This group brings together the best thinkers on energy and climate. Join us for smart, insightful posts and conversations about where the energy industry is and where it is going.


Looking Forward, Looking Back

Roger Arnold's picture
Director, Silverthorn Institute

Roger Arnold is a former software engineer and systems architect. He studied physics, math, and chemistry at Michigan State University's Honors College. After graduation, he worked in...

  • Member since 2018
  • 1,084 items added with 147,392 views
  • Dec 30, 2021

As I write this, 2021 draws to a close. A new year knocks at the door. A traditional time for taking stock, reflecting on what the past year or the past decade have brought, and what the future may hold.

Looking over articles on clean energy that I’ve posted over the years, I’m struck by a three part series on “coping with variability”. The series was published 15 years ago on the old Energy Pulse website. Energy Pulse was subsequently folded into the Energy Central platform. Its contents are archived there, and can be found by searching.

What strikes me in looking back is that so much has changed since then, and yet for the issues I wrote about, very little has changed. The issues we faced then are the issues we face now. The solutions that were available then are the solutions available now. And for the most part those solutions still go begging. 

It’s all rather disheartening. Or it could be taken that way, if one were so disposed. Yet there are glimmers of hope. I’ve included a minimally edited Part 3 of that series below, dealing with energy storage options. Most of it is as applicable today as it was when I wrote it. I’ve added footnotes in a few places where later developments have significant bearing. Then at the end, I’ve added a brief epilog. It hints at some of those glimmers of hope that I think I see.

With that, I invite readers to come along on my time machine for a short visit with a ghost of Christmas Past – or the future as it looked to me at the beginning of January, 2007.

Coping with Variability, Part 3: Energy Storage

In parts 1 and 2 (here and here), we looked at supply management and load management as mechanisms for coping with the variability of renewable energy sources. In this final part, we look at possibilities for energy storage - the ultimate solution for the long term, once the burning of fossil fuels is no longer an option. It's a large subject. We don't have space or time here to examine all the options. Instead, we'll focus on a handful of approaches that look promising for large-scale storage that would be suited for load balancing within an RBA (Regional Balancing Area).

Setting the bar

It's often noted that electricity—the flow of electrical charge through a conductor—can't be stored. What is stored is always potential energy in some other form that can be more or less easily converted to electricity.

The stored energy needn't necessarily have been put there by the use of surplus electricity. In hydroelectric dams, the energy that we tap for power is stored by the natural flow of water into the reservoir behind the dam. And even when we burn fossil fuels, we are tapping solar energy stored millions of years ago when deposits of organic material were laid down.

In fact, the main problem in developing storage solutions for surplus power today is that it's so hard to compete with fossil fuels. Coal, oil, and natural gas set a bar that is hard for the competition to clear. At the scale required for electrical load leveling, it's difficult to build any storage system that has a lower capital cost per kilowatt-hour than dispatched fossil-fueled generation. As long as fossil-fueled plants can use the atmosphere as a free dump for CO2, capital and not fuel will remain dominant in the cost of electricity. That makes building competitive systems for energy storage challenging.

But perhaps not impossible. 

Historically, pumped hydro and deep-cavern compressed air storage have been viable, when geographical conditions favor them. However, for general solutions that can be applied anywhere, we need innovation. Advances in key technologies offer some interesting possibilities.

Reengineered hydro

Among the important but less obvious advances, in that context, are those in the technologies for tunneling and excavation. The average costs per cubic meter of rock excavated or dirt moved have been slowly but steadily dropping. That has some interesting consequences for hydroelectric power – the oldest approach to large-scale energy storage.

Within limits, hydroelectric systems can be tapped for power on demand. The limits are set by factors that include installed turbine capacity, tolerance for variations in downstream river flow, and transmission capacity, as well as total annual river flow. In Part 1, hydroelectric power was mentioned as an ideal complement for wind power, because of the discretion it allows for when power is generated. Yet in many cases hydropower is used mainly or exclusively for baseload. Why is that?

