Grid Scale Supplement by DCD Magazine

Sponsored by

Contents 4. Running on gas flares A critical look at the fossil burning data centers which claim to be green 6. Powered by gravity Cranes lifting concrete blocks can store energy for data centers 8. A dvertorial: A New Paradigm for Mission Critical Power 10. Taking the nuclear option Nuclear power has low emissions. Could data centers fund a new generation of small reactors? 13. E nergy storage by air Pumping air into underground caverns could stabilize the grid

The future of energy

ecarbonizing the electric grid is like a complex game of multi-dimensional chess. Add renewable energy, and you need to consider when it is available and how it can best be used. Because renewable sources are intermittent. And while you pile in renewable sources, demand for electricity will keep increasing because of decarbonization elsewhere. If home heating shifts to heat pumps, they need power, and electric cars need to be charged. Even the hydrogen economy, sometimes seen as a replacement for other sources of power, needs electricity to power the electrolysis plants that generate green hydrogen. As a large consumer of energy, data centers have a responsibility to help in the effort to reach Net Zero. This supplement looks at them in the context of the grid. With high levels of investment and ingenuity, could data centers play a role in decarbonization?

Nuclear power - yes please?

10 13

Many environmentalists hate nuclear with a passion. But nuclear reactors have shown their worth in producing low-carbon energy at a steady rate, which renewables can never achieve. Now proponents say fossil fuel causes more deaths than nuclear ever has - and small modular reactors (SMRs) could shake off a past troubled by cost over-runs and safety fears. The nuclear industry wants data center operators to enable the new age nukes, through power purchase agreements (p10).

Gravity power The decarbonized grid will require huge amounts of energy storage to match demand cycles to capricious renewable energy supplies - storing energy when the sun shines or the wind blows. In Switzerland, one firm's lowtech route to energy storage is to raise giant concrete blocks to store potential energy. And at least one project has considered using this for data centers (p6).

Storage by air On a larger scale, another proposal suggests storing surplus energy in underground caverns, or giant tanks, by pumping them full of compressed air. There are systems out there which have been in operation for decades, and some of the proponents want data center providers to help the technology move into production (p13).

Can gas flares be green? Finally, three outfits are proposing data centers at oil wells and drilling platforms, powered by burning gas. That's not as crazy as it sounds, because oil wells currently burn off waste gas in giant flares. A handful of companies say on-site data centers could put that energy to use - albeit mostly in Bitcoin mining (p4).

A truly green grid We'll leave you with a hard thought. Data centers can try to green the grid in many ways. But they all consume precious energy. Maybe the data center that best reduces grid demand and emissions might be the one that is not built at all.

Critical Supplement 3

into the atmosphere in 2020, along with eight million tons of methane, as well as other greenhouse gases (GHGs), soot, and carcinogens. The IEA's Net Zero by 2050 scenario requires a complete end to non-emergency production gas flaring by 2030. Instead of flaring, it suggests that oil companies could use the gas for onsite power, pipe it elsewhere for use (though this needs more infrastructure), or pump it back underground to keep re-pressurize the oil well.

Peter Judge Global Editor

Data centers that run on gas flares A handful of companies are putting data centers at oil wells to feed on burning gas. Do their environmental claims stand up, or is this a front for wasteful cryptocurrency mining?

data center powered by burning natural gas might not seem a very environmentally friendly thing to do. But at least three companies are doing just this, and they claim that their schemes reduce emissions. In the US, Crusoe Energy already has 40 data centers powered by burning natural gas. In Norway, a start-up with the unwieldy name of Earth Wind & Power has been launched with a plan to offer data processing located on North Sea oil platforms. And meanwhile, Toronto-based Validus Power is offering a post-apocalyptic version of the concept, with mobile Bitcoin mining data centers in articulated trucks, parked at flaring oil wells, where they feed on burning gas and turn it into cryptocurrency.

