A review of pumped hydro energy storage

Many existing PHES systems have been developed in conjunction with a conventional river-based hydroelectric system. Two reservoirs are created, at different altitudes, but close to each other (figure 9). Often, the lower reservoir is large and located on a substantial river, while the upper reservoir is smaller, and located higher up on the same river or in a high tributary or parallel valley. Most river water passes through the system, generating electricity, and then flows on down the river. Some water is cycled between the two reservoirs to create energy storage. Typically, pumping would take place by buying electricity during times when prices are low, which is when demand is low or the availability of electricity from other sources is high (e.g. a windy and sunny day). Generation would take place during times of high demand (such as during evenings) when prices are high. This pattern of buy-low and sell-high is called arbitrage.

Zoom In Zoom Out Reset image size

An electricity transmission and distribution system is unable of itself to store significant quantities of energy. As demand rises and falls, so too must supply of electrical power to the grid, either from a matching variation in generation or through the use of storage as a buffer. A large electrical grid serving many customers will generally experience slowly fluctuating demand over a timescale of hours, because it is statistically unlikely that many customers would act in concert over a timescale of seconds or minutes.

In a traditional fossil or nuclear electricity system, nuclear and coal power stations operate continuously with little or slow (hours) variation in output. Fluctuating demand is matched by fluctuating output from peaking plant, typically gas turbines or hydroelectric power stations (including PHES). However, unexpected failure of a major generator or transmission cable can have an immediate large effect on power supply.

System failure is typically accommodated by ensuring that some additional generators are operational and ready to deliver power to the grid at short notice. For example, gas turbines and hydro generators can ramp to full power in minutes. Additionally, some non-critical loads can be quickly interrupted to ensure that energy continues to flow to other loads. To cope with supply problems on time scales of seconds, the heavy rotating mass of a generator in a fossil or nuclear power station represents rotational inertia which acts as an energy buffer.

In the future, wind and solar energy will supply most of the energy to the grid in many countries. Ideally, there will be strong interconnection of tens of thousands of wind and solar generators over millions of hectares via high voltage powerlines. Weather-related fluctuations in energy delivery will be slow (hours to days) because weather systems including cloud and wind bands move relatively slowly across the landscape. Technical failure rates of large numbers of small generators can be predicted statistically with high precision. Thus, the overall supply of PV and wind energy will be less prone to rapid unpredicted variations than in fossil or nuclear energy systems which have a small number of large generators, any one of which could fail unexpectedly.

However, failure of a high-power transmission line can cause rapid reduction in energy supply in any grid. Batteries respond very quickly (sub-seconds) to disturbances in frequency by injecting or absorbing energy, thus providing 'synthetic inertia'. The capabilities of a grid-scale battery to provide balancing support in a large system dominated by wind and solar has been demonstrated in South Australia [12].

PHES has characteristics that are well-suited for balancing large amounts of variable, inverter-based wind and PV. PHES has rapid response (from idle to full output in a time span of 20 s to a few minutes). PHES has rotational inertia if the generator is spinning, to replace the loss of the rotational inertia associated with conventional thermal generators when they retire. PHES has black start capability, meaning that an electricity system can be restarted after complete collapse of supply without the need for electricity supply to start the generators. Together, batteries and PHES can completely replace the ancillary services hitherto provided by fossil and nuclear generators [13].

Nearly all existing pumped hydro systems are river-based. In many places, there is substantial environmental and social opposition to damming or modifying more rivers. However, there are alternative methods of constructing PHES that do not require significant modification to river systems. One method is to connect closely spaced existing reservoirs using underground tunnels and powerhouses. With care, there is low disturbance at the surface. One example is the 2 Gigawatt, 350 Gigawatt-hour, Snowy 2.0 system currently under construction underground in the World Heritage Kosciuszko National Park in Australia [11].

There is large scope for off-river (closed loop) PHES systems that are located away from any significant river. Most land is not near a river, and so a survey of potential pumped hydro sites that is confined to rivers will miss most potential sites.

An off-river PHES system (figure 10) comprises a pair of artificial reservoirs spaced several kilometers apart, located at different altitudes, and connected with a combination of aqueducts, pipes and tunnels. The reservoirs can be specially constructed ('greenfield') or can utilise old mining sites or existing reservoirs ('brownfield'). Off-river PHES utilises conventional hydroelectric technology for construction of reservoirs, tunnels, pipes, powerhouse, electromechanical equipment, control systems, switchyard and transmission, but in a novel configuration.

