🏛️ Issues and Advocacy Training
Courses › Battery Energy Storage Systems: How They Work, Fail and Burn

Lesson 4 of 9 · about 16 minutes

Inside a BESS: Cells to Containers, Inverters to Fire Systems

You'll be able to

  • Name the parts of a BESS from cell to grid connection and what each does
  • Explain why cooling, gas detection and explosion control are safety systems, not accessories
  • Connect specific components to real failures in EPRI's database
  • Work out how much heat a battery must shed in a day of cycling

A utility battery looks like rows of white boxes behind a fence. Each box is a small industrial plant with its own electrical system, climate control, gas detection, fire suppression and computer network. When something goes wrong, the investigation report usually names one of those systems, and often it is not the cells.

The hierarchy

Cell Module Rack orstring Container,cabinet, room PCS(inverter) Trans-former Grid Battery management system: cell, module, string and system levels Thermal management: HVAC or liquid cooling Detection, suppression, ventilation, explosion control Energy management systemdispatch, alarms, data,link to grid operator andremote monitoring center Power flows along the top row. Control and safety systems wrap around it.
The parts of a typical BESS. Data flows up from cell sensors to the EMS; commands flow down.

Part by part

ComponentJobWhat can go wrongWhere it mattered
CellsStore energy chemicallyManufacturing defect, internal short, aging, lithium platingSurprise, Ariz., 2019: internal cell defect started cascading runaway (EPRI, citing the APS investigation)
ModulesGroup cells with sensors and thermal barriersWeak barriers let heat jump cell to cellCodes now require tests showing whether runaway spreads module to module
Racks and stringsBuild voltage; hold fuses, contactors, string BMSDC faults, loose connections, ground faultsSandia notes ground fault detection is part of standard design
Enclosure (container, cabinet, building)House and separate the racks; keep weather outWater intrusion, heat and gas trapped indoors, too little spacingEPRI's 2026 report: indoor sites overrepresented; Warwick, N.Y., 2023 failures came during a large storm
Thermal managementHold cells in a narrow, even temperature bandCoolant leaks, failed fans or chillers, cooling switched offMoorabool, Australia, 2021: service mode shut off cooling and protection (EPRI)
Power conversion systemConvert AC and DC both ways; supply voltage supportOverheating, component failure, arcingSandia Handbook Chapter 13 covers PCS design
Transformers, switchgearStep voltage up to the grid; isolate faultsOil-filled transformer fires, breaker failureEPRI now compares BESS failure rates to large oil-filled transformers
DetectionSense smoke, heat, and gases like hydrogen and carbon monoxideSensors in the wrong place, alarms that don't reach anyone2026 NFPA 855 allows early detection by smoke, thermal imaging or radiant energy
SuppressionCool or knock down flames: clean agent, water, aerosolCannot stop propagation; water leaks short batteriesSurprise 2019 (clean agent); Moss Landing, Feb. 2022 (water leak)
Explosion controlKeep gas below flammable levels or vent a blast safelyNo venting, or doors opened before gas is clearedSurprise 2019: deflagration when the door opened
EMS and communicationsDispatch, alarms, data logs, remote monitoringLost telemetry, bad commands, cyber riskMoorabool 2021: telemetry lost in service mode

Containers, cabinets and buildings

Sandia's handbook describes two common outdoor formats: walk-in shipping containers with racks along the walls, and smaller cabinets shipped fully assembled and repeated as many times as needed. Sandia stresses that spacing between racks and modules is critical to stopping fire spread when a few cells run away. The 2018 IFC that Idaho enforces treats an outdoor container people can walk into for service as a battery storage room, requires walk-in units to be entered only for inspection and repair, and requires outdoor battery areas to be secured against unauthorized entry.

The other format is a building. Moss Landing 300 and the Gateway plant in San Diego, both in EPRI's table as indoor, held hundreds of MWh under one roof. EPRI's October 2026 report found indoor facilities overrepresented among failures and sustaining more damage, and recommended fire walls or other barriers to divide large indoor sites.

Residents opposing projects point out that the outdoor container layout assumes a unit may burn and is designed to keep the fire to one box. That is accurate, and it is also how the codes are written: the 2026 NFPA 855 requires large-scale fire testing showing whether fire in one unit spreads to the next. The argument is over whether "one box burns for hours next to my road" is an acceptable design outcome. That is a siting question as much as an engineering one.

Cooling is a safety system

Sandia's handbook puts it bluntly: without proper thermal management, overheating cells degrade, malfunction or go into thermal runaway, and the main heat source is the batteries themselves. Cooling also has to keep cells close to the same temperature, because uneven temperatures make cells age unevenly, which the BMS then has to manage. That is why HVAC or liquid cooling failure shows up in incident reports, and why cooling equipment runs constantly and makes noise.

