Datacenter
Lithium-ion battery thermal runaway and fire safety field guide for data centers
Why a lithium-ion cell fire feeds itself, how the off-gas can explode before it burns, and how early detection, separation, suppression, and explosion control keep one bad cell from taking the room.
Direct answer
Thermal runaway is a self-feeding reaction where a lithium-ion cell makes more heat than it sheds, vents flammable gas, ignites, and cascades cell to cell. It reignites and shrugs off a little water, so the design relies on early off-gas detection, separation, explosion control, and water, not a clean agent alone. NFPA 855, UL 9540A, and the AHJ control.
Key takeaways
- Thermal runaway is a self-feeding reaction: a cell makes more heat than it sheds, vents flammable gas, ignites, and cascades cell to cell.
- A clean agent alone does not stop runaway; water removes the heat the cell makes itself, applied long enough to cool the affected mass.
- Off-gas detection gives the only useful margin, roughly 5 to 20 minutes between first vent gas and full runaway, to isolate, cut charge, and exhaust.
- LFP onset runs about 220 to 260 C versus NMC around 170 to 210 C; NMC burns hotter and ejects burning material, but neither chemistry is safe.
- NFPA 855 frames spacing, detection, suppression, and explosion control; UL 9540A tests fire propagation; confirm numbers against the adopted edition and the AHJ.
Thermal runaway, and why lithium-ion is a different fire problem
Thermal runaway is a self-feeding chain reaction inside a lithium-ion cell. The cell generates more heat than it can shed, the temperature climbs, the internal chemistry starts to break down, the cell vents a cloud of flammable gas, that gas ignites, and the heat shoves the next cell over the same edge. One cell becomes a module. A module becomes a rack. That cascade is the whole reason a lithium-ion battery room is built, detected, and protected the way it is.
This is not an ordinary fire, and treating it like one is how rooms get lost. An ordinary fire goes out when you cool it or smother the flame. A cell in thermal runaway carries its own fuel and, in some chemistries, much of its own oxidizer, so smothering the flame does not stop the heat. It reignites. A little water knocks the flame down and the cells keep cooking underneath. The energy is already stored in the cell, and the reaction will run until that energy is spent or carried away.
So the design problem is not the same as a sprinkler over a server. It is early detection of the vent gas before there is a flame, gas management so that gas cannot collect and explode, and containment so one unit cannot light the one next to it. The battery and energy storage types guide covers how lithium-ion compares to VRLA and flywheel storage. This guide is the failure mode: what runaway is, how it starts, and what the room has to do about it.
What is thermal runaway?
Thermal runaway is the point where a cell produces heat faster than it can lose it, so the temperature feeds itself upward with no outside help. Below that point a warm cell sheds heat to its surroundings and settles. Above it, the heat drives reactions that release more heat, which drives the reactions harder, which releases more heat. There is no equilibrium to fall back to. The cell goes until it is finished.
The sequence inside the cell is consistent across chemistries even when the violence is not. Heat first breaks down the thin layer on the anode, then the separator that keeps the electrodes apart softens and shrinks, which lets the electrodes touch and short internally. The internal short dumps the cell's stored energy as heat in seconds. The electrolyte boils and the cell vents, throwing out hot flammable gas and, in the harsher chemistries, burning particles and molten material.
The part that makes it a room problem rather than a cell problem is propagation. A single cell in runaway radiates and conducts heat into its neighbors, and if the pack was not designed and tested to stop that heat, the neighbors reach their own onset temperature and go too. The fire walks through the module cell by cell, then jumps module to module and rack to rack. Stopping that walk, by design and by tested spacing, is the job. Once a few cells are gone, you are managing the event, not preventing it.
What triggers thermal runaway
Runaway has a short list of triggers, and they sort into one idea: anything that drives a cell hotter than it can shed, or shorts it internally, can start the cascade. They are worth knowing by name because each one points at a different defense.
