Thermal Runaway
Thermal Runaway: The Battery Fire That Starts in One Cell and Ends in the Whole Pack
Thermal runaway is a self-accelerating failure: the cell makes heat faster than it can shed it, and each reaction makes the next one hotter, until the cell vents flammable gas and may ignite, at internal temperatures over 800 °C. One cell in Example Labs' 74.9 Wh pack holds 33,696 J, and heating it from 25 °C to 200 °C takes only 23% of that energy, so the BMS disconnect does not save the cell, it buys time for the other layers. The barrier is the layer that decides whether the pack survives: it must block 86% of the failing cell's heat to keep the neighbour below the 130 °C separator melt, and a ceramic fibre barrier slows propagation by more than 30×, turning 10–30 seconds into 5–15 minutes.
First, a simple picture
A campfire in a dry forest, with no firebreak.
One tree catches fire. The fire heats the trees next to it until they catch, and each new tree adds heat to the ones beside it, until the whole forest is burning, and the fire cannot be put out from inside the forest, because the fire is making its own heat faster than anything can take it away. Thermal runaway is the same shape inside a battery: one cell gets hot, the heat starts chemical reactions that make more heat, the reactions get faster as the cell gets hotter, and the cell vents flammable gas and may ignite, and then the heat reaches the next cell, and the next.
The fire is not one event. It is a chain, and the chain is the product's safety problem.
What thermal runaway is, and the ladder it climbs
Thermal runaway is a self-accelerating failure in which a lithium-ion cell's internal reactions generate heat faster than it can escape, raising temperature until the cell vents flammable gas and may ignite. It typically begins when heat generation exceeds heat removal, after an internal fault, an external short, overcharge, or physical damage raises the cell temperature past the onset point, and the exothermic reactions generate more heat than the cell can shed. The heat balance can be written simply as dT/dt = (Q_gen − Q_loss) ÷ (m × c_p), where Q_gen is the internal heat generation, Q_loss is the heat removed by conduction, convection, and cooling, m is the cell mass, and c_p is its specific heat capacity. Runaway starts the moment Q_gen exceeds Q_loss, and from that moment, the temperature climbs, the reactions accelerate, Q_gen grows, and the climb gets steeper. As temperature rises, resistance decreases and reaction rates increase. Higher reaction rates release more energy as heat. The loop cannot be stopped without external intervention, and by the time the loop is visible from outside the cell, the cell is already on the ladder.
The ladder is the same in every lithium-ion cell, and it has five rungs. The exact temperatures depend on chemistry, state of charge, cell design, and heating rate, the ranges below are typical values reported in published abuse testing:
The temperature ladder: each step is a chemical reaction that feeds the next (NMC; LFP holds to about 270 °C and releases no oxygen)
| Temperature | What happens |
|---|---|
| 60 to 80 °C | Ageing accelerates; BMS warnings sit below this |
| 80 to 120 °C | The SEI layer breaks down: the first heat-releasing step |
| 130 to 165 °C | The separator melts: internal short, no return |
| 150 to 220 °C | The NMC cathode releases oxygen: the fire needs no outside air |
| Over 300 °C | Full runaway; internal temperatures exceed 800 °C |
- 60–80 °C, accelerated ageing. Electrolyte reactions begin. The BMS warning thresholds are normally set below this. Nothing has failed yet, but the cell is ageing faster, and the SEI layer is drying out: sustained operation above 50 °C dries out the SEI and causes permanent loss of storage capacity.
- 80–120 °C, the SEI breaks down. The SEI (solid electrolyte interphase) is a thin passivation film that forms on the graphite anode during the first charges; it protects the electrolyte from the reactive anode. Its breakdown is usually the first exothermic step: it generates heat and flammable gases, and the exposed lithiated graphite reacts with the electrolyte, adding more heat. Early detection at this stage gives the BMS time to intervene.