Zoom In Zoom Out Reset image size

The energy that is stored in an off-river PHES system is usually lower than in a major river-based hydroelectric dam with similar power rating. For example, a pair of 100 Ha reservoirs with a head of 600 m, an average depth of 20 m, a usable fraction of water of 90% and a round trip efficiency of 80% can store 18 Gigalitres of water with energy potential of 24 GWh, which means that it could operate at a power of 1 GW for 24 h. This is much smaller than the Three-Gorges Dam in China (23 GW, 87 000 GWh annual energy production) but much larger than a utility-scale battery such as the Hornsdale Power Reserve in Australia (0.15 GW, 0.2 GWh) [14].

An off-river PHES system has the advantage that flood mitigation costs are minimal compared with a river-based PHES system. Heads are generally better than river-based systems because the upper reservoir can be on a high hill rather than higher in the same valley as the lower reservoir. Environmental costs of damming rivers are avoided with off-river PHES, which helps with social acceptance. The much greater number of off-river sites compared with on-river sites allows much wider site choice from environmental, social, geological, hydrological, logistical and other points of view.

Another advantage is that construction of off-river pumped hydro can be much faster than other storage methods. Bespoke engineering in mountainous river valleys is unnecessary. Work can proceed in parallel on the two reservoirs, the water conveyance, the powerhouse and the transmission. Construction timetables of 2–3 years are feasible for 10 GWh storages, although longer periods would be typical. Although an individual battery can be completed in a few months, it is typically two orders of magnitude smaller than an off-river pumped hydro system, so the true speed of installation in terms of MW month⁻¹ and Megawatt-hours (MWh) month⁻¹ are modest.

Locating good sites for PHES is not easy even for experienced hydro engineers. The first requirement is to find places where reservoirs can be constructed that store a large amount of water compared with amount of rock and other material used to construct the reservoir walls.

The second requirement is to find closely spaced pairs of sites that have large differences in altitude ('head'). The former requirement is because pipes and tunnels connecting the two reservoirs are expensive, and the latter requirement is because doubling the head doubles the storage energy volume and storage power capacity but does not double the system cost.

The use of efficient computer algorithms [15] is key to searching large areas for good sites. A global survey of greenfield off-river PHES was undertaken by the Australian National University. A total of 616 000 good sites [16, 17] around the world were found in the latitude range 60° N to 56° S. Each site comprises a closely spaced reservoir pair with defined energy storage potential of 2, 5, 15, 50 or 150 GWh. All identified sites are outside of major urban or protected areas. Each site is categorised into a cost-class (A through E) according to a cost model described below, with class A costing approximately half as much per unit of energy storage volume as class E.

For context, to support 100% renewables electricity (90% wind and solar PV, 10% existing hydro and bio), Australia needs storage [18] energy and storage power of about 500 GWh and 25 GW respectively. This corresponds to 20 GWh of storage energy and 1 GW of storage power per million people. Australia is an isolated country, and has high energy use per capita, similar to the aspirations of most countries.

The combined storage potential of the 616 000 identified sites is 23 million GWh (figure 11), which is 150 times more than required to support a 100% renewable global electricity system using Australia as a per capita benchmark. The identified sites are preferably widely distributed to support regional grids. An Atlas of off-river PHES sites is available at the Australian Government's Australian Renewable Energy Mapping Infrastructure sites at [16]. Users can pan and zoom to see detail of all 616 000 sites (figures 12 and 13).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The ANU is now working on expanding this work to a brownfield survey. The aim is to survey the vicinity of existing reservoirs and mining pits to find out whether there is a possible matching reservoir site that could be used to form a good pumped hydro pair.

The capital cost of a river-based hydroelectric system is highly dependent upon local geology, geography and hydrology. Such systems are frequently located in mountainous regions with difficult road access and lengthy distances for transmission. Environmental considerations are (hopefully) a high priority and constraint, both for the river and possibly also for surrounding mountains. The river needs to be diverted during construction, and provision must be made for major flooding during construction or subsequently. This means that each system is bespoke.

In contrast, the cost of an off-river pumped hydro system is relatively predictable. Cost models for a river-based hydro system are not applicable to an off-river system. The absence of a river eliminates major flooding as a criterion, and also eliminates the environmental consequences of flooding a river. Systems can be built in farmland with good road access and close to high voltage transmission systems. The volume of stored water (and the area of flooded land) is far smaller than a typical river-based hydroelectric system. Since there are two orders of magnitude more off-river than on-river sites in a typical landscape, it is generally straightforward to select an alternative nearby site if problems arise relating to geology, hydrology, road access, powerline easements, land ownership, indigenous rights, environmental impacts or social opposition. Most off-river sites are similar from key points of view, allowing a substantial element of 'copy and paste' to be employed in a large-scale storage construction program.

