Third-party testing of failing lithium iron phosphate (LFP) cells found that an aspirating detector alarmed on invisible pyrolysis particles long before a conventional smoke detector had anything to see.
Battery energy storage systems (BESS) are showing up everywhere: utility-scale grid support, microgrids, EV charging infrastructure, data center backup, and the list goes on. The same energy density that makes lithium-ion attractive is what makes the fire protection problem hard. Once a single cell enters an advanced failure state, heat transfer to adjacent cells can turn a cell-level event into a rack-level event, and a rack-level event into a total loss. That progression is why early fire detection in BESS matters more than in almost any other occupancy.
Now, if your detection strategy waits for visible smoke, it alerts you at roughly the moment your options narrow from prevention to containment. So the useful question isn’t “how fast does the system detect a fire?” It’s “how far upstream of thermal runaway can we get a signal that something is wrong?”
New third-party test data puts a number on that.
WHAT THE TEST OF EARLY FIRE DETECTION IN BESS MEASURED
In January 2026, an independent test house in Japan ran a series of destructive tests on LFP lithium-ion cells (3.2V/50Ah) sourced from three different mid-scale manufacturers. The test house drove the cells to failure two ways:
- Overcharge: a 15V/20A constant-voltage/constant-current charge applied until the cell’s safety vent activated.
- External heating: four 1,400W heat guns heated the cell inside a container-simulated enclosure, approximating escalating thermal conditions in an ESS container.
Throughout both methods, a Cirrus Hybrid dual-technology air sampling detector monitored the space through two sampling pipe locations: one simulating container ceiling sampling and one simulating an HVAC return-air inlet. The team monitored hydrogen (H₂) and carbon monoxide (CO) at the detector exhaust, and a general-purpose smoke and CO detector served as the reference point.
Cirrus Hybrid was the detector under test because of how it senses. Rather than waiting for visible smoke to reach a sensing chamber, it operates with a cloud chamber sensor utilizing the Wilson cloud chamber principle to detect products of combustion down to roughly 0.002 μm. Those are the particles that early thermal decomposition emits before smoke exists in any sense a human or a conventional detector would recognize. A high-sensitivity scattered-light smoke sensor runs alongside it, and the detector integrates the two outputs into a single Combined Fire Signal (CFS).
WHAT THE TESTING FOUND
In overcharge testing, the detector alarmed at the vent event – not at the smoke event.
All three manufacturers’ cells vented at approximately 25 minutes. Cirrus Hybrid alarmed within 6 to 10 seconds of vent activation in every case, with no visible smoke present at the time of alarm. Battery surface temperatures at vent ranged from 107.6 °F (42 °C) to 140 °F (60 °C).
The reason is chemistry. As internal temperature climbs, organic solvents in the electrolyte decompose and vaporize. When the vent releases, those vapors carry submicron combustion nuclei. They exist well before smoke particles do and they are exactly what a cloud chamber sensor is built to find.
Two of the three cells stabilized after venting, with no smoke and no ignition. The third kept drawing current after its vent activated, degraded further, and produced visible smoke at 35 minutes. This was still roughly 10 minutes after the Cirrus Hybrid detector had already alarmed. Even the worst case in the set delivered a warning.
In external heating testing, the early warning window opened much wider.
Cirrus Hybrid alarmed at 3 minutes on one manufacturer’s cell (surface temperature ~212 °F or 100 °C) and at 8 minutes on another (~302 °F or 150 °C). Visible smoke didn’t appear until 25 and 27 minutes respectively. That’s an early warning window of approximately 19 to 22 minutes ahead of visible smoke and before the temperature runaway that carried final surface temperatures above 572 °F (300 °C). Note what’s different here: the detector didn’t wait for a vent event. It caught the cell as it heated, well before venting.
For context, the reference general-purpose smoke and CO detector didn’t alarm until surface temperatures crossed roughly 392 °F (200 °C) and the cell was already producing visible smoke.
The test also logged what the cells gave off. At the overcharge vent stage, H₂ measured between 78 and 143 ppm and CO between 20 and 24 ppm; under external heating, concentrations climbed as high as H₂ 526 ppm and CO 146 ppm. Those numbers are a reminder that early fire detection in BESS and gas detection answer different questions; and why a complete energy storage strategy usually pairs the two.
WHY THE DETECTION WINDOW IS THE WHOLE POINT
Nineteen to twenty-two minutes isn’t interesting as a spec-sheet number. It’s interesting because of what fits inside it.
Within that window, an operator or an automated sequence can stop charging, isolate the affected rack or cabinet, verify battery management system data, start or stop ventilation, notify remote monitoring, and get responders moving with accurate information. This can all be done while the event is still confined to one cell. After thermal runaway begins, most of that list stops being available.
That reframes what early fire detection in BESS is for in an energy storage application. It isn’t a smoke detector with better sensitivity; it’s a source of decision time.
WHAT’S IN THE FULL EARLY FIRE DETECTION IN BESS WHITE PAPER
The summary above covers the headline findings. The full technical white paper from Safe Fire Detection goes considerably deeper, including:
- Complete test data tables for both methods, broken out by manufacturer — vent timing, surface temperatures at alarm, CFS progression, and measured H₂ and CO concentrations at each stage.
- Sampling location findings that help explain how ceiling-level sampling compared against low-level sampling, why particle behavior explains the difference, and where return-air sampling belongs in containerized systems with forced-air circulation.
- Staged alarm sequencing guidance showing how multiple CFS thresholds map to a progressive response, from investigation through shutdown, ventilation, and confirmed alarm.
- How gas detection and fire detection complement rather than replace each other, with the measured off-gassing data behind that conclusion.
- A design considerations checklist for engineers and owners covering container and rack geometry, HVAC airflow, pipe routing and transport time, battery management system (BMS) and fire alarm control panel (FACP) integration, and coordination with the authority having jurisdiction (AHJ).
FREQUENTLY ASKED QUESTIONS
How much earlier can aspirating smoke detection warn in a BESS?
In this third-party testing, the aspirating detector alarmed up to roughly 22 minutes before visible smoke under external heating, and about 10 minutes before visible smoke in an overcharge scenario. The exact window depends on failure mode, cell chemistry, and sampling design.
Why is early fire detection in BESS so critical?
Lithium-ion failure propagates. Heat from one failing cell can cascade to adjacent cells, escalating a cell-level event into a rack-level or total loss. Early detection provides the decision time to isolate and intervene before thermal runaway removes those options.
What is a cloud chamber sensor?
It’s a sensing technology based on the Wilson cloud chamber principle that detects products of combustion down to roughly 0.002 μm — the submicron particles released during early thermal decomposition, well before conventional smoke exists.
PROTECT THE ASSET WHILE YOU STILL HAVE OPTIONS
Come back to that detection window one more time. Everything useful an operator can do in a battery fire (isolate, ventilate, verify, respond) depends on knowing early. Waiting for visible smoke trades away the minutes that decide whether you’re preventing a loss or cleaning one up.
That early-warning advantage isn’t unique to this one test, either. Independent U.S. government research, the NRC and NIST DELORES-VEWFIRE study (NUREG-2180), found cloud chamber detection able to identify low-energy fires in the pre-flaming stage, the same sensing principle Cirrus Hybrid uses.
Space Age Electronics is an authorized distributor of Safe Fire Detection air sampling products. For application support on a battery energy storage project, or to review a detection approach with our team, download the white paper below, then let’s talk through your project site.
Discover more from Space Age Electronics Life Safety Blog
Subscribe to get the latest posts sent to your email.