- 130–165 °C, the separator melts. The polymer separator between the anode and cathode softens and melts, polyethylene near 130 °C, polypropylene near 165 °C, and the electrodes touch. Internal short circuits form, dumping the cell's stored energy as heat almost instantly. Cell voltage collapses. Crossing roughly 130–150 °C marks the no-return zone for many packs: at this point, the runaway is irreversible for the affected cell.
- 150–220 °C (NMC); 250 °C and above (LFP), the cathode decomposes and releases oxygen. This is the decisive stage, and it is where the chemistries diverge. NMC / NCA cathodes begin to decompose and release oxygen, oxygen plus heat plus flammable electrolyte makes a self-sustaining fire that needs no outside air, which is why NMC thermal runaway is so violent and so hard to extinguish. LiFePO4 (LFP) cathodes remain structurally stable to roughly 270 °C and do not release oxygen, the fire triangle is missing a leg, so the cell still gets extremely hot and vents gas, but the reaction starves instead of accelerating.
- Above 300 °C, full thermal runaway. The separator is gone, shorts multiply, electrolyte burns, and hot jets ignite accumulated gases. Internal temperatures exceed 800 °C, and the vented flammable gases, hydrogen, carbon monoxide, carbon dioxide, methane, ethylene, and vaporised carbonate solvents, plus toxic hydrogen fluoride, ignite into a jet of flame. The intense heat rapidly propagates to adjacent cells, triggering cascade failures across the pack.
LFP vs NMC: the chemistry that decides how bad it gets
The safety story of LFP was never about the electrolyte. LFP cells use the same family of organic carbonate solvents as NMC cells. It is about the cathode: no oxygen release, no self-sustaining fire.
| LiFePO4 (typical) | NMC (typical) | |
|---|---|---|
| Thermal runaway onset | ~190–270 °C | ~150–200 °C |
| Cathode oxygen release | No (stable to ~270 °C) | Yes (from ~180–220 °C) |
| Nail penetration result | Smoke and venting, rarely sustained flame | Frequently ignites |
| Peak temperature | Lower | Higher, internal temperatures exceed 800 °C |
The onset ranges are typical literature values, and exact figures vary with cell design, state of charge, and test method, treat them as comparative guidance, not guarantees for any specific product. The pipeline file's narrower ranges (LFP ~250–270 °C, NMC ~150–210 °C) sit inside these verified ranges and are consistent with them; the article uses the wider verified ranges because the lower bound is the number that matters for the safety design.
What the table means in practice: an NMC pack must be designed as though the fire can sustain itself, because it can, on the oxygen its own cathode releases. An LFP pack must still be designed as though the cell will vent violently, because it will, but the fire cannot easily sustain itself without that oxygen, so the same barrier buys more time, and the same vent path has less to carry.
Chemistry is the first safety decision, and it is made when the cell is chosen, not when the pack is built.
Five layers
Five layers of protection: each catches what the one before it missed
| Layer | What it is | What it does |
|---|---|---|
| 1. Prevent | Qualified cells, controlled charging, current protection | Stops abuse, not cell defects |
| 2. Detect | NTC on the casing, off-gas sensors | 45 °C slow, 50 °C warn, 55 °C disconnect, 60 °C shut down |
| 3. Isolate | Cell-level fusing, busbars sized for even current | Protects the other cells, not the failed one |
| 4. Limit spread | 2 to 4 mm air gap, ceramic fibre barrier | Slows spread more than 30×: 10 to 30 s becomes 5 to 15 minutes |
| 5. Control | Vent path out, fire-resistant enclosure | Contains what is left |
Safety is not one device. It is five layers, and each layer catches what the one before it missed:
- Prevent. Qualified cells, controlled charging, current protection, insulation, mechanical clearance. Prevention catches abuse, overcharge, over-discharge, external shorts, crushing, and it does not catch defects: a manufacturing defect inside a cell, or a dendrite that grows through the separator over months, starts no matter how careful the charging is.
