How to Assess BESS Levelised Cost of Storage: The PNNL Formula, What Lazard's 2025 Ranges Assume, NREL's Cost Projections, and the Chinese Utilisation Statistics, Targets and Price Mechanisms to Read Alongside Them

Short answer. Levelised cost of storage (LCOS) is the price per unit of discharged energy at which a storage project recovers all its costs — capital, financing, taxes, O&M, replacements and the electricity it buys to charge. CNWattHub reading of the formula: how much energy a plant actually discharges each year and what it pays to charge are important operating inputs to localise, and capital cost, financing, O&M, round-trip efficiency and replacement assumptions need localising too; this guide does not rank them. Published US figures — Lazard's $115–254/MWh for a 100 MW/4-hour system in 2025, PNNL's 2021 values, NREL's projections — are US reference-project numbers; Lazard's cases and NREL's ATB assume one cycle a day. For a Chinese project the same formula applies with the project's own capex, financing, annual discharge and charging price; the Chinese documents supply the context around those inputs — a preliminary national average of 1,195 equivalent utilisation hours in 2025, a 180 GW national target for 2027, a capacity-price formula for grid-side storage, and a rule that storage configuration may no longer be a precondition for approving, connecting or exporting new new-energy projects. This guide gives the formula, the US assumptions to borrow, and the Chinese context to read alongside them. No Chinese cost figure is quoted because no retrieved official source states one.

1. The formula

PNNL's 2022 Energy Storage Grand Challenge assessment defines LCOS as the price a unit of energy output from the storage asset would need to be sold at to cover all project costs including taxes, financing, O&M and others, and explains that moving from LCOE to LCOS lets storage-specific items — charging cost and augmentation or replacement of the storage block and power equipment — be included. Its equation is LCOS = [(FCR × CAPEX_PV) + fixed O&M] ÷ annual hours discharged + electricity charging cost, where the fixed charge rate is a percentage, present-value capex is in $/kW, fixed O&M in $/kW-year, and the charging cost in $/kWh discharged equals the purchased-energy price divided by round-trip efficiency. PNNL's base case used 2.80% inflation, 8.00% nominal interest, 13.00% nominal cost of equity, 50% debt, a 24.873% combined tax rate, a real WACC of 6.52%, a $0.03/kWh charging cost and a 20-year economic life for all technologies.

CNWattHub reading of the equation: the first term is an annualised cost per kW spread over the kW-hours actually discharged, so halving annual discharge hours doubles it; the second term is simply the charging tariff grossed up for losses, so a 90% round-trip efficiency turns a 0.30 元/kWh charging price into 0.33 元/kWh of cost per discharged kWh. Annual discharge and charging price are important operating inputs for a Chinese project to localise; capital cost, financing terms, O&M, round-trip efficiency and replacement or augmentation assumptions need the same treatment, and without a sensitivity calculation this guide does not rank one input above another.

2. What the US reference numbers assume

PNNL (2021 values). Battery LCOS at 1–10 MW fell roughly between $0.20 and $0.40/kWh for durations up to 10 hours, was lowest at about 10 hours, and rose steeply at longer durations because fewer cycles per day are possible; at short durations cycle-life limits force frequent augmentation or replacement, while at 10 hours all battery systems except lead-acid coalesced around $0.20/kWh. These are 2021 values and are quoted for the shape of the curve, not the level.

Lazard (June 2025). Lazard's LCOS v10.0 gives unsubsidised ranges of $129–277/MWh for a 100 MW/2-hour and $115–254/MWh for a 100 MW/4-hour utility-scale standalone system, and $319–506/MWh for a 1 MW/2-hour C&I system; it assumes 20% debt at 8% and 80% equity at 12%, excludes augmentation from capital costs, and prices charging at weighted average hourly wholesale prices. The 100 MW/400 MWh case assumes one 90%-depth-of-discharge cycle per day on 350 operating days, a 20-year life, 126,000 MWh of annual output, DC capital cost of $107–232/kWh, AC capital cost of $25–67/kW, EPC of $28–116/kWh, storage O&M of $3.0–8.0/kWh, a $27/MWh charging cost escalating at 1.97%, and round-trip efficiency of 92% to 86%. Lazard itself says the analysis is a snapshot, not a forecasting tool, excludes among other things recent tariff-related cost impacts, network upgrades and congestion or curtailment costs, and does not address social and environmental externalities or long-term residual and societal consequences that are difficult to measure, giving resource extraction, end-of-life disposal and lithium-ion safety hazards as examples. CNWattHub arithmetic on the tagged assumptions: 100 MW × 4 h × 90% × 350 days = 126,000 MWh a year, which is 126,000 ÷ (100 MW × 8,760 h) ≈ 14.4% of the system's rated-power hours.

