In this guide we look at energy storage systems, their main technologies and applications, with simple diagrams, comparison charts and a battery sizing calculator.
Quick answer: Energy storage systems take in electricity when it is plentiful and give it back when it is needed. The main technologies are pumped hydro, lithium-ion batteries, flow batteries, compressed air, flywheels, supercapacitors, thermal storage and hydrogen. They differ in power (MW), energy (MWh), duration, efficiency and speed. Their applications include shifting solar energy to the evening, fast frequency response, peak capacity and backup power. Today, lithium-ion batteries are most of the new storage, and pumped hydro still holds the most stored energy.
Why the grid needs energy storage
A power grid must match supply and demand every second. If generation falls short, the frequency drops. If there is too much, the frequency rises. For most of history, we matched the two by turning power plants up and down.
Solar and wind change this. Their output follows the weather, not the demand. Solar gives the most power at noon, but demand often peaks in the evening. So how do we move midday sunshine to seven o'clock at night?
The answer is an energy storage system (ESS): equipment that takes in electrical energy, holds it, and gives it back later. The figure below shows the main job on one sunny day.
Storage does four main jobs for the grid:
- Time-shifting. It charges when energy is cheap or in surplus and discharges when it is scarce. When this is done for price, it is called energy arbitrage.
- Frequency response. It injects or absorbs power within a fraction of a second when the frequency moves. This keeps the grid stable after a generator trips.
- Capacity. It is available at the peak hour, so fewer peaking plants are needed. It can also delay upgrades of lines and transformers.
- Backup. It keeps hospitals, data centres and microgrids running during an outage.
Two numbers describe any storage system. Power, in megawatts (MW), is how fast it can deliver energy. Energy, in megawatt-hours (MWh), is how much it holds. The ratio is the duration: a 100 MW, 400 MWh battery can run for 4 hours at full power.
Think of a water tank with a tap. The size of the tank is the energy, in MWh. The width of the tap is the power, in MW. A huge tank with a thin tap empties slowly. A small tank with a wide tap empties fast. We size both to suit the job.
Storage is growing quickly. The International Energy Agency (IEA) reports that 108 GW of batteries were added worldwide in 2025, 40% more than in 2024. The average new project now lasts about 3 hours, up from about 2 hours in 2023. By installed power, batteries (about 257 GW) have now passed pumped hydro (about 202 GW). But pumped hydro still stores far more energy, because its plants run for many more hours.
The main storage technologies
We can sort storage by the form the energy takes while it waits. It can be mechanical (height, pressure or spin), electrochemical (batteries), electrical (an electric field), thermal (heat) or chemical (a fuel such as hydrogen).
Pumped storage hydropower
Pumped storage hydropower uses two reservoirs at different heights. To charge, it pumps water up. To discharge, it lets the water fall through a turbine. The same reversible machine works as a pump-motor and as a turbine-generator.
The stored energy is E = ρ·g·V·h, where ρ is the water density, g is gravity, V is the volume and h is the head, the height difference. One cubic metre falling 400 m holds 1000 × 9.81 × 400 ≈ 3.9 MJ, or about 1.1 kWh. So big plants need millions of cubic metres. The largest plant today is Fengning in China: 3,600 MW from 12 units, with about 40 GWh of storage and a 425 m head. It became fully operational in 2024.
Pumped hydro is mature, long-lived and good for 6 to 24 hours of storage. Its weak points are geography, cost and build time. It needs suitable land, water and years of construction.
Lithium-ion batteries: LFP and NMC
A battery energy storage system (BESS) stores energy in electrochemical cells. Lithium ions move from one electrode to the other through an electrolyte as the cell charges and discharges. Lithium-ion is now the main choice for new storage, because it is efficient, fast, modular and falling in price. BloombergNEF reports that stationary storage packs fell to about $70/kWh in 2025.
Two lithium-ion chemistries matter most. LFP (lithium iron phosphate) has an iron-phosphate cathode. NMC (nickel manganese cobalt) has a cathode with those three metals. Grid storage has moved strongly to LFP.
| Property | LFP | NMC |
|---|---|---|
| Cell voltage | 3.2 V | 3.6–3.7 V |
| Energy per kg | Lower | Higher |
| Cycle life | Long; up to 10,000 cycles in some warranties | Shorter |
| Heat stability | Better | Lower; runaway starts sooner |
| Nickel, cobalt | None | Yes |
| 2025 pack price | $81/kWh | $128/kWh |
| Main use today | Grid and home storage | Electric vehicles |
Older chemistries still appear. Lead-acid is cheap and familiar, so it remains common in backup and UPS systems, but it has short cycle life. Nickel-cadmium works well in extreme temperatures but is now rare, because cadmium is toxic. Sodium-ion is a newer option with no lithium, and the IEA sees it as a lower-cost choice for stationary storage.
