Energy Storage

Energy storage is the process of capturing energy produced at one time so that it can be used at a later time, often after conversion into another form. It covers a wide range of physical and chemical methods, including batteries, pumped water reservoirs, compressed air, flywheels, supercapacitors, hydrogen, and heat stores. Because it is a carrier and converter rather than a primary source, its design is governed by energy balance and by the thermodynamic losses that occur at every conversion step.

Storage matters because supply and demand rarely coincide. Electricity grids must match generation to consumption moment by moment, and the growing share of solar and wind power, whose output varies with weather and time of day, has made this matching harder. Storage also underpins electrified transport, off-grid power, and industrial heat. Pumped hydro has long supplied most of the world's installed storage capacity, but lithium-ion batteries have become the fastest-growing form, and new chemistries and long-duration technologies are being developed to serve needs that batteries cannot meet economically.

This entry first defines the basic quantities and physical limits of storage, then surveys storage technologies by category: mechanical, electrochemical, electrical, chemical, and thermal. It then covers performance metrics and economics, grid and end-user applications, safety and resource concerns, historical development, market and policy developments, and open research problems.

Definition and basic quantities

Energy and power are different quantities, and storage systems must be described with both. Energy is the total amount of work a system can deliver, measured in joules (J) or kilowatt-hours (kWh), where 1 kWh equals 3.6 MJ. Power is the rate at which that energy moves, measured in watts (W) or kilowatts (kW). The two are linked by time: $E = P \cdot t$. The ratio $t = E/P$ gives the discharge duration at full rated power. A 100 MW system with 400 MWh of usable energy therefore lasts four hours at full output. A storage asset is characterized by both numbers because they determine different services. Power sets how quickly the system can respond to a fault or a price spike, while energy sets how long it can sustain delivery. A system can have high power and little energy, as with a capacitor, or the reverse, as with a large seasonal store.

Several quantities describe how much a system holds and how compactly. Capacity is the usable energy under specified conditions, often stated in kWh for systems and in ampere-hours (Ah) for individual cells. Gravimetric energy density is energy per unit mass (Wh/kg) and volumetric energy density is energy per unit volume (Wh/L); see energy density for the general concept. Power density is the analogous ratio of power to mass or volume (W/kg, W/L), and it is the figure that matters for vehicles and for fast-response applications. Commercial lithium-ion cells typically reach roughly 150 to 250 Wh/kg, while gasoline holds about 12 kWh/kg in chemical energy, which illustrates why fuels are hard to replace in some uses. State of charge (SoC) is the remaining energy as a percentage of usable capacity. Depth of discharge (DoD) is the fraction of capacity withdrawn in a cycle, so a cycle from full to 20 percent SoC has a DoD of 80 percent. Because wear generally rises with deeper cycles, DoD links directly to the cycle life discussed in later sections.

Storage is an energy carrier and converter, not a primary source. Primary sources such as solar radiation, wind, fossil fuels and nuclear fission supply energy that originates outside the storage system. A storage device takes energy from electricity or heat, holds it in some physical or chemical form, and returns it later with losses. It therefore moves energy in time, and sometimes changes its form, but it does not create net energy. This distinguishes storage from generation. A generator converts a primary resource into usable energy, while a storage system converts usable energy into a stored state and back. In a grid the same device is a load while charging and a source while discharging, so the classification depends on the operating mode, and the value of storage comes from timing, location and the ability to respond quickly.

Technologies serve very different time scales, and the boundaries between them are approximate. At the millisecond to second scale, electrostatic and supercapacitors and superconducting magnets respond to power-quality disturbances. Flywheels, which store kinetic energy in a spinning rotor, cover seconds to minutes. Lithium-ion batteries are most commonly deployed for one to four hours, which is the usual duration of short-term grid balancing and of most electric vehicle use. Pumped hydro and compressed air, which store gravitational potential energy and compressed gas respectively, typically cover hours to days. Flow batteries and thermal stores bridge daily to multi-day needs. Hydrogen, produced by electrolysis and later reconverted in a fuel cell or turbine, is the main candidate for weekly to seasonal storage, because its chemical energy can be held for months with little self-discharge. Chemical energy carriers and large thermal stores also have a role in seasonal heating. Each of these options is discussed in its own section, but the mapping above explains why no single technology dominates across all durations.

Thermodynamic principles and losses

Every storage device obeys the first law of thermodynamics, which requires that energy be accounted for exactly. For a store with state of charge $E(t)$, charging power $P_{\text{in}}$, discharging power $P_{\text{out}}$ and loss power $P_{\text{loss}}$, the energy balance is $$\frac{dE}{dt} = P_{\text{in}} - P_{\text{out}} - P_{\text{loss}}$$ Storage therefore creates no energy. It holds energy in a form such as chemical, electrochemical, mechanical or thermal internal energy, and every joule that leaves without being delivered as useful output appears as a loss. The round-trip efficiency is the ratio of energy recovered to energy supplied over one complete charge–discharge cycle: $$\eta_{\text{rt}} = \frac{E_{\text{out}}}{E_{\text{in}}}$$ Published values depend on the boundary chosen. AC-to-AC figures include inverters and transformers, DC-to-DC figures do not, and auxiliary loads such as thermal management pumps may or may not be counted, so comparisons are meaningful only when the definition is stated.

The second law explains why $\eta_{\text{rt}}$ can never reach 100%. Every real charging or discharging step is irreversible and produces entropy, $S_{\text{gen}} \ge 0$. The available work destroyed by this entropy production is $T_0 S_{\text{gen}}$, where $T_0$ is the ambient temperature (the Gouy–Stodola relation). Losses are therefore a consequence of irreversibility, not merely a matter of poor engineering. The second law also limits heat-based storage. Heat stored at a high temperature $T_h$ can yield work only up to the Carnot factor $\eta_C = 1 - T_c/T_h$, where $T_c$ is the temperature of the environment or heat sink. Heat stored at 800 K and rejected to 300 K can yield at most 62.5% of its energy as work, and real heat engines fall well below that. Stores that keep energy in electrical, chemical or mechanical form avoid this restriction in the storage step itself, which is one reason their efficiencies can be much higher.

Storage is also time-dependent because stored energy leaks away. In many systems the loss rate is roughly proportional to the stored amount, so that $E(t) = E_0 e^{-t/\tau}$, with $\tau$ a characteristic time constant. Batteries lose charge through parasitic side reactions and slow electronic leakage across the separator, and capacitors lose it through leakage in the dielectric and electrolyte. Thermal stores lose heat through walls, and their rate is governed by insulation and thermal-conductivity. Liquid hydrogen boils off, and flywheels lose energy to bearing friction and aerodynamic drag. Standby losses are a different category because they are fixed power draws that do not depend on throughput: power electronics, battery management systems, heaters that keep molten salts from freezing, and cooling systems. These losses weigh most heavily for small energy quantities and for long-duration stores that cycle infrequently, while resistive and electrochemical overpotential losses, which scale with current, dominate at high power. As a result, efficiency is a function of both power and time, not a single constant.

Conversion chains compound these losses multiplicatively. For a sequence of steps, $\eta_{\text{chain}} = \prod_i \eta_i$. An electricity-to-chemical-to-electricity path such as a lithium-ion battery with roughly 95% efficiency in each direction yields about 90%. A power-to-hydrogen-to-power path using electrolysis, compression or storage, and a fuel-cell or turbine can easily fall to a fraction of that, because each step contributes its own loss. An electricity-to-heat-to-electricity path is even more constrained. Resistive heating is nearly lossless, but the conversion of heat back into electricity is limited by the Carnot factor and by the efficiency of the power cycle, so the overall figure is much lower than for the electrical path. This multiplicative penalty explains why direct electrical or electrochemical storage is generally favoured for short-duration services, and why chemical and thermal routes are reserved for cases where their other advantages outweigh lower round-trip efficiency.

Pumped hydroelectric storage

Operating principle

Pumped hydroelectric storage (PHS) keeps electricity in the form of gravitational potential energy held by water. A plant has two reservoirs at different elevations. When power is cheap or in surplus, pumps lift water from the lower reservoir to the upper one, and a motor drives them. When the grid needs power, the water falls back through a turbine that turns a generator. The stored energy follows $$E = \rho V g H$$ where $\rho$ is the density of water, $V$ the volume moved, $g$ gravitational acceleration and $H$ the head, or height difference. As a rough check, 1,000 m³ of water dropped through 100 m stores about 0.27 MWh. Because the same water is reused, the system acts as a rechargeable store rather than a source. Practical sites need specific geology and a sufficient elevation difference between the reservoirs, often between 1000 and 2000 feet. The reversible machines that do both jobs are called pump-turbine units.

Share of storage and efficiency

Pumped hydro dominates grid storage, although published shares depend on the year and the metric. The International Hydropower Association says that PSH accounts for over 94% of installed global energy storage capacity. A review in a journal reports more than 95% of the world's electricity storage power (GW) capacity and 99% of the storage energy (GWh). An Argonne National Laboratory analyst, quoted by Yale E360, put the share at more than 98 percent of installed storage capacity globally, a figure from an earlier period. Absolute totals also vary: one figure cites a global installed capacity of nearly 200 GW, while another gives 182 GW by 2023. Energy stored is more stable across sources, with the IHA estimating that PSH projects worldwide store up to 9,000 gigawatt hours (GWh) of electricity.

Round-trip efficiency, meaning the electricity returned divided by the electricity put in, is commonly placed in the 70–85% range; one market report gives high efficiency (70-85%), and this figure should be treated as a typical range rather than a fixed value. Losses come from the pump, the turbine, the generator and the motors, and from evaporation or seepage in the reservoirs. Asset life is also long. A life-cycle study of a closed-loop facility assumed an 80-year lifetime without replacing reservoir liners. Civil works such as dams, tunnels and caverns generally outlast the electromechanical equipment, which can be refurbished or replaced during the plant's life.

Siting, environmental and social impacts

Geography, rather than efficiency, is the main constraint. Suitable sites require large elevation differences, substantial reservoir volumes, and appropriate geological conditions. Most existing plants are open-loop. The US Department of Energy defines open-loop projects as "continuously connected to a naturally flowing water feature", while closed-loop projects are "not continuously connected to a naturally flowing water feature." Open-loop plants can alter aquatic habitat, and ongoing operations can also have effects on water quality, specifically increased concentrations of dissolved solids, nutrients, and heavy metals.

Closed-loop designs reduce many of these impacts. A DOE-supported study found that the environmental effects of closed-loop PSH projects are generally lower — both more localized and of shorter duration — than those of open-loop PSH projects. The trade-offs are real, however. Certain specific impacts of closed-loop systems can be higher than those of open-loop systems, particularly for geology and soils and groundwater. Construction may also weigh more heavily, especially if they require the construction of two new reservoirs. Social conflict has become a growing issue. Concerns over permitting and consultation processes and conflicts over siting, water resources, and Indigenous lands are emerging more frequently, particularly in the arid western United States. Underground layouts can shrink the surface footprint. The Nant de Drance plant in Switzerland has a powerhouse cavern 600 m below ground between the Lac d'Emosson and Lac du Vieux Emosson reservoirs, and its tunnels stretch for more than 17 km.

Early history and modern role

The first pumped-storage installations appeared in Europe in the 1880s and 1890s. A plant in Zurich in 1882 used a reciprocating pump powered by the wind to store water in an upper reservoir, but it had no generator. Later Swiss installations followed, in Zurich in 1891 on the Limmat river, followed by a second installation in 1894 at Lake Maggiore, and a third one in 1899 at the Aare River. In Italy, a 1894 plant at Luino used a pump to drive a spinning mill. Early twentieth-century grids were small and isolated, and conventional hydropower stations with large reservoirs and PSH were the only utility-scale energy storage technologies available at that time. The first large plant near Schaffhausen, Switzerland, in 1909, had a capacity of 1.5 MW and used a separate pump and turbine. The first commercial plant, Niederwartha in Germany, began operating in 1929. Storage was originally used to match night-time pumping with daytime demand, and later to smooth the output of baseload plants. Its role has since widened. Today, pumped storage hydropower is still the only one that is mature, proven, and commercially available to provide large utility-scale energy storage. Modern plants increasingly use variable-speed machines, in which the pump's speed can be adjusted rather than fixed to the grid frequency. This allows operators to vary pumping power and respond more quickly to grid needs, which in turn widens the range of sites and services that a project can serve.