Often, the culprit is transmission capacity. Transmission losses on a line are proportional to the square of the power carried. When a hydroelectric dam is located far from the market it serves, there may simply be insufficient transmission capacity to deliver power at anything more than a steady average rate. Attempting to deliver a full day’s energy quota in only a few hours of peak demand could melt transmission lines not built for that level of power flow.

To serve as a dispatchable resource for handling peak loads, a hydroelectric facility needs robust transmission capacity. It must also have turbines capable of meeting peak power loads much greater than the average load. The latter is not usually a problem; hydro plants tend to be built that way as a matter of course, to accommodate large seasonal variations in river flow. However, one further requirement is less commonly met: water from the power plant must discharge into a lake or secondary reservoir that can buffer the plant’s outflow. Although occasional flooding is now recognized as necessary for the health of riparian ecosystems, it’s not good for the downstream river flow to cycle daily between a trickle and high flood.

When an adequate secondary reservoir does exist (or can be built), it becomes possible to extend the load-following and backup capacity of a hydro facility by adding pumped storage. The peak power from a hydroelectric plant augmented by pumped storage can be many times greater than the average power available from stream flow.

But what's the scope?

The conventional wisdom is that hydroelectric potential in the U.S. has been fully tapped. We can't expect its contribution to power generation to increase. Is that true?

It’s true that nobody wants to see giant dams built on the few remaining stretches of wild rivers we have left. However, with advanced tunneling technology, it may be feasible to expand hydroelectric resources without building any large new dams.

Where power generation is concerned, a dam is nothing more than a way to get water from the reservoir inlet to the power turbine without losing head. A smooth-walled tunnel would serve just as well, as long as it was large enough to allow the water to move relatively slowly. So instead of building a giant dam and flooding hundreds of square miles of river valley, one could have only two small reservoirs, connected by a tunnel. A portion of the river's flow would continue in its natural course, but the larger portion would be diverted through the tunnel for power generation.

As a point of reference for how much this type of project might cost, we can consider a recent tunneling project undertaken by the Metropolitan Water District of Southern California (MWD). The Riverside Badlands Tunnel is a 13 km segment of MWD's Inland Feeder Project. It brings 31 cubic meters of water per second from a state facility near San Bernardino to a reservoir near Hemet. The tunnel, with a finished diameter of 12 feet, cost slightly under $10 million per km. to complete. Tunneling conditions varied considerably over the length of the tunnel. Some sections were very difficult, some were easy. Overall, they were probably fairly representative of what might be encountered in a "typical" tunneling project for hydroelectric power.

The Riverside Badlands tunnel takes an approximately level course. But had it been for a hydroelectric project linking two reservoirs at different elevations, it's instructive to see what its capital cost per kilowatt would have been.

One cubic meter of water dropping 100 meters in elevation releases just under a megajoule. If the average tunnel grade were 10% (100 meters per kilometer) then 31 cubic meters per second flowing through the tunnel would represent about 30 megawatts per kilometer. At a cost of slightly under $10 million / kilometer, the tunnel's contribution to capital cost, in round numbers, would be $300 per kilowatt. The turbines and generators and other "balance of system" items would up that figure, but it's still in the ballpark for the cost of dispatchable gas-fired generation capacity. It’s well under the $1500 / kilowatt "rule of thumb" for new coal-fired plants.

It's hard to quantify how much potential there may be for reengineering of hydroelectric systems to provide backup for renewable energy. I haven't found any specific studies. If it hasn't already been done, a good preliminary step for a national renewable energy policy would be a detailed engineering survey to identify projects where upgraded transmission, construction of secondary buffer reservoirs, and diversions through tunnels could augment hydroelectric capacity to deal with supply and load variability.

Meanwhile, back on the plains

Augmented hydro is all well and good for markets located near mountains, but what about the large areas of the world that are too far from any mountains for hydroelectric power to be of use?
One option for large-scale storage might be to use conventional earth-moving equipment to construct a primary holding reservoir on the surface, then use tunneling machines to excavate a matching secondary reservoir far below the surface. The secondary reservoir would of course have no outlet, so this would be a pure pumped storage system. With large enough reservoirs, however, the amount of energy storage could be sufficient for load shifting.

