Defending a Remote BESS Facility from Wildfire: What One Utility's Design Choices Teach the Rest of Us

How an Upper Midwest utility built autonomous robotic fire protection for a remote battery storage site, using single 2,000-gallon tanks and no permanent water infrastructure.

An Upper Midwest electric utility recently built a battery energy storage system (BESS) in forested service territory, with the nearest fire department 24 minutes away. Like most owners of outdoor lithium-ion storage, it planned to isolate the equipment and let the affected container burn down. What that plan could not address were the two exposures on either side of it: a battery fire that escapes into the forest, and a wildfire that reaches the facility. The utility closed both gaps with an Autonomous Robotic Fire Suppression System (ARFSS): fixed NozzleBot Defender units with outward-facing flame detectors, engineered around a strict 2,000-gallon water limit and bought as a capital asset under its regulated rate structure.

Why burn-down is the accepted practice

Lithium-ion thermal runaway is self-sustaining. A cell in runaway generates its own heat and releases oxygen internally, so oxygen-displacing agents have limited effect and water applied outside a container rarely reaches the cells that are burning. NFPA 855 and the International Fire Code both acknowledge this, and for remote outdoor installations, authorities having jurisdiction routinely approve a defensive strategy: de-energize and isolate the system, let the affected container consume itself, and apply water only to protect neighboring containers. It is a defensible plan for the container itself, coordinated here with the local AHJ. But it says nothing about the forest beyond the fence line.

The two exposures burn-down ignores

A battery fire that becomes a forest fire. The site is surrounded by timber. A container in thermal runaway vents burning electrolyte and ejects embers intermittently for hours, and under burn-down it does so unattended in a fuel bed. The nearest department is 24 minutes away in drive time alone; add dispatch and turnout for a volunteer or paid-on-call department, and the interval between ignition in the treeline and first water is 35 to 45 minutes or longer. On a dry, windy afternoon, a wildland fire given that much undisturbed growth becomes an established head fire, and the responding crew is requesting mutual aid rather than suppressing an ignition. Utility-caused ignitions have produced some of the largest loss events in the industry's history.

A wildfire that reaches the BESS. In a regional wildfire event, firefighting resources are triaged toward life safety and populated areas, and a remote, unmanned facility sits near the bottom of that list. Embers arriving miles ahead of a flame front lodge against enclosures, cable trays, transformers, and switchgear, and radiant heat alone can compromise seals and initiate thermal runaway in healthy cells, turning a wildfire victim into a second hazardous incident. The mitigation is perimeter defense, wetting vegetation and knocking down ember ignitions as they occur, which traditionally requires people on site at exactly the moment when nobody can safely be there.

The engineered response: autonomous perimeter defense

Both exposures share one characteristic: the decisive interval is measured in minutes, and the site is unmanned. The utility specified fixed NozzleBot Defender units, autonomous suppression monitors aligned with FM Global Data Sheet 4-14, the April 2026 standard that recognizes autonomous monitor-based suppression as an engineered protection layer. It paired them with triple-infrared (IR3) flame detectors facing outward only. When a detector confirms flame in the treeline or along the perimeter, the nearest Defender trains a directed water stream on the base of the fire within seconds, 24/7/365, in any weather, with no human intervention, and notifies pre-determined personnel. An ember ignition is attacked while it is still a spot fire, making the 24-minute response the backup rather than the plan, and an approaching wildfire meets pre-wetted vegetation and active suppression at the fence line rather than an undefended asset.

Two constraints that shaped the design

A 2,000-gallon ceiling on water. The site has no water infrastructure of any kind. The conventional answer is large tank storage, typically 20,000 gallons or more, but state environmental regulations intervened: on-site vessels exceeding 2,000 gallons trigger multiple reviews, each adding months to the schedule. The design ceiling became a single 2,000-gallon tank for each NozzleBot, which rules out any area-coverage approach. A conventional monitor flowing 500 GPM exhausts 2,000 gallons in four minutes; a directed, detection-driven system flows water only when flame is confirmed, only at the point of ignition, and only for as long as flame persists. Applied that way, 2,000 gallons is enough to knock down a treeline spot fire or wet the perimeter ahead of an advancing front until tenders arrive. A disqualifying limitation became a design input.

Do not spray the batteries. The utility was adamant that the NozzleBots would not cover the containers, on both engineering and regulatory grounds: the water required to meaningfully affect a container in thermal runaway is far beyond what the site could store, and any water on the enclosures generates runoff. Every nozzle faces outward, defending the treeline, the perimeter vegetation, and the approach corridors, with none aimed at the containers.

The economics: why the utility bought rather than leased

As a regulated utility, the owner earns an authorized return of 7% on capital expenditures and nothing on operating expenses. Mobile Defender units can be rented seasonally, which suits many organizations, but here rental payments would have been pure operating expense with no return, while purchasing placed the system in the rate base as a capital asset earning the authorized 7%. ARFSS procurement, in other words, is not purely a fire protection decision: the same hardware can be a rental expense or a rate-base asset, and the right answer depends on the owner's regulatory and accounting position.

What other infrastructure owners should take from this

Remote, unmanned, high-consequence, long fire department response, little or no water: that profile describes a great deal of critical infrastructure, from pump stations and substations to compressor stations, communications sites, and switchyards throughout the wildland-urban interface. The design sequence generalizes: define both directions of exposure, let water and environmental constraints size the system rather than disqualify it, aim coverage at what needs defending, and run the buy-versus-rent analysis under your own model. Burn-down remains the right plan for the battery. Autonomous perimeter defense is what makes it a complete plan for the site.

WatchDog Robotics manufactures the NozzleBot family of Autonomous Robotic Fire Suppression Systems, including the fixed-installation Defender and the mobile Expeditionary unit, pairing IR3 flame detection and thermal imaging with directed water streams to attack fires in seconds. To discuss protection for a BESS facility, substation, pump station, or other remote infrastructure, visit WatchDogRobotics.com.