An internal short is the trigger you cannot see coming. A manufacturing defect, a metal particle left in the cell, or a lithium dendrite grown across the separator over many cycles can bridge the electrodes inside a sealed cell. There is no warning at the terminals until it lets go. This is the failure that quality control at the cell maker and conservative charging are meant to hold down, and it is why a cell that has been abused or aged hard is more suspect.
Overcharge pushes voltage and current past what the chemistry tolerates, plating lithium and heating the cell until it vents. Over-temperature does the same from the outside: a failed cooling loop, a hot room, a cell near a heat source, or a neighbor already in runaway. Physical damage, a crush, a puncture, a dropped module, or a connection arcing under a loose lug, opens an internal short directly. And a manufacturing defect can sit dormant for months before it surfaces as any of the above. The battery management system and the charging design are aimed at overcharge and over-temperature. Detection, separation, and handling discipline are aimed at the rest.
Can the vented gas explode?
Yes, and this is the hazard people miss because they are watching for flame. Before and during runaway a cell vents a mixture of flammable gases, hydrogen, carbon monoxide, methane, and electrolyte vapor among them. If that gas collects in an enclosure or a cabinet and finds an ignition source, it does not just burn. It deflagrates, a fast-moving flame front that builds pressure and can blow doors, walls, and roofs off a battery enclosure.
Real incidents have shown the order matters. Cells can off-gas for minutes before there is open fire, filling a sealed cabinet or room with fuel. If suppression or a stray spark ignites that accumulated cloud all at once, the explosion can do more structural damage than the fire would have, and it can injure anyone standing at the door. That is why the code treats explosion control as its own line item, separate from fire suppression, handled through deflagration venting or active explosion prevention.
The lesson for the field is blunt. A clean-agent dump or a suppression system that knocks the flame but lets the room keep filling with off-gas has not made the space safe. It may have set up the deflagration. Gas has to be cleared or vented, not just kept from burning, and the detection that catches the off-gas early is what gives you the chance to do it before the cloud builds.
LFP and NMC: the chemistry changes the risk
The cathode chemistry sets how soon a cell lets go and how hard it hits, so the two chemistries common in this service deserve to be told apart. Lithium iron phosphate, LFP, is the more forgiving one. Nickel manganese cobalt, NMC, packs more energy into the same space and pays for it with a lower onset and a more violent event.
The numbers vary by cell and test, so treat them as approximate. LFP cells commonly show a thermal-runaway onset in the rough range of 220 to 260 degrees C, while NMC tends to start lower, around 170 to 210 degrees C, which is a meaningfully wider margin before LFP reaches failure. Once they go, NMC burns hotter and releases more energy, with peak cell-face temperatures often reported around 1,000 degrees C or higher, far above the roughly 600 degrees C an LFP cell reaches, and NMC carries more of its own oxygen into the reaction. NMC also tends to eject burning gas, liquid, and solid material in a short violent burst, while LFP more often produces mostly hot smoke and gas. Confirm the actual figures against the cell maker's data and the UL 9540A report for the product.
This is why new data center builds lean toward LFP and why the selection sits in the battery and energy storage types guide, not here. Either chemistry can run away. LFP gives you more onset margin and a less violent event, but it still vents flammable gas, it still cascades if the pack allows it, and it still demands detection, separation, and explosion control. Safer is not safe.
| Characteristic | LFP (LiFePO4) | NMC |
|---|---|---|
| Energy density | Lower | Higher |
| Typical onset temperature (approx.) | ~220 to 260 C | ~170 to 210 C |
| Event severity | Less violent, mostly gas and smoke | More violent, ejects burning material |
| Peak cell-face temperature (reported) | ~620 C | ~800 C |
| Off-gas and energy release | Lower | Higher, carries more oxygen |
| Common data center trend | Favored for new builds | Higher energy where space is tight |
How do you detect thermal runaway early?
The earliest reliable signal is the off-gas the cell vents before it catches fire, and detecting that gas is the most useful thing the room can do. A cell heading toward runaway releases electrolyte vapor and gases minutes before there is a flame. Testing by fire-protection researchers has put that window in the rough range of 5 to 20 minutes between first off-gas and full thermal runaway. That is the time you get to act, and it is the only generous margin in the whole event.
