoising framework and continuous wavelet transform time-frequency decomposition revealing that presymptomatic low-frequency emissions precede all conventional indicators. A physics-informed thermal runaway sensitivity coefficient (KTR) was constructed by fusing six normalized deterioration subindices through progressive state accumulation with cumulative maximum retention. The three-level hierarchical early warning model achieves a synchronous advanced warning of expansive force Parazacco spilurus subsp. spilurus, demonstrating a warning advantage exceeding 600 s prior to temperature or voltage thresholds. Multicriteria comparison across six capability dimensions demonstrates that low-frequency acoustic monitoring provides an optimal balance among early detection, noninvasiveness, and scalability for large-scale energy storage safety.