- Detect. NTC (Negative Temperature Coefficient) thermistors, small sensors whose resistance falls as temperature rises, bonded directly to the cells, at the hottest points, with off-gas sensors that detect electrolyte vapours before smoke. One guide's example ladder: 45 °C reduce the rate, 50 °C warn, 55 °C emergency disconnect, 60 °C complete shutdown, and the verified ranges behind it: charging stops above 45 °C, discharge disconnects at 60 °C, and a rate-of-rise alarm fires on a temperature rise above 1 °C per minute (warning) and above 5 °C per minute (emergency disconnect). The BMS cannot directly observe every internal defect or local hot spot: it measures the casing and the busbar, not the inside of the cell, so there is a lag between the internal fault and the sensor's reading, and that lag is why the rate-of-rise alarm matters more than the absolute threshold. How many sensors: at least one per 10–20 cells as a minimum, bonded directly to the cells at the hottest points, with tighter spacing, every 2–4 cells, in high-reliability packs, where the sensors are glued to the cell casing and sometimes tucked between cells, because the inside of the pack can be far hotter than the air around it.
- Isolate. Cell-level fusing, a fuse per cell that blows when the short current exceeds its rating, cutting the failing cell out of the parallel string, and busbar geometry for even current sharing, so no cell takes more than its share. The honest limit: the fuse protects the other cells from the failing cell's current, not the failing cell from itself, the internal short is inside the cell, and the fuse cannot reach it.
- Limit propagation. A 2–4 mm air gap between cells, in published 18650 array tests, 2 mm and 4 mm gaps prevented propagation, and intumescent materials or ceramic blankets between groups of cells. The intumescent material absorbs heat through an endothermic chemical reaction and then expands, starting at a SET (start expansion temperature) of roughly 75–150 °C, sealing the gap the heat would cross. The published result that matters: in a Journal of the Electrochemical Society study of dense 18650 arrays, none of the tested mitigation strategies completely prevented propagation, but physical barriers slowed it, and ceramic fiber board was the most effective, slowing propagation by more than a factor of 30. Without a barrier, conductive transfer across the cell holder triggers catastrophic breakdown in the nearest neighbour within 10–30 seconds; with a ceramic fibre barrier, that becomes 5–15 minutes, which is where the pipeline file's "5–15 minutes per barrier" comes from, derived here from the two published figures, not asserted. The heat-flow split in that study explains why: in air, 50% of the heat flow to downstream cells came from flaming combustion of ejected materials, 20% from direct cell-to-cell conduction, and the rest from convective and radiative transfer, so the barrier must block the flame and the conduction, not just the conduction.
- Control consequences. Vent paths directing ejecta, the hot gas, the flame jet, the molten material, away from other cells, the electronics, and people, and fire-resistant materials in the enclosure. The vented gases include hydrogen, carbon monoxide, and hydrocarbons, which create explosion risk in an enclosed container if they are not managed, and the enclosure from Article #7 is where the vent path lives: a sealed box that keeps the rain out also keeps the gas in, unless the vent was designed as a path, not a hole.
Example Labs is a made-up company we use for worked examples.
DEMO
This pack is made up. Call it Example Labs' field gateway pack: the battery for the farmhouse gateway from Article #12, a 4S2P pack of eight 18650 NMC cells, 2.6 Ah each, 74.9 Wh, 270 kJ stored. One cell fails, an internal short, a dendrite that grew through the separator over months, the defect no prevention layer could catch.
The demo's arithmetic is the article's own, verified before writing, on the published figures cited above and on stated assumptions (a 45 g cell, a specific heat capacity of 1,000 J/kg·K); the pack and its failure are the worked example, not a measured product.
Step 1, The failing cell heats, and the BMS sees it late. The cell's heat capacity is 45 J per °C, and it holds 33,696 J of stored energy. A modest internal short of 10 W raises the cell at 13.3 °C per minute: the casing crosses 45 °C in 90 seconds, 60 °C in 158 seconds, and the separator melts at 130 °C in 7.9 minutes. A bad short of 50 W raises it at 66.7 °C per minute: 45 °C in 18 seconds, 60 °C in 32 seconds, the separator melts in 94 seconds. In both cases the rate-of-rise alarm fires first, 13.3 and 66.7 °C per minute are both above the 5 °C per minute emergency threshold, so the BMS disconnects on the rate, not on the absolute temperature, and it disconnects while the cell is still below the separator melt. That is the detection layer working as designed.
