BESS Fire Safety in 2026: Thermal Runaway, NFPA 855, and What the Incidents Actually Taught Us

Two consequences follow, and they are different problems:

1. Fire — if the vent gas ignites immediately.
2. Explosion — if the gas accumulates unburned inside an enclosure and then finds an ignition source.

The second one is what injures people. The 2019 McMicken incident in Arizona is the canonical case: a cell failure filled a container with flammable gas that had not ignited, and when firefighters opened the door, the resulting deflagration seriously injured responding firefighters. Nearly every gas-detection, ventilation, and explosion-control requirement in today’s codes traces back to that event.

A hard-won nuance: the industry’s shift to LFP improved fire behavior but did not solve the explosion problem. LFP vent gas is hydrogen-rich. In many ways the modern design challenge is less “how do we stop fire spreading” and more “how do we prevent a flammable gas volume from ever assembling.”

Moss Landing: the incident that reset public perception

On January 16, 2025, the Moss Landing 300 battery building at Vistra’s site in Monterey County, California caught fire. Facts worth knowing, drawn from the US EPA’s response documentation and subsequent reporting:

• The facility was a 300 MW system holding about 100,000 lithium-ion batteries (EPA’s figure) of NMC chemistry, installed indoors in a former turbine-hall building — an architecture the industry has moved decisively away from.
• EPA estimates roughly 55% of the batteries were damaged; independent studies put the range at 55–80%. The site flared up again on February 18, 2025.
• A peer-reviewed study later estimated on the order of 25 metric tons of nickel, manganese, and cobalt deposited across surrounding wetlands.
• Cleanup is still running as I write this: EPA-supervised battery removal is expected to continue into late 2026, with Phase 2 demolition of the burned building starting mid-2026.
• The cause remains under investigation at the time of writing. Anyone telling you definitively why it happened is ahead of the evidence.

The industry lesson is not “batteries are dangerous.” It’s that this specific architecture — hundreds of MW of NMC sharing one indoor air volume — concentrates consequences. A CPUC survey found the majority of California’s grid-scale fleet is the opposite design: outdoor, dispersed containers, LFP chemistry. Same industry, very different risk profile. The regulatory aftermath was real either way: California’s CPUC adopted General Order 167-C, taking oversight of BESS operations and maintenance and requiring emergency response plans to be filed with local fire departments.

The standards stack, in one map

• UL 1973 — safety listing for the battery system (racks and modules).
• UL 9540 — safety listing for the complete ESS (battery + PCS + controls).
• UL 9540A — not a pass/fail listing but a test method that characterizes thermal runaway propagation at cell, module, unit, and installation level. Its data feeds everything else. A new edition was released in the same cycle as NFPA 855 (2026), with large-scale fire testing requirements and a greater focus on system-level safety.
• NFPA 855 — the installation standard: where systems can go, how far apart, and what protection they need. It reaches projects through the fire codes (IFC, NFPA 1) as jurisdictions adopt them.
• NFPA 68 / NFPA 69 — explosion protection: deflagration venting (68) and explosion prevention through gas control (69).
• Internationally, the IEC 62933-5 series plays the system-safety and test-method roles (the UL 9540 and UL 9540A slots), and Canada layers CSA C800 over the same fire-code base — but no IEC document fills the NFPA 855 installation slot.

The fire-safety hub now has this stack as an interactive map — hover any document to see what it does and how it connects to the others, from the fire code down to the component listings. A second map does the same for Europe, where the EU Battery Regulation and the IEC 62933 family fill the frame and the NFPA 855 slot is — instructively — empty.

What UL 9540A tests at each level, and what each result feeds

UL 9540A has four test levels: cell, module, unit and installation. Its 6th edition was published on March 13, 2026. UL Solutions calls the top level of that edition “Installation Level - Large Scale Fire Test”, and says a unit-level test is still required for residential systems and for non-residential systems that use an active thermal runaway propagation prevention system (TRPPS).

For each level, the table pairs UL Solutions’ description of what is evaluated with the contents that UL’s 2020 checklist for code officials expects in the report. The last column gives the use that paragraph 9.2.1 of NFPA 855 (2026) makes of the data. The annex note to that paragraph (A.9.2.1) ties all of the fire-testing data to the decision on what fire and explosion control and prevention an installation needs.

