AI could help mitigate lithium battery fires, study finds
Fire officials and others gather outside of a Bronx supermarket the day after a fire tore through a market that fire officials are blaming on a faulty lithium-ion battery on March 6, 2023, in New York City. Spencer Platt/Getty Images
Researchers have developed an algorithm that can detect when a lithium battery begins to overheat, which they say can help inform public safety efforts to mitigate fires.
Lithium-ion batteries, lauded for their long lifespans, low maintenance needs and fast charging, have gained popularity in recent years, as they are increasingly used in electric vehicles, cell phones and micromobility devices like electric scooters. But as the battery’s use increases, so do its fire and safety risks.
Last week in Moss Landing, California, — a city about 100 miles south of San Francisco — a fire broke out at the battery storage facility Vistra, where thousands of lithium-ion batteries were being kept. The fire was contained, but an estimated 1,500 residents had to evacuate the area, and some people have since reported experiencing health issues like headaches and sore throats.
The U.S. Environmental Protection Agency earlier this week deemed the area safe from harmful air contaminants, but Moss Landing’s brush with a lithium fire underscored the growing public safety concerns of the batteries.
Fighting tech with tech could help keep lithium-ion batteries under control, a study from the National Institute of Standards and Technology suggests. Researchers have developed an artificial intelligence model that can identify the early stages of a lithium-ion battery fire — the moment when a safety valve in the battery’s hard casing breaks to release pressure from the swelling battery cell.
“This is a way to use advanced technology to reduce the risks of not only fires occurring, but also of injuries, loss of life and property damage from these devices in the future by using this novel detection method,” said Anthony Putorti, fire protection engineer at NIST’s Fire Research Division.
Once broken, the battery’s safety valve produces a hissing sound, which researchers used to train a machine learning model to develop an early warning detection system that could alert users of a potential fire, according to Wai Cheong Tam, a mechanical engineer for NIST’s Fire Research Division. The algorithm was able to detect the valve sound 94% of the time.
The findings could help inform future public safety resources and operations aimed at curbing lithium-ion battery fires and explosions, such as modernized fire alarm systems or safer building codes, Purtoti said.
Several lawmakers have moved to implement safer lithium-ion battery practices in recent years, including a proposed bill in Illinois that would prohibit the disposal of the batteries in mixed-recycling bins.
At the local level, New York City, which saw 18 deaths and 150 injuries caused by lithium-ion battery fires in 2023, started restricting the sale of mobility devices and storage batteries that do not meet certain safety standards that same year.
A garage used for storing an electric vehicle fleet, for instance, could be connected to a smart system that detects the sound of the battery’s valve breaking and alerts vehicle users or building occupants of an at-risk battery, Tam said.
There is typically a two- to three-minute window between the safety valve’s breakage and ignition, during which the battery’s temperature skyrockets, reaching as high as 2,000 degrees Fahrenheit, as the device continues to release smoke and flames, he explained.
That brief window of time is crucial, because lithium-ion battery fires burn much hotter and faster than traditional fires, Tam said. In less than 30 seconds, an entire room can be engulfed in the battery’s flames, while a traditional fire could take five to 15 minutes to take over a room.
When it comes to getting first responders to the scene quicker to manage a fire and get people to safety, “earlier detection becomes a lot more important,” Tam said.
NIST’s research is still in its early stages, with further experiments needed, he said, but the findings offer actionable information for officials to address the burgeoning risks of lithium-ion batteries.
NEXT STORY: New research: Done right, virtual tutoring nearly rivals in-person version