How Do Grid Batteries Store Electricity?

The warehouse the grid never had

Junior level — plain language, no maths

Electricity has always had one embarrassing problem: it has to be used the instant it is made. There is no shed at the back of the power station with yesterday's electricity in it. For a century that meant every power plant had to chase demand up and down, minute by minute, all day.

A grid battery is the shed. It looks like a row of shipping containers in a field, and inside are the same lithium cells that are in your phone, just millions of them. A typical one is rated 100 megawatts and 200 megawatt-hours, and those two numbers mean different things: the first is how fast it can push, the second is how much it holds. Full power for two hours. It is exactly like a bucket - how wide the tap is, and how big the bucket is.

What it does with that is simple arithmetic. Around midday, when every solar panel in the country is producing at once, electricity is cheap and sometimes worth less than nothing. In the evening, when everyone comes home and the sun has gone, it is expensive. So the battery fills up in the middle of the day and empties itself into the evening rush. It is buying low and selling high, except the thing it is trading is the electricity in your walls.

Two catches. The first is that you never get back everything you put in - about ten per cent is lost as heat on the way in and out, so ten stored units come back as nine. The second is time. Four hours of storage is easy and cheap; four days is not, and four months is a completely different problem that batteries cannot solve at any sensible price. That gap is where pumped hydro, hydrogen and a lot of unfinished argument live.

Power, energy and the spread you have to beat

Student level — the core equations

A storage asset is described by two independent numbers and their ratio. Power \(P\) in MW is what the inverter and cells can deliver instantaneously; energy \(E\) in MWh is what the chemistry holds. Their ratio \(E/P\) is the duration, and it is the single most important design choice, because power and energy are bought separately: the inverter and connection cost per MW, the cells cost per MWh. Most grid batteries land at two to four hours because that is the shape of an evening peak.

Then there is the tax on every round trip. Round-trip efficiency \(\eta = E_{\text{out}}/E_{\text{in}}\) runs 85-92% for lithium at grid scale, the losses being ohmic heating, electrochemical overpotentials, inverter conversion and the air conditioning that keeps the cells at 25 °C. That efficiency sets a hard economic threshold: arbitrage only pays if the price at which you sell exceeds the price at which you bought by a factor of at least \(1/\eta\). At 85% you need an 18% spread just to break even before you have paid for the hardware.

Which explains the daily pattern the market produces on its own. Charge in the cheapest hours - usually the solar-flooded middle of the day, sometimes overnight - and discharge into the highest-priced hours, typically the evening ramp. Making the battery bigger does not scale the revenue linearly, because each extra hour of duration forces you into the next-cheapest hour to buy and the next-least-expensive hour to sell. The marginal value of duration falls, quickly, and it falls faster as more storage arrives and flattens the spreads that made it profitable.

Finally, batteries wear out by being used. Cycle life depends strongly on depth of discharge and on the C-rate; running from 100% to 0% every day ages a cell far faster than running 80% to 20% twice. Modern warranties are written in megawatt-hours of throughput rather than in years, which tells you exactly what the manufacturer thinks the limiting variable is.

Where the losses live, how arbitrage eats itself, and the duration wall

Scholar level — full mathematical depth

01Three products, not one

Storage is habitually discussed as if it were a single commodity, but a grid buys at least three distinct things from it, on different timescales and with different physics. Sub-second: inverters can inject or absorb real power almost instantly, which is frequency containment and is limited by power rating alone. Hours: energy arbitrage and peak shifting, limited by \(E\). Days to seasons: covering wind droughts, limited by \(E\) again but at a cost per kWh that lithium cannot reach. The cost structure explains the split - power capacity is priced per kW and energy capacity per kWh, and the ratio between them decides which technology wins at which duration.

02Accounting for the round trip

The DC round trip of a good lithium cell is above 95%: ohmic loss through \(I^2R\), charge-transfer overpotential from Butler-Volmer kinetics, and a small hysteresis in the open-circuit voltage. What turns that into an AC round trip of 86-90% is everything else - inverter and transformer conversion at each direction, auxiliary loads, and thermal management, since cycle life depends steeply on cell temperature. Because ohmic losses scale as \(I^2\), efficiency falls with C-rate: the same asset is more efficient shifting energy over four hours than over one.

03Arbitrage as an optimisation, and its self-destruction

With perfect foresight, optimal dispatch is a linear program: maximise \(\sum_t p_t(d_t - c_t)\) subject to power limits, an energy balance \(\text{SoC}_{t+1} = \text{SoC}_t + \eta_c c_t - d_t/\eta_d\), and SoC bounds. The structure of the solution is simple - buy in the cheapest \(n\) hours, sell in the most expensive \(m\) - and its value is the area between the sorted price duration curve's tails. Two consequences follow. Marginal duration has sharply diminishing returns, because each added hour reaches into a less extreme part of the curve. And as installed storage grows, it flattens exactly the tails it feeds on: the spread collapses, and the last battery built earns a fraction of what the first did.

04Degradation, briefly and honestly

Two mechanisms dominate. Calendar ageing comes mainly from growth of the solid-electrolyte interphase, roughly \(\propto\sqrt{t}\) and strongly Arrhenius in temperature, which is why a hot cell at high state of charge ages fastest even sitting still. Cycle ageing adds mechanical fatigue from lattice expansion and, at low temperature or high charge rate, lithium plating on the anode - irreversible, capacity-destroying and a safety concern. This is why operators cycle within a window rather than end to end, keep cells near 25 °C, and accept a lower charge rate in winter. It also explains warranties written in throughput: the physics counts energy pushed through, not calendar years.

05The duration wall

Lithium's energy cost sits near 100-150 $/kWh installed, so a 100-hour system would cost ten times a ten-hour one. Pumped hydro's marginal energy cost is the reservoir - single-digit dollars per kWh - which is precisely why it dominates long-duration storage despite needing geography. Beyond a day or two, the economics stop being about efficiency at all: hydrogen at 35% round trip is a poor engine but a cheap tank, and for a store cycled a handful of times a year, capital per kWh dwarfs efficiency. The right question at every duration is not what is most efficient but what is cheapest per unit of energy delivered, given how rarely it will be used.

06What it does to the system

Storage's contribution to reliability - its capacity credit - starts near 100% for the first few gigawatts and falls as penetration grows, because a fleet of four-hour batteries cannot cover a six-hour peak once it has flattened the first four. The literature on very high renewable shares converges on the same shape: a large amount of short-duration storage does most of the work cheaply, and then a small, awkward, expensive residue of multi-day energy has to come from somewhere else - long-duration storage, overbuild and curtailment, interconnection, or a little firm generation. Which of those wins is an economics question, and the answer moves every time a technology's cost curve does.