How to Assess BESS Augmentation Scheduling: the Trigger (PNNL: Augment When Available Energy Falls to the Discharge Energy of the Operating DOD — 80% DOD Before the First Top-Up, 60% After It and at Each Later Return to 60% of Rated; SAM: Replace at a Capacity Threshold or on a Fixed-Year Schedule), the Sizing Rule (Add Enough That the Old 80%-DOD Discharge Is Met at 60% DOD of the Augmented System; Lazard: Build to 110% of Nameplate in Year 1 and Top Up to Nameplate as Needed at Year-1 Module Cost), the Life Logic (Project Life 20–25 Years Set by the Number of Augmentations; NREL's 15-Year, One-Cycle-a-Day Fixed O&M of 4% That 'Might Not Be Sufficient' at Higher Cycling), the Warranty Interface (Throughput-or-Years Limits; PNNL's Modelled Warranty Cost Restarting Four Years After Each Block Replacement) and the Chinese Hooks (Design-Life Assessment and Retirement Duty; Grid-Interface Tests and Filed Maintenance Plans; Inner Mongolia's In-Principle Limit of 1.5 Full Charges a Day and 0.28 Yuan/kWh Discharge Compensation; a Capacity Price Localities May Grant, Scaled by Discharge Duration)

Short answer. An augmentation schedule is a set of three decisions — when to add cells, how many, and for how long to keep doing it — and the published models make each decision explicit. Trigger: PNNL's 2022 assessment augments a lithium-ion block when its available energy falls to the discharge energy of the operating depth of discharge — 80% DOD before the first augmentation, 60% DOD after it, with further augmentations each time available energy returns to 60% of rated; NREL's SAM offers either a capacity-percentage threshold or a fixed-year schedule, and under its 'Cycle and Calendar Degradation' option models calendar and cycle degradation separately with the lower governing (its built-in NMC and LMO/LTO models add the two). Sizing: PNNL adds enough that the discharge energy the original system delivered at 80% DOD is available at 60% DOD of the augmented system; Lazard's 2025 LCOS reflects the market practice of upsizing to 110% of nameplate in year one and augmenting back to nameplate as needed at year-1 module cost. Life logic: PNNL sets project life at 20–25 years, ending when an augmentation's own end of life lands in that window, with replacement frequency equal to cycle life at the operating DOD divided by cycles per year; NREL's alternative convention folds augmentation into a fixed O&M of 4% of $/kW capital over a 15-year life at about one cycle a day, with the warning that at much higher cycling this might not be sufficient. Warranty: a product warranty can run to a throughput or a number of years, whichever comes first; PNNL's NMC schedule books a modelled warranty cost from four years after each block replacement. China: the 2021 interim rules oblige the project unit to assess and rectify when the project reaches design life or fails safety requirements and to retire it if rectification fails; NEA's 2024 notice requires performance and grid-interface tests and periodically filed maintenance and outage plans; Inner Mongolia says listed independent stations are in principle not to exceed 1.5 full-capacity charges a day and sets a 0.28 yuan/kWh compensation for 2026 discharge to the public grid; and the 2026 national notice lets localities grant qualifying grid-side independent storage a capacity price scaled by full-power discharge duration, which in CNWattHub's reading makes sustained duration at rated power a variable worth tracking in the schedule. This guide reads modelling reports, software documentation, a product warranty and Chinese regulatory texts; it quotes no project augmentation plan, EPC clause or utility-scale warranty table.

1. The trigger

CNWattHub reading: the trigger is where a project model most often hides an assumption. A model that augments "in year 7 and year 14" is using SAM's fixed-schedule logic and needs a degradation curve to justify the years; a model that augments "at 70% SOH" is using the threshold logic and needs to say whether the threshold is on available energy at the system's actual DOD (PNNL's definition) or on cell capacity. The two give different years for the same plant.

2. The sizing rule

• PNNL. The amount added takes account of the block's remaining available energy and the extra energy needed so that the discharge required at 80% DOD of the unaugmented rating is available at 60% DOD of the augmented system; the analysis assumes all racks degrade uniformly and notes that in reality some degrade faster and others last longer, and that the last block has unused energy at project end.
• Lazard. Capital cost excludes augmentation because it varies by use case with usage profile and lifespan; the O&M line carries augmentation in the years needed, priced at the original module cost, with warranty cost from year 3; the operating assumptions upsize to 110% of nameplate in year one (a 100 MWh battery starts at 110 MWh), keep the battery between 5% and 95% of usable energy, and augment to nameplate as needed on year-1 module cost.