Compressed air, flywheel and gravity storage

Diabatic and adiabatic compressed-air storage

Compressed-air energy storage (CAES) uses electricity to run compressors that pressurize air. The air is held in a reservoir and later expanded through a turbine to generate power. The main design question is what happens to the heat produced during compression. In diabatic plants, much of the compression heat is dissipated into the atmosphere through intercoolers, so the air must be reheated before it returns to the cycle. That reheating usually burns natural gas or another fuel. Because of this, the round-trip efficiency of purely diabatic designs is modest, around 40 % to 50 %, and the process emits greenhouse gases.

Adiabatic designs keep the heat instead. Heat is pulled from the air during compression, stored in a thermal store, and returned to the air before it enters the turbine. Common media include solid materials such as concrete or stone, or fluids such as hot oil or molten salt. The thermal store is what makes the cycle more efficient, and practical round-trip efficiency is generally expected to be about 70%. The largest adiabatic plant reported so far is in Huai'an, China, which consists of two 300 MW units with a 71% conversion efficiency rate, with a storage capacity of 2.4 GWh.

Liquid air energy storage (LAES) is a related approach. Air is cooled to cryogenic temperatures and stored as a liquid in insulated tanks. When power is needed, the liquid is evaporated and expanded to drive a turbine, and the cold released in the process can be reused. Because liquid air is far denser than compressed gas, LAES is more compact: one study finds that when all spatial factors affecting energy density are considered, LAES demonstrates an energy density advantage over CAES by a factor of 6.9. Its commercial deployment is still at an early stage, mostly in demonstration plants.

Caverns, salt formations and site constraints

Compressed air can be kept in several kinds of underground media, including porous rock formations, depleted natural gas or oil fields, and caverns in salt or rock formations. Above-ground storage in pipelines is possible, but such systems tend to be expensive and typically do not have a storage capacity comparable to an underground cavern.

Salt caverns are the preferred medium for commercial plants because salt is impermeable and tolerates repeated pressure cycling. Huntorf, in Germany, stores air in a salt cavern about 600 m deep, with a volume of roughly 310,000 m³ and an operating pressure between 5 and 7 MPa. McIntosh, in Alabama, uses a rock-salt layer about 450 m deep and cycles between 4.5 and 7.4 MPa. Salt is widely available in some regions; northern Europe, for example, has many salt deposits, and nearly 500 salt caverns are currently used for natural gas storage. Even so, geology is the main limit. Studies emphasize that technical feasibility has been demonstrated, but site selection remains the core bottleneck. A site needs suitable rock, tightness, and a nearby grid connection, and the cavern's volume and pressure range set the stored energy.

The two commercial plants show the trade-off between power and duration. Huntorf, which began operating in 1978, has charging and discharging power ratings of 60 and 290 MW, and it can be charged continuously for 8 h to provide 2 h of generation. McIntosh, operating since 1991, can produce 110 MW of electricity for up to 26 hours. Huntorf burns natural gas in its turbine, while McIntosh uses exhaust heat from the last stage of the gas turbine to preheat the turbine inlet through a heat exchanger.

Flywheel storage

A flywheel stores energy as the kinetic energy of a rotating mass. In a typical system, a motor accelerates the flywheel when surplus electricity is available, and when power is needed the stored kinetic energy drives a generator. The stored energy is $E = \tfrac{1}{2} I \omega^{2}$, where $I$ is the moment of inertia and $\omega$ the angular velocity, so high rotational speed is the most effective way to raise capacity. To achieve this, rotors are usually housed in vacuum enclosures and run on low-friction bearings, often magnetic, to minimize windage and bearing losses. These design choices, which are general engineering practice rather than a single standard, allow the rotor to hold its speed for minutes to hours of self-discharge-limited storage.

Flywheels are fast and powerful but store little energy. They are considered environmentally friendly short-term storage solutions because they can store and release energy quickly and efficiently, with high power density and long lifespans. That profile suits frequency regulation. A grid's frequency falls when load exceeds generation and rises when generation exceeds load, so fast-responding devices can correct imbalances within seconds. Grid-scale examples include Beacon Power's installation in Pennsylvania, which commissioned a 20 MW plant, with a carbon-fiber rotor operating at 16,000 RPM. Flywheels are often paired with batteries in hybrid systems, where the flywheel handles rapid fluctuations and the battery supplies longer-term energy.

Gravity storage

Gravity storage stores energy as the potential energy of masses lifted against gravity, with $E = mgh$. It is conceptually similar to pumped hydro, but the working mass is solid. Proposed designs include stacking heavy blocks in towers with cranes, lowering weights in disused mine shafts, and using winches to raise and lower weights in deep shafts. The energy per unit mass is low: lifting one tonne through 100 m stores about 0.27 kWh. Scaling up therefore requires large masses, tall structures or deep shafts, plus mechanical systems that tolerate frequent cycling.

Most gravity concepts remain at pilot or demonstration scale. Their advantages are a long design life, few materials that degrade, and no need for a particular water source. Their disadvantages are lower energy density than most chemical systems and a strong dependence on site geometry and construction cost. Whether they can compete at grid scale depends on demonstrated round-trip efficiency, capital cost per kilowatt-hour, and operating experience over many years, which are still limited.

Electrochemical cells: principles

Galvanic cells and redox reactions

A galvanic (voltaic) cell converts the free energy of a spontaneous redox reaction into electrical work. The reaction is split into two spatially separated half-reactions. Oxidation at the negative electrode (the anode during discharge) releases electrons into an external circuit, and reduction at the positive electrode (the cathode) consumes them. Charge balance inside the cell is maintained by ions moving through the electrolyte. In the classic Daniell cell, zinc is oxidized in one half-cell while copper ions are reduced at a copper electrode, giving a nominal voltage of about 1.1 V. The separation is the essential point: if zinc and copper sulfate reacted in a single beaker, the same chemical energy would appear only as heat. In rechargeable cells, the current direction reverses during charging, so the electrodes exchange roles in the reaction sense, while the terms negative and positive electrode remain fixed by polarity.

Cell voltage, potentials and losses

The open-circuit voltage is set by the difference between the two electrode potentials, $E_{\text{cell}} = E_{\text{cathode}} - E_{\text{anode}}$, and it links to the reaction's Gibbs free energy through $\Delta G = -nFE$, where $n$ is the number of electrons transferred and $F$ the Faraday constant. Concentrations shift the potential according to the Nernst equation:

Lithium is attractive because its standard potential is very negative, about −3.04 V against the standard hydrogen electrode. In a lithium-ion cell, a graphite anode operates near 0.1 V versus Li/Li⁺ and a cobalt-oxide cathode near 3.9 V, giving a nominal cell voltage of roughly 3.6–3.7 V.

Under load, the terminal voltage falls below this equilibrium value because of overpotentials. Activation overpotential reflects the kinetics of charge transfer at the electrode surface, described by the Butler–Volmer relation:

Here $i_0$ is the exchange current density, a measure of how readily the reaction proceeds at equilibrium, and $\eta$ is the overpotential. Ohmic losses arise from ionic resistance in the electrolyte, electronic resistance in the electrodes and contacts, and the resulting IR drop, so that roughly $V = E_{\text{OCV}} - IR_{\text{int}}$ for the ohmic part. Concentration overpotential appears when reactants cannot reach the surface fast enough, which limits high-current operation. Low temperature slows kinetics and raises resistance at once, which is why cell voltage sags and capacity drops in cold conditions.

Intercalation and conversion electrodes

Electrode materials that store charge through ion insertion are called intercalation electrodes. Guest ions enter vacancies in a host lattice with little change to the host framework. Graphite takes up lithium up to LiC₆, a theoretical capacity of about 372 mAh/g, and layered oxides and olivine phosphates such as LiFePO₄ use the same principle on the cathode side. Because the framework survives each cycle, these electrodes can sustain thousands of cycles, but their capacity is bounded by the number of available sites.

Conversion electrodes instead undergo a chemical reaction that breaks and reforms bonds and produces new phases. A metal oxide, for example, can be reduced to metal and lithium oxide, and metal fluorides form lithium fluoride. Each formula unit can exchange several electrons, so specific capacity can be much higher. The cost is large volume changes, particle pulverization, slow kinetics, and voltage hysteresis between charge and discharge.

Reversibility is what separates a rechargeable electrode from a primary one. A rechargeable electrode must return to nearly its original state, and irreversible side products consume charge. Coulombic efficiency per cycle is therefore critical: at 99.9% per cycle, about 37% of the original capacity would remain after 1,000 cycles, since $0.999^{1000} \approx 0.37$. Intercalation's small structural change favors reversibility, whereas many conversion reactions are either limited to primary use or require extensive engineering before they can be cycled.

Electrolyte and separator

The electrolyte must conduct ions while blocking electrons. An electronically conducting electrolyte would short-circuit the cell internally and cause self-discharge. It must also remain stable across the voltage window between the two electrodes; otherwise it decomposes. Lead-acid cells use aqueous sulfuric acid, nickel-based cells use alkaline potassium hydroxide, and lithium-ion cells use a lithium salt such as LiPF₆ dissolved in organic carbonate solvents. Liquid electrolytes typically reach ionic conductivities on the order of 10 mS/cm at room temperature, far higher than most solid electrolytes.

The separator is a distinct component with a mechanical rather than chemical role. It is a porous membrane, usually a polyolefin such as polyethylene or polypropylene, sometimes with a ceramic coating, that keeps the electrodes from touching while leaving ion-conducting pathways open through electrolyte-filled pores. Thinner separators raise energy density but increase the risk of short circuits. Polyethylene separators also provide a thermal safeguard: the pores close near 130 °C, which interrupts ion flow when a cell overheats. Electrolyte and separator therefore complement each other: one carries ions, the other keeps the electrodes apart.

Lithium-ion batteries

Cathode chemistries: LFP and nickel-based layered oxides

Lithium-ion cells store energy by moving lithium ions between a graphite (or silicon-containing) anode and a metal-oxide or phosphate cathode, with electrons passing through the external circuit. The cathode chemistry sets most of the trade-offs in voltage, energy density, cost, thermal stability and cycle life.

Lithium iron phosphate (LFP, LiFePO₄) uses an olivine structure in which strong phosphorus–oxygen bonds resist oxygen release at high temperature. Its nominal voltage is about 3.2 V, and it contains no cobalt or nickel. Its gravimetric energy density is lower than that of nickel-based cells, typically in the range of roughly 150–200 Wh/kg at cell level, but it tolerates deep cycling well, commonly delivering several thousand cycles to 80 % of initial capacity, and its thermal-runaway onset is higher than that of nickel-rich cells. Its weaker performance in cold conditions is a recognised limitation.

Nickel-based layered oxides, such as nickel–manganese–cobalt (NMC) and nickel–cobalt–aluminium (NCA), reach higher energy density, roughly 200–270 Wh/kg at cell level, and perform better at low temperature. The trade-offs are higher raw-material cost, dependence on cobalt and nickel supply chains, lower thermal-runaway onset as nickel content rises, and typically fewer cycles to end of life. Market shares have shifted rapidly. Industry data for 2025 indicate that LFP became the most widely deployed EV chemistry worldwide, and it accounts for most stationary storage installations, whereas NMC remains prominent in premium and long-range vehicles, particularly outside China. Pack-price comparisons depend heavily on application and region, so cost gaps should be read as indicative.

Cells, modules, packs and the battery management system

A cell is the basic electrochemical unit, containing electrodes, separator, electrolyte and a sealed casing in cylindrical, prismatic or pouch format. Cells are connected in series to reach the required voltage and in parallel to increase capacity, forming modules. Modules are mounted with cooling plates, busbars, fuses and contactors into a pack, which is then integrated into a vehicle or a containerised storage system together with power electronics.

The battery management system (BMS) monitors voltage, current and temperature of every cell or module and estimates state of charge and state of health. It balances cell voltages, enforces safe operating limits, opens contactors on faults and communicates with the charger, inverter or grid controller. Because series-connected cells differ slightly in capacity and resistance, balancing and accurate state estimation largely determine how much of a pack's nominal capacity remains usable over time.

