Capacity warranty

What it is (precise)

A capacity warranty defines a guaranteed energy-retention curve: a year-by-year schedule of minimum capacity values, stated either as a percentage of beginning-of-life nameplate or as guaranteed delivered MWh at a defined measurement point — DC or AC, and the two are different promises. For a stationary lithium iron phosphate (LFP) system, a representative curve guarantees on the order of 65 to 70 percent retained capacity by roughly year 20 (the end of a 15 to 20 year warranty term) under one full cycle per day, with the exact schedule set by the contracted use profile.

Two delivery structures dominate. Under an energy-retention warranty, capacity is allowed to fade along the guaranteed curve and the buyer oversizes at day one to hit the contracted usable energy at End of Life. Under a maintenance or augmentation warranty, the supplier adds modules or racks over time to hold a constant guaranteed energy, which converts capacity fade from an engineering problem into a scheduled supply obligation.

Either structure applies only inside a defined operating envelope: exceed the contracted annual throughput, run cells outside the specified temperature window, discharge below the agreed minimum state of charge (SOC), or skip required maintenance and firmware updates, and the supplier can void or pro-rate the claim.

Why it matters in a real grid-scale project

The capacity warranty is where degradation risk transfers from owner to supplier, and lenders price the project on it. A merchant or tolling revenue stack — whether the project sells into markets directly or rents its capacity to an offtaker under a toll — assumes a certain deliverable MWh each year; if measured capacity falls below the warranted curve, the supplier owes a remedy — added modules, cash liquidated damages, or replacement racks — which protects debt service and contracted dispatch obligations.

Independent engineers benchmark the warranted curve against the supplier's own cycle-test data and third-party degradation models, and a curve too optimistic against the chemistry's known calendar and cycle aging is a red flag, not a selling point.

The fine print also drives physical design. An augmentation-style warranty sets how much spare container footprint, DC bus headroom, and power conversion system (PCS) and transformer margin must be reserved from day one, because year-8 augmentation racks still have to connect somewhere.

And the warranty must be co-designed with the operating strategy: an aggressive two-cycle-per-day arbitrage case consumes the annual throughput allowance roughly twice as fast as a one-cycle capacity contract, so the energy management system (EMS) has to meter and log throughput against the warranty cap, and the trading desk has to treat that cap as a hard constraint rather than a suggestion.

The assumption set behind the curve

Every guaranteed curve is the output of a model run, and the contract's boundary conditions are that run's inputs written down as obligations. The first input is the duty profile: cycles per day, depth of discharge, resting behavior, and C-rate.

Change any of them and the exhibit's curve describes a machine you are no longer operating — a curve modeled for four-hour duty at 0.25C says nothing about sustained higher-rate dispatch, which is why the rate limit sits in the envelope alongside cycles and DOD. Duty is then compressed into the throughput condition: an annual allowance in MWh or equivalent full cycles that converts the whole operating year into one auditable number — the number the envelope check reaches for first.

Temperature and state of charge condition the calendar half of the fade. The envelope specifies a cell-temperature window — cell, not ambient, so compliance is read from logged battery management system (BMS) temperature history rather than the site weather record, and a thermal-management fault that lets racks sit hot for a season is a breach even if every dispatch was contractual.

State of charge enters twice: warranted capacity is defined inside the contracted SOC window, the same window that sets usable energy, and calendar fade accelerates at high SOC, so a fleet parked full between trades ages faster than the model assumed while its cycle count stands still.

Two more assumptions hide at the ends of the curve. The start point: whether the warranty clock begins at shipment, energization, or the commercial operation date decides who owns the calendar fade that accrues in transit and construction storage — months of pre-COD degradation that either sit inside the supplier's first-year step or transfer silently to the buyer.

And the remedy: the cure ladder matters less than its ceiling, so check whether liquidated damages are a genuine make-whole or a capped, sole-remedy payment that hands the residual shortfall back to the owner once the cap is exhausted. Run all eight — duty, throughput, temperature, SOC window, C-rate, measurement point, start point, remedy — against the intended operating strategy before signing; each one is a condition the supplier can hold up against a claim.

