Early-stage anomaly discrimination in lithium-ion batteries using internal pressure-temperature coupling and...

Thermal runaway is one of the most critical safety risks in lithium-ion batteries. Mitigating this risk requires not only detecting imminent failures but also effectively discriminating early-stage chemical instabilities. With the growing complexity of battery systems, it is crucial to shift from reactive, post-event alarms to proactive anomaly discrimination strategies. To meet this need, a framework is developed around temperature-and-pressure coupling to distinguish abnormal deviations within the early-stage of battery failure, enabling rapid state differentiation single window (10 points). An empirical formula pressure-temperature curve is formulated to characterize the baseline evolution of internal pressure. The framework is evaluated on a multi-source dataset dominated by lithium iron phosphate samples (52 and 300 Ah, 30 variables), featuring wide P-T distributions with temperature differences exceeding 100 °C under identical pressure conditions. The curve fitting captures normal behaviors with high precision, achieving 99th-percentile accuracy >97% and 90th-percentile performance ≥0.99. Based on these fitted baselines, anomaly indicators are derived and sensitivity analyses are performed to evaluate the robustness of early-state separation. Across three complementary validation schemes, the temperature-and-pressure model achieved false negative rate (FNR)/false positive rate (FPR) values of 4.41% ± 1.84% and 2.79% ± 2.06% for 52 Ah (95% bootstrap confidence interval (CI): 3.36%-5.61% and 1.44%-4.11%), and 3.26% ± 1.80% and 0.59% ± 0.57% for 300 Ah (95% bootstrap CI: 2.23%-4.53% and 0.24%-0.96%). Without pressure, the FNR/FPR increased by 2.66×/10.81× for 52 Ah and 3.16×/24.86× for 300 Ah. Overall, the temperature-and-pressure model supports earlier and potentially transferable warnings and is suitable for integration into battery-management systems.