Solid electrolyte interphase and degradation

During the first charges, the electrolyte is reduced at the graphite anode, whose potential lies below the stability window of conventional organic electrolytes. The reduction products form a thin passivating layer, the solid electrolyte interphase (SEI), which is electronically insulating but permits lithium-ion transport. A well-formed SEI stops further electrolyte decomposition, but it consumes lithium irreversibly, which accounts for the initial capacity loss of a new cell.

Capacity fade over cycling arises from several mechanisms. The SEI thickens and cracks, lithium inventory is lost, active material becomes electrically isolated, and electrode structures strain during repeated intercalation. Lithium plating occurs when lithium metal deposits on the anode instead of intercalating, typically at low temperature, high charge rate or high state of charge; plated lithium can react with electrolyte and may form dendrites that threaten separator integrity. Calendar aging, the loss of capacity while a cell is idle, is driven mainly by slow SEI growth, which accelerates with temperature and high state of charge. Battery management strategies such as charge-rate limits, temperature control and voltage windows aim to slow these processes.

Commercial history

The chemistry was developed through the 1970s and 1980s by M. Stanley Whittingham, who proposed intercalation cathodes, John B. Goodenough, who identified layered lithium cobalt oxide as a high-voltage cathode, and Akira Yoshino, who built a practical carbon-anode prototype. Sony commercialised the first lithium-ion cell in 1991, initially for camcorders and then for portable electronics, displacing nickel-cadmium cells in many devices. The three researchers shared the 2019 Nobel Prize in Chemistry. Through the 2000s, mobile phones and laptops drove volume growth, and cathode and cell manufacturing scaled up sharply, especially in East Asia. Electric vehicles from the late 2000s and early 2010s, followed by grid-connected storage projects from around 2015, made lithium-ion the dominant technology for both mobile and stationary storage, a position that prompted renewed attention to cost, safety, raw-material supply and recycling.

Other battery chemistries

Flow batteries

In a flow battery, the active materials are dissolved in liquid electrolytes stored in external tanks and pumped through a stack of electrochemical cells where the reactions take place. Because the stack determines power and the tanks determine energy, the two can be sized independently. Adding tank volume increases stored energy without changing maximum power, and enlarging the stack does the reverse. This decoupling is the main structural difference from sealed cells such as lithium-ion batteries, where energy and power are tied to the same electrode area.

Vanadium redox flow batteries use vanadium ions in several oxidation states on both sides of a membrane, so cross-contamination between the half-cells can be partly reversed by re-balancing the electrolyte. Zinc-bromine systems instead plate zinc onto an electrode during charging and rely on bromine complexing agents to manage the reactive species. Iron-based flow chemistries use cheaper raw materials, but as of mid-2026 one developer reports only limited deployment and still-open questions about yield, cost and efficiency.

The economics are mixed. For 4-hour-plus utility projects in early 2026, one industry analysis puts installed vanadium systems at roughly USD 400–500/kWh against about USD 230–280/kWh for LFP, while another gives a European LFP turnkey figure near USD 177/kWh, so benchmarks depend heavily on source and method. Vanadium round-trip efficiency is typically 65–85%, below the 85% or more usually cited for LFP. Costs fall as duration rises because adding energy means adding electrolyte rather than cells, which is why flow batteries are mainly proposed for storage of eight hours or longer. Vanadium prices have also swung widely in recent years, and China holds more than 70% of the vanadium supply chain, which adds financing risk.

Sodium-ion batteries

Sodium-ion cells use the same basic rocking-chair principle as lithium-ion, with sodium ions shuttling between electrodes. Sodium is far more abundant and geographically dispersed than lithium, and sodium-ion designs avoid lithium and, in most cases, graphite. The trade-off is lower energy density. CATL's Naxtra line, launched in April 2025, is rated at up to about 175 Wh/kg and is designed for operation from −40 °C to 70 °C. The company began large-scale production of sodium-ion cells in 2025 and announced broader deployment across vehicles, battery swapping and energy storage for 2026. A reported three-year, 60 GWh storage supply agreement with a Chinese integrator signals commercial intent, though its delivery schedule is not yet proven.

Analysts disagree about cost. Sodium-ion proponents emphasise cheaper materials and better cold-weather performance, while the IEA notes that the chemistries nearest to market still rely on nickel and manganese, that global supply chains are much less developed than those for lithium-ion, and that optimised LFP remains ahead on energy density, supply maturity and cost unless lithium prices rise sustainably or sodium-ion density improves significantly. Early sodium-ion adoption is therefore strongest where weight matters less, such as stationary storage and low-range vehicles.

Solid-state batteries

A solid-state battery replaces the liquid or gel electrolyte with a solid ionic conductor, typically a ceramic, a sulfide or a polymer. The motivation is threefold: a solid can, in principle, allow lithium-metal anodes and higher energy density, it removes the flammable organic solvent that contributes to thermal runaway, and it can tolerate a wider range of mechanical and thermal conditions. Solid electrolytes were studied in the nineteenth century, but practical problems delayed widespread use until renewed interest in the late twentieth and early twenty-first centuries.

The main barriers are at the interfaces. Solid electrode–electrolyte contacts are much less intimate than liquid wetting, so resistance rises and contact can be lost as electrodes swell and shrink during cycling. Cracks and voids can also form, and maintaining stack pressure becomes a design requirement. Manufacturing is equally difficult: thin, defect-free solid layers must be produced at scale and at acceptable cost. Commercial status is therefore uneven. Semi-solid, solid-liquid hybrid cells are entering vehicles, while all-solid-state cells are mostly in pilot production, with road testing by automakers and mass-production targets concentrated between about 2027 and 2030. Some commentators cite production cost premiums of several times that of conventional cells, but such estimates are contested and vary considerably between sources.

Lead-acid, nickel-based, zinc-air and iron-air chemistries

Lead-acid batteries, first practical in 1859, remain the cheapest and most recyclable rechargeable option. They are widely used for vehicle starting and backup power, but their low specific energy and limited cycle life under deep discharge restrict them in grid and mobile applications.

Nickel-based cells include nickel–cadmium, which tolerates abuse and was long used in power tools, but suffers from a so-called memory effect and cadmium toxicity. Nickel–metal hydride replaced much of its consumer market with higher energy density and lower toxicity, but it remains heavier than lithium-ion. Zinc-air cells draw their oxidant from atmospheric oxygen, giving high theoretical energy density; they are mainly used as primary cells in hearing aids, and their rechargeability is limited by zinc dendrite growth and carbonation of the electrolyte.

Iron-air batteries, which store energy in the reversible oxidation of iron, use abundant and inexpensive materials. Developers target discharge durations of 50–150 hours at a system energy cost below USD 20/kWh, but these figures remain targets rather than established performance, placing the chemistry at the early stage of long-duration deployment.

Capacitors and superconducting storage

Capacitors and superconducting magnetic devices store energy in physical fields rather than in chemical bonds or ion inventories. They respond within milliseconds and tolerate very many cycles, but they hold far less energy per unit mass than batteries.

Capacitor families

An ideal capacitor stores energy as separated charge on two conductors. The stored energy is $$E = \frac{1}{2}CV^{2}$$ where $C$ is capacitance and $V$ is the voltage across the terminals. Because energy grows with the square of voltage, the voltage ceiling matters as much as the electrode geometry. Doubling the usable voltage quadruples the energy for the same capacitance. In film and ceramic capacitors the limit is dielectric breakdown; in supercapacitors it is the electrochemical stability window of the electrolyte.

Electric double-layer capacitors (EDLCs), usually sold as supercapacitors, store charge at the interface between a porous electrode and an electrolyte. Ions gather at the electrode surface to balance the electronic charge, forming a layer only nanometres thick. EDLCs depend on non-faradaic charge separation at the electrode/electrolyte interface. Capacitance therefore scales with the accessible surface area, which is why activated carbons, carbon aerogels and nanotube assemblies dominate electrode design. Since no chemical changes take place within the electrode or electrolyte, the EDLC can have much higher cycling lifetimes compared to batteries.

Pseudocapacitive materials, typically transition-metal oxides, sulfides or conducting polymers, add a faradaic contribution. They utilize rapid and reversible Faradaic redox reactions near or on the electrode surface, which lets them access more charge per unit of area. The trade-off is that redox chemistry wears the material, so pseudocapacitors generally have a higher energy density but a shorter life cycle than EDLCs.

Comparison with batteries

The gap in energy density is large. Commercial supercapacitors typically store about 5–8 Wh/kg, compared with 50–100 Wh/kg for NiMH batteries, while lithium-ion batteries are typically in the range of 100–180 Wh/kg. The reverse holds for power. Battery power is limited by the need for lithium ions to move into and out of the bulk of electrode particles, which slows charging and discharging. Capacitors have higher power density and life cycle than lithium-ion cells, which makes them suited to short, high-current events such as regenerative braking and frequency-related peak shaving. Because of this complementarity, hybrid systems combining lithium-ion batteries and supercapacitors have been studied, with the battery supplying sustained energy and the capacitor absorbing transients.

Superconducting magnetic energy storage

Superconducting magnetic energy storage (SMES) keeps energy in the magnetic field of a direct current circulating in a superconducting coil. Because the wire has almost no resistance below its critical temperature, the current persists with very little decay. The coil must remain cryogenic, and cooling is provided by liquid helium, liquid nitrogen, or mechanical cryocoolers, depending on the superconductor. This refrigeration, together with sensors and controls, consumes power, so SMES is not loss-free. Reported round-trip efficiencies also vary by source: one overview cites more than 95%, while the same text notes that power conversion alone accounts for a substantial share of losses, so such figures should be read with care.

Its main use is power quality rather than bulk storage. SMES may be used to store and instantaneously provide electrical power to offset damaging voltage dips and sags, events that can last only a fraction of a second. For multi-hour storage, other technologies are far more economical.

Materials challenges

Several physical limits constrain capacitive storage. First, capacitance in EDLCs depends on surface area, and the energy stored per gram ultimately is limited by reliance on the double-layer mechanism, which motivates pseudocapacitive and hybrid designs. Second, because energy scales with voltage squared, electrolytes with wide stability windows (organic electrolytes instead of aqueous ones) are essential, but they bring conductivity and safety trade-offs. Third, pore structure must allow ions to reach the surface quickly; maintaining porosity is critical for rapid ion transport. Finally, self-discharge is significant. Leakage currents through the electrolyte and parasitic faradaic reactions drain the stored charge over hours to days, which limits capacitors to short-duration roles.

Hydrogen, ammonia and synthetic fuels

Power-to-gas chains

Chemical storage of electricity usually takes the form of a power-to-gas chain. Surplus electricity drives electrolysis of water, which splits water into hydrogen and oxygen. The hydrogen can be held in pressurised tanks, cooled into liquid form, or stored in underground salt caverns, which are sealed well enough to hold gas for months. When power is needed, the hydrogen is reconverted by a fuel-cell or burned in a gas turbine or engine. Hydrogen can also be combined with captured carbon dioxide to make synthetic methane or methanol, which can use existing gas pipelines and fuel infrastructure.

Hydrogen is a poor store of energy by volume at ambient conditions. Its volumetric energy density at atmospheric pressure is only about 3 Wh/L, far below that of gasoline, so it must be compressed or liquefied before it is practical. Liquefaction is especially costly because the gas must be cooled to about −253 °C, which is why one assessment puts the energy consumed in liquefying hydrogen at 30–40% of its energy content. See also liquid-hydrogen.

Round-trip efficiency and cost

Round-trip efficiency is the energy recovered divided by the energy put in. In a multistep chain it is roughly the product of the efficiencies of each stage. A simple case illustrates the point: an electrolyser at 90% efficiency followed by a power plant at 60% gives $0.9 \times 0.6 = 0.54$, or 54% at best, before any compression, storage or transport losses are counted. Reported figures for power-to-hydrogen-to-power are lower still. One compilation cites 18–46%, another estimates 34–44% for the hydrogen pathway, and grid studies often assume about 37%. By comparison, pumped hydro is typically 70–85% and flow batteries 60–80%, so hydrogen loses roughly half of the input energy that a battery would return.

