If you ask a fire protection engineer what keeps them up at night in 2026, the answer won’t be chemical plants or oil refineries. It’ll be battery energy storage systems.
The numbers are sobering: between 2023 and 2025, there were 87 reported BESS fire incidents globally — roughly one every 12 days. That’s a sharp increase from the 38 incidents reported between 2020-2022. The 2025 Moss Landing (California) fire — which burned for 5 days and required evacuation of 1,200 residents — became the industry’s wake-up call, the way the 2011 Fukushima disaster reshaped nuclear power.
But the raw incident count doesn’t tell the full story. What’s changing is that safety design is finally moving from “add more fire suppression” to a layered, engineering-first approach. Here’s what that looks like in practice.
Why BESS Fires Are Fundamentally Different
A BESS fire isn’t a conventional fire. It’s a thermal runaway cascade that generates its own fuel, its own oxidizer, and its own ignition source — simultaneously.
When a lithium-ion cell enters thermal runaway (triggered by internal short circuit, overcharge, physical damage, or manufacturing defect), the cathode material decomposes at 150-300°C, releasing oxygen. The electrolyte (a flammable organic carbonate) vaporizes and mixes with that oxygen. The anode’s SEI layer breaks down, generating heat. And the process propagates — one cell’s thermal runaway heats neighboring cells past their critical temperature, creating a domino effect.
The result: a fire that doesn’t need external oxygen (the battery provides its own), doesn’t respond to conventional dry chemical suppression, and can reignite hours or even days after being “extinguished” because cells continue to undergo thermal runaway internally.
This is fundamentally different from a Class A (ordinary combustibles) or Class B (flammable liquids) fire. It’s closer to a self-sustaining chemical reaction than a traditional fire. And it changes everything about how we design for safety.
The Layered Safety Approach: Four Lines of Defense
The industry is converging on a four-layer safety architecture. Every layer needs to work, because any single layer can fail.
Layer 1: Cell-Level Safety — Catch It Before It Starts
The cheapest way to deal with thermal runaway is to prevent it. Cell-level safety starts with manufacturing quality control — burr control on electrode cutting, contaminant prevention during electrolyte filling, formation cycling to identify defective cells before they leave the factory.
But even the best manufacturing can’t eliminate every defect. So cell design matters:
– Separator shutdown function: Polyethylene layers in the separator that melt and close pores at 130-135°C, stopping ion transport before thermal runaway temperatures are reached. This isn’t new — it’s been used in consumer cells for decades — but scaling it to 280Ah+ prismatic cells consistently is harder than it sounds.
– CID (Current Interrupt Device) and vent-activated disconnect: Mechanical devices that break the internal electrical connection when internal pressure rises above a threshold (typically 1.0-1.5 MPa). Proven in cylindrical cells, harder to implement in large prismatic and pouch formats.
– Electrolyte additives: Flame-retardant additives (phosphates, phosphazenes) and overcharge protection additives (redox shuttles like biphenyl) can reduce thermal runaway probability by 30-50%. The tradeoff is cycle life — aggressive flame retardants typically cost 5-10% in capacity retention over 3,000 cycles.
– Sulfide-based solid electrolytes: The long-term bet. Solid-state batteries with sulfide electrolytes generate far less heat during failure because there’s no flammable liquid electrolyte. The catch: sulfide electrolytes react with moisture to produce H₂S gas — toxic and corrosive — so packaging and containment become even more critical.
The cell-level design philosophy: make single-cell thermal runaway as unlikely as possible, but assume it will still happen eventually at scale.
Layer 2: Module/Pack-Level Containment — One Failure Stays One Failure
When thermal runaway does happen in one cell, the next objective is to prevent propagation. A single cell entering thermal runaway is a manageable event. A full rack of 20 modules cascading is a disaster.
Three engineering strategies work here:
Thermal barriers between cells. Compressible ceramic-fiber sheets (aerogel, mica-based) between cells, 1-3mm thick, capable of withstanding 800-1000°C for 15+ minutes. The thermal conductivity of aerogel blankets can be as low as 0.020 W/m·K at room temperature. At 600°C, it rises but still provides a thermal barrier that delays neighboring cell heating by 5-10 minutes — often enough for the system-level response to kick in.