The idea isn't quite as wild as it sounds. Over the past three decades, towns in the Chicago metropolitan area have been building a series of underground reservoirs to prevent storm run-off from flushing sewage into Lake Michigan. A recently completed segment included 8.1 miles of concrete-lined tunnel bored through bedrock. The cost for the finished project reportedly came to roughly $1.00 per gallon of holding capacity, or $265 per cubic meter. If a deep reservoir for pumped hydroelectric storage could be built for the same cost, would it be a feasible solution for energy storage?

The potential energy represented by one cubic meter of reservoir capacity depends on the elevation difference between the source and receiving reservoirs. At the 100 meters of head typical for a large hydroelectric dam, one cubic meter gives only 0.28 kilowatt-hours of energy. At $265 per cubic meter of capacity, that would come to almost $1000 per kilowatt-hour of storage capacity – too expensive for economic feasibility. However, with a tunnel-excavated receiving reservoir, head is determined by the depth at which the reservoir tunnel has been bored. That could be almost anything.

Up to a few kilometers, the depth should be largely irrelevant to excavation costs. In fact, going deep could theoretically give better tunneling conditions and reduced costs. That's because the cost of tunneling, these days, isn't in the removal of rock. The cutting wheels on modern tunnel boring machines (TBMs) make short work of even hard bedrock. The cost, rather, is in the grouting and sealing operations and special measures that have to be taken to avoid flooding and collapse. At depths of more than a kilometer, in most areas, the rock will be solid and stable, with little or no groundwater penetration.

As long as the project is large enough to amortize the cost of the main access shaft and underground assembly of a TBM, then the finished reservoir cost could come in under $200 per cubic meter. A reservoir two kilometers below the surface would allow each cubic meter of water to flow through the equivalent of twenty large hydroelectric dams in series. Even if the cost of the finished reservoir were $500 per cubic meter rather than $200, that would still be a capacity cost of under $90 / kWh.

Whether $90 / kWh of capacity is a feasible cost for pumped hydroelectric storage depends on the financing model and the capacity turnover time. With minimal maintenance, a pumped hydro facility should have an indefinite lifetime. Capital cost therefore amounts to the cost of interest on construction loans. Assuming 6%, that would be $5.40 per kilowatt-hour of capacity per year. If the average turnover were 33% of capacity per day, then delivered power would be 680 kWh / year per cubic meter of capacity. Interest on capital would then add 0.8 cents / kWh delivered.

An added cost of only 0.8 cents / kWh would be very attractive for large-scale energy storage. However, there's risk in the assumptions on which that figure is based. Underground mining operations have been conducted at depths well beyond 2 km, but as far as I know it's all been traditional "drill and blast" work. I don't know of any experience with very deep TBM operations that might serve to validate cost projections as low as $200 - $500 per cubic meter. And, of course, the cost of the excavated reservoir would only be a portion of the cost of the complete pumped-hydro storage facility – although I think it's safe to say that it would be the dominant portion. [As a side note, if a pumped hydro reservoir were to be dug at a depth of 2 km or more, it would be a great source of hot water for geothermal district heating. The geothermal gradient averages about 20 ℃ per kilometer, so at a depth of two kilometers, the temperature of rocks around the reservoir tunnel would likely be about 60 ℃ (140 ℉). - RDA]

In the end, what we're left with is an intriguing possibility. Such a system may well be feasible, but it will require more study to know. 

Compressed air

The other established technology that has been used for large-scale energy storage employs compressed air. A large volume of compressed air is stored underground, either in natural or man-made caverns or in deep saline aquifers. When power is needed, the pre-compressed air is used to feed a combustion turbine. In a conventional combustion turbine, about two thirds of the power from the output stage is taken to drive the compressor stage. Avoiding the need to drive the compressor nearly triples the power output of the turbine, relative to the fuel burned. However fuel is burned, so this is actually a hybrid storage and generation system.

To date, there are only two large CAES facilities, worldwide, that are operational. One is in Hundorf, Germany, and has been operational since 1974. The other, built in 1991, is in McIntosh, Alabama. A new and much larger system is being developed near Norton, Ohio, and a group of utility companies is proposing a combined wind farm and CAES facility near Fort Dodge, Iowa. [Alas, both of these projects died about two years after this article was published. - RDA]  There is also considerable interest in CAES for buffering wind power in west Texas.