The low efficiency has direct cost consequences. Every kilowatt-hour delivered requires about two kilowatt-hours of input in the best case, so the chain only pays when the electricity is cheap or would otherwise be curtailed, and when electrolysers run for many full-load hours to spread their capital cost. Efficiency is not the only factor, however. One grid simulation found that lowering round-trip efficiency from 48% to 37% raised total system cost by only about 1% at high wind penetration, which shows that the cost impact depends strongly on the surrounding system. Economic value also comes from selling hydrogen directly as feedstock or fuel, which is often worth more than converting it back to electricity.

Ammonia and methanol as carriers

Ammonia (NH₃) is one of the most studied hydrogen carriers. It contains about 17.7% hydrogen by mass and holds around 120 g of hydrogen per litre, which is more than liquid hydrogen does on a volumetric basis. It can be liquefied at about 1 MPa and 25 °C, or at −34 °C at ambient pressure, which avoids the cryogenic handling that liquid hydrogen requires. Its vapour pressure is similar to that of propane, and an existing industrial supply chain for fertiliser already handles large volumes. Ammonia is toxic and corrosive, however, and recovering hydrogen from it (cracking) needs heat and a catalyst, adding another conversion loss. Ammonia can also be burned directly, which is one reason it is studied as a maritime fuel.

Methanol is a liquid at ambient conditions and is easy to store and transport using existing tanks and ships. It carries about 12.5% hydrogen by mass, and its volumetric energy density is close to that of ammonia. It is toxic, and it requires a carbon source, so its climate benefit depends on the origin of the carbon. In one comparative study of micro gas turbine systems, the round-trip efficiency was 42.6% for hydrogen, 40.7% for methanol and 38.6% for ammonia, with the authors noting that efficiencies could improve with better electrolysis and generation.

Realistic use cases

Synthetic fuels are most credible where their particular properties matter. Seasonal storage is the clearest case, because the chain can hold energy for months with little self-discharge, which batteries cannot do economically. Industrial feedstock is another: hydrogen and ammonia already serve refining and fertiliser production, so renewable versions can displace fossil-based supply without new end uses. Heavy transport is a third, particularly shipping and long-haul trucks, where battery mass and slow charging are disadvantages. In these roles the low round-trip efficiency is accepted as the price of storing large quantities of energy cheaply over long periods.

Thermal energy storage

Thermal energy storage holds heat or cold for later use. It is the oldest form of storage discussed in this entry: a masonry wall warmed by daytime sun and a block of ice cut in winter both do the same job. The simplest case is sensible heat storage, in which a medium changes temperature without changing phase. The stored energy is $Q = m\,c_p\,\Delta T$, so design comes down to mass, usable temperature swing and specific heat. Water is the common medium for tanks in hot-water and district-heating systems because its specific heat is high and it is cheap. Rock and gravel beds suit air-heating systems, where their low cost outweighs their low energy density. For high temperatures, molten salts serve as the storage medium in concentrating solar power (CSP) plants.

In a typical two-tank CSP arrangement, the salt is solar salt, roughly 60% sodium nitrate and 40% potassium nitrate. It melts near 220 °C and is kept liquid between about 290 and 565 °C. Heat collected in the solar field warms salt drawn into a hot tank, and during discharge the hot salt feeds steam generators that drive a steam-turbine. Molten salt storage in CSP plants reached about 21 GWh of electrical capacity worldwide at the end of 2019, and one review reports that adding storage can raise a plant's capacity factor from around 25% to nearly 70%. Conversion back to electricity is limited by the temperature difference across the heat-engine, following carnot-cycle bounds, so hotter storage gives better returns.

Latent heat storage uses phase transitions instead of temperature change. Melting absorbs a large amount of heat at a nearly constant temperature. Ice absorbs about 335 J/g on melting, while water takes up only about 4 J/g for each kelvin of warming, so water must warm by roughly 84 K to absorb the same heat as one melt of ice. Engineers therefore use phase-change materials (PCMs), which fall into organic classes such as paraffins and fatty acids and inorganic salt hydrates. Salt hydrates offer high latent heat and low cost but can corrode, separate, or supercool. One patent claims PCMs store five to fourteen times more heat per unit volume than water, masonry or rock. In buildings, ice storage is the most established application: chillers make ice during off-peak hours and the melt supplies cooling at peak times. Because the chiller must produce brine well below 0 °C, its efficiency is lower than that of a chilled-water store, which is one reason PCMs with melting points in the comfort range are under study for walls and ceilings.

Thermochemical and sorption storage relies on reversible reactions. Energy is absorbed to split a compound or to desorb a sorbate; the separated components can be held at ambient temperature with little loss, and recombining them releases heat. Because the energy sits in chemical bonds or sorption binding rather than in sensible or latent heat, the energy density can be much higher. One solid-gas review reports values above 2000 kJ/kg, which it describes as 10–20 times those of sensible and latent stores, although volumetric comparisons across studies are unreliable because test conditions vary. Water–LiCl systems have been estimated near 253 kWh/m³. Obstacles remain: energy density falls with repeated cycling, coefficients of performance are often low, sorbents can crystallize, and laboratory prototypes are still small. Recovery efficiencies in demonstrations, such as a seasonal magnesium-sulphate battery, have been reported around 65%.

Applications follow the heat load. District heating networks use large insulated water tanks, pit stores and borehole fields to shift heat between seasons or daily peaks. Buildings combine hot-water tanks, ice and PCM stores with heat pumps. Industrial processes use steam accumulators, molten salt loops and stores that capture waste heat. Heat storage competes with electricity storage mainly where the demand is thermal. Converting electricity to heat and back loses most of the energy, whereas storing heat directly avoids the second conversion. Electricity storage remains more valuable when the grid needs flexible power, while heat stores are generally cheaper for space heating, hot water and process heat, although they lose heat to self-discharge over time.

Performance metrics and economics

Operating metrics

Round-trip efficiency (RTE) is the share of input energy that can be recovered after a full charge–discharge cycle, $\eta_{\text{RT}} = E_{\text{out}} / E_{\text{in}}$. It captures conversion losses and, over longer periods, standby losses, and it sets how much of the purchased electricity can be sold again. In one cradle-to-gate assessment of residential lithium-ion systems, a round-trip efficiency of 82–89% was used. Pumped hydro and most electrochemical systems fall in a broadly similar or lower range, depending on design and power level.

Cycle life is the number of full charge–discharge cycles a system can complete before its capacity falls below a set threshold, commonly 80% of its initial value. Calendar life is the time-based equivalent: ageing continues when a battery is idle, and it accelerates at high temperature and high state of charge, so a system can wear out before it has cycled much. Depth of discharge (DoD) is the fraction of capacity drawn per cycle. Shallower cycles usually extend cycle life, so DoD trades usable energy per cycle against longevity. Degradation rate, expressed as capacity or energy loss per year or per cycle, determines when replacement or augmentation is needed and how much energy remains to be sold in later years. Together these parameters fix the number of revenue-earning cycles a asset can deliver, which is why they dominate project value.

Levelized cost of storage

The levelized cost of storage (LCOS) is the discounted lifetime cost of a storage asset divided by the discounted energy it discharges. It is directly comparable to the levelized cost of electricity (LCOE) for generation technologies. In simplified form:

Here $I_t$ is investment, $O_t$ operation and maintenance, $C^{\text{ch}}_t$ the cost of charging electricity, $R_t$ replacement or augmentation, $E^{\text{dis}}_t$ discharged energy, and $r$ the discount rate. Because most costs are paid up front, the metric is highly sensitive to three assumptions:

• Cycling frequency. More cycles spread the fixed capital over more delivered energy. The lowest LCOS is achieved at maximum utilisation across the discharge durations and cycle counts modelled in one projection.
• Discount rate. A higher rate penalises the capital-heavy profile of storage. One comparison of lithium-ion and vanadium redox flow batteries found that the system LCOE could be lower for VRB when the discount rate is less than 5%, showing that rankings can reverse.
• Lifetime and degradation. Longer life and slower fade lower the cost per delivered unit, while the charging price enters the numerator, so the efficiency loss effectively raises the cost of each unit sold.

LCOS also has limits. It is a cost measure, not a revenue measure, and for services that value power rather than energy, practitioners normalise cost per unit of power capacity instead.

Battery pack prices and learning effects

Battery prices have fallen steeply. BloombergNEF's 2025 survey put the volume-weighted average lithium-ion pack price at $108/kWh, making them 93% lower than in 2010. Stationary-storage packs were priced at $70/kWh in 2025, a 45% decrease from 2024, and the survey reported $99/kWh for battery-electric vehicle packs. Prices are revised between editions, so the year of each figure must be stated; the 2024 survey, for example, reported a higher level. BloombergNEF had also forecast a further decline of about 3% for 2026, to roughly $105/kWh.

The decline reflects a learning effect, in which unit cost falls by a roughly constant fraction with each doubling of cumulative production. Contributing factors include larger factories, higher yields, better cell designs, and the shift toward lower-cost chemistries. Recent price moves also reflect cyclical overcapacity: in 2025, higher metal costs did not increase cell or pack prices, partly because producers offset them through LFP adoption, long-term contracts and hedging. Learning-curve estimates therefore need careful interpretation, since part of a year-on-year drop may be temporary competition rather than sustained technical progress.

Life cycle assessment and energy payback

Life cycle assessment (LCA) accounts for the emissions and energy used in mining, materials, cell and pack manufacturing, operation, and end-of-life treatment. Results depend strongly on the system boundary and assumptions. In a comparison of five lithium-ion chemistries, most chemistries have embodied carbon footprints of around 200 kg CO2e per kWh of useable storage capacity, which corresponds to a much smaller figure per unit of lifetime energy delivered when batteries cycle daily. The same study found that energy delivered on energy invested is 2–4 when only the battery's own energy is counted, but drops to 0.54–0.66 once the electricity used for charging is included. Energy payback is therefore sensitive to whether charging electricity is counted, and comparisons across technologies should state their boundaries and grid assumptions explicitly.

Grid services and system operation

A storage asset connected to the grid can serve several purposes, and the time scale of each service determines which hardware is suitable. Services fall roughly into three groups: sub-second responses to disturbances, regulation and reserves lasting seconds to minutes, and energy shifting over hours. Because every service draws on the same state of charge and power headroom, operators must decide how much capacity to hold back for each one.

Energy arbitrage means charging when wholesale prices are low and discharging when they are high, so revenue equals the price spread minus round-trip losses and degradation cost. Peak shaving applies the same logic to demand: the battery discharges during the highest-load intervals, reducing the peak that a utility or large customer must serve and, in commercial settings, the demand charges. Capacity firming pairs storage with variable renewable output so that the combined plant delivers a more predictable and schedulable profile. These services are paid through different channels: arbitrage through energy prices, peak reduction through tariffs or bilateral contracts, and firming through capacity payments. In many capacity markets, the payment covers only the share of the asset's output that the market operator accredits, and that share depends on how long the asset can sustain its output. In the United States, FERC Order 841 (2018) directed regional operators to remove barriers that kept storage out of energy, capacity and ancillary-service markets.

Frequency regulation keeps grid frequency near its nominal value by following a dispatch signal that moves output up and down, while contingency reserves stand ready to replace a lost generator. Batteries can respond to frequency deviations within milliseconds, which makes them well suited to these fast markets. Frequency regulation accounted for 81.7% of grid scale stationary battery storage market revenue in 2025, according to one market-research estimate, though such figures vary by source. Ancillary markets are small, and the revenue they offer tends to fall once enough fast-responding batteries qualify. GB BESS earned 87% of revenue from frequency response services across 2020 to 2022 (Modo Energy, 2025). They earn 33% now on a gross basis (Modo Energy, 2026). The revenue types also differ in their wear. Arbitrage pays for energy moved, so it requires cycling, whereas frequency response pays for capacity held available, so the battery must be charged and discharged to gain revenue in one case, while in the other revenue depends on the time that power capacity is available. Regulation services still add throughput, which accelerates degradation.

