This post will review some basic facts about energy storage in general, and pumped storage in particular, because these concepts will come up throughout the later sections.
First, I have to talk about how energy storage should be quantified. (The way it’s usually done is misleading, at best.) An energy storage plant functions like a rechargeable battery. At full charge, it holds a certain amount of energy. At full discharge, it has given all of that energy up. So the amount of energy the battery can store is one important number. The other is how fast the battery can discharge, or release, its stored energy (and how fast it can charge, or take energy back in; the charge rate and the discharge rate may be significantly different, or about the same). So we need two (or sometimes three) numbers to describe the capabilities of a rechargeable battery, for purposes such as comparing it to a different, competing battery. One number is not enough.
Grid-scale storage works the same way, so to define “how big” a storage plant is, we need at least two numbers. The first is the energy stored. The second is the maximum power, or rate at which it can release that energy into the electrical grid.
Power is energy per unit time, so if we know the stored energy and the power, we can compute the time it will take to go from full charge to full discharge. For example, a battery might store 2 hours’ worth of energy, meaning that it can produce its peak power for 2 hours before it’s fully discharged and has to stop. This is often called duration. It’s not independent of power and energy—energy is power times duration, so if you know two of them, you can compute the third.
What many publications do is quote a single number: peak power. They might write about a new deal for “200 megawatts of storage.” It would make sense to say a generating plant produces 200 megawatts of power; but a storage plant is not the same as a generating plant—it cannot be described by one number, and peak power is the less meaningful number if you insist on quoting just one. A capacitor that will fit in your hand can produce gigawatts of power—for an insignificant amount of time. Energy is a much harder hill to climb. There is nothing that will fit in your hand that can store a gigawatt-hour of energy and then give it back. Stored energy is what matters most—and that number is almost never mentioned in the press release!
Another common, but also misleading way to describe the size of a storage plant is how much energy it’s expected to provide to the grid over a certain period, such as a year. You might read, “this plant will provide 10 gigawatt-hours of electricity per year.” That does not mean the plant can store that much energy all at once. In fact, it says nothing at all about how much energy it will store, because the number is based on certain projections of how much charging and discharging the plant will do, which depends on arbitrary assumptions about things like the weather and nearby demand in the coming year. It’s really just a way of attaching a big-sounding number to the press release.
So, a given storage system might produce a lot of power for a short period of time, or a little power for a long period of time, or any combination at all. High-power, short-duration storage is useful in many ways, such as handling the first seconds of unexpected outages. On the other hand, outages of sun or wind power can be quite long—days or weeks—and so long-duration storage is needed to supplement those energy sources. Long-duration storage at high power levels, at reasonable cost, is what we need in order for renewable energy to really start dominating the grid.
Units of Measurement
I’ll briefly review some of the units that are relevant to hydroelectric applications. Power is measured in watts (W). 1,000 watts is a kilowatt (kW), and in utility-sized doses, 1 million watts is a megawatt (MW), 1,000 megawatts is a gigawatt (GW), and 1,000 gigawatts is a terawatt (TW).
Energy is measured in joules (J), but it’s common to use units of power times duration, as we discussed above: 1 watt-hour is the energy it takes to provide one watt of power for one hour. At larger scales, we have kilowatt-hours (kWh), megawatt-hours (MWh), gigawatt-hours (GWh), and terawatt-hours (TWh).
In the metric system, one liter of water has a mass of one kilogram (kg). A 1 kg mass weighs 9.8 newtons (N), meaning it feels that much force pulling it downward.
One cubic meter of water is 1000 liters and has a mass of 1000 kg, for which another name is 1 ton (that is, a “metric ton,” sometimes spelled “tonne”). An Imperial unit that’s used for large volumes of water is the acre-foot, the amount of water that will cover one acre of area one foot deep.
Hydraulic head, or the elevation difference between two bodies of water, is a distance, in meters or feet. The hydrostatic pressure at the bottom of a column of water is proportional to hydraulic head. In Imperial, the pressure is about 0.4 psi of pressure per foot of head. In metric, that’s 0.088 atmosphere per meter.