If that doesn’t already sound scary, Validus calls this the Mad Maxx Mobile Power Fleet. Stranded assets All these offerings - and their environmental claims - are based on the fact that a lot of natural gas goes to waste. Oil wells produce gas as well as oil, but can’t always harvest the gas. Instead through so-called “production flaring,” unwanted gas is simply burnt off. The International Energy Agency (IEA) says it's a big problem, with oil producers burning off 142 billion cubic meters of natural gas in 2020 - roughly as much as the entire natural gas demand of Central and South America. The chief offenders are Russia, Iraq, Iran, the US, and Algeria. Flaring contributes to the greenhouse effect, putting some 265 million tons of CO2

DCD Supplement • datacenterdynamics.com

Burning for the planet All three offerings put containerized data centers at oil rigs and oil wells. The kind of data they process, and the customers they can serve, depend on their reliability and network bandwidth. Crusoe Energy is the farthest advanced, with 40 flare-powered data centers in action across North Dakota, Montana, Wyoming, and Colorado, at wells operated by companies including Devon Energy, Kraken Oil & Gas, and Enerplus. In April, Crusoe raised $128 million in Series B funding to commercialize the service, launching it under the brand Digital Flare Mitigation. The energy will be used to run computing on Island Cloud, an Nvidiapowered GPU-centric hardware platform. Crusoe’s Flare Mitigation data centers are modular steel structures with security access controls, dust-proof vestibules, redundant HVAC cooling, and fire suppression systems. Customers pay for power at flat monthly rates, and can have a power at 12.5kW-25kW per rack, or 250kW per module. With multiple containers, the service can scale up to megawatts. The system has a dual-feed design with an uninterruptible power supply (UPS). Crusoe offers connectivity to anywhere on the public Internet, at rates from 1Gbps to 100Gbps, with a round-trip latency of less than 60ms to anywhere on the US continent, 160ms to Europe, and 180ms to Asia. “Customers are never far from their data, leveraging extensive connectivity through a high-availability, high-bandwidth fiber optic network with microwave backup,” explained a Crusoe spokesperson. “This network, operated by Crusoe, delivers 99.95 percent service availability.” Crusoe’s current data centers are made by custom fabricator Easter-Owens, but other partners could be used in future. Norway’s Earth Wind & Power (EW&P) is not so advanced. The company, led by Ingvil Smines Tybring-Gjedde, a former Minister of National Public Security and Deputy Minister of Petroleum and Energy in the Norwegian Government, was founded early in 2021. Although EW&P says it aims to harness power wasted at all sorts of sites, including

Grid Scale Supplement solar farms and wind farms, it’s clearly most focused on the oil industry, and has strong links with executives who harnessed Norway’s North Sea Oil boom. EW&P is funded by PetroNor executives Knut Søvold and Gerhard Ludvigsen, but also has backing from Valinor Energy Group, and board members from Equinor, AGR, and renewable firms SustainSolar and Norsk Vind. EW&P has ordered a 1.8MW high performance computing (HPC) data center container from Germany’s Cloud & Heat, a company with a history of using liquid cooling to reuse waste heat from data centers. The 40 ft Cloud & Heat module contains 144 water-cooled servers, and can be used in temperatures between -30°C and +48°C. It will be delivered in the second quarter of 2022. EW&P hasn’t said where it will go, but the company announced a pilot project with Rapid Oil Production in September. Rapid Oil Production and EW&P will conduct a three-month feasibility study of power-to-data-center projects off-shore in the North Sea, with Rapid installing Cloud &Heat’s containerized data centers on Rapid’s Fyne Field oil rigs in the North Sea. “Off-shore North Sea is a key market for us given the scale of the opportunity,” commented EW&P CEO Tybring-Gjedde at the announcement. “We look forward to working with RPO and other innovative companies which share our goal of further reducing emissions as we transition to a netzero world.” Toronto-based Validus says it has been operating its mobile Bitcoin mines for the last five years, across many different industries with excess energy. “With this system, we can generate a profit while eliminating harmful greenhouse gases,” says Validus. “This technology can be deployed rapidly, anywhere, at scale and remain fixed for decades or redeployed within 24 hours notice.” How does this help? But how can burning gas reduce emissions? All the providers claim that using flares for data processing actually reduces total emissions, because gas is burnt more efficiently in a generator, converting more of the methane to carbon dioxide. Crusoe, for instance, claims that its mitigation service reduces methane emissions by around 98 percent, and this reduces CO2-equivalent emissions of