Flow batteries
A flow battery keeps its energy in two tanks of liquid electrolyte. Pumps push the liquids through a cell stack, where the reaction happens. The most common type is the vanadium redox flow battery. The big advantage is that power and energy are separate. The stack size sets the MW, and the tank size sets the MWh. So long durations of 4 to 12 hours are easy. Flow batteries also cycle many times with little wear. The downsides are lower efficiency, about 60 to 75%, and more equipment per kWh.
Compressed air energy storage (CAES)
Compressed air energy storage uses surplus power to compress air into a sealed underground space, often a salt cavern. To discharge, the air is released and expanded through a turbine. The two older plants, Huntorf in Germany and McIntosh in the USA, heat the air with natural gas before expansion. Their efficiencies are about 42% and 54%. Newer adiabatic plants store the heat of compression and reuse it, with no gas. The 300 MW, 1,500 MWh Yingcheng plant in China, in full operation since early 2025, reports about 70%.
Flywheels
A flywheel stores energy as spin. A heavy rotor turns at high speed inside a vacuum, on low-loss bearings. The energy is E = ½·I·ω², where I is the moment of inertia and ω is the speed in radians per second. A motor-generator on the shaft speeds it up to charge and slows it down to discharge.

Flywheels respond in milliseconds and survive a huge number of cycles. But they usually run for seconds to about 15 minutes, and they slowly lose energy to friction while they wait. A well-known plant at Stephentown, New York, uses 200 flywheels to give 20 MW and 5 MWh for frequency regulation.
Supercapacitors
A supercapacitor stores energy in an electric field. Charges gather at the surface of porous carbon electrodes, with no chemical reaction. The energy is E = ½·C·V². This gives very high power, efficiency above 90% and hundreds of thousands of cycles or more. But it stores little energy per kilogram, so it suits bursts of seconds. Typical uses are smoothing fast power swings, regenerative braking and support for batteries in hybrid systems.
Thermal energy storage
Thermal energy storage keeps energy as heat or cold. There are three methods:
- Sensible heat: we raise the temperature of a material that stays solid or liquid, such as water, rock or molten salt.
- Latent heat: we melt or freeze a phase change material (PCM), such as wax, salt hydrate or ice. It stores a lot of heat at an almost fixed temperature.
- Thermochemical: a reversible chemical reaction stores the heat. It is dense and can hold energy for a long time, but it is still mostly in research.
Concentrated solar power plants store molten salt heat for hours, so they can make power after sunset. Buildings make ice at night and use it for air conditioning in the day. Heat stored for later use as heat loses little. But turning electricity into heat and back into electricity loses most of it, as in any heat engine.
Hydrogen
Hydrogen storage uses an electrolyser, a device that splits water into hydrogen and oxygen with electricity. The hydrogen is stored as compressed gas, in salt caverns or in other forms. Later, a fuel cell or a turbine turns it back into electricity. Hydrogen can hold energy for weeks or months, so it suits seasonal gaps. The cost is efficiency. Tests at the US National Renewable Energy Laboratory (NREL) gave about 28 to 35% from electricity back to electricity.
Comparing storage options by efficiency alone. Hydrogen loses most of its input energy, but it can hold energy for months at low cost per MWh. A supercapacitor is very efficient, but it empties in seconds. The right choice depends on the duration and the job.
Comparing the technologies
The first thing to compare is round-trip efficiency (RTE): the energy we get back divided by the energy we put in. A battery with 85% RTE gives back 85 MWh for every 100 MWh of charging. The rest becomes heat in the cells, the inverter and the transformer. The figure below shows typical ranges.
Real data agrees with these ranges. In 2019, US utility-scale batteries averaged 82% round-trip efficiency, and pumped storage averaged 79%, according to the US Energy Information Administration (EIA). NREL's Annual Technology Baseline uses 85% for new lithium-ion plants. A BESS also uses some power for cooling and controls, so the RTE at the grid meter is a little lower than the RTE of the cells.
The second thing to compare is the size of the job: the power and the duration. The map below places each technology on both scales.
