Advanced adiabatic compressed air energy storage (A-CAES): round-trip efficiency, duration and LCOS

Lower limit in guards.

Unit multiplier in the CRF, guards and Customized switch.

Protects divisions.

Percent to fraction.

Fraction to percent.

MW to kW.

MWh to kWh.

Hours in a non-leap year, the clock that caps the cycles.

Base of the 'per decade' cost drop (a tenfold size increase).

Identifier of the low end of the Wolf and Budt 2014 low-temperature A-CAES range (52 %).

Identifier of the high end of the Wolf and Budt 2014 low-temperature A-CAES range (60 %), the default.

Identifier of the DOE 2023 upper bound for adiabatic CAES (70 %).

Low end of the published round-trip efficiency of low-temperature adiabatic CAES (thermal store below about 200 C, hot water or oil), the closest published match to a hot-water A-CAES plant of the Hydrostor type.

High end of the Wolf and Budt range and the default case. Hydrostor prints no round-trip efficiency for Silver City or Willow Rock; a 60-65 % figure quoted in the press is not used.

Upper bound quoted by DOE for adiabatic CAES; an optimistic case, not a measured plant value.

Market range upper bound only; no plant measurement cited

Diabatic CAES baseline of PNNL (natural gas reheating). Shown as a reference only: a diabatic plant burns gas at discharge, so its 52 % is not comparable one-to-one with an adiabatic plant.

Market range DOE 2023: diabatic 46-54 %

Ceiling of one full equivalent cycle a day used by PNNL for the LCOS. The model also applies the clock: charge plus discharge time must fit in 24 h.

Compressor electrical input / rated discharge power. Set at 1 because the cost basis prices compressor and turbine on the same kW; Hydrostor does not publish the compressor rating. With RTE below 100 %, recharging then takes duration / RTE, so above about 8 to 10 h a daily cycle no longer fits in 24 h.

Market range no published compressor rating for Silver City or Willow Rock

Power-block cost at the 100 MW reference: turbomachinery, balance of plant and EPC. DOE 2030 estimate for a diabatic CAES, transposed to the adiabatic plant (evidence medium); the heat exchangers and thermal store sit in the storage block, calibrated on Silver City.

Market range Howitt 2021 power island 890 $/kW at 40 MW, 725 $/kW at 1,000 MW (narrower scope)

Converts the PNNL/DOE cost basis to EUR. No dollar year is printed for it; PNNL-33283 carries its CAES estimates from the 2020 report, so the 2020 rate is used (the 2021 rate, 1.1827, would lower the power block by 3.4 %). No inflation correction between this basis and the 2022 AUD Silver City cost: the model mixes the two and says so.

Converts the Silver City cost (AUD, project start August 2022) to EUR.

Converts the 2025 Willow Rock loan figure to EUR.

Size exponent of the CAPEX: cost per kW multiplied by 0.92 for each tenfold increase in size, applied by PNNL to both the power block (in MW) and the storage block (in MWh). The model applies it to both blocks; at fixed duration the total cost per kW then follows (P / 100 MW)^(0.963788 - 1). PNNL fits this on 100 and 1,000 MW; below 100 MW it is an extrapolation, flagged by an output.

Market range Howitt 2021 power-island points imply about 13.6 % per decade over 40-1,000 MW (computed in an output)

Reference size of the scale law, the size of the cited 1,153 $/kW power-block cost.

Total project cost of Silver City (Broken Hill), the single cost anchor of the storage block.

Rated power of Silver City, catalogue product and calibration point.

Storage energy of Silver City: the reference size of the storage-block scale law, because the storage block is calibrated at this size.

Width of the band around the 1,600 MWh calibration point inside which the storage block is read as anchored. Outside it the 8 %/decade slope is transposed from the PNNL salt-cavern scaling, and the output flag says so.

Rated power of Willow Rock, catalogue product, used only for the convergence check.

Storage energy of Willow Rock (8 h), used only for the convergence check.

Loan principal only, before equity: a lower bound of the project cost, from the press, not from a cost breakdown.

Market range conditional commitment, not a final cost

Independent adiabatic CAES estimate (ADELE, Germany), 2020 USD, used as a benchmark only. PNNL treats it as a high value and excludes it from its average.

Low end of the ADELE power range.

High end of the ADELE power range.

Duration of the ADELE estimate.

Mid-point of a published range (ADELE 300-500 MW).

Fixed O&M at 100 MW. Converted at the 2020 rate like the rest of the PNNL/DOE basis, without inflation adjustment.

Fixed O&M at 1,000 MW. The model interpolates log-log between the two PNNL points; the exponent is deduced from them, not printed.

Size of the second PNNL fixed O&M point.