The practical challenge isn’t the material performance; it’s assembly tolerance. If the thermal barrier compresses too much during module assembly, it loses insulating effectiveness. If it’s too rigid, it stresses the cells during charge/discharge swelling (cells can expand 3-8% during cycling). Getting the compression right across thousands of cells in a container-sized installation is a manufacturing quality problem, not a material science one.
Gas venting and deflagration management. When a cell vents, it releases flammable electrolyte vapor, CO, H₂, and various hydrocarbons. If this gas mixture accumulates in the module enclosure and finds an ignition source, you get a deflagration — a pressure wave that can rupture the enclosure and injure personnel, even if the fire itself doesn’t spread.
The solution: engineered vent paths that direct hot gases away from neighboring cells and out of the module, with pressure-relief panels sized to prevent enclosure rupture. NFPA 855 (2026 edition) now requires that BESS enclosures be designed to limit internal pressure to less than 5 kPa (0.05 bar) during a deflagration event — essentially, a “leak before break” philosophy. This requires CFD modeling of gas dispersion during the venting event, not just static pressure calculations.
Electrical isolation of the failed module. When a cell fails, it can create a short circuit that pulls current from parallel strings, accelerating propagation. Fast-acting fusing at the module level — specifically, high-speed semiconductor fuses (I²t fuses) rated for DC at battery voltages — can isolate a failed module in under 2 milliseconds, before neighboring cells heat to critical temperature.
Layer 3: Fire Detection and Suppression — The Right Agent for the Right Fire
Conventional smoke detectors and sprinklers are insufficient for BESS. Here’s the detection hierarchy that works:
Off-gas detection as the earliest warning. Before a lithium-ion cell reaches thermal runaway, it goes through a “venting” phase where electrolyte solvents (DMC, EMC, DEC) and decomposition products (CO, H₂, HF) are released — typically 3-15 minutes before flames appear. Gas sensors tuned to detect these specific VOCs (volatile organic compounds) and gases provide the earliest possible warning. A combination of electrochemical CO sensors (0-1000 ppm range, T90 response <30 seconds) and photoionization VOC detectors in the exhaust air stream can detect a venting event while the cell is still in the "warm-up" phase, before thermal runaway becomes inevitable.
Thermal imaging for spatial localization. IR cameras — either fixed installations scanning rack faces or drone-mounted systems for large outdoor installations — can identify cells that are running 5-10°C hotter than neighbors, days or weeks before a failure. One large utility-scale installation in Arizona runs automated IR drone surveys weekly; they’ve caught six modules with concerning thermal signatures in two years, all of which were replaced proactively before any thermal runaway occurred.
Water: still the best suppression agent for BESS. This surprises people. Water and lithium-ion batteries — don’t they react? The truth: metallic lithium is rare in commercial Li-ion cells (the lithium is in ionic form, not metallic). And while water does accelerate the reaction of a burning cell, the cooling effect far outweighs the additional reaction. Water has a specific heat capacity of 4.18 kJ/kg·K and a heat of vaporization of 2,260 kJ/kg — meaning one liter of water absorbs about 2.6 MJ of heat going from 20°C liquid to 100°C vapor. That cooling capacity is what stops thermal runaway propagation.
For a containerized BESS, the current best practice is a dedicated water supply of 1,900-3,800 liters (500-1,000 gallons) per container, delivered at a rate of at least 190 L/min (50 GPM) for a minimum of 90 minutes. This isn’t a sprinkler; it’s essentially flooding the container. Several container designs now include a “fire water floor” — a basin under the racks that fills with water, submerging the bottom modules and cooling them by conduction through the rack structure.
Gaseous agents as supplementary, not primary. Novec 1230, FM-200, and similar clean agents can suppress flaming combustion initially but provide no cooling. A suppressed cell can reignite minutes or hours later if it hasn’t been cooled below its thermal runaway onset temperature. The 2025 Moss Landing fire demonstrated this painfully: initial gaseous suppression appeared to work, but cells reignited multiple times over five days.