greenhouse gases by 63 percent. This is based on the fact that methane is 80 times more potent than CO2 as a greenhouse gas. EW&P also told DCD that harnessing the gas will ensure a fuller combustion, so less methane will be emitted. Take that with a pinch of salt, though. Methane is much less stable than CO2. It only lingers in the atmosphere for 12 years before breaking down to CO2 and water, while CO2 remains for centuries. The providers also state that the data centers do not enable any new oil exploitation. Nicolas Roehrs, CEO of EW&P’s partner Cloud & Heat told DCD: "It's not intended to create new oil and gas sites. We, and especially EW&P, are offering a way for existing oil and gas fields to reduce their carbon footprint and to decarbonize the complete production process in this industry." Who are the customers? There’s one more big question to ask, however. What are these data centers doing? The best solution for the planet would probably be to pipe the gas elsewhere, or else generate electricity and use it to create green hydrogen. Instead, many of these remote data centers will be used for cryptocurrency mining, an activity with little benefit, whose impact on the planet is already widely criticized. Bitcoin mining takes cheap electricity and uses it to produce cryptocurrency, a speculative asset with no inherent value or links to real assets. The blockchain technology behind it is explicitly designed not to be scalable, and each Bitcoin transaction uses an increasing amount of computational power. Estimates of Bitcoin's power use vary, but it’s reckoned to consume as much electricity as Argentina. Validus’ Mad Maxx, with trucks rolling up to suck on burning gas flares, is exclusively designed for Bitcoin: “There are two industries that have garnered a bad reputation due to their carbon footprint or consumption of energy: oil producers and Bitcoin mining,” the company says. “But what if we could use the undesirable carbon output from oil producers and provide low-cost energy for Bitcoin mining?” Mad Maxx’s Bitcoin mining rigs don’t offer high bandwidth connections or massive

The IEA has set out a Net Zero by 2050 scenario, which requires a complete end to non-emergency production gas flaring by 2030

reliability, but Bitcoin customers don’t care about normal data center service issues. They simply want to burn as much power as possible, as fast as possible, in the race to solve the math which generates new coins. Norway’s EW&P is more cagey about the intended customers for its eventual service, but has pretty clear links with Bitcoin mining. The Cloud&Heat container data center it uses was co-designed by Bitcoin mining company BitFury, and EW&P’s website contains some very positive material about Bitcoin mining. EW&P's "Fact Center" quotes a prediction from Fortune Business Insights that blockchain will reach $69 billion by 2027, and one from Gartner that it will generate $3 trillion by 2030, and may be running 20 percent of the global economic infrastructure by then (an unlikely prediction, considering how much energy would be needed to run 20 percent of the world’s economic infrastructure on the inefficient blockchain network). EW&P seems to have moved away from the mining emphasis in a more recent downloadable brochure, which lists several other applications such as 3D rendering, simulation, and image processing, all of which can be done at remote sites w- but only if there is a good data connection and the power is not too intermittent. For Cloud & Heat’s Roehrs, it seems that Bitcoin is just a step on the way to creating useful computing, by enabling that feasibility study: “The final goal will be to implement and setup HPC sites,” he told DCD. “Since this is the first container DC we will start with some blockchain applications.” Real value Compared to these two, Crusoe is reaching a more varied portfolio of customers. Existing users include Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Lab (MIT-CSAIL), Folding@ Home, and OpenCV, a library of computer vision tools. Thanks to its strong networking and reliable service, Crusoe is able to serve “a wide range of Cloud Computing applications”, a spokesperson told DCD, explaining that services which only have basic satellite connectivity are limited to crypto mining. “We provide pure data center services and cloud computing,” said the spokesperson. “Crusoe Cloud can offer high-performance GPU-based computing power with low latency, reliable networking, high-speed data transfer and a carbon-negative emissions footprint at low costs. This has applications across Deep Learning, 3D Rendering, Computer Vision, Scientific Modelling and more.” Perhaps gas flares are more than hot air after all.

Grid Scale Supplement 5

Peter Judge Global Editor

he concept of a gravity battery is not new. Grandfather clocks and cuckoo clocks have been ticking away since the 1650s, powered by a weight that drives the mechanism as it falls. More recently the GravityLight from SolarAid harnessed the same principle to create off-grid light for people in developing countries. Fill a bag with rocks, lift it, and let it drive a dynamo as it falls, creating electricity to produce light. Large scale gravity storage usually makes use of pumping water uphill and then allowing it to drive a hydoelectric power station. But if you don’t have a large lake available, you could use a crane to raise a concrete block, then recover the potential energy stored, by letting it fall. It’s an idea that hasn’t been used widely, but as the grid decarbonizes, the world needs to find ways to store the surplus energy produced by renewables in times of