The table sums up typical durations and response times. Response time is how long the system takes to go from idle to full output.
| Technology | Typical duration | Response | Best at |
|---|---|---|---|
| Supercapacitor | Seconds | ms | Power bursts |
| Flywheel | Seconds–15 min | ms | Frequency, UPS |
| Lithium-ion | 1–4 h (up to 8 h) | ms–1 s | Most grid uses |
| Flow battery | 4–12 h | Seconds | Long daily cycling |
| Pumped hydro | 6–24 h | Seconds–minutes | Bulk shifting |
| Compressed air | 4–24 h | Minutes | Bulk shifting |
| Thermal | Hours–days | Minutes | Heat, CSP plants |
| Hydrogen | Days–seasons | Seconds–minutes | Long gaps |
In short, we use fast devices for seconds, batteries for hours, and bulk mechanical or chemical storage for long periods. Some projects combine two, such as a battery with a supercapacitor or a flywheel. This is a hybrid energy storage system: the fast device takes the sharp spikes, and the battery takes the energy.
How a battery storage system is built
Battery storage is now the most common type of new storage, so it helps to know what is inside. A grid BESS is built up in layers, from a single cell to a grid connection. The figure below shows the chain.
- Cell. The basic unit. An LFP cell has a nominal voltage of about 3.2 V. Large prismatic cells for grid storage hold a few hundred amp-hours each.
- Module. A group of cells in series and sometimes in parallel, with sensors for voltage and temperature.
- Rack. Modules in series to reach a DC voltage of about 1,000 to 1,500 V, with a switch, fuses and a rack controller.
- Container or enclosure. Many racks with cooling (HVAC or liquid cooling), fire detection and suppression. A 20-foot container now often holds about 3 to 6 MWh.
- PCS (power conversion system). A two-way inverter. It turns DC into AC to discharge and AC into DC to charge. It also controls reactive power, so it can support the grid voltage.
- Transformer. It steps the inverter's low AC voltage up to medium voltage, for example 33 kV, and the plant's substation may step it up again.
Two control layers run the plant. The battery management system (BMS) measures each cell's voltage, current and temperature. It balances the cells and opens contactors if anything goes outside safe limits. The energy management system (EMS) is the plant's brain. It decides when to charge and discharge, based on prices, grid signals and the state of charge.
Key ratings: MW, MWh, C-rate and DoD
A storage datasheet uses a handful of ratings. We met power and energy already. The others tell us how hard the battery works and how long it lasts.
- Power (MW) and energy (MWh). Usually the power is at the AC grid connection. The energy may be quoted at the DC side, the nameplate, or as usable energy at the grid. We always check which one.
- C-rate: the charge or discharge power divided by the energy capacity. A 1C rate empties the battery in one hour. A 0.25C rate takes four hours. Low C-rates are gentler on the cells.
- State of charge (SoC): how full the battery is, from 0 to 100%.
- Depth of discharge (DoD): the share of the capacity used in a cycle. A battery cycled between 5% and 95% SoC has a 90% DoD.
- Cycle life: how many full cycles the battery can do before it reaches its end-of-life capacity, often 70 to 80% of new. Shallower cycles usually give more cycles.
- Degradation: the slow loss of capacity with cycles, time, heat and high SoC. Owners deal with it by oversizing at the start or by augmentation, adding modules later.
- Self-discharge: energy lost while the system waits. It is small for lithium-ion, but larger for flywheels.
NREL's cost baseline assumes a 15-year life for utility-scale batteries. The plant keeps its rated capacity over that time, with augmentation costs included in the yearly running costs.
Depth of discharge is like a fuel tank we never run completely dry or fill to the very brim. We lose a little range. But the engine, in this case the cell, stays healthy for many more years.
Worked example: sizing a battery for solar
A 2 MW solar plant wants to move its output to the evening. The goal is 2 MW for 4 hours after sunset. What size of battery do we need? We assume 85% round-trip efficiency and a 90% depth of discharge. The four cards below follow the steps.
1Energy at the grid
We want 2 MW for 4 hours. That is 8 MWh delivered at the grid connection.
2Add discharge losses
Round-trip efficiency is 85%, so each direction is about √0.85 ≈ 92%. The battery must release 8.7 MWh.
3Divide by depth of discharge
We only cycle 90% of the battery. So the nameplate energy is 8.7 ÷ 0.9 ≈ 9.6 MWh.
4Check C-rate and ageing
The C-rate is low, which is easy on the cells. To still give 8 MWh at 80% health, start at 12 MWh or plan augmentation.
The result: a battery of about 9.6 MWh nameplate with a 2 MW inverter. To charge it, the solar plant must supply 8.7 ÷ 0.922 ≈ 9.4 MWh. A 2 MW plant on a clear day makes roughly 2 MW × 5 to 7 full-sun hours, so 10 to 14 MWh. This means the battery takes almost all of a sunny day's output. On cloudy days, it may charge only partly, or it may charge from the grid if the rules allow it.
