Compressed Air Energy Storage Model Spans 279 to 682 MW Output

A storage plant that can flex harder

A new modeling study offers a fix for one of compressed air energy storage's oldest problems: a plant can only take in and send out power within a fairly narrow band. The design, called a multi-temperature graded compressed air energy storage system (MTG-CAES), stores heat at three separate temperature levels. By choosing which heat source warms the air before it enters the expanders, operators can change how much power the plant produces.1 The authors, Gen Li, Huan Guo, Yujie Xu, Haisheng Chen and colleagues, published the work in the journal Carbon Neutrality.1 The paper's online listing is dated June 22, 2026.2

The headline numbers are large. Under baseline assumptions, the modeled plant can output anywhere from 279.17 MW to 682.36 MW and can absorb anywhere from 210.44 MW to 511.32 MW.2 The authors report an exergy efficiency of 70.47% and a round-trip efficiency of 72.34%.2 They also stress that these are theoretical upper limits under idealized conditions, not measurements from a working plant.12 That caveat matters for everything that follows.

How the three temperature levels work

Conventional compressed air storage is valued for its large capacity, long storage duration and long working life. Its limited power range, though, makes it awkward to pair with wind and solar output that changes from minute to minute.1 If renewable supply falls outside the range a plant can absorb, operators have to run equipment outside its design point or curtail generation.1

During charging, the MTG-CAES design uses wind and solar electricity to run a two-stage adiabatic compressor and then a single-stage isothermal compressor. Together they push air into a tank at 10 megapascals, and thermal oil captures the compression heat.1 The design also uses fast wind fluctuations that the compressors cannot follow. Instead of throwing that energy away, resistive heaters turn it into heat, which is stored alongside heat from solar collectors.2

During discharging, the heat is used in three steps. The hottest oil, heated by the sun, warms the air for the first expansion stage and leaves at about 244°C. That oil then mixes with stored compression heat of a similar temperature to drive the second stage, and leaves at about 168°C to supply the third.1 Valves let operators switch between the three levels in any order, so output can follow grid demand.1 At design conditions, the three levels deliver 568.63 MW, 484.50 MW and 398.82 MW, with exergy efficiencies of 88.70%, 75.58% and 62.21%.1 Combining those three steps with flow control between 0.7 and 1.2 times the design rate is what produces the wide headline range.2

The idea of splitting heat handling into three parts is not new in itself. A US patent issued in April 2025 describes a compressed air system with three points where heat is taken out during compression and three matching points where it is put back during expansion.5 The difference is the goal. That patent aims to keep the three heat-exchange stages at roughly equal temperatures so they can share common handling equipment.5 The new study deliberately runs its three stages at different temperatures so that each produces a different power level.1

The trade-offs in the fine print

The sensitivity analysis shows the design is a balancing act. A higher compressor pressure ratio raises the temperature of recovered heat and improves efficiency, but it narrows the power range. Very high ratios also make compressors harder and more expensive to build.1 A hotter solar input works the other way: it widens the power range but lowers round-trip efficiency, and thermal oil does not handle long periods at high temperature well.1 Sending more of the heating work to surplus wind broadens the charging range but costs efficiency, because that energy goes from electricity to heat and back to electricity.1

The model also leaves out several real-world losses. It is a steady-state MATLAB simulation that ignores pipe pressure losses, heat loss and air leakage.1 The authors identify the compression-expansion equipment and the heat exchange system as the largest sources of exergy loss.1 On cost, they estimate the extra heat tanks would add about 3% to equipment spending, since a single tank is typically about 1% of the total. They note that whether the wider range pays off depends on market rules and policy.1

How the 72% figure compares with real plants

The best way to judge the study is to compare its idealized figures with plants that are actually running or close to construction.

The strongest benchmark is in China. The Huai'an salt cavern project in Jiangsu province uses two 300 MW units for 600 MW of capacity and 2,400 MWh of storage, with conversion efficiency of roughly 71%.13 It stores heat in molten salt and pressurized hot water instead of burning fuel, and its first unit reached full load in December 2025.13 A second report confirms the plant is fully grid-connected and in commercial operation, not a demonstration.18

Hydrostor's advanced compressed air design is more conservative. The company's president has said it reaches around 60% to 65% round-trip efficiency.19 Traditional compressed air plants usually recover less than half the energy put in.12 Reference material puts the expected practical efficiency of adiabatic storage at about 70%.4

My reading: the study's 72.34% figure is plausible as a ceiling, because a large operating plant already reports about 71%. The real contribution is not efficiency. It is the claim that a single plant could deliver more than twice its minimum output while keeping efficiency close to what the best plants already achieve. That claim has not yet been tested in real equipment.

Why this matters for grid storage

The study comes as compressed air storage moves from concept to construction. Hydrostor's Willow Rock project in Kern County, California, is rated at 500 MW and 4,000 MWh, enough for eight hours of discharge. It signed a 50 MW offtake agreement with California Community Power and is expected online in 2030.1216 The California Energy Commission granted its final permit in December 2025, and the company planned to break ground in mid-2026.14 In Australia, Hydrostor's 200 MW, 1,600 MWh Silver City project won grid connection approval from AEMO and Transgrid.19 In the UK, EnergyPathways has started front-end engineering design on a 300 MW, 55.2 GWh offshore salt-cavern project in the East Irish Sea linked to offshore wind, with a final investment decision targeted for 2028.20

The MTG-CAES design speaks to a weakness these projects share. Most are built to deliver a steady amount of power for a long time. A plant that can also ramp across a wide band could compete in faster-moving markets such as peak shaving, load-following and ancillary services, which the authors identify as potential revenue.1 That matters for business reasons. Long-duration storage deployment grew 49% in 2025 to more than 15 GWh, but falling venture investment and cheaper lithium batteries are hurting the economics.12

The battery and minerals question

Compressed air is not replacing lithium-ion, and the coverage broadly agrees on that. Lithium-ion systems will stay central to storage, but they typically last only a few hours and rely on critical minerals.11 A 2024 BloombergNEF comparison put 8-hour compressed air at $293 per kWh against $304 for 4-hour lithium-ion arrays, with thermal storage cheaper still at $232.11 Commentary on the Huai'an plant argues that storage built from rock and steel can reduce pressure on the lithium, cobalt and nickel supply chains that electric vehicles also depend on, while complementing batteries rather than replacing them.18

The coverage diverges on emissions. The Huai'an project expects to avoid about 600,000 tonnes of CO₂ a year by displacing coal.13 Older compressed air plants burned fossil fuel during discharge, which a 2023 study described as environmentally unfriendly.3 Designs that store and reuse compression heat, including the new one, remove that combustion step entirely. That is not carbon capture in the technical sense, but it means there are no flue gases to capture in the first place.

What to watch next

The authors name transient behavior, coordinated control and economic performance as their next areas of work.1 Those are exactly the areas where a design built on valve switching and three oil circuits is most likely to run into trouble. Until a pilot plant shows how quickly the system can move between temperature levels and what that costs in wear and lost efficiency, MTG-CAES remains a promising blueprint. It is not yet a proven product.