Off-gas detection sniffs for those early vent gases at the cabinet or room level and alarms while you can still do something useful: shut down and isolate the affected unit, cut its charge, start exhaust, and clear people. Smoke and heat detection, by contrast, only triggers once there is already a fire, which means the cascade is usually underway. Off-gas detection is the difference between catching a single cell and fighting a rack. FM Approvals listed the first lithium-ion off-gas detection system in 2025, a sign the technology has matured from research into something an AHJ will recognize.
Cell-level temperature and voltage monitoring through the battery management system is the other early signal, watching for the rapid temperature rise or voltage collapse of a failing cell. The two layers complement each other: the BMS sees the electrical signature inside the pack, off-gas detection sees the chemical signature in the air. Confirm which detection the listing and the hazard analysis require, and verify both actually alarm and act at commissioning, because a detector that no one wired to a trip is a light on a panel, not a defense.
The battery management system as first line of defense
The battery management system, the BMS, is the electronics that watches every cell or group and keeps the pack inside its safe window. It is the first line of defense against the triggers it can see, which are overcharge and over-temperature, and a healthy BMS stops most events before any of the later layers ever get tested.
What it does is concrete. It balances cells so one does not get overcharged while the string looks fine at the terminals. It monitors voltage and temperature per cell or per group and trips on over-voltage, under-voltage, over-temperature, or over-current. It opens contactors to isolate a pack that has crossed a limit, and it talks to the charging system so the charge backs off before a cell is pushed past its edge. UL 1973, the safety standard for stationary battery systems, evaluates the BMS and its protection functions as part of listing the pack.
The BMS does not stop an internal short from a dendrite or a defect, because that failure happens inside a cell faster than any controller can open a contactor. That is the gap the other layers cover: detection catches the off-gas, separation stops the spread, suppression and explosion control handle what gets out. Treat the BMS as the layer that prevents the preventable, and never as the only layer. At commissioning, prove its trips fire. A protection setpoint nobody verified is an assumption, and assumptions are what burn.
What is NFPA 855?
NFPA 855, the Standard for the Installation of Stationary Energy Storage Systems, is the document the fire code points to for how a battery energy storage system is sized, spaced, detected, suppressed, and ventilated. If you are installing lithium-ion storage in a data center, this is the standard that frames the fire-safety design, alongside the International Fire Code that adopts much of it.
It covers the items this guide walks through: separation between units and from walls, limits on stored energy per fire area, required detection, suppression, explosion control, and a hazard mitigation analysis that ties the package together for the specific installation. The 2026 edition tightened several of these, moving toward a near-universal hazard mitigation analysis and leaning harder on large-scale fire testing rather than letting small installations skip the rigorous analysis through capacity exemptions.
Do not memorize the spacing, the energy limits, or the section numbers and carry them job to job, because they shift between editions and the jurisdiction amends them. The standard names the framework. The adopted edition, the local amendments, and the AHJ name the actual numbers for your project. Cite NFPA 855 because it governs the point, then go confirm the specific requirement against the edition the jurisdiction has adopted and whatever the AHJ has decided on top of it.
What is UL 9540A?
UL 9540A is the test method for evaluating thermal runaway fire propagation in battery energy storage systems, and it is the data the AHJ leans on to decide whether an installation is safe and how it has to be protected. It is a test method, not a pass-fail listing. It produces the evidence: how a given product behaves when a cell is forced into runaway, whether the fire propagates, what gas it makes, and how hot it gets.
The method runs in levels, from a single cell up to the full installation. Cell level checks whether one cell's runaway spreads. Module level looks at propagation across a full module. Unit level evaluates the complete enclosure, often with suppression disabled, to see the worst case. Installation level runs the real-world arrangement with the protection in place. Each level that passes without propagation can let the designer relax a requirement at the next level up, which is exactly how the tested data feeds the spacing and protection decisions.