Level	What the test evaluates	What the report should record	What the data feeds
Cell	Thermal runaway behavior of a single cell, and the composition and flammability of its vent gas	Venting temperature, thermal runaway temperature, gas data	Gas production at cell level (9.2.1)
Module	Whether runaway spreads from cell to cell; heat and gas release rates; risk of ignition or deflagration	Thermal runaway temperature, whether propagation occurred, peak and convective heat release rates, peak smoke release rate, gas data	Propagation potential at module level (9.2.1)
Unit	Whether fire spreads between modules; heat and gas release; deflagration and re-ignition	Temperatures on target units and walls, heat flux on walls and target units, hydrocarbon, CO, CO₂ and H₂ volumes before and after flaming, any explosion, flying debris or re-ignition	Propagation potential between ESS units (9.2.1); the test basis for reduced separation (see the spacing section below)
Installation	5th edition: mainly the effectiveness of the fire protection system. 6th edition: a large-scale test with a challenging fire scenario and an assumed post-deflagration condition, judging enclosure design, separation distances and, for indoor systems, building suppression	Not covered by the 2020 checklist	Propagation potential between ESS units (9.2.1), through the large-scale test; test objectives in Annex G.11 of NFPA 855 (2026)

What changed in NFPA 855 (2026 edition)

The 2026 edition was released in September 2025, and it’s the most consequential revision yet:

• Hazard Mitigation Analysis becomes the default. Earlier editions required an HMA only in defined cases, such as projects above certain stored-energy thresholds. An HMA — a formal engineering risk analysis covering thermal runaway initiation, propagation, and consequences — is now the expected baseline for essentially all in-scope ESS installations, with exemptions reserved for well-characterized chemistries such as lead-acid and aqueous nickel-based systems.
• Large-scale fire testing (LSFT) alongside UL 9540A. Before its 6th edition, UL 9540A let testing stop at the level where propagation was contained, which can leave a data gap for large installations; for residential units that rely on testing to justify spacing under 3 ft, TIA 23-1 to the 2023 edition of NFPA 855 already required the complete UL 9540A (or equivalent) report, including cell, module and unit level (15.13.1.1). Where another section requires it, the 2026 edition calls for large-scale fire testing alongside UL 9540A (9.2.1), to show that a fully involved fire in one enclosure does not spread to its neighbors (Annex G.11).
• Explosion control shifts from venting to prevention. Installations must now incorporate explosion control designed per NFPA 69 — active prevention, including combustible concentration reduction (CCR) systems that keep gas concentrations below ignitable levels — or a performance-based alternative backed by installation-level testing. Deflagration venting per NFPA 68 is no longer accepted as the primary strategy on its own. It's a philosophical shift: don't manage the explosion, prevent the gas cloud.
• Thermal runaway propagation prevention (TRPP) is defined as an active method with new requirements, including documenting piping compliance (ASME B31.1/B31.3) as part of the UL 9540 listing.
• Detection options broadened. Early fire detection may now use smoke detection, thermal-imaging detection, or radiant-energy detection installed per NFPA 72 — a more flexible menu than the previous edition allowed.
• Emergency response gets teeth. A compliant Emergency Response Plan must address mitigation, preparedness, response, and recovery, be reviewed annually, and include annual refresher training with the AHJ notified.
• The scope table was rebuilt. Chemistries are now listed explicitly (rather than lumped into an "other technologies" bucket), which reduces ambiguity for newer entrants like sodium-ion and gives AHJs clearer thresholds.

The fire-testing text sits in Section 9.2 of the 2026 edition. Where another section calls for it, paragraph 9.2.1 asks for a representative ESS to be tested to UL 9540A and by large-scale fire testing, to collect data on gas production at cell level, propagation potential at module level and propagation potential between ESS units. The annex note (A.9.2.1) says manufacturers, system designers and AHJs use the data to decide what fire and explosion control and prevention an installation needs, and it points to Annex G.11 for the objectives of the large-scale test. In the 2023 edition, fire and explosion testing was referenced at 9.1.5.

Paragraph 9.2.1 names no procedure for the large-scale test. UL describes UL 9540A as the only consensus standard NFPA 855 explicitly cites for large-scale fire testing, and its 6th edition contains one. CSA Group proposed, in tentative interim amendment log 1852 (comments closed October 17, 2025), to name the procedure in Section 9.7 of CSA/ANSI C800:25 as the large-scale test in 9.2.1. Whether NFPA issued it is not confirmed here; check the amendments in force for the edition your AHJ enforces.

Worth watching: NFPA is developing NFPA 800, a Battery Safety Code intended to cover battery hazards across the whole lifecycle; it may eventually absorb some of what NFPA 855 covers today.

Adoption reality check: most jurisdictions still enforce codes based on the 2020 or 2023 editions via IFC 2021/2024. California adopted fire-code updates based on NFPA 855 (2023) effective January 1, 2026, and the State Fire Marshal has signaled early adoption of the 2026 edition targeted for July 2027. Always confirm which edition your AHJ actually enforces before you design to one.

How test data turns into spacing and explosion control

Spacing

The 2020 edition’s default, as NFPA staff presented it in 2021, arranged ESS in groups of up to 50 kWh spaced 3 ft from other groups and from walls (4.6). Other arrangements were allowed on the basis of what the 2020 edition called large-scale fire testing (4.1.5), which meant UL 9540A testing; the separate large-scale fire test came with the 2026 edition. Remote outdoor locations were exempt. Test data is what moves a layout off that default. UL’s 2020 checklist for code officials lists what a unit-level report has to show before a reviewer relies on it:

• the test started at the maximum operating state of charge, and the outcome did not rely on the battery management system, fans or coolant pumps operating;
• temperatures on the target units stayed below the venting onset temperature measured in the cell test;
• wall temperatures rose no more than 97 °C above ambient, unless the system is intended only for noncombustible installations;
• for non-residential systems and outdoor ground-mounted residential systems, heat flux at the egress path stayed at or below 1.3 kW/m²;
• the test layout matched the intended separation from walls and from other units.