Black start is the procedure by which network operators restore supply after a total or partial shutdown of the transmission system, and most nations require that black-start capability be built into the power grid. A black-start unit needs its own power source that does not depend on the external grid, and this can be an on-site generator, battery storage, or other energy storage systems capable of providing the initial power required for startup. A notable example is the Sentinel Energy Center in California, where a 17-megawatt (MW)/35-megawatt-hour (MWh) system is used to restart a gas turbine after a total loss of power. Battery black start is limited by the finite energy of the battery, so sizing depends on the inrush of auxiliary motors and the time needed to bring the plant back.

Conventional generators provide inertia through their rotating masses, which resist changes in speed and help keep frequency steady. System inertia has typically been provided by synchronous generators in thermal, nuclear, and hydroelectric power plants. "Their large rotating masses acted as shock absorbers that smoothed out any disturbance," said Aleasoft. Most inverter-based resources, however, are grid-following: they lock onto a voltage and frequency set by others and inject power in step with it. As thermal plants retire, grids are left with fewer providers of inertia and system strength. Grid-forming inverters reverse this relationship. Grid-forming batteries use advanced inverter controls to establish and regulate voltage and frequency at their connection point, which lets them support stability through services like voltage regulation, frequency control, and synthetic inertia. Such equipment is commercially available in battery storage systems, and Australia's National Electricity Market has contracted grid-forming batteries to provide voltage support and synthetic inertia. Their inertia is software-defined and bounded by inverter current limits and the energy available in the battery, so it complements rather than replaces rotating machines.

Storage can also defer or relieve network investment. Transmission and distribution lines and transformers are sized for peak loads that occur only a few hours a year, and upgrading them is costly and slow. A battery sited near a constrained feeder can discharge during local peaks, postponing an upgrade, or charge when the line is uncongested and discharge when it is overloaded. The value of such deferral depends on the forecast load growth and on whether a multi-day peak could exhaust the battery.

Operators schedule storage with mathematical optimization. The problem typically combines state-of-charge dynamics, price or market forecasts, network constraints, and a penalty for degradation. Linear and mixed-integer programs are common, although binary charge and discharge modes make the problem harder to solve. One study formulated a linear optimisation approach to account for local energy system participation in the wholesale day-ahead electricity market and multiple frequency response services. In practice, dispatch is often re-solved in a rolling horizon so that it adapts as forecasts change. A throughput-based or depth-of-discharge-based degradation cost is added so that the scheduler only cycles the battery when the expected price spread or service revenue exceeds that cost. Machine learning forecasts and stochastic methods are increasingly used to handle price uncertainty.

Renewable integration and long-duration storage

Variability across timescales

Output from variable renewables changes on daily, seasonal and multi-year scales at once. Solar output peaks in the middle of the day and drops off sharply to zero at dusk, and its seasonal swing is much larger than its daily one. In a heavily solar-based grid, winter output can fall to about a tenth of summer production. Wind follows a different pattern: it is more variable over days and weeks, and its seasonal decline is gentler than solar's. Some variation stretches over years, when some weeks, months, or years are just much less windy. Because solar and wind peak at different times of year, combining them smooths some of this variation but does not remove the seasonal gap.

These patterns create two problems. When output exceeds what the grid can absorb, the surplus is curtailed. In one case, the rate of variable renewable curtailment reached as high as 20% in northwest area of China during the winter. When output falls short, other sources must cover the gap, and in a system with little firm capacity this can mean long periods of deficit. Planners respond with overbuilding, which is installing more IRES capacity than is needed to meet demand, combined with curtailment. A European modelling study found that this combination, with hydrogen used for seasonal storage, reduces overbuilding requirements.

Duration and suitability

Storage duration is the time a system can discharge at its rated power, which equals energy capacity divided by power capacity, $t = E/P$. A typical lithium-ion installation is built for about four hours, which covers daily peaks. Definitions of longer durations vary. One classification divides storage into short-duration (less than 8 hours), medium-duration (8 hours to 24 hours), and long-duration or multi-day (more than 24 hours). The U.S. Department of Energy uses a different threshold: storage systems capable of delivering electricity for 10 or more hours in duration. The DOE also splits longer storage into inter-day (10 hours to 36 hours), multi-day/multi-week (36 hours to 160 hours), and seasonal shifting (over 160 hours). Regulators also differ: the California Public Utilities Commission categorizes an LDES as an energy resource capable of lasting 8 hours maximum discharging.

Economics, not just physics, determine which technologies fit each duration. Lithium-ion economics drop off at roughly 8 to 12 hours, which is where many analysts place the start of long duration. Medium-duration options such as pumped hydro, compressed air and flow batteries are well suited to fill these 8- to 12-hour reliability shortfalls. For longer durations the cost structure matters more. Costs are split between energy capacity (per kWh) and power capacity (per kW). One scenario analysis found that energy capacity cost must be less than or equal to US$10 kWh^-1 for long-duration storage technologies to fully displace alternatives. Hydrogen's projected energy cost is low, but its power-side costs are high and its round-trip efficiency is below 50 percent, so hydrogen pathways face a different cost problem from batteries.

Candidate long-duration technologies

Candidates fall into several groups. Among electrochemical options, flow batteries store energy in external electrolyte tanks, so energy and power scale separately. Hydrogen routes use electrolysis to produce hydrogen, store it, and convert it back through a fuel-cell or turbine. Thermal stores hold heat for later conversion, and mechanical options include pumped hydro, compressed air and gravity systems. The LDES Council's broad definition covers storage that stores energy and then dispatches it as power, heat or cooling for extended periods of time, ranging from 8 hours to days, weeks or seasons. Because the category describes duration and role rather than one technology, LDES describes the duration and role of energy storage rather than one specific technology.

Open questions: how much is needed and how it is valued

There is no agreed answer to how much long-duration capacity a system needs. The answer depends on the generation mix, transmission, and the cost of overbuilding. Some studies find that seasonal storage becomes economic only at low costs. One model of the Western U.S. interconnection found that seasonal operation of storage becomes cost-effective if storage capital costs fall below US$5 kWh^−1. Other work notes that existing studies are limited because of modeling simplifications such as the exclusion of longer duration storage and the exclusion of multi-nodal transmission networks, which means results can differ widely between models.

Valuation raises further questions. The value of storage at longer durations depends on assumptions. One analysis notes that the answer depends heavily on the number of hours at which storage can charge at no cost. NREL work treats seasonal value as the remaining benefit once daily shifting benefits are counted, and it questions the common assumption that charging would mostly use cheap curtailed energy. Network value is also unresolved: network-level operation should be evaluated over full-year horizons to capture long-term and seasonal storage operation, which few studies have done with nonlinear grid constraints. Finally, benefit-cost analysis depends on methodological choices, including what discount rate should be used to value future costs relative to present costs. These choices, together with the differing duration thresholds in regulation, mean that policy targets for long-duration storage are not yet based on a common valuation framework.

Transportation and end-user applications

Electric vehicles as mobile storage

An electric vehicle carries a traction battery that can, in principle, act as a distributed store. Because vehicles spend most of their lives parked, planners treat their batteries as a possible pool of flexibility. Using them this way requires bidirectional charging. The charger or the vehicle must convert the battery's direct current into alternating current when power flows out, and perform the reverse conversion on charging. Unidirectional smart charging only shifts when charging occurs, and it is generally regarded as gentler on the battery. Discharging adds cycles, so the extra wear depends on chemistry, depth of discharge and the control strategy, and the net effect on battery life remains an active research question.

Two main modes are distinguished. In vehicle-to-grid (V2G) operation, a car exports energy to the grid when demand is high or when a service such as frequency response is needed, and imports when supply is in surplus. A single car offers little and unpredictable capacity, so V2G matters at system scale only when many vehicles are aggregated and their owners are compensated. Deployment is constrained by hardware and standards. Only certain models and chargers support bidirectional flow, and the connector standards used for this purpose are not uniform across markets.

In vehicle-to-home (V2H) operation, the car supplies a house instead of the grid. The most visible benefit is backup power during an outage. Some systems include a home hub with an islanding contactor that disconnects the house from the grid so that the vehicle can run it safely, and a meter at the grid connection can command the vehicle to discharge just enough to offset household imports. Owners can also use stored energy during expensive peak hours.

Residential storage with rooftop PV

A home battery paired with rooftop photovoltaics typically stores midday surplus generation and releases it in the evening, when the household draws more power than the panels produce. Its economic value depends on the tariff. Under time-of-use pricing, the spread between cheap overnight electricity and expensive early-evening electricity creates an arbitrage opportunity, and in many markets the highest prices fall on summer weekday afternoons and evenings.

Energy management software decides when to charge and discharge. It forecasts solar output, household load and tariff schedules, tracks the battery's state of charge, and keeps a reserve for backup when required. Modern systems can also respond to utility signals. The result is a trade-off between self-consumption, bill savings, export rules and battery wear, which the software must balance under constraints set by the installer and the owner.

Commercial and industrial uses

Commercial and industrial customers often pay demand charges based on their highest power draw within a billing period. A battery that shaves a short peak can therefore reduce a bill far more than its stored energy would suggest. The same installations provide backup power for critical loads, and they can ride through brief interruptions while generators start.

Data centers have become a major new source of demand. The IEA projects that 20–25 GW of batteries could be installed in data centers globally by 2030, which would allow these facilities to behave more like grid assets than passive loads. Batteries can smooth the rapid swings of AI training workloads, and some developers install storage to secure faster grid connections by guaranteeing demand response. Because grid connection queues can be lengthy, some operators pair gas generators with batteries as a bridge source. Projections vary considerably between forecasters, so the scale of this market should be read with caution.

Off-grid systems, microgrids and rural electrification

A microgrid is a local network of generation, storage and loads that can operate separately from the main grid. Storage is often the element that replaces diesel generation. A remote First Nations community on Harbledown Island, for example, combines 100 kW of PV, a 125 kW storage inverter, 440 kWh of batteries and three diesel generators, with a controller that prioritises solar and keeps generator loading low. In a California desert water project, developers planned solar-plus-storage microgrids at each well, including a 693 kW / 2,664 kWh battery; the developer projected lower energy costs and a large cut in emissions, though these are forecasts rather than measured outcomes.

Utilities are also using such systems to avoid expensive line extensions. NV Energy proposed a solar-and-battery microgrid to replace a diesel system supplying a mountain town, and it argued that combining solar, batteries and ruggedized overhead lines would save money compared with burying lines, while reducing wildfire-related outages. In the events sector, portable batteries and solar units are increasingly rented to replace diesel generators for festivals and emergency response.

Diesel retains advantages in fast start-up and long firm supply, so most current designs are hybrids in which generators remain as backup. Sizing storage for multi-day periods with little sunlight is the main technical challenge in fully renewable off-grid systems.

Safety, resources and environmental impact

Thermal runaway and propagation

Lithium-ion cells rarely enter thermal runaway by chance. Lithium ion battery cells have a very low chance to randomly enter a thermal runaway condition. Electrical, thermal or mechanical abuse can trigger it, and because cells contain large amounts of active chemical material, the reaction is irreversible once it starts (irreversible thermal runaway events can be triggered under abnormal operational conditions, including electrical, thermal, and mechanical abuses).

The first visible hazard is the vent gas. When the internal burst disc opens, the cell releases gas that can be flammable and sometimes burning, on the order of 50 litres at about 1400 °C within two seconds. Depending on age, chemistry and state of charge, the gas can also be corrosive, mainly because of hydrogen fluoride, and it carries conductive carbon and metal particles (depending on cell age, chemistry, and state of charge, might be corrosive as well primarily due to hydrogen fluoride. In addition, conductive particulates, such as carbon and metallics are being carried by this gas.).

Propagation is the system-level danger. Without protective measures, runaway can pass from cell to cell inside a module or pack and end in fire or explosion (thermal runaway propagation may occur within the module/pack due to a lack of protective measures, ultimately leading to fires and explosions). Enclosed housings without dedicated venting can trap the gas and accelerate spread (can constrain these gases, potentially promoting thermal runaway propagation). Simulations also show that burning vent gas in the gaps between cells can carry heat to neighbours, and the location of the vent hole helps determine whether propagation occurs (hot combustion products are rapidly transported in gaps between cells, potentially leading to self-sustained thermal runaway propagation to adjacent cells. The vent hole location is identified as an important parameter). Cathode chemistry matters as well: in tests on NMC-811 cells the vent gas ignited, and high temperatures ruptured cell side walls, which can disrupt the intended venting path (side wall ruptures also happen due to the high temperatures above the melting point of the aluminum cell can).

