Does a Half-Charged EV Warn You Sooner?

In the thermal channel, yes. In the gas channel, no. A lower state of charge pushes thermal runaway onset far higher — 300.8°C at 3% SOC against 203.3°C at 100% in NCA 18650 testing — which buys real thermal head-room. It also produces dramatically less vent gas. The two detection channels move in opposite directions with charge, and a threshold set on a nominal SOC is wrong at both ends.

What charge does to the thermal signature

Charge state changes almost every number that matters, and not by a little. Karmakar et al., testing 3.25 Ah cylindrical 18650 NCA cells and reporting in the Journal of The Electrochemical Society, tracked the three canonical temperatures across a charge sweep. Self-heating onset (T1) sat around 87°C at 3%, 33% and 66% SOC and dropped to 72.7°C at 100%. Runaway onset (T2) fell much harder: 300.8°C at 3% SOC, 236.1°C at 33%, 216.3°C at 66% and 203.3°C at 100%. Peak cell temperature (T3) climbed the other way — 300.8°C at 3% against 759.6°C at 100% — and total heat generated went from 8.5 kJ to 27.4 kJ, with mass loss rising from roughly 10% to about 35%.

The thermal head-start shrinks as charge rises

The useful figure is not either temperature on its own but the gap between them. T1 marks where a cell begins generating heat faster than it sheds it; T2 marks where that becomes unrecoverable. Everything a detector can do happens in between. On the JES data that window is about 213°C wide at 3% SOC, 150°C at 33%, and roughly 130°C at 66% and 100% — it collapses by a third across the first two-thirds of the charge range, then flattens.

A narrower window is not simply less time; it is less time in a regime where the cell is also heating faster, because self-heating rate rises with SOC as well. So a fully charged pack gives the shortest thermal approach to runaway and then delivers an event 2.5 times hotter at the peak and over three times larger in released energy. The detection consequence is blunt: whatever margin a thermal threshold assumes, it assumes it against the highest-SOC pack on the deck, not the average one.

The gas channel runs the other way

Everything that makes a high-SOC cell dangerous also makes it loud. Work published in Batteries on SOC and gas generation during runaway found that raising SOC lowers the critical temperature for rapid self-heating, advances the appearance of the characteristic gases, and increases measured chamber gas concentrations by roughly 2.1 to 2.8 orders of magnitude. The reaction-kinetics explanation is that highly lithiated graphite at high SOC drives stronger electrolyte reduction, which is what changes the gas-evolution pattern rather than merely scaling it.

That inverts the difficulty. The pack that leaves the least thermal margin is the one a vapour-phase channel picks up soonest and most clearly. The pack that gives generous thermal head-room is the one whose gas signature may sit near the noise floor of an electrochemical sensor that has been drifting at sea for months. Neither channel is reliably the early one — which channel leads depends on a variable nobody on the bridge can read.

The deck's charge state is undeclared

The cargo with the strictest charge control is the cargo that is not driving itself aboard. Loose lithium-ion cells and batteries shipped under UN 3480 must travel at a state of charge not exceeding 30% of rated capacity, with anything higher requiring competent-authority approval. Vehicles are governed differently: IMDG Amendment 42-24, mandatory from 1 January 2026, retired the generic UN 3171 for lithium-powered vehicles in favour of UN 3556, and Special Provision 961 still exempts vehicles carried on a designated ro-ro from the code's provisions where the battery is protected against short circuit and is not leaking.

The result is that a 60 kWh traction pack rolls aboard at whatever charge the shipper, OEM delivery process or terminal happened to leave it at, while a crate of far smaller cells beside it is capped at 30%. No manifest field carries vehicle SOC, no loading document requires it, and there is no published international rule fixing it. Detection has to assume the full range is present on every sailing.

Tuning for a charge state you cannot read

Each channel gets tuned by a different extreme. The gas channel must be set for the weakest emitter it will ever face — a low-SOC pack venting orders of magnitude less than the test cases most sensor datasheets are characterised against — which drives sensitivity and argues against thresholds derived from high-SOC laboratory work. The thermal channel must be set for the narrowest window it will ever face, which is the fully charged pack, and that lands on dwell more than on magnitude: a shorter permitted persistence before trip, not a lower delta.

Suppression and response planning take the third extreme. Sizing boundary cooling or a drencher zone on average charge state understates the job by a factor of three in released energy and by more than 450°C in peak cell temperature against a full pack. The rule mirrors the chemistry axis — tune for the weakest signal, size for the worst outcome — except that on the charge axis the weakest signal and the worst outcome sit at opposite ends of the same range, so no single reference SOC satisfies both.

• Gas channel: set sensitivity against low-SOC vent volumes, not the high-SOC cases most cell test data reports.
• Thermal channel: set dwell against the high-SOC window (~130°C between self-heating and runaway), not the ~213°C available at low charge.
• Suppression sizing: assume 100% SOC energy — 27.4 kJ per cell and a ~760°C peak in the NCA reference data.
• Fusion logic: do not assume a fixed channel ordering; which channel trips first is a function of an SOC the ship cannot see.

Sources

• 1. Karmakar, A., Zhou, H., Vishnugopi, B.S., Jeevarajan, J.A. & Mukherjee, P.P. — "State-of-Charge Implications of Thermal Runaway in Li-ion Cells and Modules", Journal of The Electrochemical Society 171(1) 010529 (2024), doi:10.1149/1945-7111/ad1ecc. 3.25 Ah 18650 NCA cells; T1 87.69 / 86.46 / 86.93 / 72.67°C, T2 300.75 / 236.14 / 216.26 / 203.29°C and T3 300.75 / 566.52 / 713.70 / 759.61°C at 3 / 33 / 66 / 100% SOC; total heat 8.5 kJ to 27.4 kJ; mass loss ~10% to ~35%.
• 2. "Impact of State of Charge on Gas Generation Characteristics During Thermal Runaway of Lithium-Ion Batteries and Early Warning Strategy Research", Batteries 12(7):241 (MDPI). Higher SOC lowers the critical temperature for rapid self-heating, advances characteristic gas appearance, and raises measured chamber gas concentrations by approximately 2.1–2.8 orders of magnitude; attributed to stronger electrolyte reduction by highly lithiated graphite.
• 3. "Study on the Multi-Stage Evolution of Thermal Runaway and the Flammability Threshold of Gas Generation in Lithium Iron Phosphate Batteries Based on SOC Gradient" — LFP-specific SOC-gradient work on vent-gas flammability thresholds.
• 4. IMDG Code — UN 3480 lithium-ion cells and batteries restricted to a state of charge not exceeding 30% of rated capacity, higher SOC subject to competent-authority approval; Amendment 42-24 mandatory 1 January 2026, retiring generic UN 3171 for lithium-powered vehicles in favour of UN 3556; Special Provision 961 exemption for vehicles carried on designated ro-ro ships.
• 5. Internal cross-references: chemistry-axis thresholds in "LFP vs NMC detection thresholds" and trip-condition mechanics in "Setting Anomaly Thresholds: ΔT and Dwell".