Step 2, The disconnect does not save the cell. The BMS cuts the power, and the charger stops feeding the fault, but the internal short is inside the cell, and the cell still holds its own energy. Heating the cell from 25 °C to 200 °C, past the NMC cathode's oxygen release, takes 7,875 J, only 23% of the cell's stored energy. The cell has five times what it needs, even after the BMS cuts the power. The disconnect buys time for the other layers. It does not cure an internal short. The cell is lost in every version of this demo, and the demo is honest about that, because a safety design that promises to save the failing cell is a design that has not understood the failure.
Step 3, The neighbour, and the barrier's number. The failing cell releases its 33,696 J. The neighbour must stay below 130 °C, the separator melt, or at least below 150 °C, the NMC onset. If the barrier lets through 10% of the failing cell's energy, the neighbour rises 75 °C, to 100 °C, safe. If it lets through 25%, the neighbour rises 187 °C, to 212 °C, the neighbour runs away. If it lets through 50%, the neighbour rises 374 °C, to 399 °C. To keep the neighbour below 130 °C, the barrier must let through less than 14% of the failing cell's energy, it must block 86% or more of the heat. That is the barrier's number, and it is the number the whole pack is designed around: not "a barrier," but a barrier that blocks 86% of the heat from a cell at over 800 °C, while the neighbour sits millimetres away.
Step 4, The timing, with and without the barrier. Without a barrier, the nearest neighbour breaks down in 10–30 seconds. With a ceramic fibre barrier, propagation slows by more than 30×, so 10–30 seconds becomes 300–900 seconds, 5–15 minutes. Five to fifteen minutes is time to act, or time to walk away. Ten to thirty seconds is neither. And the isolation layer matters here too: the cell-level fuse has already cut the failing cell out of the parallel string, so the other cells are not feeding current into the fault, the only heat the neighbour receives is the heat the barrier lets through.
Step 5, The vent path, and what happens if it is missing. The failing cell vents, hydrogen, carbon monoxide, hydrocarbons, hydrogen fluoride, and the vent path directs the ejecta out of the enclosure, away from the other cells, the electronics, and the person standing next to the gateway. If the vent path is blocked, or if it points at the electronics, the gases ignite inside the enclosure, and the enclosure from Article #7, the sealed box that keeps the rain out, becomes the container that holds the explosion. The pack loses one cell, the gateway stops, and the product survives. That is the best outcome available once a cell has an internal short, and it is the outcome the five layers were built to deliver.
The decision test
Four questions. They sort the packs that are safe to ship from the packs that are not:
- Which chemistry, and was it chosen for safety, or for energy density? NMC: the fire can sustain itself on the cathode's own oxygen, so the barriers and the vent path carry the safety case. LFP: the cathode releases no oxygen, so the same barrier buys more time, but the cell still vents violently, and the pack still needs all five layers. If the chemistry was chosen for energy density alone, the safety design starts behind.
- Does the BMS see the failure early enough, and does it act on the rate, not just the temperature? One sensor per 10–20 cells at the hottest points, bonded directly to the cells, is the minimum; every 2–4 cells is the high-reliability spacing. The rate-of-rise alarm, above 1 °C per minute to warn, above 5 °C per minute to disconnect, is what catches the fast failure, because the absolute threshold is always late for a cell that heats at 66.7 °C per minute.
- Is there a barrier between the cells, and does it block 86% of the failing cell's heat? A 2–4 mm air gap, a ceramic fibre barrier, an intumescent layer, and the number that proves it: the barrier must let through less than 14% of a failing cell's stored energy to keep the neighbour below the 130 °C separator melt. If the barrier was chosen by thickness alone, without the heat-transfer number, the pack has a barrier and no safety case.