The cost of a hydroelectric system comprises six elements:

• (a)  
  Planning and approvals
• (b)  
  Construction of reservoirs
• (c)  
  The water conveyance: tunnels, pipes, aqueducts
• (d)  
  The powerhouse including pump/turbine, generator, switchyard and control
• (e)  
  Access: roads, electricity transmission and water (for off-river systems)
• (f)  
  Operations and maintenance over the life of the system.

The capital cost of an energy storage system has two components: an energy cost ($ GWh⁻¹) and a power cost ($ GW⁻¹). Sometimes these components are conflated into a single number (e.g. $ GW⁻¹) by using a fixed storage time such as 6 h. This can sometimes be useful when comparing similar systems but is misleading when comparing different systems such as batteries and pumped hydro. A battery typically has a storage time of 1 h; i.e. it can operate at full power for one hour. Thus, a 1 h battery with a power of 0.1 GW has an energy storage of 0.1 GWh. In contrast, a 1 GW off-river pumped hydro system might have 20 h of storage, equal to 20 GWh.

Planning and approvals are generally easier, quicker, and lower cost for an off-river system compared with a river-based system.

The cost of storage energy ($ GWh⁻¹) primarily relates to the cost of reservoir construction. The cost of constructing an off-river reservoir includes moving rock to form the walls, a small spillway and a water intake. Other significant costs could include road access, water access, lining the bottom of the reservoir to mitigate water leakage and placing evaporation suppressors on the water surface. Forming the walls is usually the dominant cost and can be approximated by the cost of moving rock ($ m⁻³). The amount of energy stored in a hydro system is proportional to the head and to the usable water volume of the reservoirs. The important reservoir metrics are (a) the head and (b) the ratio of water impounded to the rock required to form the reservoir walls. Doubling the head or doubling the water/rock (W/R) ratio both approximately halve the effective cost of energy storage ($ GWh⁻¹).

The cost of storage power ($ GW⁻¹) primarily relates to the cost of the water conveyance and the powerhouse. Additionally, transmission is sometimes a significant cost depending on distance to a high voltage powerline.

The expensive component of the water conveyance is the high-pressure pipe or tunnel that spans most of the altitude difference between the reservoirs. This pipe or tunnel is preferably short and steep. The water conveyance can include an aqueduct or low-pressure pipeline to minimise the length of expensive high-pressure pipe or tunnel if the local geography permits. A doubled power rating requires a pipe/tunnel with larger cross-sectional area to convey double the volume of water per second, but the cost increases less than proportionally. A doubled head halves the flow of water for a given power rating, thus reducing the cost of the pipe/tunnel. Thus, large head, large power rating and small effective separation of the reservoirs reduce the water conveyance cost in terms of $ GW⁻¹.

The powerhouse cost in terms of $ GW⁻¹ benefits from doubled power rating; although the volume of water flow per second doubles, the cost of the pump/turbine increases less than proportionally. Doubled head is desirable because water flow halves for a given power output, allowing a smaller pump/turbine to be used (albeit with higher pressure rating).

In summary, the energy cost ($ GWh⁻¹) is minimized by having large head and large ratio of usable water volume to volume of rock needed to form the upper and lower reservoirs [17]. The power cost ($ GW⁻¹) is minimized by having large head and large average slope of the pressure pipe or tunnel. As with many engineering enterprises, systems with larger power and energy are cheaper per unit than smaller systems.

Access to roads, water supply and transmission lines is highly site dependent. Since there is a wide range of sites to choose between, such costs can be minimized.

Annual operation and maintenance costs plus major refurbishments after 20 and 40 years cost about 1% of the initial capital cost each year. This corresponds to about 20% of the annualised capital cost assuming 60 year lifetime and 5% real discount rate.

Figure 14 shows the indicative capital cost of 1 GW off-river pumped hydro storage systems [17]. The importance of large head (500 m and above), large slope and large W/R ratio is illustrated. Systems with large energy storage volume cost more than smaller systems, but not proportionally so. The capital cost of high-quality systems with large storage volumes, head, W/R ratio and slope converge to similar numbers because the 1 GW powerhouse emerges as the dominant cost.