sunshine or strong wind, to be used when those renewable resources aren’t available. There are a few companies with pilot gravity battery projects, and they point to benefits that include a much smaller footprint than pumped hydro, and no requirement for a hilly area. Swiss giant lego The most striking gravity battery is Energy Vault’s 110m-high system of cranes in the Swiss city of Ticino (pictured on these pages). It lifts and stacks 35-ton blocks, to store a total of 35MWh of energy. The blocks are more environmentally friendly than concrete: they’re made from

wasted dirt and a special polymer. And the operation is more complicated than you might think: blocks have to accelerate, and then decelerate before they make a landing. Energy Vault’s crane has six arms, so one block can accelerate while another decelerates, CEO Robert Piconi told IEEE Spectrum. In Scotland, Gravitricity is working on a prototype 250kW system using 50-ton blocks, but it plans to go much bigger, raising and lowering 5,000-ton weights in abandoned mine shafts. The benefit of those weights is they only need to move a few centimeters per second to generate MW of power - enough

Canadian architect WZMH proposed data centers invest in their own lift system and concrete blocks, specifically to recover energy wasted

DCD Supplement • datacenterdynamics.com

Grid Scale Supplement to balance the grid if energy supplies falter. In Russia’s Skolkovo Technopark, Energozapas - backed by one billion roubles ($13bn) - has built a prototype Lifted Weight Storage (LWS) system, basically a set of lifts carrying containers of compressed earth up and down. The system is 80m tall, and has a capacity of 2.5MWh. If it catches on, the company hopes to build a massive system by 2025, which will be 300m tall, hold 1GWh of energy and deliver it at a rate of 1GW. Lift Renewable Energy is a startup with an idea reverses the scheme. Buoyant gas containers are held on the bottom of the ocean by a winch which allows them to rise when energy is needed. Lift says the idea can store GWhs of energy, and is space-saving because it doesn’t use any dry land, while most of the population of the world is near enough to the sea to benefit from the energy. What about data centers? None of these projects have specifically targeted data centers, but given the importance of data centers to society, and their rapid growth, it’s only a matter of time before one is located near a gravity battery trial. And in 2020, Canadian architect WZMH actually proposed data centers could invest in their own lift system and concrete blocks, specifically to recover energy wasted when data center generators are tested. Data center generators are for backup at data centers, as well as at other facilities like hospitals, but they are fired up every

If the gravity battery has merits, however, maybe it could fit in as part of the system, between the generator and the batteries month to test them. WZMH says this produces energy which normally goes to waste. With the help of Toronto's Ryerson University, WZMH designed a system to use that energy to raise concrete blocks in an elevator. The idea is to raise the concrete block during the generator test, then when the block falls, use that energy to charge batteries, which can provide green energy to surrounding buildings. "This is a concept. It's part of a research and development project we are working on," said WZMH principal Zenon Radewych in an email to DCD. "We believe it has a strong chance of being a product." Data center power experts contacted by DCD weren’t so sure. For one thing, generator tests rarely go up to full power - and if energy really was being wasted in the amounts suggested by WZMH, then it would be evident in things heating up. For another thing, WZMH suggests using batteries to store the energy from the raised block. Why raise the block in the first place, when that energy could have been sent straight to the batteries? And in fact, data centers all have a big room full of batteries already, right next to the generator. If there really is waste energy from the generator, then surely it could be used to charge them up. If the gravity battery has merits, however, maybe it could fit in as part of the system, between the generator and the batteries, providing a small amount of energy to smooth transitions, much like flywheels. Gravity batteries in real life If all this sounds somewhat theoretical, consider this: you can help charge and discharge a gravity battery in real life every time you use an elevator. Many commercial lift systems recover some of the energy released by the downward journey, to reduce the energy needed on the upward path. Otis, for instance, says its ReGen system can save up to 30 percent of the electricity used by the lift. If you walk up the stairs and take the lift down, you’re making a net contribution.

Grid Scale Supplement 7

the past decade. Extreme weather is causing frequent damage to our electrical system, costing Americans and the economy billions each year. Potential points of failure are

A New Paradigm for Mission Critical Power

becoming harder for utilities to predict and for

Changing the Way Businesses Look at Their Critical Infrastructure

power.”

most businesses, an unexpected power outage can cause disruptions that ripple through their entire value chain. This has caused a fundamental shift in thinking from, “what is the cost of power” to “what is the cost of not having Operational risks that come from a loss of power are among the highest for mission critical facilities, and operators must heavily

ver the course of the

based on the needs and requirements of its

invest in solutions that ensure a high-quality

20th century, the U.S.

routine operations. When drilling down into the

energy stream is constantly available to power

electric grid was built as

profile of a specific facility, understanding the

their operations. Special focus must be placed

a one-way value chain

difference between demand and consumption

on meeting minimum requirements across

from fuel supply to

is key to defining its energy characteristics.

several key energy components – utility power

end-user consumption.