With any storage method, if you put in 100 units of energy, you will get less than 100 units back. How much you get back depends on the method. While efficiency is an important and interesting topic, it is often talked about as if it were a first-order measure of goodness. That this is not the case is shown by the example of California, which now has so many solar panels and wind turbines that a great deal of energy has to be thrown away, through curtailment, because there’s no use for it at the time it’s available. A system that would store some of that energy at low cost, and return even 30% of the stored energy at a different time when it’s needed and unavailable, would be an unquestioned winner, in spite of its low efficiency.
What is the right size for an energy storage facility? It’s complicated. Suppose you have a solar and wind farm, and you want to power a city with it. The power you get from your solar panels and wind turbines varies with the time of day, the season, and the weather. You need storage to keep the power flowing when sun and wind aren’t producing enough. Keep in mind that the power demand also varies, for many reasons (such as high air conditioning use on hot days).
On an ideal day (a 24-hour period, that is), it’ll be sunny for all of the daylight hours, and windy enough the rest of the time to meet your city’s power needs (assuming your solar/wind farm is big enough). So on ideal days, you don’t need any storage at all.
There will also be “nearly ideal” days, when, for example, you get through the day on solar, but in the evening there’s a large A/C demand spike, the solar output is falling, and it’s not windy. So you might need an hour or two of extra power from storage. Or there might be some windless hours at night—the demand might not be that large, but you need storage to provide it.
Then there will be other days that are less and less ideal, all the way to the worst-case scenario: multiple days in a row when perhaps it’s overcast, so solar power is far below peak, and there’s no wind. You’ll need a lot of storage to cover that.
The point is that if you start with no storage at all and start adding units of storage, each new unit of storage gets used less than the ones before it. Let’s make up some numbers. Suppose there are 20 ideal days in an average year—days when you’re fine without any storage at all. The first unit of storage will sit idle on those 20 days, and will be called into service on all other days, so it will work about 345 days a year.
Now let’s say that of those 345 days, the first unit of storage gets the job done for 20 “near-ideal” days when you only need storage for brief periods. That means the next unit of storage will only be called upon for 325 days. If those two units take care of another 20 days, the next unit is only needed for 305 days. And so on.
Taken to its conclusion, consider the worst situation, where there’s insufficient wind and solar power for, say, a week. By definition, this is very rare, like a 100-year flood; it may happen only once over the lifetime of the storage facility. So all the storage units that are needed for that event, and only that event, will only ever get used once in their entire service life.
This is unfortunate. All the storage units cost the same, assuming they’re all identical modules. Yet some of them are used all the time, and some of them are basically always sitting there idle. The more storage you have, the worse the cost/benefit ratio gets.
This is why relatively costly storage, such as batteries, is never built to even try to meet power needs 100% of the time. It would cost way too much. The solution is to fall back on other power sources, typically natural gas. The capital cost of a natural gas plant, per unit of available power, is much less than batteries of the same power rating. The duration of the gas plant is infinite as long as there’s gas in the pipeline. And of course, when the natural gas plant runs, it’s emitting CO2, but that isn’t so bad, if it’s only called upon for that once-per-decade run of very unfavorable weather.
So, sizing a storage plant appropriately is not simple. The one certainty, in our current utility model, is that blackouts and brownouts are unacceptable (though this is likely to change in the future). That means there must be a backup power supply with effectively infinite storage at the lowest possible capital cost. This will continue to be natural gas for some time (as I discussed in The Gas Battery). The right amount of more expensive, but carbon-free storage sitting “ahead of” the gas plant will be a complex function of weather, demand, how long it takes to build, and who pays for it all.
Overcapacity and Curtailment
One stratagem being used now in the utility-scale renewables world is overcapacity, that is, building out solar and wind farms that are significantly larger than what it would take to supply peak demand in ideal conditions. This makes sense in the current environment in which solar and wind costs per watt have plummeted, but storage remains expensive. Overcapacity works because there are many hours in the year when wind and solar are not zero, but are marginal. During one of those hours, an “ideally”-sized wind and solar capacity would fall short, but an overbuilt system, in spite of marginal solar and wind conditions, can get the grid through that hour without a brownout. (Many of those hours are in the late afternoon, when sun power is waning but demand is spiking. Or on breezy but not windy nights when you can get through if you have enough wind turbines.)