Zoom In Zoom Out Reset image size

In many countries, no pumped hydro scheme has been constructed for many years (if ever). It is not unusual in the early stages of an industry for costs to be higher than expected. However, rapid cost reductions are possible as companies quickly learn to do things better.

Several items must be accounted for when calculating the cost of storage:

• (a)  
  The capital cost of the system
• (b)  
  The annual cost of operations and maintenance, including periodic major refurbishments
• (c)  
  The amount of energy sold by the storage each year and the price received relative to the price paid for the energy sent to the storage
• (d)  
  Losses in the energy storage cycle
• (e)  
  The operational lifetime
• (f)  
  The real discount rate.

The discount rate reflects the fact that people value money in the hand more highly than the promise of future money—which is why interest is charged on money that is loaned. A risky investment uses a higher discount rate. Almost all the costs of a pumped hydro system are up front, similar to a solar or wind power station, but unlike a gas power station where most of the costs are for fuel.

A typical real (after subtracting inflation) discount rate for a low-risk investment is 5%. Pumped hydro, solar and wind energy system costs are sensitive to the discount rate while gas and coal power systems are sensitive to changes in fuel prices. For a hydro system with a lifetime of 60 years, real discount rates of 1% or 12% approximately halve and double the levelized cost of storage respectively relative to a discount rate of 5%.

If one system cycles twice as often per year compared with another then the capital cost is spread over twice the volume of sales and the levelized cost of energy storage is approximately halved.

Figure 15 shows the levelized cost of storage for a range of parameters assuming 5% real discount rate, 60 year operational lifetime and 180 or 360 cycles yr⁻¹. The levelised cost of storage in this context means the average difference between the purchase price of energy used to pump water to the upper reservoir (which is set by the external market and assumed to be $40 MWh⁻¹ in this example calculation) and the required selling price of the energy from the storage. The required selling price is higher than the purchase price to cover the capital and operational costs over the system lifetime and the round-trip storage energy losses.

Zoom In Zoom Out Reset image size

In a real pumped hydro storage income from arbitrage may be highly non-uniform, with a large proportion coming from very high prices during occasional stress periods for the electricity network, such as during heat waves (caused by air conditioning) or supply failures elsewhere in the network. Revenue from ancillary services may also be important, including payments for provision of black start and rapid start capability, voltage and frequency maintenance, rotational inertia of the generator (to cover sub-second glitches), and network insurance (maintaining full upper reservoirs over long periods).

Detailed analysis is required to calculate the amount of storage required to support an electricity system that depends mostly on variable wind and solar PV.

Solar PV and wind energy comprise two thirds of net new generation being constructed around the world. In some countries they comprise nearly 100% of generation power capacity additions. They are both variable energy sources, with power output rising and falling in response to the sun and the wind. When small amounts of PV and wind are added to an electricity system, the existing storage and fast-response gas generators can ensure that the system remains stable. As more and more PV and wind are added, additional measures will eventually be required.

Methods of ensuring that energy supply and demand in an electricity system is balanced on every time scale from sub-seconds to months include the addition of storage; the addition of high voltage transmission to smooth out local weather and demand fluctuations by importing and exporting electricity; and management of demand to reduce peak demands and respond quickly to supply interruptions.

A recent study [18] examined the amount of storage required to support a 100% renewable electricity system in Australia which derives 90% of its energy from variable wind and solar and 10% from existing hydro and bioenergy sources. The analysis matched historical supply and demand for every hour of the year over many years by including sufficient solar and wind generation, storage and transmission. Australia is an industrialized country with high per-capita consumption of electricity by world standards (10 MWh person⁻¹ yr⁻¹). It is isolated from its neighbors, has good solar and wind resources, and is installing solar and wind faster per capita than any other country. It is a global pathfinder [4] for adapting its electricity network accordingly.

Australia lies mostly in the sunbelt (lower than 35 degrees of latitude) where there is low seasonal variation of the solar resource and no cold winter. About three quarters of the global population lives in the sunbelt, including countries in which most of the growth in population, energy consumption and Greenhouse emissions is occurring. Australia represents a state that much of the world's population might aspire to and reach over the next few decades.

Australia's National Electricity Market spans about 1 million km² in the eastern and south eastern parts of the continent. Strong long-distance transmission was found to minimize the amount of required storage. Broadly speaking, the study concluded that the required storage power and storage energy are 1 GW and 20 GWh per million people respectively. The amount of energy storage required is similar to the average daily electricity consumption (27 GWh d⁻¹ per million people).