Engineering responses work at several levels. These include flame-retardant pack cases combined with active cooling, designs that isolate hot rows and columns of cells, dedicated vent channels that route gas out of the pack, and ventilation rates and outlet positions chosen to reduce explosion risk (increasing ventilation rate can decrease the gas concentration and shorten the duration of battery pack under explosion). Fire suppression is harder than in most fuel fires because a cell can reignite after apparent extinction; this is a general engineering observation rather than a settled rule, and practice varies by manufacturer.

Hydrogen and compressed-gas hazards

Hydrogen, used in power-to-gas storage, is hazardous in several ways. It has a wide flammability range in air, low ignition energy, a nearly invisible flame, and it leaks through small openings more readily than natural gas. Compressed tanks, often at hundreds of bar, store large amounts of pressure energy, so rupture and projectile risks must be managed. Some metals become embrittled in hydrogen service. Mitigation relies on gas detection, ventilation, pressure-relief devices, separation distances and leak-tight materials. Ammonia and methanol carriers add toxicity and flammability concerns. Siting and installation are governed by codes and standards, including NFPA 2 (the Hydrogen Technologies Code) in North America, ISO and IEC standards for electrolysers and refuelling, and pressure-vessel codes such as ASME Boiler and Pressure Vessel Code. Local building and fire codes add further requirements. (This paragraph is based on general engineering practice; no source was retrieved for it.)

Critical minerals and supply concentration

Cobalt is the most concentrated supply case. USGS estimates that the DRC produced 230,000 tonnes of mined cobalt in 2025, about 73% of the 310,000-tonne world total. Most cobalt is a by-product of copper or nickel mining (Cobalt is primarily mined as a by-product of copper or nickel mining, with notable exceptions in Congo and Morocco.), and China continues to lead the world in production of refined cobalt. Supply policy has been volatile: after prices fell, the DRC halted exports and later replaced its ban with strict quotas (it was betting that controlling 70 percent of global supply would force Western powers and China to negotiate on African terms).

Lithium, nickel and graphite raise different concerns. Lithium is extracted from hard-rock and brine operations, and brine evaporation is water-intensive in arid regions. Nickel production has expanded rapidly in Indonesia, where Indonesia recently emerged as the second-largest producer of cobalt, accounting for 9.8% of global production, because of the rapid expansion of its nickel industry, and the expansion has raised land-use and forest concerns. Graphite anodes depend heavily on Chinese processing. Mining, refining and processing are linked to water use, habitat loss, and, in artisanal cobalt mining, documented hazards and child-labour concerns. These environmental and social issues are real but vary greatly by site, and the figures above are a snapshot rather than a comprehensive audit.

Recycling, second life and regulation

Recycling recovers metals from end-of-life packs. Three main routes exist. Hydrometallurgy dissolves black mass in acids and can achieve high recovery rates, while pyrometallurgy smelts at roughly 1400–1700 °C to burn off organics and produce metal alloys, and direct recycling aims to keep cathode structures for reuse. Reported recovery rates vary by source. One industry glossary claims that hydrometallurgy (chemical dissolution using acids, achieving up to 98% recovery rates), while a market report puts hydrometallurgical lithium recovery at approximately 80% for lithium — well above the 50–60% lithium recovery typical of thermal routes. Official data are more conservative in aggregate: the IEA reports that rates surged to over 40% for nickel and cobalt and to 20% for lithium in 2023, measured relative to theoretically available feedstock.

Regulation is driving change. In the European Union, the Battery Regulation sets recycled-content minimums and recovery targets. Sources differ on the exact years: one describes targets of 80% lithium recovery and 95% cobalt/nickel recovery rates by 2031, while another cites Commission rules under which 90% for cobalt, copper, and nickel; 50% for lithium by 2027 apply. In the United States, the United States lacks comprehensive federal battery legislation, though several states have introduced Extended Producer Responsibility (EPR) laws covering EV battery collection and battery recycling.

Second-life reuse competes with material recovery. Packs with remaining capacity may serve stationary storage, while damaged or degraded packs go to metal recovery. Second-life deployment is growing from a small base, with one industry estimate placing it at about 25-30 GWh in 2025 toward 330-350 GWh by 2030, though such projections are not independently verified. Open problems include grading state of health, traceability, safety testing and common standards for reused modules.

Historical development

The history of electrochemical storage starts with the voltaic pile. In 1800, Alessandro Volta described a stack of alternating metal discs separated by cloth soaked in brine. Unlike the Leyden jar, which held static charge briefly, the pile sustained a continuous current. Within months, William Nicholson and Anthony Carlisle used it to split water by electrolysis. Humphry Davy later showed that the pile's output depended on chemical reactions rather than on contact between dissimilar metals alone, and he used the current to isolate sodium and potassium. These experiments linked the battery to the emerging science of electrochemistry, and they established the electrolyte and electrode as the core elements of any cell. The pile was a primary cell, however. Its reactions were not reversible, so it could not be recharged.

The first rechargeable battery came in 1859, when the French physicist Gaston Planté built a cell from two lead sheets separated by cloth and immersed in dilute sulfuric acid. Most sources date the invention to 1859, though one gives 1860. Planté's design could be restored by passing a reverse current through it, which made it the first secondary battery marketed for commercial use. Its practical importance grew after Camille Alphonse Faure introduced pasted lead-oxide plates in 1881, which were easier to manufacture in quantity. The chemistry, known as the lead-acid battery, is still used for starting engines and for backup power, though its low energy density limits other uses. Pumped water storage developed along a separate track. Around the turn of the twentieth century, Alpine countries began operating reversible hydro schemes that pumped water uphill when electricity was cheap, and this approach expanded through the early decades of the century as utilities built larger grids. This early form of pumped hydro storage remained the main large-scale storage method for most of the century.

Rechargeable alkaline chemistries followed in 1899, when the Swedish engineer Waldemar Jungner invented the nickel-cadmium battery. The cell combined a nickel-based positive electrode with a cadmium negative electrode in an alkaline electrolyte, and it tolerated abuse far better than lead-acid. Jungner also tried replacing cadmium with iron, but the nickel-iron cell suffered from poor charging efficiency and hydrogen gassing, and he abandoned it. Thomas Edison later developed a nickel-iron battery in the United States, and his company made it from 1903 to 1932. Nickel-cadmium technology matured slowly. Porous sintered electrodes arrived in 1932, and sealed designs followed in the late 1940s. Sealed cells powered portable tools and electronics, and they were common in satellites and rockets during the 1960s and 1970s. Nickel-metal hydride (NiMH) batteries were commercialized in 1991, and they replaced most nickel-cadmium cells in consumer devices. Cadmium's toxicity then led to restrictions in the European Union.

Lithium-based cells were the next major step. Early work by Stanley Whittingham on layered titanium disulfide cathodes and by John Goodenough on lithium cobalt oxide cathodes made high-voltage cells possible. Akira Yoshino then built a practical cell with a carbon anode, and Sony commercialized the first lithium-ion battery in 1991. The three researchers shared the 2019 Nobel Prize in Chemistry. Lithium-ion cells first dominated portable electronics, then powered electric vehicles from the late 2000s. Their energy density and cycle life made them the preferred chemistry for mobile storage.

The move from niche use to grid infrastructure took place mainly in the 2010s and 2020s. Battery pack prices fell by roughly an order of magnitude over the period, driven by manufacturing scale, gains in cell chemistry, and the growth of electric vehicle production, which spread learning effects across the supply chain. Lower costs made it practical to pair batteries with solar and wind generation, and large utility-scale projects such as the Hornsdale Power Reserve in South Australia (2017) demonstrated that lithium-ion systems could provide fast frequency response and other services to the electrical grid. Policy support, falling costs, and the growing share of renewable energy in power systems together shifted storage from a specialized technology to a standard component of power-system planning.

Markets, deployment and policy

Global deployment figures

Deployment figures depend heavily on the source and the metric, so each should be read with its publication year. BloombergNEF's October 2025 outlook expected about 92 GW (247 GWh) of new storage in 2025, excluding pumped hydro, roughly 23% more in power terms than in 2024, and a further rise of about 33% in 2026 to around 123 GW (360 GWh). Its cumulative projection reaches about 2 TW (7.3 TWh) by 2035, roughly twelve times the 2024 total. The International Energy Agency's Global Energy Review 2026, published in April 2026, put 2025 battery additions at 108 GW, 40% above 2024, with cumulative battery capacity about eleven times its 2021 level. A later BNEF count reportedly reached about 112 GW (307 GWh) for 2025. The differences arise from coverage, the treatment of small behind-the-meter installations, and whether power or energy is the headline unit. Pumped hydro is usually reported separately, so totals that include it are not directly comparable with battery-only figures.

Most of the growth is utility-scale. The IEA estimates that about 80% of 2025 battery additions were grid-connected projects. Duration is also lengthening: developers still favour two-hour systems, but four-hour and longer designs are increasingly common. Lithium iron phosphate chemistry accounted for about 90% of deployments, a point treated in more detail in the lithium-ion section of this entry.

Regional leadership

China is the largest market by a wide margin. Its National Energy Administration-based tally showed 136 GW (351 GWh) of operational new-type storage at the end of 2025, up 84% in one year and more than twice the country's 66 GW of pumped hydro. The China Energy Storage Alliance (CNESA) counted a narrower 144.7 GW of new storage in the same period, and gave a broader partial total of 213.3 GW when all storage was included, so the baseline matters. Ember reports that China held more than half of global battery storage capacity by the end of 2025, and the IEA attributes nearly 60% of 2025 additions to China. In June 2026 Beijing raised its 2030 target for new storage to 300 GW, though a frequently cited 370 GW figure comes from an industry white paper rather than from the official plan.

The United States added about 19 GW of batteries in 2025, around 60% more than in 2024, and BNEF expects the United States and China to remain the two largest markets through 2035. Europe added about 6.2 GW, slightly fewer than in 2024, but its utility-scale additions more than doubled to about 4.6 GW. Germany, the United Kingdom, Australia, Canada, Saudi Arabia and Sub-Saharan Africa are expanding quickly, and the IEA highlights accelerating adoption in the Middle East. India, Southeast Asia, Italy and Latin America are expected to scale up later in the decade.

Policy instruments

Early growth in China relied on mandates. New wind and solar projects were conditionally required to pair with batteries, typically sized at 10–20% of the renewable capacity. The mandate delivered capacity but often low utilisation. Renewable co-located batteries lagged standalone systems in annual cycling, and policy reform followed. Document 136, issued in early 2025, ended the co-location requirement for new projects and shifted revenue toward market mechanisms such as peak shaving and frequency regulation. A late-2025 analysis noted that China still lacked a capacity compensation mechanism for storage, but in January 2026 Document 114 extended national capacity remuneration to standalone battery systems.

Other instruments follow similar logic. Capacity payments reward availability rather than energy throughput. Procurement tenders, such as government auctions in India, fix volumes and prices for specific projects. Tax credits, which have been a major support in the United States, are sensitive to shifting federal policy. Storage mandates and market-design reforms, including spot-market integration and the valuation of ancillary services, are the subject of most regulatory work today. China's planning has also moved toward reliability metrics: its latest plan sets firm-capacity requirements for new wind and solar plants and measures storage by its contribution to system adequacy, not only by installed power.

Supply chains and trade

Cell manufacturing is highly concentrated. BNEF estimated that China produced around 557 GWh of stationary storage cells in 2025, a figure based on its own estimates rather than on reported production data. This concentration links storage deployment to trade policy. BNEF lowered its 2025 forecast partly because of United States tariffs on trading partners, particularly China, and because of changes in US policy direction. Governments that wish to reduce exposure have turned to domestic content rules, diversified sourcing and support for local production. Each of these measures has trade-off costs, and they can slow deployment if component prices rise. Energy-security concerns therefore sit alongside cost considerations in most national storage strategies.