CNWattHub reading: these are two different sizing philosophies — top up to a discharge requirement (PNNL) versus overbuild and top up to nameplate (Lazard) — and the price assumption differs too: Lazard's year-1 module cost for later augmentations is conservative if module prices fall and the reverse if they rise. The uniform-degradation assumption PNNL flags is the one to challenge in a real plant, where the schedule has to be built rack by rack from measured data, not from one fleet curve.

3. The life logic

CNWattHub reading: a Chinese independent-storage model that assumes an ATB-style "4% FOM covers augmentation" should be checked against its own dispatch assumption, because the NREL authors tie that number to roughly one cycle a day and say it may fail at higher cycling — and Inner Mongolia's rule contemplates up to 1.5 full-capacity charges a day in principle for listed stations. The PNNL method has the opposite property: it recomputes the replacement frequency from cycles per year, so a higher dispatch shortens the interval whenever cycle life rather than calendar life is the binding term in its schedule.

4. The warranty interface

• Throughput or years. Sungrow's Australian residential HV battery warranty 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 a standard capacity test, after which the warranty terminates; the table gives 40.32 MWh for the 9.6 kWh model, and Sungrow requires internet access to performance data and logs to assess the battery's condition.
• Modelled warranty cost. PNNL's NMC schedule books a warranty cost every year starting four years after each DC block replacement — a cost-modelling convention, not a statement of any contract's terms.

CNWattHub arithmetic: 40.32 MWh over 9.6 kWh is 4,200 full-capacity equivalents; spread over ten years that is about 1.15 per day, so a residential battery cycled harder than that reaches the throughput limit before the ten years. The point for a utility-scale schedule is structural, not numerical — the augmentation plan has to be read against the throughput and time limits of whatever warranty covers each block, and the cost model should say how warranty on added blocks is treated, as PNNL's schedule does for replacements.

5. The Chinese hooks

CNWattHub reading: the Chinese texts do not prescribe an augmentation schedule, but they set its boundary conditions. Where a locality grants the capacity price, its formula uses full-power continuous discharge duration, so a station's ability to sustain rated power for its rated duration becomes relevant to that revenue, which argues for tracking duration at rated power alongside energy when setting the trigger. The Inner Mongolia in-principle limit bounds annual cycles for listed stations — by CNWattHub arithmetic, about 548 full-capacity charges a year if 1.5 a day held every day — which is the kind of cycles-per-year input PNNL's method needs to turn cycle life into an interval. And Article 21 means a station that defers augmentation past design life must document an assessment, not just a decision.

6. An assessment sequence built from the sources

1. Trigger definition. Available energy at actual DOD, cell SOH, or fixed years; threshold stated.
2. Degradation basis. Calendar and cycle terms stated and how they combine (SAM's user option takes the lower; its built-in NMC and LMO/LTO models add them); rack-level not fleet-level data.
3. Sizing rule. Top-up to discharge requirement or to nameplate; overbuild in year one if any; module price assumption for later years.
4. Dispatch consistency. Cycles per year used in the schedule versus the dispatch the revenue model assumes and the regulatory cap.
5. Life end. How the last augmentation's residual energy is treated at project end.
6. Warranty fit. Throughput and years per block; how warranty on added or replaced blocks is treated in the cost model.
7. Revenue link. Full-power duration maintained for the capacity price; discharge volume for per-kWh compensation.
8. Compliance. Design-life assessment and filed maintenance plans; operating indicators per GB/T 36549-2018.

Scope limits

This guide reads PNNL's 2022 assessment (section 6.1.5 and Table 6.5), NREL's 2025 cost-projection update and ATB page, Lazard's June 2025 LCOS notes, the SAM battery-life help page, one residential product warranty, the 2021 interim project-management rules, NEA's 2024 grid-connection notice, Inner Mongolia's November 2025 notice, the 2026 capacity-price notice and Q&A, and the GB/T 36549-2018 record (text not read). It quotes no project augmentation plan, EPC or LTSA clause, utility-scale warranty SOH table, provincial capacity-test rule or Chinese cost figure; the US figures are modelling assumptions, not Chinese project data. Not engineering, warranty or investment advice.