The drawback of overcapacity is that once the oversized system is online, there are also many more hours per year when more wind and solar power is available than the demand—there just isn’t anyplace on the grid that can take that power. The solution is “curtailment”: some solar panels are disconnected, and some wind turbines are feathered so they don’t spin. Curtailment isn’t a terrible thing, and it doesn’t hurt the equipment. But it does mean diminishing returns on capital invested. So there’s a complex tradeoff in a solar/wind-plus-storage utility setup. More storage means less curtailment, because the surplus power has a place to go, but also means diminishing returns on the additional storage. The optimum plan can only be found by modeling the whole system, using real data, and running many scenarios on a computer to search for the best outcome.
Renewable energy supply (and demand, for that matter) has both predictable and unpredictable aspects. At one extreme, the lack of sunlight at night is a very predictable phenomenon. So is the seasonal variation in solar intensity. Weather causes fluctuation in solar collection, always to the downside of potential (weather never makes the sun unusually bright). Wind varies too, generally with wider swings—for the most part, the full range from dead calm to a turbine-threatening storm is theoretically possible most of the year, though wind is definitely more predictable in some places (such as offshore) than in others. Weather forecasting can predict renewable energy well enough to be quite useful to grid managers, getting more accurate as the time delta of the forecast gets smaller.
Demand also fluctuates unpredictably around predictable basic patterns. A common example is the evening spike in air-conditioning load in hot climates. It can make sense for storage system design to take advantage of these known patterns. For example, given a pumped storage system where charging rate and discharging rate aren’t necessarily the same—such as a ternary system where turbine costs and pump costs are decoupled—it might make sense to build in a fast discharge capacity, to cover that evening demand spike, but a slower charge capacity that matches the slower and steadier wind and solar input across the rest of the day and night.
How much power and how much energy are we talking about for grid energy storage? Let’s look at three pumped hydro stations (two in service, one proposed) in increasing order by energy stored. These three are only a small sample of the large number of pumped hydro facilities that are in service or proposed worldwide. I chose them to show something of the size range (Bath County is considered the largest in the world by power, though a different site, Raccoon Mountain, gets the prize for energy):
2. Dinorwig Power Station:
Location: Snowdonia, Wales
Energy Stored: 9.1 GWh
Peak Power: 1800 MW
Duration at full power: 6 hours
3. Bath County Pumped Storage Station:
Location: Virginia, USA
Energy Stored: 24 GWh
Peak Power: 3000 MW
Duration at full power: 8 hours
And, for comparison, the largest battery grid power storage facility ever built to date, a project built by Tesla, Inc. using their (Panasonic-designed) lithium-ion batteries:
Hornsdale Power Reserve:
Location: South Australia
Energy Stored: 0.185 GWh (185 MWh)
Peak Power: 100 MW
Duration at full power: 3 hours
(Note: this facility is divided in two parts. One portion is reserved for short-term (sub-10-minute) uses, and is not available for long-term storage. The remaining 90 MWh can run for 3 hours, but at a power of only 30 MW.)
If we define a “Hornsdale” as 185 MWh of energy, the smallest pumped hydro storage plant on our list (Gordon Butte) stores the energy of 18 Hornsdales; the largest, Bath County, stores 130 Hornsdales. No wonder pumped hydro is agreed to be the gold standard for long-duration storage. (And is this one reason the press releases talk about power instead of energy?)[UPDATE 8/25/20: per this link, “The Gateway Energy Storage project, operated by grid infrastructure developer LS Power, holds the title now with a storage capacity of just over 230 megawatts in commercial operation. Still, it is expected to rise to 250 MW by the end of August.” (As usual, no mention of the energy stored, only the power.)
“LS Power has increased the battery’s capability, and the Gateway project can charge or discharge 230 MW for one hour, but it is expected to increase to 250 MW, which makes it the most massive battery in the world. According to pv magazine, LS Power has not yet shared the Gateway project’s MWh capacity.” Well, actually you just did share it, 1 hour duration at 250 MW implies 250 MWh. This is 1.4 Hornsdales, and still only about 1/14th the size of the smallest pumped storage we looked at, Gordon Butte. So it hardly seems worth the trouble to convert from Hornsdales to “Gateways”.]