On January 16, 2025, white smoke rose from a corner of the Moss Landing Power Plant on the coast of Monterey County, California. The source was "Moss 300," a 300 MW-class lithium-ion battery energy storage facility operated by Vistra Corp. The facility housed roughly 100,000 lithium-ion battery cells. Firefighters arrived on scene, but because flammable gas may have been accumulating inside the enclosure, they could not carry out the standard procedure of opening the doors and applying water. The fire burned for two days. According to a post-incident report by the U.S. Environmental Protection Agency (EPA), about 55% of the batteries were damaged, and roughly 1,200 nearby residents were placed under evacuation orders for more than 24 hours.
This was the fourth fire at the facility since 2019.
What Moss Landing demonstrated is that fires at lithium-ion battery storage facilities are fundamentally different from conventional building fires. Not only is extinguishing them difficult, but the very act of opening the enclosure to check what is happening inside can itself trigger an explosion. This paradox has shifted the safety design philosophy for battery storage facilities from "extinguishing the fire" to "preventing its spread."
The Expansion of Renewable Energy Made "Giant Batteries" Inevitable
Behind the integration of lithium-ion battery energy storage systems (BESS) into power grids lies a timing mismatch inherent to renewable energy. Solar power peaks during the day, while wind power often generates large amounts of electricity at night or during periods of low demand. Without a way to bridge the gap between generation and demand, renewable energy must be curtailed.
Battery storage has rapidly expanded as the solution to this problem. According to the International Energy Agency's (IEA) 2026 edition of the Global Energy Review, newly installed battery storage capacity worldwide reached 108 GW in 2025, a 40% increase from the previous year. Cumulative installed capacity stands at approximately 270 GW (Wood Mackenzie estimate)—11 times the scale seen in 2021. China accounts for roughly 60% of the global total, while the United States installed 19 GW, up about 60% year-over-year.
The composition of battery chemistries has also shifted. Lithium iron phosphate (LFP) batteries now account for about 90% of new installations (IEA, 2025 data). While LFP is inferior to nickel-manganese-cobalt (NMC) chemistry in energy density, it offers greater thermal stability, making it well-suited to the frequent charge-discharge cycles of stationary storage. As discussed below, this choice of chemistry is directly linked to fire risk.
Thermal Runaway: A Three-Stage Chain Reaction Starting From a Single Cell
To understand the fire risk of BESS, one must first understand what happens inside a single battery cell.
Thermal runaway in lithium-ion batteries begins when heat generation inside a cell exceeds its capacity to dissipate that heat. Triggers vary widely, including internal short circuits, overcharging, external heating, and mechanical damage. Once initiated, a chain of three exothermic reaction stages unfolds inside the cell.
The first stage is the decomposition of the solid electrolyte interphase (SEI) film on the negative electrode surface. The SEI film begins to decompose at around 90–120°C, releasing gases such as ethylene and carbon dioxide along with heat. The activation energy for this stage is relatively low, around 50–70 kJ/mol, making the reaction easy to initiate. In the second stage, once the SEI film is lost, the active material of the negative electrode reacts directly with the electrolyte, generating even more heat, with an activation energy of about 80–120 kJ/mol. In the third stage, the positive electrode material decomposes and releases oxygen, which then reacts with the electrolyte to generate additional heat. The decomposition onset temperature of the positive electrode material varies significantly by chemistry: approximately 180°C for NMC and above 240°C for LFP.
This temperature difference is one reason LFP has become the dominant chemistry for stationary storage. LFP's higher thermal runaway onset temperature provides a greater margin before a chain reaction occurs. However, "less likely to occur" does not mean "will not occur."
In a single, isolated cell, thermal runaway remains a localized event. The problem is that BESS units pack thousands to tens of thousands of cells closely together. As the National Fire Protection Association (NFPA) points out, thermal runaway in one cell can heat adjacent cells, potentially propagating in a chain reaction. At Moss Landing, this propagation extended across 55% of the entire facility.
Invisible Danger: Gas Accumulation Even Without Visible Flames
Another factor that sets lithium-ion battery fires apart from conventional fires is gas.
As materials inside a cell decompose during thermal runaway, large quantities of flammable gases—such as hydrogen, carbon monoxide, and ethylene—along with toxic gases like hydrogen fluoride are released. The UK's National Fire Chiefs Council (NFCC) has warned that this process "can rapidly generate large volumes of toxic and explosive vapors."
Inside a sealed BESS enclosure, these gases can accumulate. If flammable gas reaches a certain concentration and encounters an ignition source, deflagration or explosion can occur. In a 2019 BESS fire in Surprise, Arizona, gas ignited when firefighters opened the enclosure, injuring several firefighters. It remains one of the few documented cases of human casualties from a BESS incident in the United States.