For large lithium-ion installations, UL 9540A data is effectively the price of admission with the AHJ, and the 2026 direction pairs it with large-scale fire testing for the harder cases. The current edition of the test method was revised in 2026 to align its large-scale method with the NFPA 855 guidance. Ask for the UL 9540A report on the specific product, read what it actually demonstrated, and confirm with the AHJ which test level and what additional testing they require. The report on a similar product is not the report on yours.
Separation and spacing to stop the cascade
Separation is the physical defense against propagation: keep enough distance or a rated barrier between units that one in runaway cannot light the next. It is the layer that turns a single-unit event into a single-unit event instead of a room loss, and it is where NFPA 855 and the tested data meet.
A commonly cited starting point is a minimum separation on the order of 3 ft between battery units or racks, and between racks and walls, for lithium-ion arrays above a threshold stored energy. Treat that as a default, not a law. The standard allows that distance to be reduced where UL 9540A or large-scale fire testing shows runaway will not spread between units at a closer spacing, and it can be replaced by a rated fire barrier where the layout demands it. The point is not the number. The point is proven non-propagation, whether by distance, by barrier, or by tested design.
Where this goes wrong in the field is units packed tighter than the tested arrangement to save floor space, or a fire barrier specified on the drawing and value-engineered out before install. Both defeat the one layer that limits the size of the event. Confirm the separation against the project's UL 9540A basis and the adopted NFPA 855 edition, and confirm with the AHJ before anyone compresses the layout. The spacing you can prove is the spacing you have.
Can you put out a lithium battery fire with water or a clean agent?
Water is the workhorse for a lithium-ion fire, because it does the one thing that matters: it carries heat away. A clean agent or an inert gas is not, on its own, enough, because it knocks down the visible flame without cooling the cells, and the cells reignite. This is the single most important suppression fact to get straight, and the one most often gotten wrong.
Thermal runaway is driven by heat the cell makes itself. Stop the flame and the heat keeps building, so the moment the agent clears, the cell reignites and the next one goes. Clean agents, carbon dioxide, and inert gases remove flame from the fire triangle but leave the heat, which is exactly the part of a runaway you have to remove. Water removes heat, and it takes a lot of it, applied long enough to pull the temperature down through the whole affected mass and hold it there. A short knockdown is not enough. The flame coming back is the tell.
There are real complications, which is why this hedges to the listing and the AHJ. The unit may still be energized, so flooding it with a conductive agent before it is isolated puts responders at shock risk. Stranded energy inside damaged cells can reignite the pack hours later even after the flame is out. The current direction in this field combines fast cooling with explosion control rather than betting on any single agent, and dual-agent and immersion approaches are being tested. Whatever the product, confirm the suppression strategy against the listing, the UL 9540A data, and the AHJ. The one thing that is settled: a clean agent alone does not stop runaway.
Ventilation and explosion control for the off-gas
The flammable off-gas is its own hazard, separate from the fire, and the code handles it with two paths: prevent the explosion or vent it. NFPA 855 requires explosion control, and it points to NFPA 69 for explosion prevention and NFPA 68 for deflagration venting. They solve the same problem from opposite ends.
NFPA 69 is the prevention side. Gas detection drives mechanical ventilation that exhausts the off-gas and keeps the concentration in the enclosure below a fraction of the lower flammable limit, commonly cited around 25 percent of the LFL, so a flammable cloud never forms in the first place. NFPA 68 is the relief side. If a deflagration happens anyway, sized vents, panels, doors, or roof elements that open at a set pressure, let the pressure out along a planned path instead of through the walls. Many designs use both, prevention as the first defense and venting as the backstop for the case prevention misses.
This is where the lithium-ion room and the lead-acid room part ways, and the contrast is worth holding onto. A VRLA battery room handles steady hydrogen off-gassing with continuous ventilation sized to dilute a slow, predictable release, which is the subject of the battery room ventilation and hydrogen safety guide. A lithium-ion room is managing a sudden burst of mixed flammable gas during an event, which is why it adds deflagration venting and detection-driven exhaust on top of any normal ventilation. Same idea, flammable gas, different release, different design. Verify the explosion-control basis against the adopted code and the AHJ.