Buying "8 MWh" because 2 MW × 4 h = 8 MWh. After losses and the DoD limit, an 8 MWh nameplate battery delivers only about 8 × 0.9 × 0.92 ≈ 6.6 MWh. It runs out after about 3.3 hours.
Use the calculator below to try other cases. Watch how a lower efficiency or a smaller DoD makes the battery bigger.
Losses are split evenly between charging and discharging, so each direction is √(round-trip efficiency). Auxiliary loads such as cooling are not included.
Safety and standards
Lithium-ion cells hold a lot of energy in a small space, and their electrolyte can burn. The main hazard is thermal runaway: a cell heats up, the heat speeds up internal reactions, and the reactions make even more heat. The cell then vents flammable, toxic gas and may catch fire. The heat can spread to the next cell.
Thermal runaway can start with overcharging, a short circuit, physical damage, a manufacturing defect or high outside heat. LFP cells are more stable than NMC, but no lithium-ion chemistry is free of risk. In January 2025, a fire at the Moss Landing plant in California burned for about four days. That 300 MW phase used NMC cells inside a former turbine hall, not separate containers, so the fire could spread.
Several standards cover BESS safety:
- UL 9540 is the safety standard for the complete energy storage system.
- UL 9540A is a fire test method. It tests at cell, module, unit and installation level, to see whether thermal runaway spreads. Its 6th edition makes the large-scale fire test always required and takes effect on 1 January 2027.
- NFPA 855 is the US installation standard. It sets spacing, fire detection and suppression, explosion control and emergency planning. Its 2026 edition adds large-scale fire testing and makes a hazard mitigation analysis the default for most projects.
- IEC 62933 is the international series for grid storage systems, including safety. Many countries also use local fire codes.
Local rules vary, so each project checks which editions the local fire authority enforces.
Applications of energy storage systems
Storage now works at every level of the power system, from the transmission grid to a single house.
- Frequency response and reserves. Batteries and flywheels react faster than any power plant. The Hornsdale Power Reserve in South Australia, now 150 MW and 193.5 MWh, can respond in about 100 ms.
- Energy arbitrage and time-shifting. Plants charge at low prices and discharge at high prices. The IEA reports that over 90% of new battery projects now do arbitrage.
- Renewable integration. Storage firms up solar and wind, captures energy the inverter would otherwise clip, and smooths fast ramps. About 30% of new utility-scale batteries in 2025 were built next to renewable plants.
- Capacity and network deferral. A battery at a busy substation can cover the peak and delay a costly upgrade of a line or transformer.
- Commercial and industrial sites. Peak shaving lowers the demand charge, which is set by the highest power the site draws. Time-of-use shifting moves energy use to cheaper hours. Storage also improves power quality for sensitive machines.
- Backup power. UPS batteries and flywheels ride through short outages in hospitals and data centres. Larger systems can keep critical loads running for hours.
- Microgrids. In a microgrid, storage holds the voltage and frequency when the site runs on its own, called islanded mode. It is key for remote communities, campuses and disaster recovery.
- Homes. Home batteries store rooftop solar for the evening and give backup. Many homes linked together can act as a virtual power plant.
The trend is towards longer durations and new chemistries, such as sodium-ion, iron-air and solid-state cells. Second-life electric vehicle batteries are also used in stationary storage. The IEA estimates that tripling renewable capacity by 2030 would need about 1,500 GW of battery storage worldwide.
Key takeaways
- Energy storage systems take in electricity when it is plentiful and give it back later. They time-shift energy, support frequency, provide peak capacity and give backup.
- Power (MW) and energy (MWh) are separate ratings. Their ratio is the duration, such as a 4-hour battery.
- Round-trip efficiency ranges from about 90% for supercapacitors and flywheels, through 82–90% for lithium-ion and 70–82% for pumped hydro, down to about 30–40% for hydrogen.
- Fast devices cover seconds, batteries cover hours, and pumped hydro, compressed air and hydrogen cover long durations.
- A BESS is built from cells, modules, racks, a PCS and a transformer, with a BMS for protection and an EMS for control. LFP now dominates grid storage.
- Size a battery with losses, depth of discharge and ageing: 2 MW for 4 h needs about 9.6 MWh nameplate, not 8 MWh. Safety rests on layered protection tested to UL 9540A and NFPA 855.
Next, we suggest looking at how storage fits with renewables and microgrids. The guides below go deeper.
Related guides on Electrical Blog:
Integration of renewable energy