Physical service life of a CAES plant (caverns and turbomachinery). Exported to project mode as the asset lifetime; the standalone LCOS annualises over the 20-year economic life instead, as PNNL does.

Salt-cavern cost only, without thermal store or a mined hard-rock cavern. Shown next to the storage block, not used as its default: the A-CAES storage block (mined cavern, hydrostatic compensation, thermal store, development) is about 24 times higher.

Adiabatic power island at 40 MW, the only cost point below 100 MW. Narrower scope than the 1,153 $/kW power block (no full BOP and EPC).

Adiabatic power island at 1,000 MW. With the 40 MW point it gives an independent slope check of the 8 %/decade law.

Size of the small Howitt power-island point.

Published LCOS of diabatic CAES at 1,000 MW and 10 h, a benchmark only: different technology (gas-fired reheating, salt cavern) and different scope.

Low end of the DOE-reported range of CAES stored-electricity costs. The cost year of the underlying studies is not given; converted at the 2020 rate for an indicative comparison only.

High end of the DOE-reported range; indicative comparison only, cost year not given.

PNNL charging price, converted to EUR/MWh in an output; it sets the default of the charging price input (26.3 EUR/MWh).

Floor of the scaled CAPEX clamp. The model gives 1,490 to 2,850 EUR/kW over 50-1,000 MW x 4-12 h, so the floor never binds inside the sliders.

Ceiling of the scaled CAPEX clamp; never binds inside the sliders.

Default real WACC of the standalone LCOS. Real, because no cost line is escalated.

Default annualisation period. Shorter than the 60-year calendar life; PNNL uses 20 years for every technology so that LCOS values compare.

Default charging price, a US planning value, not a European market price. The slider stops at 50 EUR/MWh: above it the LCOS leaves the sourced range.

Default request: one full cycle a day, the PNNL ceiling.

Not used: the model defines the storage energy as the energy actually delivered per cycle (power x duration) and counts full equivalent cycles, so no depth-of-discharge factor applies. A cavern held at constant pressure by hydrostatic compensation delivers its working volume.

Not used: the standalone LCOS carries no interest during construction (simplification of PNNL Eq. 1). Project mode holds the schedule and the financing.

Storage block of A-CAES: mined hard-rock cavern with hydrostatic compensation, thermal store, development and everything in the Silver City total project cost that the power block does not cover. Computed in the model from the Silver City figures (node storage_block_capex_calibrated_eur_per_kwh), so the CAPEX at 200 MW / 8 h equals Silver City by construction.

Market range PNNL salt cavern alone 2.7-13.8 $/kWh (not comparable: no heat store, no mined cavern)

Origin of the size exponent 0.963788 = 1 + log10(0.92).

Benchmark only: diabatic CAES (gas-fired reheating, salt cavern) at 1,000 MW / 10 h, a cheaper technology than a mined-cavern A-CAES. Same simplified LCOS family (PNNL Eq. 1).

Benchmark only: low end of the DOE-reported range. The cost year of the underlying studies is not stated, so the comparison is indicative.

Benchmark only: high end of the DOE-reported range; cost year not stated, indicative.

Benchmark of the CAPEX per kW, for a 10 h plant: compare with the model at 400 MW / 10 h (output), not at the current configuration. Scope: total plant estimate.

Benchmark of the CAPEX per kW for an 8 h plant: loan principal only, before equity, so a lower bound of the project cost; 2025 USD.

LCOS of the PNNL-33283 type (Eq. 1), simplified: (CRF x CAPEX + fixed O&M) / energy delivered + charging price / RTE. No interest during construction, no tax, no revenue, no replacement within the 20-year economic life. Constant money, so the WACC is real.

The round-trip efficiency is a sourced system-level value, not derived from a compressor/expander chain: Hydrostor publishes no efficiency and no component split, so the model does not invent one. The PNNL and DOE test values (52 %) are for diabatic CAES and are shown as references only.

Charge time = duration / (RTE x charge power ratio). With the compressor rated at the turbine power, a full cycle fits in 24 h up to about 8.2 h (RTE 52 %) to 9.9 h (RTE 70 %); beyond, the achieved cycles fall below 365 and the LCOS rises. PNNL's own Eq. 2 applies the same clock.

Two parts of the slider range sit outside the sourced domain and are flagged: below 100 MW (the PNNL scale law rests on 100 and 1,000 MW; only the 40 MW Howitt power-island point touches that range), and storage energies more than a factor 2 from 1,600 MWh (the storage block rests on one calibration point).

Project mode chains the charging electricity (inlet, priced by the project) and the discharged electricity (main outlet). Fixed O&M is exported undiscounted; no variable O&M is booked, so charging is never paid twice. The CAPEX is a total project cost (battery_limit fully_financed), so project mode adds no buildup on top.