The storage requirements for a particular country would need to be determined by detailed calculations. An approximate rule of thumb for the amount of storage needed to support a large-area electricity network with high levels of variable solar and wind is 1 d (24 h) of energy consumption. This allows the day-night cycle of solar energy output to be accommodated. This storage could be a combination of pumped hydro and batteries. Demand management is equivalent to storage and can also contribute. Furthermore, occasional spillage of solar or wind electricity on sunny and windy days when storages are full is economically preferable to overbuilding storage to absorb all the generated energy. Typical spillage of solar and wind electricity in optimized hourly-resolution modelling studies is 5%–25% [18].

The area of land required for the upper and lower reservoirs per GWh of storage is about 12 hectares for an off-river pumped hydro system with a head of 400 m, generation efficiency of 90%, usable water volume of 85% and average water depth of 20 m.

Taking an energy storage volume requirement of 27 GWh per million people (the one-day-storage rule of thumb estimated above), this corresponds to 3 m² person⁻¹, which is about the same area as a queen-sized bed.

The land flooded for off-river pumped hydro is relatively small and can avoid sensitive areas. For example, the electricity storage needs of a million-person city could be provided by an off-river PHES system with a power rating of 1 GW and one day of storage that floods 3 km² of land located away from any river and outside environmentally sensitive areas. This is vastly less than the area required to deliver an equivalent amount of energy from conventional hydropower with commensurately less impact. This is due to the much shorter storage durations (hours rather than months) and the frequent recycling of the reservoir volume. Because most countries have a large over-abundance of potential sites, if one site is problematic then an alternative site could be selected.

A solar PV system in a typical sunbelt location will generate about 160 GWh km⁻² yr⁻¹, assuming a system capacity factor of 18%, panel efficiency of 20% and land coverage by the panels of 50%. If average electricity consumption is 10 MWh person⁻¹ yr⁻¹ (similar to Australia and Singapore) then the area of land required for the panels is 60 m² person⁻¹. This is 20 times larger than the area of land required for the supporting PHES system estimated above.

Floating PV could be located on pumped hydro reservoirs provided that the floats are designed to accommodate turbulence and rapid fluctuations in water depth. In the case of off-river pumped hydro reservoirs, the reservoir area per person is only 5% of the per capita area requirement to achieve 100% solar electricity. Floating PV will provide the additional benefit of reducing evaporation in arid regions.

In summary, the land area required for off-river PHES systems to support high levels of variable solar and wind generation is relatively small and can be selected to minimise sensitive land.

The volume of water required per GWh of energy storage is about 1 Gigalitre for an off-river pumped hydro system with a head of 400 m and generation efficiency of 90%. Doubling or halving the head halves or doubles the water requirement respectively.

Taking an energy storage volume requirement of 27 GWh per million people (the rule of thumb estimated above), this corresponds to 27 kl person⁻¹. This water is not actually consumed—it cycles indefinitely between the upper and lower reservoirs. If a fleet of off-river PHES systems were constructed over a 25 year period to support the development of a solar and wind electricity system, the amount of required water for the initial fill would be 3 l person⁻¹ d⁻¹.

After completion and filling of the reservoirs, evaporation losses need to be replaced. In many places, annual rainfall and evaporation over the reservoir area are approximately balanced. Evaporation suppressors (small plastic objects floated on a reservoir to reduce wind speeds and evaporation rates) can tip the balance in favour of rainfall in arid regions. In a dry country such as Australia, pan evaporation in the regions that are prospective for off-river pumped hydro (along the Great Dividing Range) is about 1600 mm yr⁻¹ [19]. Evaporation from lakes in hot climates is about two thirds [20] of the pan evaporation rate (i.e. 1100 mm yr⁻¹). Since rainfall in these regions is about 750 mm yr⁻¹, the annual difference between evaporation and rainfall is about 350 mm yr⁻¹. Since the required combined area of upper and lower reservoirs per person (calculated above) is 3 m² person⁻¹, the evaporation rate is about 3 l person⁻¹ d⁻¹.

In summary, the amount of water required for the initial fill and to replace evaporation is about 3 l person⁻¹ d⁻¹. This is similar to the amount of water used by a person in 20 s of a typical daily shower.

An electricity system based mainly on wind, solar and PHES rather than coal fired power stations will benefit from the absence of water loss in cooling towers, which is an order of magnitude larger per person [21].