CAES systems of the types built at Hundorf and McIntosh have some compelling advantages for providing spinning reserve, regulation, and other ancillary grid services. Startup is quick, and output can be throttled rapidly over a fairly wide range without losing much efficiency. However, as energy storage solutions, these systems are mediocre. The McIntosh facility, with newer technology, is the more efficient of the two. It uses about 0.7 kWh of off-peak electricity and 4500 Btu of gas to generate 1 kWh of output [from Denholm P, Kulcinski GL, as reported in this paper from NREL].

If one ignores the 0.7 kWh of off-peak electricity, 4500 Btu per kWh of output looks like an extremely good heat rate. The newest and most efficient combined cycle gas turbine generators have a heat rate around 5680 Btu / kWh, which is a thermal efficiency topping 60%. However, looking at it another way, the 4500 Btu of fuel that the McIntosh CAES plant uses to produce 1.0 kWh of electricity could have produced 0.79 kWh in a new combined cycle gas turbine. So the 0.7 kWh of off-peak electricity used for compression is only buying 0.21 kWh of additional output. That's an effective round-trip efficiency for electricity-in to electricity-out of only 30%.

Although the stored energy is shifted from off-peak to premium on-peak power (regulation capacity), that level of loss makes it hard to justify the investment in wind farms feeding the system. More to the point, with 79% of its output deriving from combustion of fuel, this system will never allow wind and solar to become our predominant energy sources. For that, we will need something better.

Adiabatic CAES

To be fair, when the McIntosh facility was designed and built, natural gas was still cheap and expected to remain so. Renewables were virtually unknown. The primary aim of the facility would have been responsive power generation, not high round-trip energy storage efficiency. No doubt with some tweaking of control parameters and flow rates, it could be operated with a larger input of off-peak electricity and a smaller input of fuel. The effective round-trip efficiency might then be raised to 50%; the tradeoff is that it would run through its compressed air storage more quickly, and be able to supply peak power for a smaller number of hours each day.

In Europe, a group of researchers are studying a different approach. It eliminates fuel consumption altogether, so that the system functions as a pure storage facility. They call it Advanced Adiabatic CAES (AACAES).

The idea for AACAES is that, in the charge cycle, hot compressed air is passed through a counter-flow heat exchanger before being sent to its storage cavern. The compressed air transfers its heat to a thermal storage fluid, which then enters a well-insulated storage tank of its own. In the discharge cycle, both flows are reversed, and the cool compressed air from the storage cavern recovers most of the heat it gave up before being stored. The now-hot compressed air exiting the heat exchanger drives an expander turbine to generate power.

If the heat exchanger and thermal storage work well, then efficiencies approaching those achieved with pumped hydroelectric storage should be possible. However the requirements are challenging. The counter-flow heat exchange system must span a temperature range from ambient to 650º C, and store a large volume of 650º fluid for the better part of a day or more without significant temperature loss.

Isothermal CAES

Another approach to improved storage efficiency is isothermal CAES. In an isothermal process, temperature remains constant. Isothermal compression minimizes the work required to compress a gas, while isothermal expansion maximizes the work extracted when expanding it. The problem is that the usual means to achieve near-isothermal compression and expansion are inherently slow. That means high cost in capital equipment, relative to throughput.

There is a potential way to achieve quasi-isothermal compression and expansion with high throughput, but as far as I know, it's unproven. The gas is injected into a stream of water or other carrier liquid, and then the gas-liquid foam is compressed or expanded. The liquid constitutes at least 90% of the mass of the foam. During compression, it absorbs heat from the tiny gas bubbles entrained throughout its volume; during expansion, it supplies heat to the expanding bubbles. The result is that the temperature of the gas remains nearly constant.

