How to Assess Battery Storage Degradation and Augmentation Cost: What Drives Capacity Fade, How Laboratory Models Schedule Augmentation, and What a Warranty Will Be Judged On

Short answer. A battery storage project's usable energy falls with time and use; the owner either lets it fall, buys capacity back ("augmentation"), or pays a supplier to guarantee it. The public laboratory texts agree on what drives the fall (temperature, state of charge, depth of discharge and cycling rate) and show two ways of costing the response: PNNL's 2022 model schedules explicit augmentations down to a 60% floor over a 20–25-year life, while NREL's 2025 cost work folds augmentation into a fixed O&M charge of 4% of capital cost at one cycle a day over 15 years. This guide gives five questions to put to any degradation or augmentation assumption. It quotes no project curve, availability figure or augmentation price; the laboratory figures are US laboratory assumptions.

1. What is the capacity promise measured against?

The PNNL/Sandia performance protocol (PNNL-22010 Rev 2, April 2016) defines energy capacity stability as the energy capacity at any point in time as a percentage of the initial energy capacity, and stored energy as the product of rated power and discharge time at rated power. IEC 62933-2-1:2017 deals with unit parameters and testing methods of electrical energy storage systems, with the storage devices themselves outside its scope. In China the battery-level standard is GB/T 36276-2023 (电力储能用锂离子电池), a current recommended national standard issued 28 December 2023 and implemented 1 July 2024 under the China Electricity Council. This guide cites it from its information page.

Why the definition matters: NREL's 2017 life-prediction paper says end of life is typically defined when only 70–80% of beginning-of-life capacity remains under nameplate conditions, and NREL's 2023 aging study names 80% remaining capacity as the common end-of-life metric while listing 30% increased DC internal resistance, 50 mV increase in average voltage and 1.5% loss of energy efficiency as other monitoring metrics. A capacity promise that does not say which metric, which test conditions and which reference (initial measured or nameplate) is not yet a testable promise.

2. What drives the fade?

NREL's 2017 paper says lifetime varies with thermal environment and with how the battery is charged and discharged; side reactions occur even in storage and accelerate with high temperature and high state of charge; and the industry mainly uses semi-empirical models that extrapolate accelerated aging tests, its own model fitting capacity fade with 1.4% RMS error and resistance growth with 15% RMS error across nine conditions from 0 °C to 55 °C, with uncertainty quantification still needed. NREL's 2023 study of large-format commercial NMC and LFP cells (>50 Ah, tier 1–2 manufacturers) ran calendar-aging tests varying temperature and state of charge and cycling tests varying temperature, C-rates, depth-of-discharge window and duty cycle; it says degradation is temperature-sensitive and cell temperature depends on cell design and fixturing, so life predictions may need a thermal model, and that resistance variance increases dramatically beyond 10% capacity loss.

PNNL's 2019 cost characterisation compiles cycle life at 80% depth of discharge, says cycle life is a function of depth of discharge, and defines calendar life as the maximum life when the system is not operated — in operation, whichever of calendar or cycle degradation runs faster sets the overall life, and calendar life depends strongly on conditions such as ambient temperature. The IEA's 2024 batteries report defines cycle life as the number of charge and discharge cycles a battery can complete before losing performance, and warns that the increased dispatch frequency of value stacking can accelerate degradation and that some services may be mutually incompatible by design.

CNWattHub editorial method: for any degradation curve offered, ask for the four inputs behind it — average cell temperature, resting state of charge, depth-of-discharge window and cycles per year — and whether the curve was produced for the duty cycle the project will actually run.

3. Does the cost model buy capacity back or let it decline?

PNNL's 2022 Energy Storage Grand Challenge assessment reports laboratory data in which LFP reached 2,400 cycles and NMC 1,520 cycles at 80% depth of discharge before available energy fell below 80% of rated energy; it assumes a linear decline to 60% of rated energy for both chemistries and sets end-of-life at 60%; and after the first augmentation it sets depth of discharge going forward to the energy corresponding to 60% of rated energy, augmenting again each time energy falls to 60% until a 20–25-year plant life is reached. Its augmentation methodology uses a variable 20–25-year project life set by the number of DC storage-block augmentations or replacements, which depend on calendar and cycle life; replacement frequency equals cycle life at the depth of discharge divided by cycles per year; blocks run at an average 80% depth of discharge at the start and 60% after the first augmentation; the augmentation quantity covers the energy needed so that the original block's discharge energy at 80% is available from the augmented block at 60%; and the cost is the DC block $/kWh price discounted to present value.