Research frontiers and open problems

Solid electrolytes and lithium-metal anodes

Lithium metal is attractive as an anode because its theoretical specific capacity is very high and its reduction potential is the lowest of any element. Pairing it with a solid electrolyte is a leading route to higher-energy cells, partly because conventional organic liquid electrolytes are flammable and can let dendrites puncture the separator. Solid electrolytes fall into polymer, oxide, sulfide, halide and composite families, and their main obstacles lie at interfaces: grain-boundary impedance, poor contact with electrodes, volume changes during cycling, and space-charge effects. An earlier assumption that solid electrolytes were impermeable to lithium metal has been shown to be wrong under a range of compositions and cycling conditions, and lithium filament growth through them is now a central mechanistic question.

On the metal side, the solid electrolyte interphase (SEI) that forms on lithium is uneven and unstable, which leads to non-uniform deposition, dendrite growth, and a risk of short circuits and thermal runaway. Current work pursues artificial SEI layers and composite anodes that incorporate polymers, ceramic particles or carbon nanotubes. Even so, modified interphases often fail after long cycling, and coulombic efficiency remains too low for commercial use, while much of the work is still confined to laboratory cells. Understanding of the SEI's structure and evolution is itself incomplete, and in-operando characterization and AI-assisted analysis are proposed as ways forward.

Sodium and multivalent chemistries

Sodium-ion cells use an element that is abundant and avoids lithium, cobalt and nickel, but the larger and heavier sodium ion generally lowers gravimetric energy density, so research focuses on layered and polyanionic cathodes and on hard-carbon anodes. Multivalent systems based on magnesium, calcium, zinc or aluminium can transfer two or three charges per ion and may offer higher volumetric capacity. The obstacles are slow ion diffusion through host lattices, the scarcity of electrolytes with wide stable windows, and reversible metal plating and stripping at the anode. Progress here is still largely at the materials-discovery stage.

Economics of long-duration storage

Market participation is still dominated by short-duration assets of about two to four hours, and existing capacity markets are generally built around needs of four hours or less. Some markets have begun to reward duration. PJM, for example, gives storage a higher capacity credit as its duration increases, up to ten hours. Cost trajectories matter too: analysts expect an eight-hour lithium-ion system in 2050 to cost roughly what a four-hour system costs today, yet lithium-ion remains unlikely to be cost-effective for multi-day durations.

Valuation is the harder open question. The total value of storage tends to rise with the share of variable renewables, but whether longer durations gain relative value depends on region and grid mix. Tenders written around a familiar short-duration product can exclude longer-duration technologies before their system value is compared, and selection driven by initial capital cost can undervalue assets with longer lives and lower replacement costs. Policy experiments include cap-and-floor schemes and long-term procurement contracts. NREL researchers have called for an open-source evaluation framework that quantifies contributions to resource adequacy, reliability and resilience, a need that reflects the gap in planning and operations models for multi-day uncertainty.

Machine learning and state estimation

A battery management system cannot measure state of charge or state of health directly, so it infers them from voltage, current and temperature. Recent studies report strong results: one machine-learning state-of-power model, using an LSTM network, was validated across temperatures from -20 to 40 °C and reported at least threefold better accuracy than equivalent-circuit and electrochemical models. The same authors note that earlier machine-learning studies had rarely been validated over wide temperature ranges or tested on embedded hardware. Generalization is the central difficulty: a model trained on one chemistry or set of conditions may not transfer to others. Field evidence is growing; one state-of-health framework was tested on a year of data from 700 vehicles, with a maximum relative error of 4.5%. By contrast, a residential photovoltaic tool was tested on simulated data, with the simulation serving as ground truth, which illustrates how hard it is to validate against real operating conditions. Aging datasets are often produced at a single controlled temperature, so transfer to field temperatures and load profiles remains uncertain, and models must also cope with unlabelled data arriving under shifting distributions.

Sourcing, recycling, second life and interoperability

Several issues remain unresolved at the system level. Recycling yields depend on heterogeneous cell formats and chemistries, and the operating history of returned packs is often incomplete, which complicates sorting and recovery of cathode metals. Second-life reuse requires reliable grading of retired cells, and rapid methods such as impedance spectroscopy combined with machine learning are being studied for that purpose. Common standards for pack grading, performance warranties and data exchange between batteries, inverters and energy management software are still being developed. Supply-chain traceability for lithium, cobalt, nickel and graphite is likewise an open governance problem, because few common frameworks yet link material origin to the environmental and labour records used in life-cycle assessments.