NREL (2025). NREL's 2025 projections give 4-hour lithium-ion capital costs of $147, $243 and $339/kWh in 2035 and $108, $178 and $307/kWh in 2050 (low, mid, high; 2024 dollars), and note that a $/kWh cost converts to $/kW by multiplying by the 4-hour duration. NREL's 2025 Annual Technology Baseline expresses utility-scale battery cost as battery pack cost ($/kWh) × duration + balance-of-system cost ($/kW), assumes roughly one cycle per day so that a 4-hour device has an expected capacity factor of 16.7% and a 2-hour device 8.3%, and notes that degradation depends on usage and systems may need replacement within the analysis period.

3. The Chinese context

Utilisation. At the NEA press conference of 30 January 2026, new-type storage in operation at end-2025 was put at 136 GW / 351 GWh, average duration 2.58 hours (up 0.30 h in the year), independent storage 51.2% of cumulative capacity and lithium-ion 96.1%; the 2025 national equivalent utilisation of new-type storage was 1,195 hours (preliminary; up nearly 300 hours on 2024), 1,175 hours in the State Grid area and 1,294 hours in the China Southern Grid area. CNWattHub note: the Chinese 'equivalent utilisation hours' statistic and Lazard's modelled annual output are defined differently and are not converted into each other here; for a Chinese project, take the annual discharge input from the project's own location and dispatch regime rather than from a one-cycle-a-day assumption.

Dispatch and targets. The 2025–2027 storage action plan (发改能源〔2025〕1144号, dated 27 August 2025) targets 180 GW or more of new-type storage by 2027 with about 2,500亿元 of direct project investment; in regions without continuous spot operation it asks dispatchers to optimise the merit order of regulating resources and increase storage dispatch to reduce frequent deep cycling and start-stop of coal units, while regions with continuous spot operation dispatch by market results; and it calls for improving capacity price mechanisms for storage, establishing reliable-capacity compensation in an orderly way and forming reasonable charging and discharging prices through markets. In the NEA's Q&A, the official put national new-type storage at 73.76 GW at end-2024 and said more than 100 GW is expected to be added within three years to reach 180 GW or more by end-2027. The NEA's 2024 notice on grid connection and dispatch distinguishes dispatcher-called storage (independently metered, run by market results or dispatch instructions) from plant-own-use storage, and requires market-participating storage to be run by clearing results and others by dispatch instruction.

Revenue mechanisms. 发改价格〔2026〕114号 lets provinces grant a capacity price to grid-side independent new-type storage that serves system security and has not taken part in paired-storage arrangements, based on the local coal capacity price scaled by a peak-capability ratio (full-power continuous discharge duration ÷ longest annual net-load peak duration, capped at 1) and adjusted for market progress and system need, under list management. And 发改价格〔2025〕136号 prohibits making storage configuration a precondition for the approval, grid connection or export of new new-energy projects. CNWattHub reading: a capacity payment enters the LCOS comparison as revenue, not cost; the practical effect is to lower the energy price a station must earn per discharged kWh, and the formula rewards longer duration up to the province's net-load peak length.

4. How to run the numbers for a Chinese project

CNWattHub editorial method, not a sourced rule. Keep PNNL's structure, replace the inputs:

1. Capex per kW: use the project's own quotation; the US DC/AC/EPC split in Lazard's table shows which items to ask for.
2. Fixed charge rate: build it from the project's own debt share, interest rate, cost of equity, tax rate and economic life; PNNL's table shows the parameters needed.
3. Annual discharge hours: derive them from the project's own expected annual discharge under its dispatch regime (market results where spot runs continuously; dispatch instructions elsewhere) divided by its rated power, rather than assuming one cycle a day; the national equivalent-utilisation statistic is context, not an input.
4. Charging cost: the charging tariff divided by round-trip efficiency; Lazard's efficiency band of 86%–92% is a reasonable sensitivity range.
5. Augmentation: Lazard excludes it from capex; PNNL includes replacements — decide which convention you are using before comparing.
6. Revenue side: add any provincial capacity price under 114号 and ancillary-service income; LCOS alone does not tell you whether the project pays.

Reading table

Scope limits

US figures are US reference-project values from one national laboratory assessment (2021 values), one investment-bank analysis (Lazard's own estimates, June 2025) and one laboratory projection; they are not Chinese costs. No Chinese capex, LCOS, spot spread or provincial capacity price is quoted. Not engineering, financial or investment advice.