- Does the vent path point out, and has the propagation been tested, not assumed? The vent directs the ejecta away from the other cells, the electronics, and people, and the pack's propagation behaviour has been measured at cell, module, unit, and installation level under UL 9540A, because the test data is what the authority having jurisdiction reads to decide the spacing. If the propagation was assumed, the pack has a vent and no evidence.
If the answers are yes, yes, yes, yes, the pack is safe to ship, and the review is a final check, not a rescue. If any answer is no, the job is a battery-safety review, and the honest thing is to do it before the pack ships, not after the first customer asks what happens if one cell fails. And if the answer to question 3 is "we have a barrier, but we don't know what it blocks", the problem is not the barrier.
The problem is that the safety case was never written down, and the review starts by writing it.
"But LFP is safe, so we don't need the layers", and two more fair doubts
"But LFP is safe, so we don't need the layers." LFP is safer, and safer is not safe. LFP cells can enter thermal runaway: the SEI still breaks down at 80–120 °C, the separator still melts at 130–165 °C, and the electrolyte still vents violently. What LFP removes is the cathode oxygen, the leg of the fire triangle that makes the fire self-sustaining, so the same barrier buys more time and the same vent path has less to carry. The layers still matter, because the cell still gets extremely hot, still vents toxic gas, and still heats its neighbours. Choosing LFP is the first safety decision. It is not the only one.
"But our BMS disconnects at 60 °C, so the pack can't run away." The BMS disconnects the external current, the charger, the load, and the internal short is inside the cell. Heating the cell from 25 °C to 200 °C takes 23% of the cell's stored energy, and the cell has five times what it needs even after the disconnect. The disconnect buys time for the other layers. It does not cure the fault. A safety case that ends at the BMS is a safety case that has not followed the failure past the disconnect.
"But we already passed UN38.3, so the pack is certified safe." UN38.3 is a transport test, it proves the cell or battery can be shipped safely by air, sea, or land, under altitude, thermal cycling, vibration, shock, external short, impact, overcharge, and forced discharge. It is not a product-safety approval, and a battery that has passed transportation testing is not automatically approved for stationary installation. The pack still needs UL 1973 for the battery system, UL 9540 for the complete system, UL 9540A test data for the propagation behaviour, and NFPA 855 for the installation, and the chain does not shortcut: no UL 1973 on the battery, no UL 9540 on the system, no NFPA 855 compliance, no approval from the authority having jurisdiction, no installation.
Sources
- Electrical Equipment Factory, "Thermal Runaway Battery Guide: Causes, Temps & Protection"
What we used
Thermal runaway is a self-accelerating failure in which a cell's internal reactions generate heat faster than it can escape; begins with SEI breakdown near 80–120 °C and accelerates once the cathode decomposes, at roughly 150–220 °C for nickel-rich NMC and above about 250 °C for LFP; heat balance dT/dt = (Q_gen − Q_loss) / (m × c_p); temperature ladder: 60–80 °C accelerated ageing (BMS warnings below this), 80–120 °C SEI decomposes, 120–135 °C polyethylene separator softens and melts, ~165 °C polypropylene melts, 150–220 °C (NMC) / 250 °C+ (LFP) cathode decomposes and releases oxygen, venting releases flammable gas, peak several hundred °C with NMC generally higher than LFP; vent gas composition hydrogen, carbon monoxide, carbon dioxide, methane, ethylene, vaporised carbonate solvents; UL 9540A tests propagation at cell, module, unit, and installation level, and its results feed the spacing and protection decisions required by NFPA 855
- Autoven, "Understanding Thermal Runaway in EV Batteries"
What we used