PNNL keeps augmentation separate from fixed O&M and warranty: fixed O&M is 0.43% of direct capital cost for the storage block, balance of system and power equipment plus a major power-equipment refurbishment of 30% of power-equipment cost every 10 years, giving about $4.59/kW-year for a 10 MW, 4-hour LFP system and $5.11/kW-year for NMC; LFP is not expected to need an additional warranty purchase (warranty terms of 10–15 years or 10,000 cycles at 80% DOD are cited), whereas NMC is assumed to need one, about $3/kWh-year in 2021, from four years after purchase and after each augmentation.

NREL's June 2025 cost-projection update takes the other route: all operating cost goes into fixed O&M at one cycle per day (variable O&M zero); the main difference between published fixed-O&M values is whether augmentation is included; NREL adopts a high-end value of 4% of the $/kW capacity cost for a 4-hour battery on the assumption that it counteracts degradation so the system performs at rated capacity through its life, and warns that at a much higher cycling rate this may not be sufficient. It selects a 15-year lifetime — the lower end of published values and a conservative choice because no existing 4-hour battery has yet reached 15 years — and 85% round-trip efficiency. NREL's 2025 Annual Technology Baseline page says the same: fixed O&M includes augmentation so the LFP-based system operates at rated capacity through its 15-year life, at 4% of capital cost in $/kW; cost and performance assume about one cycle per day (a 4-hour device at 16.7% capacity factor, a 2-hour device at 8.3%); and degradation is a function of usage, so systems might need replacement during the analysis period.

Reading the two together (CNWattHub derivation, not a sourced figure): a block cycled once a day at 80% depth of discharge would, on PNNL's laboratory cycle counts of 2,400 (LFP) and 1,520 (NMC) cycles to 80% of rated energy, reach that point within roughly 6.6 and 4.2 years respectively. That is a conditional illustration on laboratory figures, not a reading of any particular model: it suggests that a project model assuming rated capacity for 15 years at one cycle a day under similar conditions needs augmentation, replacement or reserve capacity somewhere, but it does not show that the model's O&M line already contains that cost. CNWattHub editorial method: ask where the model carries augmentation, replacement and reserve (oversized) capacity — in fixed O&M (as NREL states for its own 4% assumption), in a separate augmentation or replacement schedule, in a warranty purchase, or in capacity installed at commissioning — how much each line carries, and whether any of it has been left out.

4. Who pays, and when?

The split between warranty, O&M and augmentation decides who pays. PNNL's schedule places augmentation with the owner's capital plan, charges a warranty purchase for NMC after year four and after each augmentation, and treats power-equipment refurbishment as a 10-year O&M item. NREL's approach puts augmentation in a fixed annual O&M that in effect assumes performance is guaranteed over the lifetime. A supplier's performance warranty is the third vehicle: as one published example, Sungrow's Australian HV battery warranty (V7.0, December 2023) guarantees at least 60% of initial capacity over ten years from installation (no more than 126 months from manufacture) or until a minimum energy throughput is reached, whichever comes first, under stated standard capacity-test conditions, and requires internet access to performance data and log history to assess the battery's condition. That text is a residential product warranty, cited only for its structure: a capacity floor, a time limit, a throughput limit and a data condition.

Chinese rules add an end-of-life duty: the NEA's 2021 project rules require the project unit to monitor operating status in real time and, when the project reaches its design life or no longer meets safety requirements, to assess and rectify, and to retire the project and report to the filing authority if rectification fails. An augmentation plan that extends economic life past the design life stated in the filing should say how that duty is met.

5. What data will the judgement rest on?

NREL's O&M best-practices report treats capacity degradation as a key performance metric, says that for lithium-ion batteries rising internal resistance can point to capacity loss from several electrochemical mechanisms, that the goal of health assessment is to tell premature degradation (charging approach, temperature outside acceptable ranges) from normal end-of-life loss so that reserve accounts can be adjusted on revised useable-life assumptions, and that temperature is an important input, with tools such as BLAST or SAM able to estimate its impact. It also says the O&M plan should include battery replacement procedures and that replacement strategy depends on topology and technology — in a VRLA lead-acid string, for example, resistance imbalance between new and older cells can make the new cells degrade faster than scheduled. CNWattHub editorial method: write into the contract who logs cell temperature, resting state of charge and throughput, at what resolution, and who holds the record, because the warranty example above makes data access a condition of the warranty itself.

Assessment checklist

Scope limits

Laboratory and standards texts only. The cycle counts, O&M percentages, lifetimes and warranty structure are US laboratory assumptions or one manufacturer's published terms, labelled as such; no Chinese project curve, availability figure or augmentation price was found in a public source. GB/T 36276-2023 and IEC 62933-2-1 are cited from information and abstract pages. The year figures in section 3 are CNWattHub's arithmetic on PNNL's table values under the stated conditions (one cycle a day at 80% depth of discharge); they do not show what any cost model's O&M line contains. Not engineering, procurement or investment advice.