A grid battery is sold on two numbers that people routinely confuse. Power, in megawatts, is how fast it can move energy. Energy, in megawatt hours, is how much it can hold. Divide the second by the first and you get duration, which determines which markets the asset can serve and, indirectly, how long its cells will last.
The scale is no longer niche. BloombergNEF counted 112 GW and 307 GWh of storage added worldwide in 2025, up 48 per cent on 2024, with China taking 54 per cent of additions and the United States 16 per cent. The US installed a record 58 GWh in 2025 according to American Clean Power and Wood Mackenzie, and Europe added 13.5 GW and 26.4 GWh according to the European Market Monitor on Energy Storage published by LCP Delta and Energy Storage Europe in June 2026.
The stack, from cell to grid
Cells to modules to racks. Individual cells, almost always lithium iron phosphate in new build (about 90 per cent of 2025 additions, per BloombergNEF), are assembled into modules, and modules into racks with their own battery management electronics and rack-level protection.
Enclosure. Racks sit in a container or a purpose-built building, with thermal management, gas detection, fire detection and suppression, and increasingly deflagration venting or explosion prevention. Energy density per enclosure has risen fast: Tesla's Megapack 2 XL holds about 3.9 MWh, with the roughly 5 MWh Megapack 3 entering production in late 2026, CATL's TENER, launched in April 2024, packs 6.25 MWh into a 20-foot container, and Sungrow's PowerTitan 3.0, announced in September 2025, reaches 6.9 MWh in the same footprint. A single modern container therefore stores as much energy as an entire early-2010s grid battery project.
Power conversion system. A bidirectional inverter converts between the direct current of the batteries and alternating current for the grid. It also implements the grid-support behaviour the connection agreement demands.
Transformer and switchgear. Voltage is stepped up to the medium or high voltage of the connection, through protection equipment.
Point of interconnection. The metered boundary. Every commercial guarantee that matters is measured here, not at the battery terminals.
Control layer. A site energy management system dispatches against market signals or operator instructions, coordinates with the battery management system's limits, and reports to a SCADA layer. Where the asset is co-located with solar or wind, the coupling can be alternating current, with separate inverters, or direct current, sharing an inverter and capturing clipped generation that would otherwise be lost.
Where the efficiency goes
Round-trip efficiency is quoted far too loosely, and the measurement point changes the answer by several percentage points.
| Loss | Roughly where it arises | Notes |
|---|---|---|
| Cell resistive loss | Charging and discharging | Scales with the square of current, so higher C-rate systems lose more |
| Power conversion | Each direction through the inverter | A few per cent each way, and worse at low part-load |
| Transformer | Each direction | Small but continuous, including no-load loss |
| Auxiliary consumption | Cooling, heating, controls, lighting | The variable one; in hot climates cooling can dominate the auxiliary budget |
Cell-level round-trip efficiency for LFP is high, typically above 95 per cent. System-level alternating current to alternating current efficiency at the point of interconnection is meaningfully lower: modern projects typically deliver 85 to 90 per cent, vendor datasheets quote beginning-of-life figures a few points above that under favourable conditions, and EIA operating data have put the average across the installed US fleet near 82 per cent. Efficiency also degrades as resistance rises with age. When comparing supplier quotations, confirm the measurement boundary, the ambient temperature assumed, the C-rate and whether auxiliary load is included. Two systems quoted at different boundaries are not comparable, and the difference is worth real money across twenty years.
Duration, duty cycle and degradation
Duration and C-rate are reciprocals, so a four-hour system runs at 0.25C and a one-hour system at 1C. That distinction drives the whole degradation profile. The market has moved toward longer durations: the early British fleet was built at one to two hours for frequency response, while new build in the large US markets, California and Texas above all, now clusters at four hours.
| Revenue stream | Typical duration | Duty character | Degradation driver |
|---|---|---|---|
| Frequency response and fast reserve | 0.5 to 1 hour | Frequent shallow cycles, rapid direction changes, high state of charge dwell awaiting events | Resistance rise, calendar ageing at high state of charge |
| Energy arbitrage | 2 to 4 hours | One to two deep cycles per day | Throughput-driven capacity fade |
| Capacity market and reserve | 4 hours plus | Rarely called, held available | Almost pure calendar ageing |
| Renewable firming and co-location | 2 to 8 hours | Weather-driven, irregular | Mixed, with seasonal temperature exposure |
| Grid congestion relief | Varies by constraint | Site-specific, often predictable | Throughput |
An asset held at 90 per cent state of charge for months waiting for a scarcity event is ageing hard without earning a single cycle. Operators sometimes discover this only when the first capacity test comes in below the warranty curve. State of charge management during idle periods is an operational lever with real value, and it needs to be written into the dispatch strategy rather than left to the default.
Warranties and augmentation
Storage warranties are throughput instruments. A typical structure guarantees a capacity retention curve subject to a cap on annual energy throughput, a maximum number of equivalent full cycles per year, and an operating envelope covering temperature and state of charge. Exceed the envelope and the guarantee weakens or lapses. Degradation has also become a headline selling point: CATL launched TENER in April 2024 marketing zero capacity fade over the first five years of use, a claim that shifts the augmentation calculation if the warranty schedule actually backs it. This is why the dispatch optimiser and the warranty document need to be read together, which happens less often than it should.
Because capacity fades, a project contracted to deliver a fixed energy for twenty years has two options.
Overbuild at the start, installing more capacity than contracted so that the faded system still meets the obligation at end of life. Simple, and it carries the cost up front.
Augment during life, adding racks or replacing modules at planned intervals. Cheaper in present value terms, and it depends on cells of a compatible specification being available years later, on the enclosure and conversion equipment having headroom, and on the site having physical space reserved.
Mixing old and new cells in one string creates its own problems, so augmentation is normally done at rack or block level with separate control, rather than by inserting new modules among aged ones. Whichever route is chosen, it is a financial close decision. Retrofitting space, thermal capacity and electrical headroom into a built site is expensive.
Safety and the standards that apply
For a stationary installation the relevant framework typically includes UL 9540 for the system, UL 9540A propagation test data for the specific product, NFPA 855 for installation requirements including spacing and separation, IEC 62619 for cell and battery safety in industrial applications, and the IEC 62933 series for energy storage systems generally. Jurisdictions differ, and the authority having jurisdiction has the final word.
Two points that matter more than the certificate list.
Ask for the UL 9540A large-scale test report, not the summary. What matters is whether propagation stopped at cell, module or unit level, and what gas volumes and heat release were recorded, because those numbers drive spacing, enclosure design and the fire service response plan.
Involve the local fire service before commissioning. Battery incident response is defensive cooling and exclusion, potentially for many hours, with re-ignition a documented possibility after apparent extinction. A service that has walked the site and understands the strategy makes better decisions than one meeting the technology at two in the morning.
Grid code behaviour
Inverters have historically been grid-following, synchronising to an existing grid waveform. As synchronous generation retires, system operators increasingly require grid-forming capability, where the inverter establishes voltage and frequency itself and can provide synthetic inertia and fault current contribution.
This is a procurement issue rather than a footnote. Grid-forming capability affects inverter selection, control software, headroom reservation and sometimes battery sizing, since providing inertia requires energy to be available in both directions. Requirements are tightening in several markets, so a system specified today for a twenty-year life should be checked against where the grid code is heading rather than only where it stands.
Informational and educational content only. Not professional, financial, legal, or engineering advice.