References

1. Information Study for Pumped Storage Hydropower Sitingenergy.wsu.edu
2. The world’s water battery: Pumped Storage Hydropower and the clean energy transitionhydropower.org
3. Global Atlas of Closed-Loop Pumped Hydro Energy Storage - ScienceDirectsciencedirect.com
4. For Storing Electricity, Utilities Are Turning to Pumped Hydroe360.yale.edu
5. Pumped storage hydropower explained: how it works and why it mattersPumped storage hydropowerhydropower.org
6. Pumped Hydro Storage Market - Market Size, Sustainable Insights and Growth Report 2025-2032datamintelligence.com
7. Pumped Hydro Storage Market Outlook 2026-2034intelmarketresearch.com
8. Life Cycle Assessment of Closed-Loop Pumped Storage Hydropower in the United Statespubs.acs.org
9. Photovoltaics beyond the panel: system-level net energy under grid and firming constraintsarxiv.org
10. A Comparison of the Environmental Effects of Open-Loop and Closed-Loop Pumped Storage Hydropower (Technical Report)osti.gov
11. Pumped Storage Hydropower in the United States: Emerging Importance, Environmental and Social Impacts, and Critical Considerations - Karambelkar - 2025 - WIREs Water - Wiley Online Librarywires.onlinelibrary.wiley.com
12. New DOE report on environmental effects of pumped storage hydropowerrenewableenergyworld.com
13. Pumped Storage Hydropower in the United States: Emerging Importance, Environmental and Social Impacts, and Critical Considerationspdxscholar.library.pdx.edu
14. Giant Pumped Storage Hydropower Facility Opens in Switzerland - Newseepower.com
15. A Review of Technology Innovations for Pumped Storage Hydropower April 2022publications.anl.gov
16. Pumped-Storage Hydroelectricity - an overviewsciencedirect.com
17. Pumped storage power plant - Hydro News 32andritz.com
18. Compressed Air Energy Storage I CAESsiemens-energy.com
19. The principle of adiabatic compressed air storagenfp-energie.ch
20. Compressed-air energy storageen.wikipedia.org
21. Assessment of the Huntorf compressed air energy storage plant performance under enhanced modificationsresearchgate.net
22. System and method of using a compressed air storage system with a gas turbineimage-ppubs.uspto.gov
23. Compressed Air Energy Storage-Part I: An Accurate Bi-linear Cavern Modelarxiv.org
24. A review of compressed air energy storage (CAES) technology in salt caverns and its technical assessment in China - ScienceDirectsciencedirect.com
25. An Accurate Bilinear Cavern Model for Compressed Air Energy Storagearxiv.org
26. Review and prospect of compressed air energy storage systemlink.springer.com
27. AnyBolt: Flywheel Energy Storage: Principles, Development, and Future Prospectsanybolt.com
28. Applications of flywheel energy storage system on load frequency regulation combined with various power generations: A review - ScienceDirectsciencedirect.com
29. A review of flywheel energy storage systems: state of the art and opportunitiesarxiv.org
30. Flywheel system using wire-wound rotorimage-ppubs.uspto.gov
31. LFP to dominate 3TWh global lithium-ion battery market by 2030 -energy-storage.news
32. LFP vs NMC batteries: Complete fleet owner's guide 2025evinfrastructurenews.com
33. Beyond NMC batteries: Supply chain issues for emerging battery technologiesiea.org
34. Why LFP Became The Dominant EV Battery Chemistry In 2025insideevs.com
35. LFP vs NCM Market Outlook 2026: How Battery Chemistry Is Shiftingiatlas.blog
36. LFP vs NMC Batteries: 2025 Comparisonevtech.news
37. LFP vs NMC Battery: Which Is Better for Your EV in 2026?motorwatt.com
38. Vanadium Redox Flow Battery Costs 2026: Price Drivers, Comparisons & Future Trends - Yakeclimateyakeclimate.com
39. Vanadium flow batteries 2026: where VRFBs beat lithium — Earth Energy Logearthenergylog.com
40. Vanadium Flow Batteries in Energy Storage Systems - GreenEdge Solutionsgreenedge-solutions.com
41. Vanadium Flow Batteries vs Lithium-Ion: The Cost Shiftdiscoveryalert.com
42. ESS Tech, Inc. - Form 10-Q - FY2026sec.gov
43. CATL confirms 2026 large-scale sodium-ion battery deployment in multiple sectorscarnewschina.com
44. Sodium Ion Battery Commercially Viable: What Industrialists Need to Know in 2026tocco.earth
45. A closer look at CATL's new sodium-ion battery - Energy Storageess-news.com
46. Sodium-ion battery momentum grows, but challenges remain - IEAiea.org
47. Solid State Batteries in 2026: From Hype to Adoptionidtechex.com
48. When Will Solid-State Batteries Enter Commercial Production?interactanalysis.com
49. Solid-state batteryen.wikipedia.org
50. Solid-State Battery Hype vs. Reality: 2026 Editionbonnenbatteries.com
51. Double-layer and pseudocapacitance types of electrochemical capacitors and their applications to the development of hybrid devicesresearchgate.net
52. Diatomaceous energy storage devicesimage-ppubs.uspto.gov
53. Frontiersfrontiersin.org
54. Comparative Analysis of Lithium-Ion Batteries and ...odr.chalmers.se
55. Partially and fully surface-enabled transition metal ion-exchanging energy storage devicesimage-ppubs.uspto.gov
56. Superconducting Magnetic Energy Storage Guideankersolix.com
57. Fundamentals of superconducting magnetic energy storage systemselectronics360.globalspec.com
58. Refrigeration of superconducting magnet systemimage-ppubs.uspto.gov
59. Nanoscale manganese oxide on ultraporous carbon nanoarchitectureimage-ppubs.uspto.gov
60. Partially and fully surface-enabled metal ion-exchanging energy storage devicesimage-ppubs.uspto.gov
61. Surface-mediated cell-driven power tools and methods of operating sameimage-ppubs.uspto.gov
62. Electrical Double Layer Capacitance - an overviewsciencedirect.com
63. Pseudocapacitors - an overviewsciencedirect.com
64. Power-to-Hydrogen-to-Power: Technology, Efficiency, and ...oxfordenergy.org
65. Hydrogen technology faces efficiency disadvantage in power storage racespglobal.com
66. Power-to-gas for long-term energy storage - Energy for Growth Hubenergyforgrowth.org
67. Assessment of power-to-power renewable energy storage based on the smart integration of hydrogen and micro gas turbine technologies - ScienceDirectsciencedirect.com
68. Annual Benefit Analysis of Integrating the Seasonal Hydrogen Storage into the Renewable Power Gridsarxiv.org
69. Hydrogen Energy Storage - an overviewsciencedirect.com
70. Liquid Ammonia for Hydrogen Storage - Ammonia Energy Associationammoniaenergy.org
71. Limitations of Ammonia as a Hydrogen Energy Carrier for the Transportation Sectorpubs.acs.org
72. Energy Efficient Transshipment of Ammonia: A Numerical Study of Various Terminal Concepts - Pfeifer - 2025 - Energy Technology - Wiley Online Libraryonlinelibrary.wiley.com
73. Production of ammonia as potential hydrogen carrier: Review on thermochemical and electrochemical processes - ScienceDirectsciencedirect.com
74. Liquid Ammonia in Hydrogen Storage: Energy Density Evaluationeureka.patsnap.com
75. Power-to-Xen.wikipedia.org
76. Molten Salt Storage for Power Generation - Bauer - 2021 - Chemie Ingenieur Technik - Wiley Online Libraryonlinelibrary.wiley.com
77. Molten Salts - an overviewsciencedirect.com
78. Review of commercial thermal energy storage in concentrated solar power plants: Steam vs. molten salts - ScienceDirectsciencedirect.com
79. Phase-change materialen.wikipedia.org
80. Phase change materials for thermal energy storagectc-n.org
81. Building conditioning technique using phase change materials in the roof structureimage-ppubs.uspto.gov
82. Latent heat storage module and latent heat storage apparatusimage-ppubs.uspto.gov
83. Refrigeration device uses phase change material for cooling airimage-ppubs.uspto.gov
84. Experimental study and synergistic performance analysis of phase change material assisted cold thermal storage system for energy efficient air cooling - ScienceDirectsciencedirect.com
85. State of the art on solid–gas sorption based long-term thermochemical energy storage - ScienceDirectsciencedirect.com
86. Sorption Heat Storage - an overviewsciencedirect.com
87. Frontiersfrontiersin.org
88. Experimental investigation on a thermochemical seasonal sorption energy storage battery utilizing MgSO4-H2O - PMCpmc.ncbi.nlm.nih.gov
89. Review of Technologies and Recent Advances in Low-Temperature Sorption Thermal Storage Systemsdoi.org
90. Comparative life cycle assessment of lithium-ion battery chemistries for residential storage - ScienceDirectsciencedirect.com
91. Projecting the Future Levelized Cost of Electricity Storage Technologies - ScienceDirectsciencedirect.com
92. Lifetime coststorage-lab.com
93. Levelized Cost of Energy for PV and Grid Scale Energy Storage Systemsarxiv.org
94. BNEF: Lithium-ion battery pack prices fall to $108/kWh, stationary storage becomes lowest price segment - Energy Storageess-news.com
95. BNEF: Why Global Battery Prices Are Expected to Fall in 2026 - Bloombergbloomberg.com
96. 2024 Lithium-Ion Battery Price Surveybnef.com
97. Grid Scale Stationary Battery Storage Market Size, Forecasts Report 2026-2035gminsights.com
98. How does a battery energy storage system make money? - Researchmodoenergy.com
99. Revenue stacking for behind the meter battery storage in energy and ancillary services markets - ScienceDirectsciencedirect.com
100. Benefits of Battery Storage-Based Black-Start Capabilitypowermag.com
101. Black Start Capabilityumbrex.com
102. Battery Energy Storage System (BESS) With Black Startburnsmcd.com
103. Batteries with grid-forming inverters crucial to keep 100%-renewables grids functioning - Energy Storageess-news.com
104. ​The rise of grid-forming batteries in the NEM - Researchmodoenergy.com
105. Grid-Forming Inverters - GridTechPediagridtechpedia.inl.gov
106. Grid-forming technology and its role in the energy transition - Energy-Storage.Newsenergy-storage.news
107. ©2022 ESIG. All rights Reserved. Grid Forming Battery Storage Julia Matevosyanpubs.naruc.org
108. Frequency Regulation and Battery Throughputgridcog.com
109. Geophysical constraints on the reliability of solar and wind power worldwidenature.com
110. Why Wind and Solar Need Natural Gas: A Realistic Approach to Variabilityitif.org
111. Capacity Expansion of High Renewable Penetrated Energy Systems Considering Concentrating Solar Power for Seasonal Energy Balancearxiv.org
112. Firm wind and solar photovoltaic power with proactive curtailment: A European analysis - ScienceDirectsciencedirect.com
113. Long-Duration Energy Storage: What Is It, Why Do We Need It, and When Will It Get Here? - Clean Energy Groupcleanegroup.org
114. Long-Duration Energy Storageenergy.gov
115. Reliability Value of Long-Duration Energy Storage against Extreme Events in High-Renewable Grids: A Full-Year AC-OPF Assessmentarxiv.org
116. 1 ENERGY STORAGE Beyond short-duration energy storageosti.gov
117. Long Duration Energy Storage Council (LDES Council)ldescouncil.com
118. Long Duration Energy Storage: Benefits & Technologiesziontechnologies.co.nz
119. The value of long-duration energy storage under various grid conditions in a zero-emissions futurenature.com
120. The value of long-duration energy storage under various grid conditions in a zero-emissions futurencbi.nlm.nih.gov
121. Exploring the Future Energy Value of Long-Duration ...osti.gov
122. Energy Storage Benefit-Cost Analysis A Framework for State Energy Programscesa.org
123. Benefit Analysis of Long-Duration Energy Storage in Power Systems withdocs.nrel.gov
124. Long-Duration Energy Storage (LDES) Definition - Renewable Energy Glossaryenergea.com
125. Vehicle-to-grid (V2G) and vehicle-to-home (V2H)gridx.ai
126. What’s Bidirectional Charging and Which EVs Offer It?cars.com
127. Review of Electric Vehicle Charging Technologies, Configurations, and Architecturesarxiv.org
128. What Is Bidirectional Charging for EVsgreenlancer.com
129. Vehicle-to-Home (V2H)gmenergy.gm.com
130. Bidirectional EV Chargers Review - V2G & V2H — Clean Energy Reviewscleanenergyreviews.info
131. Bidirectional EV charging explained - V2G, V2H & V2L — Clean Energy Reviewscleanenergyreviews.info
132. The 300GWh boom in demand for AI data centre battery storage - Energy-Storage.Newsenergy-storage.news
133. Battery Storage For Data Centers In 2026: FEOC Compliance, FERC Co-Location, And The Deals Getting Done Nowlegal500.com
134. Data centers are finding a surprising way to deploy batteriesfortune.com
135. Data Center Battery Storage Statistics 2026: Capacity, Cost, Chemistry & the BESS Infrastructure Race - Axis Intelligenceaxis-intelligence.com
136. Solar+storage microgrids replacing diesel at remote California water projectrenewableenergyworld.com
137. Utilities Are Building Remote Grids Where Power Lines Can’t Reach - Yale E360e360.yale.edu
138. Harbledown Island Off-grid Microgridgenerac.com
139. Replacing “Stinky Diesel Generators” with Mobile Microgrids for Concerts, Live Events, Festivals and Firefightingmicrogridknowledge.com
140. Battery cell pack thermal runaway mitigationimage-ppubs.uspto.gov
141. Experimental investigation of the influence of venting gases on thermal runaway propagation in lithium-ion batteries with enclosed packaging - ScienceDirectsciencedirect.com
142. Thermal Runaway Propagation in Li-ion Battery Packs Due to Combustion of Vent Gases - IOPscienceiopscience.iop.org
143. Thermal runaway propagation in automotive lithium-ion batteries with NMC-811 and LFP cathodes: Safety requirements and impact on system integration - ScienceDirectsciencedirect.com
144. Modeling venting behavior of lithium-ion batteries during thermal runaway propagation by coupling CFD and thermal resistance network - ScienceDirectsciencedirect.com
145. DRC Cobalt Production - 70% of Global Supply - Lobito Corridorlobitocorridor.com
146. Cobalt facts - Natural Resources Canadanatural-resources.canada.ca
147. The DR Congo’s Cobalt Power Move - New Americanewamerica.org
148. Lithium-Ion Battery Recycling Guide: Process, Technology & Market Analysisbenchmarkminerals.com
149. Li-ion Battery Recycling Market Size, Share & Growth Reportmarketresearchfuture.com
150. Executive summaryiea.org
151. Battery Recycling Market Size, Share and Forecast, 2026-2033coherentmarketinsights.com
152. EV Battery Recycling for Investment Recovery: The 2026 Playbook for Second-Life, Resale, and ROI - Invrecoveryinvrecovery.org
153. The History of Electrochemistry: From Volta to to Edisonelectrochem.org
154. Voltaic Pilessparkmuseum.org
155. History of the batteryen.wikipedia.org
156. The Voltaic pile - BatteryIndustry.netbatteryindustry.net
157. Gaston Plant%C3%A9en.wikipedia.org
158. Lead%E2%80%93acid batteryen.wikipedia.org
159. History Of The Lead Acid Battery - Suzuki Batterysuzukibattery.sg
160. Gaston Planté and his invention of the lead–acid battery—The genesis of the first practical rechargeable batteryresearchgate.net
161. Nickel%E2%80%93cadmium batteryen.wikipedia.org
162. Waldemar Jungneren.wikipedia.org
163. Nickel%E2%80%93iron batteryen.wikipedia.org
164. Edison Storage Battery Companyen.wikipedia.org
165. Nickel Cadmium Battery - an overviewsciencedirect.com
166. Battery Universitybatteryuniversity.com
167. BNEF forecasts record 2025 for global energy storagepveurope.eu
168. Global energy storage deployments on track for record year in 2025, BNEF says - Energy-Storage.Newsenergy-storage.news
169. US energy storage market looks resilient amid global growth: BNEFutilitydive.com
170. Global Energy Storage Boom: Three Things to Knowabout.bnef.com
171. Global Battery Storage Capacity Jumps 40% in 2025, IEA Reports - Ecofin Agencyecofinagency.com
172. Technology: Battery storageiea.org
173. Battery storage is scaling up and taking on a larger system roleiea.org
174. Battery Storage Statistics 2026: The Complete Data Referencebitsfrombytes.com
175. From scale to system꞉ navigating the next phase of China's battery storageember-energy.org
176. China puts battery storage at center of renewable energy strategy - Energy Storageess-news.com
177. China’s Batteries Move Beyond Capacity Scale-up: Utility-Scale Storage Could Have Shifted 23 TWh More Clean Power in 2025 - CleanTechnicacleantechnica.com
178. At nearly 200GWh, China's new energy storage deployment rate hit record high in 2025 - Energy-Storage.Newsenergy-storage.news
179. Thirtyfold Growth in Five Years! From China to the World, New Energy Storage Unlocks a Trillion-Yuan Market — China Energy Storage Allianceen.cnesa.org
180. The maturation of China's domestic energy storage marketsolarplaza.com
181. China's Battery Storage Market Shifts from Scale to Utilisationdiscoveryalert.com
182. China Targets 180 Gigawatts of Battery Storage by end of 2027 - IERinstituteforenergyresearch.org
183. Research Progress on Solid-State Electrolytes in Solid-State Lithium Batteries: Classification, Ionic Conductive Mechanism, Interfacial Challenges - PMCpmc.ncbi.nlm.nih.gov
184. Research Progress on Solid-State Electrolytes in Solid-State Lithium Batteries: Classification, Ionic Conductive Mechanism, Interfacial Challengesncbi.nlm.nih.gov
185. Challenges in Lithium Metal Anodes for Solid-State Batteriespubs.acs.org
186. Progress in Modeling and Applications of Solid Electrolyte Interphase Layers for Lithium Metal Anodes - PMCpmc.ncbi.nlm.nih.gov
187. Advancing Metallic Lithium Anodes: A Review of Interface Design, Electrolyte Innovation, and Performance Enhancement Strategies - PMCpmc.ncbi.nlm.nih.gov
188. Recent Advances in Solid-Electrolyte Interphase for Li Metal Anode - PMCpmc.ncbi.nlm.nih.gov
189. Recent progress in understanding solid electrolyte interphase on lithium metal anodepnnl.gov
190. Long-Duration Energy Storage Grid Integration-Valuation ...docs.nrel.gov
191. Exploring the Future Energy Value of Long-Duration ...docs.nrel.gov
192. Exploring the Future Energy Value of Long-Duration ...docs.nlr.gov
193. Long-Duration Energy Storage Has a Market Problemenergycentral.com
194. What makes a ‘priority market’ for scaling long-duration energy storage? - Energy-Storage.Newsenergy-storage.news
195. Long-Duration Energy Storage Grid Integration-Valuation Framework and Incentive Gaps - National Renewable Energy Laboratoryresearch-hub.nrel.gov
196. Battery state-of-power estimation: A machine learning battery model with numerical searching approach - ScienceDirectsciencedirect.com
197. Machine Learning Techniques for Battery State of Health Prediction: A Comparative Reviewdoi.org
198. Overview of Machine Learning-Enabled Battery State Estimation Methodsresearchgate.net
199. Machine learning estimation of battery state of health in residential photovoltaic systems - ScienceDirectsciencedirect.com
200. Dataset for rapid state of health estimation of lithium batteries using EIS and machine learning: Training and validationncbi.nlm.nih.gov
201. GPT4Battery: An LLM-driven Framework for Adaptive State of Health Estimation of Raw Li-ion Batteriesarxiv.org
202. Recent Progress in Understanding Solid Electrolyte Interphase on Lithium Metal Anodes (Journal Article)osti.gov
203. Exploring the Future Energy Value of Long-Duration Energy Storage: Article No. 1751 - National Laboratory of the Rockiesresearch-hub.nlr.gov