Capacity fade

What it is (precise)

Capacity fade is the permanent reduction in the charge a cell can store, distinct from temporary capacity loss caused by low temperature or high-rate discharge, and distinct from efficiency losses, which waste energy per cycle without shrinking the tank.

Mechanistically, in the LFP chemistry that dominates stationary storage, the leading cause is loss of cyclable lithium inventory as the solid-electrolyte interphase (SEI) on the graphite anode continues to grow and consume lithium, with secondary contributions from lithium plating at low temperature or high charge rates and from active-material loss. Because these mechanisms permanently sequester lithium, the lost capacity does not return on the next cycle.

Engineers split fade into two superimposed pathways: Cycle aging, which scales with energy throughput, depth of discharge and C-rate, and Calendar aging, which proceeds even at rest and accelerates with high state of charge and elevated temperature.

A grid asset that sits at high SOC in a hot climate fades from calendar effects alone, independent of how much it is cycled. The reference point matters: cell-level fade is measured in DC Ah or kWh under a defined test protocol, while the number the owner actually contracts is usable AC energy at the point of interconnection, net of PCS, transformer and auxiliary losses.

Why it matters in a real grid-scale project

Capacity fade is a commercial problem before it is a physics problem. A BESS is financed against a contracted deliverable energy at the POI, for example a 4-hour, 100 MW / 400 MWh resource. As cells fade, raw MWh at the rack falls; if usable energy at the POI drops below the contracted value, the owner faces capacity-payment shortfalls or liquidated damages.

The two standard countermeasures are oversizing (installing more MWh at BOL so the system still meets the guarantee after years of fade) and Augmentation (adding fresh capacity later by installing extra racks or containers, or replacing aged modules).

Fade also silently shortens duration. At constant nameplate power, a 100 MW / 400 MWh system that has faded to 80% retention holds only 320 MWh, a 3.2-hour asset, unless extra BOL energy or augmentation backfills the gap. In markets where capacity accreditation depends on demonstrated duration, that slippage directly cuts capacity revenue.

Vendor Capacity warranty terms are therefore written around the expected fade curve, typically guaranteeing minimum retained capacity each year over a 10-to-20-year term, conditioned on operating limits: temperature, SOC window, annual throughput or cycles, and C-rate.

The fade model agreed at financial close drives the augmentation schedule, the spare-container footprint reserved on the site plan, and the auxiliary load of the thermal management system that slows fade. An optimistic fade estimate propagates straight into project IRR and into missed or early augmentation triggers, which is why lenders' independent engineers scrutinize the degradation curve as hard as any electrical study.

Typical values and standards

End of Life for stationary cells is conventionally defined at ~70% of BOL capacity, and grid assets are routinely operated below the 80% figure often cited for EV packs because augmentation backfills the deficit.

Modern LFP cells are typically rated on the order of 6,000-12,000 full equivalent cycles to that threshold under specified laboratory conditions, commonly 25 °C, moderate C-rate and a defined DoD. Real fleet fade is usually 1.5-3% of BOL per year in the early years for a daily-cycled LFP system, with calendar aging alone contributing very roughly 1-2% per year at moderate temperature and SOC.

The fade curve is non-linear. A relatively fast initial drop in the first year, loosely 2-4%, is followed by a slower, near-linear decline, which is why warranty tables quote a separate Year-1 figure. Late in life some cells exhibit a "knee" where fade accelerates, so extrapolating the linear mid-life slope to Year 20 is not conservative. Temperature is the strongest lever: as a rule of thumb, aging rates increase steeply with cell temperature (a rough doubling per 10 °C is often used for screening), which is why HVAC or liquid-cooling setpoints are a warranty condition, not a comfort choice.

No safety code governs fade; it is a performance and warranty matter settled by contract and by capacity-test procedures. Performance testing of industrial lithium cells is covered by IEC 62620, and system-level capacity tests are defined in the supply agreement or per owner test protocols.

The safety framework that constrains the same design sits alongside it: UL 9540 certifies the ESS product, UL 9540A provides the thermal-runaway propagation test data, NFPA 855 governs installation, and NFPA 68/69 cover deflagration protection. Operating choices made to slow fade, such as tighter SOC windows and lower temperatures, interact with these thermal-management requirements, so fade mitigation is never optimized in isolation.

How it shows up in specs, studies and contracts

On a cell or DC-block datasheet, fade appears as a cycle-life claim, for example "≥8,000 cycles to 80% SOH at 25 °C, 0.5C/0.5C, 100% DoD". Every qualifier matters: change the temperature, C-rate, DoD or the EOL threshold and the number is not comparable across vendors. In the supply agreement it appears as a year-by-year capacity retention table (the Capacity warranty), with remedies, exclusions and an annual or biennial capacity test at a defined C-rate, temperature and rest protocol whose measured value settles warranty claims.

A working engineer should check five things: whether retention percentages reference nameplate or usable energy; whether the measurement point is DC at the rack or AC at the POI; what throughput cap (MWh/yr or cycles/yr) voids or de-rates the guarantee; how the degradation curve feeds the revenue model and the Augmentation plan; and whether the EMS enforces the SOC and temperature limits the warranty assumes.

In interconnection and market paperwork, fade shows up as the degradation-adjusted capability the plant can register, and operating teams track it as SOH trending in the SCADA historian against the warranty curve.

Common pitfalls

The most common confusion is between fade and Round-trip efficiency. RTE is energy lost per cycle as heat; fade is the shrinking of what the battery can hold at all. A plant can have excellent RTE and severe fade, or vice versa, and both erode revenue through different lines of the model. A second trap is comparing cycle-life numbers quoted to different EOL thresholds: 10,000 cycles to 70% SOH is a weaker claim than 8,000 cycles to 80%, not a stronger one.

System-level fade is also worse than the cell average suggests. Capacity in a series string is limited by its weakest element, and rack-to-rack imbalance means usable system energy can sit 1-3% below the average cell SOH unless the BMS balancing and Availability of all racks are maintained. Finally, do not treat the vendor curve as a measurement: it is a warranty floor negotiated under assumed conditions. Actual fade depends on the site's dispatch profile and thermal history, which is why owners run periodic capacity tests instead of trusting the model.