This illustrates that an enclosure showing no visible flames from the outside is not necessarily safe—gas may be accumulating inside. Determining whether firefighters should open an enclosure requires assessing internal gas concentration, ventilation conditions, and explosion risk.
Failure Rates Down 99%—So Why Did Moss Landing Still Happen?
Are battery storage fires becoming more frequent? The answer depends on whether one looks at absolute numbers or failure rates.
According to the Battery Energy Storage System Failure Event Database maintained by the Electric Power Research Institute (EPRI), the failure rate of grid-scale BESS worldwide, measured per unit of cumulative installed capacity, dropped by 99% between 2018 and 2025. However, the absolute baseline and comparison values (such as annual failure counts per cumulative GWh) are not explicitly disclosed in EPRI's published materials. This decline reflects the resolution of early-generation design flaws and manufacturing issues over successive generations of technology. Between November 2017 and September 2024, 90 fire and explosion incidents at battery storage facilities were recorded worldwide. South Korea accounted for the largest share at 34 incidents (37.8%), many of them concentrated in early-generation NMC-based systems.
However, a lower failure rate does not mean the risk has disappeared. Because the absolute volume of installed capacity is growing so rapidly, the total number of incidents will not reach zero. Moreover, as Moss Landing demonstrated, once propagation begins, the scale of damage can extend across an entire facility.
An analysis published by the Clean Power Association in March 2025 found that among 35 large-scale BESS fires recorded in the United States between 2012 and 2024, roughly half of the cases for which the system's operational age was known occurred within six months of commissioning. Additionally, 69% of incidents occurred during operation, with causes more often traced to system integration, installation, or assembly issues rather than the battery chemistry itself.
| Indicator | Previous Generation (c. 2018–2020) | Current Generation (2024–2026) |
|---|---|---|
| Dominant chemistry | NMC (nickel-manganese-cobalt oxide) | LFP (lithium iron phosphate) |
| Positive electrode decomposition onset temperature | ~180°C | 240°C or higher |
| Failure rate (per cumulative capacity) | Baseline (2018) | Down 99% vs. 2018 (2025, EPRI) |
| Safety test standard | UL 9540A editions 1–4 (cell, module scale) | UL 9540A 6th edition (full-scale burn + intentional ignition) |
| Coordination with fire services | Primarily reactive response | Joint planning from the design stage (NFCC guidance) |
UL 9540A's 6th Edition Ushers in an Era of "Evaluate by Burning"
Among the international standards governing battery storage facility safety, the most significant change is unfolding in UL 9540A. Its full title is "Test Method for Evaluating Thermal Runaway Fire Propagation in Battery Energy Storage Systems." The 6th edition was published on March 13, 2026.
Through the 5th edition, UL 9540A evaluated thermal runaway propagation at three scales: cell, module, and unit. But this testing approach had limits. Measuring the behavior of individual cells or modules cannot predict what happens when thousands of cells are arranged in their actual configuration. Spacing, ventilation, enclosure structure, and the presence or absence of thermal management systems all influence whether propagation succeeds.
The biggest change in the 6th edition is the formal incorporation of the Large-Scale Fire Test (LSFT). This test involves igniting a fire within a complete BESS unit with all detection and suppression systems disabled, then evaluating propagation to adjacent units or the surrounding building. In addition, if flammable gas release is confirmed during cell- or module-level testing, a test is now mandatory in which the gas is intentionally ignited to evaluate the risk of explosion and fire propagation.
Furthermore, the 6th edition introduces pass/fail criteria for the first time. Previous editions lacked clear pass/fail determinations; test data was voluntarily disclosed by manufacturers, and regulators and fire services interpreted it on a case-by-case basis. Some manufacturer claims of having "passed UL 9540A" were, strictly speaking, inaccurate under this earlier framework.
The 2026 edition of NFPA 855 (Standard for the Installation of Stationary Energy Storage Systems) incorporates this two-stage evaluation—thermal runaway testing plus full-scale burn testing—as a mandatory requirement. UL 9540A test data is also used to establish minimum separation distances between units and maximum allowable capacity (in kWh) per fire compartment.
Detecting Trouble Before Temperatures Rise: Early Warning via Mechanical Signals
Preventing propagation requires detecting anomalies and isolating the affected unit before the first cell reaches thermal runaway. Conventional monitoring detects abnormalities in voltage, current, and temperature, but these signals only change significantly once thermal runaway has already progressed substantially.
