🔌 Grid-Scale Storage: Batteries, Pumped Hydro, and Beyond

What you’ll learn

1. The Grid Has No WarehouseEstablish that the grid has no inventory, and that frequency is the signal that stands in for one.Every other commodity industry has a shed; electricity has never had one, and has operated on instantaneous balance since 1882. The frequency is the inventory reading: generators are physically spinning, so a shortfall is covered in the first moments out of their momentum and they slow, dragging frequency below 50 or 60 Hz. Operators defend that number in layers — inertia in milliseconds, governors in seconds, dispatch in minutes — each buying time for the next. That layered structure is the course's argument: storage is not a product but an answer to 'for how long?'
2. Seconds: The Storage We Never Meant to BuildIntroduce inertia as the grid's accidental, shortest-duration storage — and explain why it is quietly being retired.Rotating generator mass meets the definition of storage exactly: energy held physically and released automatically. It is the fastest storage there is and among the smallest — a grid's whole rotating kinetic energy would run it for seconds. What it provides is not energy but time: it limits RoCoF, the rate of change of frequency, so that governors and reserves can act before protection thresholds are crossed. Solar has no rotating mass and wind is decoupled by power electronics, so inertia is falling as a side effect of the energy transition — which is why South Australia now installs synchronous condensers, and why fast frequency response from batteries became the first service grid batteries were paid for.
3. Every Storage Question Is a Question About TimeGive the reader the two tools that make every storage argument legible: duration, and the power/energy distinction.There is no 'storage' category — there are five duration problems sharing one word: milliseconds (inertia, solved for a century and now being un-solved), seconds (fast frequency response), minutes (regulation), hours (the evening shift, being solved fast by cost), days (expensive and awkward), and seasons (unsolved). Underneath sits one distinction: power in MW is the tap, energy in MWh is the tub, and they are bought separately — duration is one divided by the other. Different technologies have wildly different tap-to-tub cost ratios, and that ratio decides which durations each can serve.
4. Pumped Hydro: 95% of It, and Nobody KnowsCorrect the field's biggest blind spot: pumped hydro is essentially all of the world's storage, and geography is its only real constraint.Wikipedia records that as of 2020 pumped storage 'accounted for around 95% of all active storage installations worldwide, with a total installed throughput capacity of over 181 GW' and over 1.6 TWh; as of 2023 that 181 GW still exceeded all batteries combined at about 88 GW. The IHA figure cited there is 200 GW and 9,000 GWh — about 45 hours' worth, a ratio that turns out to be the point of the duration chapter. It works because it has no chemistry to degrade and scales to 3,600 MW (Fengning), 3,003 MW (Bath County) and 2,400 MW (Guangdong). It is limited because gravity is weak: roughly 1 gigalitre and 10 hectares per GWh, which demands a specific geography and a decade of permitting.
5. What a Battery Farm Actually SellsShow through Hornsdale what grid batteries are actually paid for — speed, not stored energy.Hornsdale was commissioned on 1 December 2017 at 100 MW / 129 MWh — barely an hour — and expanded in September 2020 to 150 MW / 193.5 MWh. It was never a solar-shifting machine. Its money came from frequency control, where a battery beats a turbine decisively: milliseconds rather than seconds, and equally easy in both directions. Wikipedia records it reduced grid service costs to AEMO by 90% in its first four months, saved an estimated A$40 million by end-2018 and $116 million in 2019, with FCAS costs falling from $470/MWh to $40/MWh. When Loy Yang's 560 MW unit tripped in December 2017, Hornsdale did not replace the loss — it arrested the frequency's fall and bought time.