Thermal runaway is a self-reinforcing cycle: the cell generates more heat than its thermal management can dissipate, excess heat accelerates reactions, which generate more heat; four progressive stages, with early detection at stages 1–2 giving the BMS time to intervene; hydrogen and other off-gases begin accumulating inside the cell and pack, detectable by gas sensors before any external signs appear; the separator melts, the safety vent opens, and at that point the runaway is irreversible for the affected cell; temperatures exceed 800 °C; thermal runaway during charging alone accounts for one-third of total EV accidents
- Manly Battery, "Battery Thermal Runaway 101: Temperature Milestones, Risks, and Controls"
What we used
The temperature ladder from SEI breakdown (~90–120 °C) to separator softening and melt (~130–180 °C), cathode oxygen release and solvent oxidation (>150 °C), binder breakdown with sharp pressure rise (>200 °C), then flaming combustion above ~300 °C; polyethylene softens near ~135 °C, polypropylene near ~166 °C; layered NMC starts releasing lattice oxygen roughly 180–220 °C, LFP holds longer (>260 °C); combined reactions can exceed ~600 J/g; peak skin temperatures exceed ~600 °C
- LJY Energy, "LiFePO4 Thermal Runaway Explained: The Four Stages"
What we used
Thermal runaway is a self-accelerating chain reaction; four stages: SEI breakdown (~80–120 °C), separator meltdown (~130–165 °C), cathode decomposition (NMC/NCA release oxygen from ~200 °C; LFP remains stable to roughly 270 °C and does not release oxygen, the fire triangle is missing a leg), then electrolyte ignition; onset table: LFP ~190–270 °C, NMC ~150–200 °C; nail penetration: LFP smoke and venting, rarely sustained flame; NMC frequently ignites; UL 9540A measures propagation from cell → module → unit → installation, with gas analysis at each level
- IOPscience (Journal of the Electrochemical Society, 2020), "Passive Mitigation of Thermal Runaway Propagation in Dense 18650 Lithium Ion Cell Assemblies"
What we used
Dynamics, heating rates, gaseous emissions, and energetics of thermally-induced propagation in dense arrays of 12–15 fully charged 18650 cells; 5 mm gaps between select rows and columns, and physical barriers (double-layer stainless steel, intumescent material, ceramic fiber board); none of the tested mitigation strategies completely prevented propagation, but the barriers slowed propagation speed, and ceramic fiber board was the most effective, slowing propagation by more than a factor of 30; in air, 50% of heat flow to downstream cells was associated with flaming combustion of ejected materials, 20% with direct cell-to-cell conduction; Lopez et al. and Zhong et al. prevented failure propagation in cylindrical cell arrays with 2 and 4 mm inter-cell gaps respectively
- Sentiention, "Thermal Runaway Propagation Mitigation Strategies in Off the Shelf Packs"
What we used
An initiating NMC811 21700 cell at full charge releases between 650 and 850 °C of localised thermal energy in under three seconds (accelerating rate calorimetry, 2023, fifty cells, adiabatic); without sufficient interstitial resistance, conductive transfer across the cell holder triggers catastrophic breakdown in the nearest neighbour within ten to thirty seconds; aerogel blankets compressed to 1.2 mm sustain thermal conductivity below 0.04 W/m·K during a 750 °C trigger event; inter-cell spacing below 1.0 mm fails because convective heat transfer and radiative coupling overwhelm the thin air pocket within two seconds of cap breach; prismatic cells transfer significant heat across wide faces with aluminium walls 0.6–1.0 mm thick
- Justia Patents (US 7,781,097 and US 20100075213), "Cell thermal runaway propagation resistance using an internal layer of intumescent material"
What we used
Intumescent material absorbs excess heat via an endothermic chemical reaction; once the activation (SET) temperature is reached, 75–150 °C, or 100–200 °C, or 200–300 °C depending on the material, the layer begins to expand; the layer delays the onset of thermal runaway and provides significant resistance to propagation
- AmpLogic, "Thermal Alerts: Battery Overheating Warnings & Safety"
What we used
Thermal alerts trigger when internal thermistors detect cell temperatures exceeding safe thresholds, typically 45 °C during charging or 55 °C during discharging; the BMS derates operating current or trips internal DC contactors; NTC thermistors attached directly to cell busbars and internal module walls report temperature data to the central BMS board; at 41 °C the BMS tapers current; past 50 °C during charging or 60 °C during discharging, the BMS opens its primary high-voltage contactors; sustained operation above 50 °C dries out SEI layers and causes permanent capacity loss