The last check is the one that binds a site plan. NFPA’s own annex wording on it (A.15.13.1.4) entered the residential chapter of the 2023 edition through TIA 23-1, issued on August 25, 2023 and effective September 14, 2023. It says a configuration that departs from the tested size, separation, cell type or energy density should be accepted only if it is shown to give equivalent results, and it names stacked units, units on a mezzanine and back-to-back units as layouts a test may not have covered. Treat the tested spacing as the limit of what the report supports.

Explosion control

The cell and module data also size explosion control. NFPA staff summarized the 2020 edition’s exhaust ventilation provision (4.9) as 1 cfm/ft² and a design to keep flammable gas under 25% of the lower flammable limit (LFL). NFPA’s annex ties that provision to flammable gas released during normal charging and discharging (A.4.12, as revised by TIA 20-2). Explosion control applied where an ESS could exceed 25% of the LFL (4.12). In the 2026 edition, the technology table for electrochemical ESS (Table 9.7.6) sends exhaust ventilation to 9.7.6.1 and explosion control and prevention to 9.7.6.7, as corrected by NFPA errata 855-26-1 of October 13, 2025.

A worked illustration shows why the report matters. For a well-mixed enclosure at steady state, the exhaust flow Q that holds the gas concentration at a fraction f of the LFL is approximately Q = G ÷ (f × LFL), where G is the gas release rate. The cell test supplies the LFL and the module and unit tests supply G. The following inputs are assumed for illustration and come from no test report: LFL = 8% by volume, a peak release G = 5 L/s, and f = 0.25.

With these inputs, Q = 5 L/s ÷ (0.25 × 0.08) = 250 L/s, about 530 cfm.

The release in this illustration comes from thermal runaway, which the same annex (A.4.12) places under the explosion control provisions of 4.12. A real design also has to account for how the release changes over time, detector response time, fan start-up, backup power and imperfect mixing. The illustration is a reason to read the report closely. It is not a sizing method.

Clause locators by edition

Clause numbers move between editions, so cite a number only with its edition. The locators below come from NFPA documents: a 2021 staff presentation of the 2020 edition and TIA 20-2 to that edition, TIA 23-1 for the 2023 edition, and errata 855-26-1 and proposed TIA 1852 for the 2026 edition. A dash means the locator for that edition was not confirmed from an NFPA document.

Requirement	2020 edition	2023 edition	2026 edition
Hazard mitigation analysis	4.1.4	—	—
Fire testing	4.1.5, “large-scale fire testing” (UL 9540A); the basis for non-default arrangements under 4.6	9.1.5, fire and explosion testing	9.2 (9.2.1); large-scale test objectives in Annex G.11
Size and separation	4.6: groups of up to 50 kWh, 3 ft apart	15.3.1: residential units 3 ft apart unless fire and explosion testing documents less	—
Exhaust ventilation	4.9	—	9.7.6.1 (electrochemical ESS)
Explosion control	4.12	—	9.7.6.7, explosion control and prevention

What good design looks like in practice

Across projects, the pattern that satisfies both the codes and common sense:

• Outdoor, dispersed, non-occupied enclosures with propagation resistance demonstrated by UL 9540A data (unit level under the 5th edition, installation level under the 6th) and large-scale fire testing.
• Off-gas detection tied to system shutdown, plus combustible-gas concentration control so a flammable volume never assembles.
• Spacing and site layout consistent with the tested listing, with clear fire-apparatus access and water supply planned for exposure protection.
• A real emergency response plan, walked through with the local fire department before energization — not a PDF that lives in a drawer.
• Response tactics that accept a burning enclosure may be allowed to burn out while exposures are protected — a decision that always belongs to the fire service on scene.

FAQ

Are LFP batteries safe from thermal runaway? Safer, not safe. Harder to ignite and less energetic in failure — but hydrogen-rich vent gas keeps explosion prevention the central design problem.

Does NFPA 855 (2026) apply to my project? That depends on what your jurisdiction has adopted. Many AHJs still enforce earlier editions through their fire codes. Ask first — designing to the newest edition is often wise anyway, since it’s where regulation is heading.

Should communities fear BESS projects? The honest answer: modern outdoor LFP container plants designed to current codes carry a very different risk profile than the legacy indoor NMC design that burned at Moss Landing. Scrutiny is fair; conflation is not.

---

Fire safety, UL 9540A test data, and how safety flows into permitting and insurance get full modules in my Grid-Scale BESS: Complete Guide.