A 2026 paper published on arXiv, "Regime-Aware Physics-Guided Early Warning of Lithium-Ion Battery Thermal Runaway Using Thermo-Mechanical Signals," attempts to close this time gap using mechanical signals. The research team integrated five types of signals—temperature, voltage, force applied to the cell, deformation, and state of charge (SOC)—into a machine learning model designed to detect early signs of thermal runaway.
Cross-validation across 30 mechanical abuse tests (including conditions at SOC 10%, 50%, and 90%) showed an average warning lead time of 15.6 seconds—a 69.6% improvement over the best baseline model (a temporal convolutional network, or TCN), which achieved 9.2 seconds. The detection success rate was 0.92, the experiment-level false alarm rate was 2.7%, and the F1 score was 0.89.
Notably, the force signal contributed most significantly to detection. Removing the force signal shortened the lead time by 60.3%. In the early stages of thermal runaway, gas generated inside the cell causes the casing to expand, damaging internal barriers and separators. This mechanical change appears before temperature rises sharply. However, the research team itself acknowledges that these results apply specifically to mechanical abuse conditions, and that in scenarios involving overcharging, external heating, or aging degradation, voltage, impedance, or gas concentration may prove to be more effective indicators.
A separate study from the Beijing Institute of Technology monitored surface stress in 280 Ah LFP cells and reported that detecting the transition point from exponential to linear stress increase enabled prediction of thermal runaway more than 30 minutes in advance. The ability to detect this before the safety vent opens carries significant practical implications for real-world operation.
These studies remain at the laboratory stage. Integrating them into operational BESS units will require addressing challenges such as sensor durability, cost, and the reliability of systems capable of simultaneously monitoring thousands of cells.
Building Fire Services Into the Design: The NFCC's Ten Principles
The UK's NFCC has published guidance for coordinating with fire and rescue services from the planning stage of grid-scale BESS projects. Its core principles are organized into ten items covering facility siting, access route provisioning, fire water supply, fire spread prevention, accurate information sharing, and emergency planning.
The NFCC places particular emphasis on detection and ventilation design. It calls for automatic smoke, gas, and radiant heat detectors, along with continuous flammable gas monitors, to be installed in every BESS cabinet and enclosure, with alarms, system shutdown, and full ventilation triggered automatically upon gas detection. Installation of deflagration panels is also listed as an option.
The underlying philosophy of this guidance aligns with the lessons of Moss Landing: rather than responding after firefighters arrive on scene, the design stage itself must guarantee that a single initial cell failure does not escalate into a facility-wide incident. The NFCC states that "a properly managed lithium-ion system can be operated safely," while acknowledging the risks of thermal runaway, toxic and explosive vapors, fire, and explosion.
Why the Cause of the Moss Landing Fire Still Isn't Known
More than a year and a half after the Moss Landing fire, as of mid-2026, Vistra Corp. still describes the cause as "undetermined." The company continues its investigation with the help of multiple outside experts, but identifying the cause from components damaged by extreme heat has proven extremely difficult.
Removal of damaged batteries is ongoing. According to Moss Landing's response site, more than 34,000 intact battery modules have been removed, discharged, and transported to approved recycling facilities. Removal of batteries from the damaged portion of the building is expected to be completed in the latter half of 2026. Vistra has abandoned plans to restart Moss 100 and is currently considering resuming operations at Moss 350.
The undetermined cause remains an important caveat in discussions of BESS safety. Even as design improvements and stricter standards move forward, if the root cause of a specific incident cannot be identified, the possibility that the same failure mode lurks in other facilities cannot be ruled out.
Safety Becomes a Built-In Component of Storage Infrastructure
The deployment of battery storage shows no sign of slowing. BloombergNEF projects annual installations of 158 GW in 2026, reaching 308 GW by 2036. According to IEA analysis, the annual deployment of battery storage has already surpassed the historical peak of gas-fired power generation additions (approximately 107 GW in 2002).
Amid this expansion, fire safety is shifting from being an "additional measure" to becoming a core component of infrastructure design. UL 9540A's 6th edition full-scale burn testing, the mandatory requirements in NFPA 855, the NFCC's planning guidance, and research into mechanical-signal-based early detection—all of these converge on a single principle: assume individual cell failures will happen, and contain their impact.
The questions that remain are clear. Can mechanical-signal-based early detection function at the scale of monitoring thousands of cells simultaneously in operational BESS? Will LFP's thermal stability advantage hold up over 20-plus years of operation? Will the cause of the Moss Landing fire ever be identified? Whether battery storage safety can scale at the same pace as battery storage deployment itself—the success of the energy transition may hinge on the answer to that question.