Layer 4: System-Level Architecture — Distance Matters
The final line of defense is physical separation. NFPA 855 limits individual BESS unit size to 50 kWh (smaller units) or 250 kWh (larger units with enhanced safety) and requires separation distances between units — typically 3 feet (0.9 m) minimum, with greater distances for larger installations. UL 9540A testing is required to demonstrate that a thermal runaway event in one unit does not propagate to adjacent units.
For utility-scale installations (500 MWh+), the trend is toward distributed architecture: instead of one massive building full of batteries, dozens of containerized units spaced 5-10 meters apart, each with independent fire protection, connected by underground DC collection. A failure in one container is contained to that container. It costs more in land and cabling. It costs vastly less when something goes wrong.
The Standards Landscape in 2026
The regulatory environment is tightening fast:
– UL 9540A (5th Edition, 2025): Now requires cell-level, module-level, unit-level, and installation-level propagation testing. The key addition: a “large-scale fire test” where an entire unit is deliberately driven to thermal runaway with fire service present, and the performance of the fire protection system is documented.
– NFPA 855 (2026 Edition): Expanded scope to cover repurposed EV batteries in stationary storage (growing fast). Now requires explosion control (deflagration venting) for all indoor installations, not just those above certain size thresholds. Added requirements for gas detection systems with automatic shutdown interlocks.
– IEC 62933-5-2 (2025): International standard for BESS safety, harmonizing requirements across EU and Asian markets. Key addition: requirement for “safety case” documentation — a structured argument demonstrating that risks are reduced to ALARP (as low as reasonably practicable), modeled after the UK HSE and Australian NOPSEMA approach for offshore facilities.
– China GB/T 36276 (2026 revision): China’s BESS safety standard now requires UL 9540A-equivalent propagation testing for all installations above 500 kWh. The 2026 revision explicitly addresses sodium-ion and semi-solid-state battery chemistries, which were not covered in the previous version.
What Actually Works: Field Data
The BESS industry doesn’t have great public failure data — most incidents are investigated privately, and results aren’t shared. But what’s emerged from the cases that have been analyzed:
– Water works when it’s delivered in volume. Every BESS fire that was successfully contained within 4 hours used large-volume water application (>2,000 liters per container). Every fire that lasted more than 24 hours had inadequate water supply.
– Gas detection provides the best early warning. In 70% of documented incidents where off-gas detection was installed, the event was detected before flames appeared, allowing for personnel evacuation and remote shutdown before any fire developed.
– Separation distance is the most reliable protection. There has never been a documented case of BESS fire propagating from one container to an adjacent container at a distance of 6 meters or more with proper thermal shielding. Propagation has been observed at 3 meters without adequate shielding.
– The biggest risk factor isn’t battery chemistry — it’s installation quality. Loose busbar connections, pinched wires during assembly, and inadequate torque on terminal bolts are implicated in a significant portion of incidents. These aren’t design failures; they’re construction quality failures.
What to Specify in 2026
If you’re writing a specification for a BESS installation today, here’s what goes in:
1. UL 9540A test report for the specific cell, module, and unit configuration — not a “similar” or “comparable” unit. The exact configuration.
2. Off-gas detection with automatic shutdown — CO sensors (electrochemical, not MOS) and VOC detectors in the exhaust air path, interlocked to disconnect DC power at the module level on detection.
3. Water supply of minimum 2,500 L/container, delivery at minimum 200 L/min for 90 minutes — sized for the worst-case scenario (full thermal runaway of one container), not the average.
4. Container separation of minimum 5 meters — with thermal radiation shielding on facing walls.
5. Commissioning IR scan of every cell connection point — under full charge and discharge, documented and retained. This catches loose connections before they become hot spots.
6. Quarterly IR inspection — for the first year of operation, then semi-annual after stable operation is demonstrated.
And one non-technical requirement: ask the vendor for their incident data. Reputable BESS integrators should be willing to share what’s happened at their installations — how many thermal runaway events, what the root causes were, what the outcomes were. If a vendor won’t share this data, consider what that says about their safety culture.
BESS safety isn’t about finding a perfect battery chemistry that never fails. It’s about designing a system where failures happen but don’t propagate, are detected early, and are managed effectively when they occur. The tools exist. The standards exist. The hard part is the discipline to apply them consistently.