6. Round-Trip Efficiency, HonestlyPut round-trip efficiency in its place: the most quoted and least decisive number in storage.Every store is a leaky bucket, and the losses are mostly heat — resistance and inverters in a battery, friction and hydraulic losses in a pumped-hydro plant. Wikipedia's quoted ranges: pumped hydro 75–85% (its own PSH article says 70–80%), vanadium flow 60–75%, advanced CAES 60–80%, LAES 50–70%, green hydrogen about 41%; Huntorf measured 42% and McIntosh about 27%. But efficiency is a fraction of the input and says nothing about what the input cost — a plant charging at negative prices is being paid to take it, so losing a quarter of something free is not a problem. What decides things is cost per useful cycle over the asset's life, where a 75% reservoir with no cycle count beats a 90% battery rated at 400–1,200 cycles.
7. The Duration WallExplain the duration wall as arithmetic, not technological failure — and reframe every technology as a tap-to-tub ratio.Going from four hours to eight is cheaper than double, because the tap — inverters, land, connection, permits — is already bought. But the fixed base stops helping almost immediately: at a hundred hours you are buying 25× the cells and the cost is essentially linear, while those cells sit idle almost all year, and a battery earns by cycling daily. Hence the four-hour default. Reframed as ratios: lithium-ion has a cheap tap and expensive tub; pumped hydro an expensive tap and near-free tub (hence 200 GW against 9,000 GWh); flow batteries buy the separation explicitly at the cost of 60–75% efficiency; hydrogen has a near-free tub and a brutal tap at about 41%.
8. The Exotics, and How to Be Fair to ThemAssess CAES, molten salt and gravity storage fairly — and build the habit of asking whether a number is measured or specified.Huntorf (1978, 290 MW, 580 MWh, 42%) and McIntosh (1991, 110 MW for 26 hours, 2,860 MWh, about 27%) are the entire utility-scale CAES fleet. The poor efficiency is thermodynamics — compression heat is lost to the rock and expansion chills the machinery, so both plants reheat by burning gas, which is why McIntosh's honest accounting includes 1.2 MJ of gas per MJ delivered. Adiabatic CAES would fix this at a quoted 60–80%, but that is a specification and remains unbuilt. Molten salt at a CSP plant is genuinely elegant because it never converts to electricity in the first place. Gravity towers are real physics with the wrong shape: their tub — supposed to be the cheap part — is the expensive part.
9. The One Nobody Has SolvedBe honest that seasonal storage is unsolved, give hydrogen a fair hearing, and hand the reader a checklist.Seasonal storage needs enormous energy capacity, cycled once or twice a year, with negligible self-discharge and near-zero cost per unit stored — close to the opposite of every technology in this course, all of which earn by cycling. Hydrogen is the only right-shaped candidate: a salt cavern is nearly free and does not self-discharge. But Wikipedia gives green hydrogen roughly 41% round-trip, and that is the compound of three conversions in series, not a defect awaiting an engineer — so you must generate about 2.5 summer units to deliver one winter unit, using machinery that runs a few weeks a year. The actual plan is overbuilding, interconnection, a rarely-run gas fleet, and demand shifting. The reader's checklist: how many hours; what does the tub cost and what does the tap cost; is that number measured or specified; how many have been built; and what is it really selling?

Questions this course answers

Grounded in trusted sources

• Wikipedia — Grid energy storage: https://en.wikipedia.org/wiki/Grid_energy_storage
• Wikipedia — Pumped-storage hydroelectricity: https://en.wikipedia.org/wiki/Pumped-storage_hydroelectricity
• Wikipedia — Hornsdale Power Reserve: https://en.wikipedia.org/wiki/Hornsdale_Power_Reserve
• Wikipedia — Compressed-air energy storage: https://en.wikipedia.org/wiki/Compressed-air_energy_storage
• Wikipedia — Lithium-ion battery; Flow battery; Gravity battery
• Wikipedia — Inertia (electrical grid); Rate of change of frequency; Synchronous condenser; Utility frequency
• Wikipedia — Thermal energy storage; Concentrated solar power; Seasonal thermal energy storage
• International Hydropower Association, via Wikipedia — worldwide pumped storage of 200 GW power and 9,000 GWh energy

Every Wunder lesson is built from real, reputable sources — never invented.

Related Science courses