VRLA hydrogen versus lithium-ion runaway: different hazards
VRLA and lithium-ion are both battery fire-safety problems, but they fail in different ways and the designs do not transfer. Treating a lithium-ion room like a lead-acid room, or the reverse, is a common and dangerous shortcut.
A flooded or VRLA lead-acid string makes hydrogen on charge, a steady, predictable release. The hazard is accumulation in a confined space, and the answer is continuous mechanical ventilation that holds the room below the flammable limit, backed by hydrogen detection. There is also the acid and the DC shock hazard, but the explosive risk is the slow build of hydrogen. That whole problem is covered in the battery room ventilation and hydrogen safety guide.
A lithium-ion cell does not sit there making a little gas. It is stable until it is not, and then it dumps heat and a burst of flammable gas in minutes and tries to take its neighbors with it. The answer is early off-gas detection, separation to stop the cascade, water to cool, and explosion control for the off-gas burst. Ventilation alone, which is most of the lead-acid answer, is only one piece of the lithium-ion answer. If you walk into a lithium-ion room expecting the hydrogen-room playbook, you will have ventilation and no detection, no separation basis, and no plan for the burst. Different hazard. Different defense.
| Aspect | VRLA / flooded lead-acid | Lithium-ion |
|---|---|---|
| Primary hazard | Hydrogen off-gassing on charge | Thermal runaway, cell to cell |
| Release pattern | Slow, steady, predictable | Sudden burst during an event |
| Main defense | Continuous ventilation and H2 detection | Off-gas detection, separation, water, explosion control |
| Suppression | Conventional | Water to cool; not clean agent alone |
| Governing detail | Ventilation rate below LFL | NFPA 855 plus UL 9540A data |
Commissioning and acceptance: prove every layer works
Commissioning a lithium-ion battery room is where the layers stop being a drawing and become a tested defense, or where you find out they were never wired to do anything. Most of what fails in a real event fails because a layer was assumed at acceptance rather than proven. The job at startup is to make each layer prove itself.
Verify the off-gas detection alarms and that the alarm actually does what the sequence says, starts exhaust, cuts charge, isolates the unit, and notifies. Verify the BMS trips on over-voltage and over-temperature, not just that it reports values. Confirm the suppression system activates and that its water supply or agent quantity matches the design and the listing. Confirm the explosion-control path: that detection-driven ventilation runs and that any deflagration vents are clear and unobstructed. Confirm the separation as built matches the UL 9540A basis the design was approved on, because field crowding is easy and quiet.
Write the sequence of operations down and test it end to end, detector to action, with the AHJ's witness where required. The failure to avoid is a stack of devices that each work in isolation but were never proven to act together. A detector that alarms to a panel nobody monitors, a trip that was bypassed during fit-out and never restored, a vent blocked by a later cable tray. Find those at commissioning. The event is the wrong time.
The first-responder pre-plan
A lithium-ion battery event does not fight like a normal fire, and the fire department needs to know that before they arrive, not while they are standing at the door. A pre-plan worked out with the local responders is part of the safety case, and on larger installations the AHJ will expect one.
The differences that drive responder tactics are specific. The unit may stay energized during the fire, so applying water or any conductive agent before it is isolated is a shock hazard. The off-gas can deflagrate, so opening a smoke-filled enclosure can ignite the accumulated cloud. The accepted approach for many lithium-ion events is to cool and contain, protect what is around it, and in some cases let the affected unit burn down under control rather than chase reignition, because the stored energy will run its course. And the event can reignite hours later from stranded energy, so the scene is not clear when the visible fire is out.
The pre-plan should give responders the site layout, where the disconnects are and how to confirm isolation, where the off-gas detection and explosion vents are, the chemistry installed, water supply, and the expectation that overhaul and monitoring run long after knockdown. Industry first-responder guidance for battery storage incidents exists and is worth handing to the local department in advance. The worst outcome is a crew applying a structure-fire playbook to a battery room that needed a different one.