A possible configuration for an isothermal CAES system works as follows:

  • There are two columns of fluid, one ascending from, and one descending to, a deep underground pressure chamber;
  • To pump gas into the deep chamber, gas bubbles are injected into the descending column. As the entrained bubbles are carried downward, they are compressed by the increasing weight of fluid in the column above them;
  • In a stilling pool at the bottom of the descending column, the bubbles rise and separate from the fluid.
  • The fluid, now free of entrained bubbles, exits the chamber via the ascending fluid column.
  • To discharge gas from the chamber, the flows are reversed. Bubbles of compressed gas from the chamber are injected at the base of the fluid column, and expand isothermally as they are carried upward in the fluid column.

The column with the entrained bubbles is less dense than the column of bubble-free fluid. When the low-density column is descending (carrying compressed air into the chamber), energy must be supplied by pumps to maintain its flow against the pressure of the heavier column of ascending fluid. Conversely, when the low-density column is ascending (carrying compressed air to the surface), the lower pressure at its base, relative to chamber pressure, provides a head for power generation.

A quasi-isothermal CAES system of this type closely resembles a pumped hydroelectric storage system. There's an important difference, however: the energy capacity per cubic meter of storage is approximately six times larger than for a pumped hydroelectric system. That makes the system much less sensitive to the excavation cost of the underground storage chamber. If deep tunneling costs are under $500 per cubic meter, then the cost of storage capacity is under $15 per kWh. At that cost, high turnover is not needed, and it becomes realistic to talk about banking an entire week's worth of energy production.

The solar thermal connection

It's possible to use the entrained bubble system in "one-way" mode, as an efficient means to deliver compressed air to the deep storage chamber. The compressed air would then be delivered straight to some type of heater before driving an expansion turbine. The heating allows the expander turbine to deliver more power than was used to compress the gas.

If the heating is supplied by combustion of natural gas, then the system is a more or less conventional CAES system. It gains from the more efficient compression and from the constant pressure of its air supply as water replaces compressed air during discharge, but still burns fuel. 

As an alternative to burning fuel, heating for the compressed air could be supplied by concentrated solar thermal energy – either direct or stored. It would be an alternative to steam or Sterling engines for solar-thermal power generation, with the advantage of much higher power capacity from operating with pre-compressed air. Net thermal efficiencies should be similar.


This series has looked at supply management, load management, and energy storage as means for coping with the inherent variability of renewable energy sources. It appears that:

  1. Existing mechanisms of supply management for coping with daily load variations are also sufficient to cope with any levels of RE that are likely to be reached within at least the next five years;
  2. Power pricing policies that support and encourage the development of responsive loads are desirable irrespective of RE accommodation. They will help to level load curves and enable better utilization of capital assets. At the same time, they will enable substantially higher levels of RE penetration than what could be economically accommodated through supply management alone;
  3. Feasible technologies do exist for long-term energy storage at levels that would ultimately allow for the elimination of most fossil-fueled power generation. Further research and actual pilot projects are needed to establish the best designs and economic feasibility of these systems.

What I find most interesting, in terms of a national energy policy, is how much commonality there is in research needs and regulatory policy, independent of what our energy sources are to be. As far as I can see, there is no looming fork in the road where we must decide between RE vs. nuclear power, for example. Nuclear, the ultimate in reliable baseload generation, has as much to gain from responsive loads and energy storage solutions as a system that relies heavily on wind and solar energy.

The real divide is rather in how seriously we take the threat of CO2-related global warming. Do we have the will to impose a serious carbon tax and cut our dependence on fossil fuels, or will we hang on to the current model and "business as usual" to the bitter end? Time will tell.


In the years since this article was first published, the specific cost of new wind and solar installations has fallen dramatically. Installed capacity has advanced – though not at the rate necessary for RE to be the basis for timely decarbonization of the world’s energy supply. 

The main problem has been that as penetration has grown beyond the levels that can be easily accommodated by supply management, the value of added RE capacity has fallen. Measures for demand management and technologies for large scale energy storage have not developed nearly fast enough to uphold the value of intermittent renewables. Wholesale electricity prices often fall to zero or below zero, which the market interprets as a signal of oversupply. That makes it hard to attract investment for new capacity.

The good news is that the problems are finally starting to receive serious attention. There’s growing recognition of the need to address the economic limitations of intermittent renewables at high penetration levels. Solutions are likely to involve policy changes for pricing electricity, along with stronger incentives for development of responsive loads. But those are large topics; subjects for the New Year.