Worked example: how much heat a battery makes

The invented Ridgeview battery delivers 400 MWh per day at 85 percent round-trip efficiency.

Energy bought to deliver 400 MWh: 400 / 0.85 = about 471 MWh
Energy lost:                      471 - 400  = about 71 MWh per day

If charging and discharging take 8 hours combined:
  71 MWh / 8 h = about 8.9 MW of heat, on average,
  during those hours

Not all of that heat lands in the battery enclosures (some is in inverters and transformers), but several megawatts of heat must be pulled out of the site on a busy day. That load is what the chillers, fans and pumps are sized for, and what neighbors hear.

Detection, suppression and explosion control

Sandia lists the ways systems detect thermal runaway: cell voltage changes seen by the BMS, swelling detected by deformation sensors, temperature from thermocouples or fiber optics, and gas sensors. The most common suppression medium has been a gaseous clean agent such as Novec, which works mainly by cooling. Sandia notes two limits: as energy density has risen, clean agents have struggled to fully stop propagation, and any suppression that stops flames can let combustible gas build up instead of burning off. That is why makers added carbon monoxide and other gas sensors, and why designs include deflagration vent panels that channel a blast upward.

The Surprise, Arizona explosion on April 19, 2019 is the textbook case. UL FSRI's report lays out the timeline: smoke alarm and clean agent discharge at about 4:55 p.m.; hazmat team entries finding dangerous hydrogen cyanide and carbon monoxide; the door opened at about 8:01 p.m.; deflagration at about 8:04 p.m. Four firefighters were seriously injured. FSRI noted the system was commissioned before the first draft of NFPA 855 and complied with the codes in force then. Its recommendations included remotely accessible gas monitoring and explosion prevention.

The 2021 and later IFC editions require explosion control for walk-in units and rooms under Section 911, which points to deflagration venting (NFPA 68) or explosion prevention by ventilation and gas detection (NFPA 69). Idaho's statewide 2018 edition has no such explosion control section for batteries. It handles gas through room ventilation sized to keep flammable gas under 25 percent of its lower flammability limit (or hydrogen under 1 percent of room volume), and, where a hazard mitigation analysis is required, a test that released gas not create an explosion hazard for occupants or responders. Lesson 8 covers what that means locally.

Common mix-ups

  • Claim: A battery's fire suppression system will put out a battery fire. What the evidence shows: Suppression can slow or knock down flames, but Sandia notes clean agents struggle to stop propagation in dense systems, and EPRI's record shows suppression leaks causing failures.
  • Claim: The cells are what fail. What the evidence shows: EPRI's root cause work found cells and modules were the failed element in only 3 of 26 classifiable incidents; controls and balance-of-system parts made up most of the rest.
  • Claim: Cooling is about battery life, not safety. What the evidence shows: Sandia ties poor thermal management directly to thermal runaway risk.

Reporting angle

  • Request the construction documents and site plan: enclosure type (walk-in or not), number of units, spacing between units and to property lines, and location of the PCS and transformers.
  • Ask what suppression is installed (clean agent, water, aerosol, none) and whether the design relies on letting a unit burn while protecting neighbors.
  • Ask what gas detection is installed, which gases it measures, where alarms go and whether firefighters can read the gas levels remotely before approaching.
  • Ask whether units have deflagration venting or exhaust ventilation, and to which standard.
  • Ask what happens to cooling and monitoring during maintenance or "service mode," the condition at the 2021 Moorabool fire.

Research angle

Check yourself

1. At the 2019 Surprise, Arizona explosion, what did the container's clean agent suppression system do?

2. What converts a battery's direct current to the grid's alternating current?

3. Why does Sandia say some systems now include carbon monoxide and other gas sensors alongside suppression?

4. What went wrong at Vistra's Moss Landing system on February 13, 2022, according to EPRI's entry?

5. A battery cycles 400 MWh out per day at 85 percent round-trip efficiency. About how much energy is lost as heat each day?

6. What does a deflagration vent panel do?

Terms in this lesson

Module
A packaged group of cells with its own sensing, the building block stacked into racks.
Rack or string
Modules wired in series to reach a working voltage, usually with fuses, contactors and a string-level BMS.
Power conversion system (PCS)
The bidirectional inverter that charges the battery from AC and discharges it to AC.
Energy management system (EMS)
Site software that decides when to charge and discharge and talks to the grid operator, the PCS and the BMS.
Clean agent
A gaseous fire suppressant such as Novec 1230 that cools and interrupts flames without water.
Deflagration venting
Panels or openings that release the pressure of a gas ignition in a controlled direction; NFPA 68 is the standard for it.

Full glossary

Sources