Stranded energy and reignition
Stranded energy is the charge left in damaged cells after a fire, and it is why a lithium-ion event is not over when the flame goes out. A cell can be partly destroyed and still hold energy at a hazardous level, and residual heat inside the modules can push it back into runaway hours after the visible fire is dead.
This is the trap that catches people who treat the knockdown as the end. The flame is out, the smoke clears, the crew steps back, and the pack reignites from the heat still trapped in the cells or from a damaged cell that was simmering the whole time. It is the same reason electric-vehicle fires reignite on the tow truck and in the yard. The energy is in the chemistry, and it leaves on its own schedule.
The practical answer is overhaul and monitoring that run long after the fire appears out. Keep cooling the affected mass, watch the temperature with a thermal camera, and keep the area controlled and monitored for an extended period, hours, not minutes, before anyone calls it safe. This belongs in the pre-plan and in the operations procedure, because the people on site after the responders leave are the ones who own the reignition. Do not energize, do not crowd it, and do not declare it clear on the strength of no visible flame.
Monitoring and trending the cells
The drift that ends in runaway usually shows up in the data first, so cell temperature and voltage trending is how you catch a failing pack before it fails. A single cell creeping warmer than its neighbors, a group whose voltage no longer balances, a charge that takes longer than it used to, these are the early tells, and they show up in the trend long before any alarm.
The BMS already measures this per cell or per group. The value is in keeping the history and looking at it, not just watching the live screen. A cell that runs two or three degrees above its row, week over week, is telling you something a single snapshot hides. Trend the temperatures, trend the cell-balance spread, trend the internal resistance if the system reports it, and flag the outliers for inspection before they become the cell that vents.
Capture it in a record you can actually pull up later, a field tool like FieldOS, the cabinet logs, or the building management system, so the trend survives a shift change and a personnel turnover. The point of the data is to act on it. A cell that has been drifting for a month is a planned replacement on a quiet day. The same cell ignored is the off-gas alarm at 2 a.m. Trending is cheap. The event is not.
Maintenance and inspection
A lithium-ion battery system is not install-and-forget, and the maintenance is aimed squarely at the triggers. Loose connections, failing cooling, blocked detection, and aging cells are the things that turn a stable pack into a runaway, and each one is something an inspection catches.
Walk the cabinet and check the connections, because a loose or corroded lug heats under load and a hot joint is both a fire source and a sign of trouble. Confirm the cooling is doing its job, that fans run, that liquid loops hold flow and temperature, and that nothing is recirculating hot air back into the pack, because over-temperature is a trigger you can prevent with a clean filter and a working fan. Check that the off-gas detectors are in service and within calibration, that the explosion vents are clear, and that nothing has been stacked or routed in front of them since the last visit.
Inspect for the physical and the aging. Look for swelling, discoloration, leakage, or a cell that reads off from its neighbors. Pull the trend data and act on the outliers rather than waiting for an alarm. Confirm the BMS protection settings have not been left bypassed after service. The discipline is the same as any critical-power maintenance: the failures announce themselves early to someone who looks, and they stay hidden from someone who does not.
Why the AI buildout raises the stakes
The AI data center boom is pushing rack power and energy density up fast, and with it the size of the lithium-ion UPS and BESS plant behind the load. More stored energy in a room means a larger potential event, which means more code scrutiny, not less.
The 2026 direction in NFPA 855 reflects this, moving toward a near-universal hazard mitigation analysis, leaning on large-scale fire testing, and tightening the energy that can sit in a single fire area. Designs that would have slipped under a capacity exemption a few years ago now carry the full analysis. Expect the AHJ to ask for UL 9540A data, an explosion-control basis, and a first-responder pre-plan on installations that used to get waved through, and expect the bar to keep moving as the industry catches up to densities that only showed up recently.
Common mistakes
- Relying on a clean agent or inert gas alone, which knocks the flame but not the heat, so the cells reignite.
- No early off-gas detection, so the first signal is smoke or fire and the cascade is already running.
- No NFPA 855 separation basis or UL 9540A data, so propagation between units was never proven to be stopped.