- Zbotic, "Battery Temperature Monitor: Thermal Runaway Prevention"
What we used
35–45 °C accelerated ageing (lifespan halves for every 10 °C above 25 °C); 45–60 °C critical, reduce charge rate and alert; above 60 °C danger, disconnect immediately; charge limits: below 0 °C stop charging (lithium plating), 0–10 °C at 0.2C maximum, 10–45 °C normal, above 45 °C stop charging; discharge limits: below −20 °C stop, −20–0 °C limit to 0.5C, 0–60 °C normal, above 60 °C disconnect; rate-of-rise alarm: above 1 °C/minute warning, above 5 °C/minute emergency disconnect; NTC thermistors are the industry standard for battery monitoring
- Tefoo Energy, "Engineering Standards for Lithium-Ion Battery Protection"
What we used
NTC thermistor arrays inhibit charging outside 0–45 °C and enforce hard discharge shutdowns between 60 °C and 90 °C; protection ICs isolate charging at 4.25–4.40 V per cell and disconnect loads at 2.30–2.75 V per cell; cell balancing voltage delta 10–30 mV
- Zero Motorcycles Owner's Manual (2014 SR)
What we used
The BMS will not allow discharge above 60 °C; above 43 °C the charger reduces charge current; above 55 °C the BMS will no longer allow charging
- PowerPaul, "BMS Fundamentals Guide for Lithium Batteries" (Daly BMS)
What we used
If battery temperature rises above 60 °C, the BMS stops both charge and discharge; below 0 °C it stops charging; below −20 °C it stops discharge; SOC is a calculation by coulomb counting
- Gladway Battery, "How does the temperature control system in home energy storage lithium batteries work?"
What we used
NTC thermistors glued to every 2–4 cells in the pack, right to the cell casing and sometimes tucked between cells, because the inside of the pack can be far hotter or colder than the air around the unit; the BMS checks temperature every second against the safe range (usually 0–45 °C for charging, −20–60 °C for discharging)
- LJY Energy, "LiFePO4 Battery Safety Certifications: What Buyers Should Ask For"
What we used
UL 9540A is a test method, not a certification: no UL 9540A certificate, no UL Mark, no pass/fail verdict; it measures how thermal runaway propagates from a single cell, to a module, to a full unit, to a multi-unit installation, and characterises the gases released and heat produced; engineers, fire departments, and Authorities Having Jurisdiction use that data to decide installation requirements such as spacing under NFPA 855; standards map: UN38.3 transport test, UL 1973 stationary batteries, UL 9540A thermal runaway fire propagation data at four levels, UL 2580 EV batteries, IEC 62619 industrial lithium cells/batteries
- CMX Battery, "Battery Certifications We Hold"
What we used
UN38.3 is a transport test, not a product-safety approval: altitude, thermal cycling, vibration, shock, external short, impact/crush, overcharge, forced discharge; IEC 62619 covers external short circuit, thermal abuse, overcharge, drop, and (in later revisions) thermal propagation; UL 1973 covers the battery system for stationary, motive auxiliary, and light electric rail use; UL 9540A does not issue a "pass", it produces a report the AHJ reads to decide clearances, sprinkler demand, and unit spacing
- Polinovel BESS, "BESS Safety Standards: UL, IEC, UN 38.3"
What we used
UL 9540A sixth edition published March 13, 2026, effective January 1, 2027; a battery that has passed transportation testing is not automatically approved for stationary installation; UN38.3, IEC 62619, UL 1973, UL 9540, UL 9540A, and NFPA 855 are complementary documents, not a simple ranking
- Avantis Energy, "Stationary Battery Standards: IEC 62619, UL 1973, UL 9540 & UL 9540A"
What we used
IEC 62619 covers safety of lithium cells and batteries for industrial use; UL 1973 addresses battery systems for stationary applications; UL 9540 certifies complete energy storage systems; UL 9540A is a test method for fire propagation during thermal runaway, not a certification mark; the test is conducted at four levels, cell, module, unit, installation, and the results inform fire-separation and suppression requirements