Typical values and standards

Representative terms in current utility-scale contracts: warranty horizons of 10 to 20 years, with 15 and 20 year offerings increasingly standard for LFP; guaranteed End of Life retention of roughly 65 to 70 percent of nameplate for energy-retention products; and stated cycle assumptions, commonly around 300 to 365 equivalent full cycles per year at a defined depth of discharge.

Fade allowances typically front-load a larger first-year step — often around 2 to 4 percent — followed by roughly 1 to 2 percent per year, though the exact shape is chemistry- and duty-specific. Calendar fade and cycle fade are warranted together through a single curve.

Verification leans on a Reference Performance Test: a full charge-discharge capacity measurement at a contractually defined temperature, C-rate, and SOC window, with agreed correction methods and a baselined beginning-of-life value. Test methodology for ESS performance is standardized in the IEC 62933 series, which many contracts reference.

UL 9540 system safety certification and UL 9540A fire-propagation test data supporting NFPA 855 installation requirements are project prerequisites settled in permitting and procurement; warranty validity is conditioned separately, on staying inside the operating envelope and avoiding unauthorized modifications — a thermal-management change is the classic listed void condition. LFP dominates utility-scale guarantees because its flatter, more predictable fade curve supports longer warranty terms than NMC (nickel manganese cobalt) at comparable risk.

How it shows up in specs, studies and contracts

The warranty appears as an exhibit to the battery supply or long-term service agreement: a degradation table, a defined operating envelope, a test protocol, and a remedies section — read all four together. The degradation table is only meaningful once you know the measurement point — DC terminals versus AC at the point of interconnection (POI), where the figure is net of PCS, transformer, and auxiliary losses — the baseline (nameplate or as-measured commissioning capacity), and whether the guarantee covers total or usable energy in the SOC window.

A warranty quoted against nameplate when the commissioning test came in 3 percent over nameplate is quietly weaker than the same percentage against the as-measured baseline: the overage becomes the supplier's free degradation headroom.

Working engineers meet the warranty at three moments: at commissioning, when the capacity test establishes the beginning-of-life baseline; annually or every two years, when Reference Performance Tests generate the compliance record; and at claim time, when logged throughput, temperature history, and SOC excursions from the battery management system (BMS) and EMS decide whether the claim survives the envelope check.

Questions to ask before signing: who performs and witnesses the tests, how ambient and C-rate corrections are applied, whether degradation between scheduled tests is interpolated or ignored, and what the cure sequence is — repair, replace, augment, or pay liquidated damages, and in what order.

The capacity warranty sits alongside, but is distinct from, the round-trip efficiency (RTE) warranty and the availability guarantee, and all three interact: a system down for warranty repairs may be excused from availability targets, and an RTE shortfall can masquerade as a capacity shortfall if the test protocol is sloppy about where energy is measured.

Financiers additionally scrutinize the credit behind the promise — a 20-year warranty from a thinly capitalized subsidiary is worth less than a 15-year warranty backed by a parent guarantee — so warranty review is as much counterparty analysis as engineering.

Common pitfalls

The classic trap is conflating the warranted curve with constant deliverable energy. An energy-retention warranty explicitly permits the system to fade to its End of Life floor; holding contracted MWh at the POI over the project life requires either day-one overbuild or a planned augmentation program with the footprint, bus voltage compatibility, and PCS headroom to support it.

Augmentation carries its own trap: cell formats and rack designs evolve quickly, so a year-10 augmentation may involve products that no longer match the original containers — contracts should address form-factor obsolescence and mixing old and new capacity behind one PCS.

Measurement ambiguity is the other recurring dispute generator. If the contract does not pin down test temperature, C-rate, rest periods, auxiliary-load treatment, and the exact metering point, both parties can produce defensible but conflicting capacity numbers.

Equally common is silent warranty erosion from operations: an EMS dispatch change that raises average SOC or cycling depth can push the asset outside the envelope without anyone noticing until a claim is denied. Treat the warranty envelope as an operating specification — logged, alarmed, and reviewed — not a document that only resurfaces when degradation becomes a problem.