Matt Chester's picture
Matt Chester on Dec 31, 2021

Do we have the will to impose a serious carbon tax and cut our dependence on fossil fuels, or will we hang on to the current model and "business as usual" to the bitter end? 

We may have the will, but the question always comes to do the elected leaders have the gusto to do so. I'm remaining optimistic, but I have to admit the patience is waining 

Nathan Wilson's picture
Nathan Wilson on Jan 2, 2022

"Nuclear, the ultimate in reliable baseload generation, has as much to gain from responsive loads and energy storage solutions as a system that relies heavily on wind and solar energy."


There are two issues with this:

- Sure there is a certain amount of variation in the power system that comes from variation in demand.  But in most locations, solar and wind would increase the net variation by 2-3 times.  So grids dominated by variable renewable are much more dependent on storage and flexible load solutions.

- high temperature nuclear technologies (e.g. sodium, molten salt, or helium cooled) can be easily coupled to thermal energy storage using "solar salt".  Thermal energy storage for several hours is cheaper than batteries (likely to be in the $1/Watt range), long lasting, and easily recycled/refurbished.  Today the ingredients for solar salt come from salt mines, but the chemical components (sodium, potassium, and oxygen) come from seawater and air, so we can never run out.


The other consideration is that California's uniquely mild climate results in flat demand versus seasons, hence baseload sources (or sources which combine with storage to make baseload) are a good fit.  But for most of the country, there are strong seasonal peaks (for heating and cooling).  If the nuclear baseload generation is made large enough to cover the peaks, then for most of the year, there will be excess capacity and the storage will be idle.   So in the nuclear-rich case, it might be better to over-build the nuclear and add dispatchable syn-fuel production, rather than adding storage. 


Similarly, with windpower, as the generation is over-built to produce syn-fuel with excess energy, there is less opportunity for storage since there will be less time that the supply falls short of the demand. 


Deployment of battery EVs with night-time charging is another factors that reduces the demand variation and thus the need for storage in grids dominated by wind and especially nuclear.


Basically, solar power is the dominant driver for storage.  Currently solar is behind windpower in the US, and is on-track to stay that way.  So it seems that storage will struggle to compete.  Several studies from NREL also project a very small role for storage, across a mix a low-carbon grid scenarios.


That said, it was a useful round-up of some energy storage ideas.  Thanks for bringing up the exceptionally poor efficiency of CAES; too often CAES gets listed as an option without that being mentioned.  As to the deep-underground hydro stuff, that strikes me as a hostile environment for people, so the question is how does the maintenance get done?  Working in hostile environments can be viable when high-value equipment and hydrocarbons are at stake, but storage needs to be cheap, so I'm skeptical. 

Roger Arnold's picture
Roger Arnold on Jan 19, 2022

There's no question that solar PV, as you say, is the dominant driver for storage today. I.e., the energy resource that is most strongly dependent on storage for viability. But 15 years ago, when I wrote that nuclear had as much to gain from responsive loads and energy storage solutions as wind and solar, I was talking about national energy policy going forward. I saw a lot of commonality in needed research and regulatory policy, independent of whether nuclear or intermittent renewables ended up being the main path we relied on for decarbonization.  Storage similar to what we are implementing today would have allowed nuclear to economically supply a larger fraction of total demand without having to reduce output at off-peak hours.

The same would be true for development of responsive loads (demand-side management), had we actually managed to do much of anything in that domain. It shocks me how little progress there has been in that area.

Roger Arnold's picture
Thank Roger for the Post!
Energy Central contributors share their experience and insights for the benefit of other Members (like you). Please show them your appreciation by leaving a comment, 'liking' this post, or following this Member.
More posts from this member

Get Published - Build a Following

The Energy Central Power Industry Network is based on one core idea - power industry professionals helping each other and advancing the industry by sharing and learning from each other.

If you have an experience or insight to share or have learned something from a conference or seminar, your peers and colleagues on Energy Central want to hear about it. It's also easy to share a link to an article you've liked or an industry resource that you think would be helpful.

                 Learn more about posting on Energy Central »