Grid Scale Energy Storage Explained

Grid scale storage takes electricity off the grid when it is plentiful and puts it back when it is scarce. It does not create energy, and it loses some in the round trip, so it only makes sense when the value of electricity changes enough between the two moments to pay for the loss and the hardware. Almost all of the world’s storage capacity is either pumped hydro, which has been built since the 1920s, or lithium ion batteries, which have gone from rare to routine since about 2018.

The two dominant technologies

Pumped hydro uses two reservoirs at different heights. Cheap electricity pumps water uphill; when prices rise, the water falls back through turbines. It is mechanically simple, lasts for decades, and can be built at enormous scale, with sites storing many hours or even days of output. Its limits are geography and time: a project needs the right terrain, water rights, and usually a decade of permitting and construction. Existing schemes are visible in national statistics as a slightly odd part of the hydro category, and some countries report them separately from river and reservoir hydro.

Lithium ion batteries are containerised, fast to build and can sit almost anywhere, including inside cities and on existing substation land. A typical project reaches full output in under a second, runs for two to four hours, and can be installed in a year or two. Costs have fallen steeply, and battery fleets are now large enough to shift the shape of whole markets, most visibly in California, where storage charges on midday solar output and discharges into the evening peak. The effect shows up clearly in the hourly pattern on our CAISO grid page.

Other technologies exist and some are being built: compressed air in underground caverns, flow batteries whose energy and power can be sized independently, gravity and thermal systems, and hydrogen produced by electrolysis and burned or fed to fuel cells much later. None yet operates at anything like the scale of the first two.

Duration is the thing to ask about

Storage is described by two numbers that people often confuse. Power, in megawatts, is how fast it can charge or discharge. Energy, in megawatt hours, is how much it holds. Divide the second by the first and you get duration: a 100 MW, 400 MWh battery is a four hour battery.

Duration decides what a project can do.

• Seconds to minutes: frequency response and reserve, described in our note on ancillary services.
• One to four hours: shifting a solar peak into the evening, shaving the daily maximum, avoiding the most expensive hours of the day.
• Eight hours to a day: moving output from a windy night into the following working day, or from a weekend into a Monday.
• Days to weeks: covering a prolonged still, dark, cold spell. This is a different problem, and batteries are a poor answer to it.

Cost does not scale neatly with duration. Adding hours to a battery means adding cells, the most expensive part, so a twelve hour battery costs roughly three times a four hour one. Pumped hydro, compressed air and hydrogen behave differently, because adding hours mostly means a bigger reservoir or cavern rather than more machinery, which is why the discussion about long duration storage keeps returning to those options.

How storage earns money

Most projects stack several revenue streams.

Arbitrage. Buy low, sell high. In markets with hourly or half hourly prices, storage charges in the cheapest hours and discharges in the dearest. This is the intuitive case, and it is usually not enough on its own. Arbitrage also has a self limiting quality: every battery that charges at midday raises the midday price a little, and every battery that discharges in the evening lowers the evening price a little, so the spread that funded the project narrows as more projects are built.

Ancillary services. Frequency and reserve products pay for availability rather than energy, which suits a device that is mostly idle. These markets are small and saturate quickly.

Capacity. Where a capacity market or similar mechanism exists, storage can be paid to be available at times of system stress, usually with its contribution discounted according to duration, because a two hour battery is less useful than a gas plant in a long emergency.

Avoided network spending. A battery sited at a constrained substation can defer the cost of rebuilding a line. This value is real but hard to trade, so it usually appears through a network company owning or contracting the asset directly.

What storage cannot do

Three limits are worth stating plainly.

It cannot manufacture energy. If a system is short of generation over a month, storage makes the shortage worse, because of round trip losses. Storage solves timing problems, not supply problems.

It cannot ride out a long lull cheaply. A week of low wind and thick cloud across a large region, the condition German speakers call Dunkelflaute, requires energy measured in hundreds of gigawatt hours per day of shortfall. Batteries at that scale are prohibitively expensive today, which is why systems keep firm capacity, interconnection to other regions, or long duration options in reserve for those weeks.

It cannot move power across space. Storage shifts electricity in time within one place. Moving it between places is what transmission does, and a battery cannot substitute for a missing line except in the narrow sense of easing a local peak.

Why storage complicates the statistics

Storage sits awkwardly in generation data. Some statistical agencies report battery and pumped hydro output as generation, some net it off demand, and some report it as a separate category that can be negative when the fleet is charging. Country level statistics on this site follow the compiler’s own categories, and hourly grid data follows the operator’s. That is one more reason the two layers are kept apart rather than merged, as our methodology explains. When you compare a storage figure between two sources, check first whether you are looking at what went in, what came out, or the difference.