A battery fire looks sudden. Most of the time, it isn't. Long before a cell vents flame, it has usually been drifting away from healthy behaviour — quietly, measurably, for days or weeks. The hard part was never the warning. It was knowing how to read it.
What thermal runaway actually is
A lithium-ion fire is not really an explosion in the everyday sense. It is a self-reinforcing exothermic cascade. Inside a healthy cell, the chemistry is balanced: charging and discharging generate some heat, and the cell sheds it about as fast as it builds. Thermal runaway begins when that balance breaks — when a cell crosses a temperature or voltage threshold and its internal reactions start producing heat faster than it can dissipate.
From there the steps are well understood. The solid-electrolyte interphase (SEI) — the thin protective film on the anode — begins to decompose, releasing more heat. The electrolyte breaks down and produces gas. The polymer separator that keeps the electrodes apart softens and melts, allowing a direct internal short. The cell vents hot, flammable gas, and that gas can ignite. Each stage feeds the next, which is why, once it is truly underway, runaway is so difficult to stop.
The myth of the sudden fire
Because the final seconds are violent, it is tempting to treat battery fires as random — bad luck, a freak defect, nothing anyone could have foreseen. Field data tells a different story. When failed packs are examined after the fact, investigators very often find that the troublesome cell had been behaving abnormally for some time. The fire was the end of a process, not the whole of it.
By the time a battery is on fire, the interesting part of the story is already over. The signal worth catching arrived weeks earlier.
That gap between the first measurable drift and the actual event is the opportunity. It is also why monitoring matters: a fire is binary, but the road to it is gradual, and gradual things can be watched.
The quiet signals that come first
A cell heading toward failure tends to leave a trail. The precursors are individually subtle, which is exactly why they are missed — but together they form a recognisable pattern.
- Internal micro-shorts. Lithium dendrites or small separator defects can create tiny internal short circuits long before they become catastrophic.
- Slow self-discharge. A cell that quietly loses charge while resting is often venting energy through an internal fault.
- Voltage divergence. In a healthy pack, cells track each other closely. A cell that begins to drift apart from its neighbours is a classic warning sign.
- Impedance rise. Growing internal resistance points to degrading electrodes or interfaces.
- Coulombic-efficiency drift. When the charge returned no longer matches the charge put in, something is consuming it through side reactions.
- Mechanical swelling and off-gassing. Cells often physically expand, and frequently vent gases — hydrogen, carbon monoxide, electrolyte vapour and other volatile organic compounds — before any flame appears.
That last point is worth dwelling on. Off-gassing is one of the most reliable early indicators precisely because it happens ahead of ignition. A cell that is venting measurable gas has already told you it is in trouble.
How early detection works
Modern battery monitoring looks for these signals in combination rather than relying on any single one. Battery management systems track the cell-to-cell voltage spread across a pack, watching for the outlier that no longer matches the group. Periodic impedance measurements reveal slow internal degradation that voltage alone would hide. Dedicated gas sensors inside an enclosure can detect venting before a temperature sensor ever responds. Strain gauges and pressure sensors pick up swelling.
The deeper principle is comparison. A raw temperature or voltage reading means little in isolation. What matters is how each cell compares against what normal looks like — ideally, against what normal looks like for that exact cell, not a generic average. A small temperature asymmetry or a minor efficiency drift only becomes meaningful when you know it is a departure from that cell's own established behaviour.
Why a genomic baseline sharpens the picture
This is where a genomic approach changes the maths. If you know a specific cell's fingerprint from the moment it was manufactured — its materials, its structure, its expected electrochemical signature and how it should age — then an anomaly does not have to be large to stand out. It only has to deviate from that cell's predicted curve.
Measured against a fleet-wide average, an early-stage fault can hide inside normal manufacturing spread for weeks. Measured against the cell's own genomic baseline, the same fault separates from the noise far sooner. The warning window widens not because the sensors improved, but because the reference became precise.
What a weeks-ahead warning is worth
Detection only matters if it buys time to act — and a warning that arrives weeks early gives operators real options. A suspect module can be electrically isolated or de-rated so it carries less stress while a response is planned. Replacement can be scheduled into normal maintenance rather than forced by an emergency. An EV fleet or a grid-storage site can avoid both the unplanned outage and, far more importantly, the dangerous failure.
The goal is not to predict fires for their own sake. It is to make sure the rare cell heading for trouble is found, flagged and removed while it is still just an anomaly on a chart — long before it becomes an incident. Thermal runaway is hard to stop once it begins. Catching the drift that precedes it is not.