- No deflagration venting or explosion prevention, so the flammable off-gas can collect and explode.
- BMS protection setpoints assumed rather than verified, or left bypassed after service.
- Units packed tighter than the tested arrangement, or a fire barrier value-engineered out of the layout.
- No first-responder pre-plan, so a structure-fire playbook gets applied to a battery event.
- Calling the scene clear when the flame is out, ignoring stranded energy and reignition hours later.
What to document
The fire-safety case for a lithium-ion room is only as good as the record behind it, because the AHJ, the insurer, and the next engineer all need to see that each layer exists and was proven. Capture what each layer does, what it was tested to, and the standard or data it rests on.
Record the chemistry and the UL 9540A report it was approved on, the separation as built against that basis, the detection types and their commissioning results, the suppression design and water supply, the explosion-control path, the BMS protection settings as verified, and the first-responder pre-plan. When a layer rests on a tested exception, a reduced spacing justified by 9540A data, write down the data it rests on so the next person does not have to guess.
| Layer | What it does | Note to record |
|---|---|---|
| Cell chemistry | Sets onset and event severity | LFP or NMC, with the UL 9540A report |
| BMS protection | Trips on over-V, over-T, isolates | Setpoints, verified at commissioning |
| Off-gas detection | Catches vent gas before fire | Type, listing, and alarm action proven |
| Separation | Stops cell-to-cell cascade | Spacing or barrier vs the 9540A basis |
| Suppression | Cools the cells | Water supply or agent, design basis, AHJ |
| Explosion control | Manages the off-gas burst | NFPA 68 venting or NFPA 69 prevention |
| First-responder pre-plan | Right tactics, isolation, reignition | Filed with the local department |
Field checklist
Want this checklist to run itself on every job — with photo proof and a signed record crews can hand the customer? That's FieldOS.
Standards and references
NFPA 855, the Standard for the Installation of Stationary Energy Storage Systems, is the framework for spacing, stored-energy limits, detection, suppression, explosion control, and the hazard mitigation analysis. It is adopted and amended through the International Fire Code and the jurisdiction, so the specific separation distances, energy limits, and section numbers depend on the edition in force and the AHJ. Cite it for the framework, then confirm the numbers against the adopted edition.
UL 9540A is the test method for thermal runaway fire propagation, run at cell, module, unit, and installation levels, and its report is the data the AHJ uses to set or relax the protection requirements. UL 9540 lists the energy storage system itself, and UL 1973 covers the battery system including the BMS protection functions. Pair the 9540A data with NFPA 855 because one proves how the product behaves and the other sets what the installation has to do about it.
Explosion control points to NFPA 69 for prevention, detection-driven mechanical ventilation that holds the off-gas below the flammable limit, and NFPA 68 for deflagration venting that relieves a deflagration along a planned path. The VRLA hydrogen side of battery-room safety lives in the battery room ventilation and hydrogen safety guide, and the storage-technology selection lives in the battery and energy storage types guide. Across all of it, hedge the spacing, the stored-energy limits, and the section numbers to the standard, the edition, and the AHJ, and hold the three things that do not bend: early off-gas detection, a proven separation and UL 9540A basis, and water plus explosion control rather than a clean agent alone.
Units, terms, and definitions
Lithium-ion fire safety carries its own vocabulary, and the same idea shows up under different names across a spec, a test report, and a code section.
Thermal runaway is the self-feeding heat reaction. Off-gas, sometimes called vent gas, is the flammable mixture a cell releases before and during the event. Deflagration is the fast flame front through that gas that builds pressure. Stored energy is given in kilowatt-hours, kWh, per unit and per fire area. Onset temperature is in degrees C. The lower flammable limit, LFL, sometimes called the lower explosive limit, LEL, is the gas concentration below which the mixture will not ignite, and ventilation targets a fraction of it.
