Grid Battery Fires – Failure Data
The people who respond to battery incidents are the last to be shown the statistics about them. A reading of the record, for the crews.
The record
Battery fires arrive at a fire station as folklore before they arrive as fact: everyone has heard the stories, few have seen the dataset. There is one. EPRI maintains a database of stationary storage failure incidents compiled from public reports, cataloguing events back to 2011, and the 2024 analysis with PNNL and TWAICE found reported failure rates fell 97 per cent between 2018 and 2023, to roughly 0.2 incidents per gigawatt deployed. The events are real, and per gigawatt of installed storage, they are becoming rare.
What the record says about the job
Three findings matter operationally. First, timing: among incidents where the system’s age was known, 72 per cent happened during construction, commissioning or the first two years of a site’s life, so a newly energised site in your patch deserves different attention from a five-year veteran. Second, cause: the most common root causes sat in integration, assembly and construction, and most failures involved components around the cells, controls and balance-of-system, rather than the chemistry itself. Third, the knowledge gap is official: of 81 catalogued incidents, only 26 have public root-cause explanations, which means the pre-incident planning information crews want often doesn’t exist in public at all.
What this doesn’t establish
This is the stationary storage record, not EV fires, which have their own data and their own tactics. The database compiles publicly reported events against its own inclusion criteria, and its authors note it does not capture every relevant incident, so it is a record, not a census. And a falling rate is a statistic, not a guarantee about the site in your district.
The Ledger
Fire services rewrote their doctrine for every technology the last century delivered, and did it mostly from each other’s incident reports. Batteries deserve the same treatment, and the raw material exists, it’s just sitting in an engineering database instead of a training package. My read is that the first storage operators to walk their local crews through the actual failure record, what breaks, when, and what the site does about it, will buy more community goodwill than any glossy safety page.
So here’s the question: has anyone shown your local fire service the data on the site in their patch?
Sources
EPRI, PNNL and TWAICE, Insights from EPRI’s BESS Failure Incident Database: Analysis of Failure Root Cause, 2024, for the failure rates, the 81 incidents, the 26 with assigned root causes, the integration and construction finding and the early-life figure among incidents with known system age. EPRI BESS Failure Incident Database documentation for scope and inclusion criteria.



