As battery energy storage systems (BESSs) become more common across portable, grid-scale, and commercial applications, the possibility of catastrophic fire incidents has become an important topic in fire protection engineering and industrial safety.
Due to the improvement of BESS design from previous incidents, the number of failures per GWh have decreased dramatically1. However, predicting and mitigating these events continues to be a complex challenge. Most currently existing technologies for BESSs protect against thermally-initiated runaways. As lithium-ion cells are heated, their internal pressure increases due to the evaporation of volatile electrolyte solvents as well as early decomposition reactions. Some cells feature an internal safety device used to relieve this pressure upon reaching a particular threshold, in a process known as cell venting. The presence of volatile gases from cell venting serves as an early indicator of impending thermal runaway, and gas-detection systems are commonly installed in BESSs to detect these gases and take appropriate preventive and mitigative actions2. These actions may include, but are not limited to, isolating affected cells, activating a ventilation system to remove flammable gases, activating alarms to alert nearby personnel, and if necessary, engaging fire suppression systems3. The types and concentrations of gases produced by cell venting are highly dependent on the cell type, state of charge, and system design, so failure prevention through these means may still be challenging. Cell-level testing under a variety of conditions can provide additional insight by revealing useful relationships between external conditions and potential markers of impending runaway. Furthermore, kinetic modeling of suitable test data can provide a means of estimating the time until peak thermal runaway as well as support the prediction of runaway propagation between cells. These additional data and analysis can help to guide more timely and informed responses to thermal excursions in BESSs, supporting continued reductions in future system failures.
1. https://storagewiki.epri.com/index.php/BESS_Failure_Incident_Database/
2. https://www.frontiersin.org/journals/physics/articles/10.3389/fphy.2025.1615872/full/
3. https://www.statx.com/whitepaper/fire-suppression-battery-energy-storage-systems/
To better understand thermal behavior leading up to runaway, FAI conducted a series of controlled thermal-abuse tests on 18650 cells using the VSP2 Adiabatic Calorimeter and a custom “test can.” The test can provided a closed volume containing the cell where the gas compositions and pressures could be controlled and later sampled during tests. Although the primary goals were to capture and quantify cell vent gases as well as assess the capabilities of new experimental equipment, the tests also revealed detailed early-stage thermal behavior. The vessel installation is depicted in Figure 1, and the full experimental arrangement is displayed in Figure 2.
These tests studied the effects of varying temperature ramp rates on the thermal response of LG INR18650 MJ1 cells, particularly in the region between cell venting and peak thermal runaway. All cells were tested at 100% state of charge (SOC) and inerted with either nitrogen or helium. The ramp rates for Tests A, B, and C were 4°C/min, 1°C/min, and ~0.27°C/min, respectively. Thermally-initiated and electrically-initiated cell runaway were previously explored and discussed in FAI’s newsletter issues Winter 2023, Vol. 30, Issue 1, Process Safety News and 2024, Vol. 31, Issue 3, Process Safety News, which provide baseline comparisons for the present work.
Comparison of Thermal Abuse Data
The use of differing ramp rates yielded notable results in three main areas: the onset of observable self-heating, the temperature at which cell venting occurred, and Arrhenius-like self-heating leading up to and following cell venting.
The temperature rise rate data leading up to and shortly after cell venting is compared for each test in Figure 3.
Based on the data, the temperature at which self-heating could be detected decreased with a lower constant ramp rate. Given the similar observed onsets in Test B and C, the sensitivity gained by lowering the ramp rate eventually diminishes as the detection threshold of the VSP2 is reached, indicating a practical limit to how much additional early-stage behavior can be revealed. In Test B and C, the observed self-heating rates increased in an Arrhenius fashion prior to cell venting. Although common battery testing standards like UL9540A prescribe a ramp rate of 4°C/min to 7°C/min, these rates may therefore conceal early self-heating behavior. This discrepancy has important implications for storage, transportation, and installation of cells. However, these findings suggest an opportunity to extract temperature-dependent kinetics to better predict behavior in real applications.
Another observation from Figure 3 is the significant difference in cell venting temperatures. In experimental data, the opening of the cell relief device is indicated by a brief but rapid increase in external pressure with a corresponding temperature drop. In the temperature rise rate data, cell venting occurs at a range of approximately 132°C to 141°C. Like the observed self-heating onset, the temperature at cell vent decreases with decreasing temperature ramp rate. It is hypothesized that the longer overall heating time with lower ramp rates allows more time for internal decomposition reactions to occur, producing more gaseous products and increasing pressure at lower temperatures. While the specific mechanism requires further confirmation, this finding has practical implications for BESSs. Cell venting can release flammable gases and vapors in a confined space, which could present a significant ignition and explosion hazard without early intervention. Given that many early warning systems for BESSs utilize gas detectors to detect an imminent thermal runaway event, even slight differences in the onset of gas release may translate to meaningful variability in the available warning time.
In all three tests, the temperature rise rates largely converged following cell venting, and the cells continued to self-heat in a somewhat Arrhenius manner up until the onset of peak thermal runaway. Temperature rise data following cell venting is plotted in Figure 4, illustrating how post-vent behavior can be compared across varying heating conditions.
Similar to the behavior seen with cell venting, the data shows that the observed transition to peak thermal runaway occurred at lower temperatures as ramp rates were decreased. Future experimental configurations and methods aimed at deriving kinetics could be explored to model the self-heating behavior in the period between cell venting and peak thermal runaway.
Conclusion
In summary, three thermal abuse tests were performed on LG INR18650 MJ1 cells using the VSP2 and a custom test can enclosure. Lowering the thermal ramp rates below those commonly specified in testing standards shifted the onsets of sustained self-heating, cell venting, and peak thermal runaway to lower temperatures. Precise tracking by the VSP2 revealed that the temperature rise rates, both before and after venting, increased in an Arrhenius fashion, suggesting the possibility for kinetic fits in the future.
These experimental findings have potential applications for fire protection engineering approaches to BESS safety, where they may help support earlier detection of thermal runaway in certain system configurations, as a complement to existing approaches. Many BESS safety systems rely on the detection of flammable gases at ell venting, but variability in cell venting temperatures creates an uncertain window for emergency response. In contrast, early self-heating data can be used to design systems that can estimate the time until cell venting, providing additional time to implement appropriate prevention and mitigation strategies, such as system shutdown, inerting, or rapid cooling. Furthermore, kinetic models following cell venting, when paired with heat transfer models can be used to predict the time to peak thermal runaway as well as potential propagation for various heating scenarios, helping to support planning of evacuation and containment procedures.
In the future, additional work should expand characterization of self-heating behavior across a broader range of ramp rates, gas headspaces, states of charge, and cell chemistries, supporting the development of more predictive and proactive safety strategies. Integrating these findings into protective system designs for BESSs may enable more fine-tuned responses to a variety of heating scenarios and contribute to reducing the likelihood of future failure incidents. In parallel, continued engagement with standards-developing organizations is important so as not to inadvertently limit the detection of early signs of thermal runaway in future designs. In current practice, fire protection engineers, manufacturers, system designers, and users should remain attentive to the full progression of cell thermal runaway rather than relying on singular events to drive safety decisions. Viewed in this context, the present results offer a complementary perspective that can help inform ongoing safety improvements.