- Thermal runaway
- A self-feeding reaction where a cell makes more heat than it sheds, vents, ignites, and cascades
- Off-gas / vent gas
- The flammable mixture a cell releases before and during runaway, detectable minutes before fire
- Deflagration
- A fast flame front through accumulated off-gas that builds pressure and can damage the enclosure
- Stranded energy
- Charge left in damaged cells after a fire that can drive reignition hours later
- LFP / NMC
- Lithium iron phosphate, higher onset and less violent; nickel manganese cobalt, higher energy and lower onset
- BMS
- Battery management system, which balances cells and trips on over-voltage, over-temperature, and over-current
- LFL / LEL
- Lower flammable or explosive limit, the gas concentration below which the mixture will not ignite
FAQ
What is thermal runaway in a lithium-ion battery?
Thermal runaway is a self-feeding reaction where a cell makes more heat than it can shed. The temperature climbs, the chemistry breaks down, the cell vents flammable gas and ignites, and the heat drives the next cell over the edge. It cascades cell to cell and reignites, which is why it is not an ordinary fire.
Can you put out a lithium battery fire with a clean agent?
No, not on its own. A clean agent or inert gas knocks down the flame but does not cool the cells, so the heat keeps building and the cells reignite when the agent clears. Water is the workhorse for lithium-ion because it removes heat, applied in quantity and held long enough to cool the affected mass.
What is NFPA 855?
NFPA 855 is the Standard for the Installation of Stationary Energy Storage Systems. It frames the separation, stored-energy limits, detection, suppression, explosion control, and hazard mitigation analysis for battery installations. The specific distances and limits depend on the adopted edition and local amendments, so confirm them against the edition in force and the AHJ.
What is UL 9540A?
UL 9540A is the test method for evaluating thermal runaway fire propagation in energy storage systems, run at cell, module, unit, and installation levels. It produces the data the AHJ uses to set or relax protection requirements. Large lithium-ion installations effectively need a UL 9540A report on the specific product to get approved.
Can lithium-ion battery off-gas explode?
Yes. A cell vents flammable gas, including hydrogen and carbon monoxide, before and during runaway. If that gas collects in an enclosure and ignites, it deflagrates, a fast flame front that builds pressure and can blow out doors, walls, or a roof. That is why explosion control, through deflagration venting or prevention, is its own design layer.
How much separation does NFPA 855 require between battery units?
A commonly cited starting point is around 3 ft between units and from walls for lithium-ion arrays above a stored-energy threshold, but that is a default. The standard allows a closer spacing where UL 9540A or large-scale fire testing proves runaway will not spread. Confirm the actual requirement against the adopted edition and the AHJ.
Is LFP safer than NMC for thermal runaway?
LFP is more forgiving. It has a higher onset temperature, roughly 220 to 260 C against about 170 to 210 C for NMC, and a less violent event with less heat and off-gas. But LFP still vents flammable gas and still cascades if the pack allows it, so it needs the same detection, separation, and explosion control. Safer is not safe.
Why does a lithium-ion battery fire reignite hours later?
Stranded energy. Damaged cells can hold charge at a hazardous level after the flame is out, and residual heat inside the modules can push the pack back into runaway hours later. The scene is not clear when the visible fire is dead. Keep cooling, monitor with a thermal camera, and hold the area controlled for an extended period.
How do you detect thermal runaway before there is a fire?
Off-gas detection sniffs the electrolyte vapor and gases a cell vents minutes before it ignites, testing has shown a window of roughly 5 to 20 minutes. It alarms while you can still isolate the unit, cut charge, start exhaust, and clear people. Cell temperature and voltage trending through the BMS is the complementary early signal.
What is the difference between a lithium-ion and a lead-acid battery room hazard?
A VRLA lead-acid room handles steady hydrogen off-gassing with continuous ventilation. A lithium-ion room is managing thermal runaway, a sudden burst of heat and flammable gas that cascades cell to cell. The lithium-ion answer adds early off-gas detection, separation, water cooling, and explosion control on top of ventilation. Same flammable-gas idea, different release, different design.
People also ask
Codes cited in this guide
This guide is written and reviewed against the published standards below. Always confirm the current adopted edition with the authority having jurisdiction.