A ‘critical load’ is a portion of electricity supply

supply, back-up generation, UPS modules, and

Such infrastructure design created cascading

that powers infrastructure directly related to an

cooling systems – which presents a unique

vulnerabilities whereby a failure of any

organization’s ability to operate. Infrastructure

optimization challenge in structuring those

one component can result in disruption of

that warrants this status must either be kept

components to maximize the economic,

service to end users. This same infrastructure,

running when main power supply fails or be

reliability, and sustainability benefits.

inherently prone to failure, now faces

powered down in an orderly manner to prevent

heightened demands from digitization and

system crashes, data loss or corruption, and life

Changing the Paradigm of Reliable Power

a rising frequency and intensity of natural

shortening hardware damage.

Diesel generators have been the status quo

disasters. This has created an unprecedented

The term ‘Mission Critical’ is defined as

solution for power disruptions for decades.

risk landscape. Recognizing this new age of

something that is vital to an organization’s

However, they are monolithic machines

risk is one thing. Having a strategy to mitigate

ability to function. Operating a facility under

without inherent redundancy and produce

it is another. Doing so requires a deeper

these conditions necessitates stronger

over 40 toxic air contaminants during

understanding of the issues and how to address

emphasis on criteria that ties directly into the

operation, including a variety of carcinogenic

them from the top down.

structural elements of a building’s design.

compounds. What’s more, since they are idle

The Mission Critical Facility

The Resiliency Challenge

testing to ensure they can be available when

Every building has a unique energy “fingerprint”

U.S. grid outages have increased 60% over

needed. Further, the availability of diesel fuel

assets, they needlessly consume fuel while

DCD Supplement • datacenterdynamics.com

Bloom | Advertorial

during an extended outage and the reliability of diesel engines to operate continuously for long periods of time introduce significant degrees of risk. Technologies like solar and wind are great for their sustainability profile but due to their

Diesel generators have been the status quo solution for power disruptions for decades. However, they are monolithic machines without inherent redundancy and produce over 40 toxic air contaminants during operation.

inherent intermittency, cannot practically solve reliability challenges. Their very nature requires

vulnerabilities of conventional transmission

critical challenges and take control of their

some sort of energy storage and today’s

and distribution lines by generating power

energy future with onsite generation that scales

technology does not cost effectively support

onsite, where the electricity is consumed, and

to meet present and future business needs.

the massive load shifts from day to night, and

operates at very high availability due to its

From healthcare facilities, to data centers, to

certainly not from summer to winter.

modular and fault-tolerant design. Because

manufacturing facilities, and more, Bloom

Bloom Energy Servers are in continuous

provides hundreds of megawatts of clean,

onsite fuel cell microgrids have emerged as a

operation, there is no risk associated with cold

reliable, AlwaysON power.

viable alternative to traditional architectures,

starts and load transfers.

Resilient distributed baseload solutions like

providing UPS-quality, mission critical

The power quality delivered by Bloom

generation capacity and the enhanced

Energy Servers is comparable to best-in-class

reliability that utility infrastructure has

UPS systems deployed in data centers and other

struggled to provide.

mission critical facilities. We utilize state-ofthe-art PWM (pulse-width modulation) inverter

Introducing AlwaysON Microgrid Platform

technology for conversion of fuel cell DC

Bloom Energy’s AlwaysON Microgrid platform

power to 480V AC power. The waveform of the

is specifically designed to address the key

current supplied to the customer is generated

challenges of traditional mission critical power

by a sophisticated multi-level current-source

infrastructure: reliability, energy efficiency,

inverter control scheme. The high inverter

operational cost reduction, phased use of

switching frequency supplemented by a high-

capital, emissions reduction, and air and water

performance filter in each inverter module

sustainability.

means that harmonics are virtually eliminated.

As a critical, always-on solution, our

Bloom is a trusted onsite energy partner,

microgrid platform can operate alongside

supporting a wide array of mission critical

a main grid, but independently of it during

operations around the world. Companies

a power outage. Our technology avoids the

deploy our proven technology to solve their

About Bloom Energy Bloom Energy’s mission is to make clean, reliable energy affordable for everyone in the world. Bloom’s product, the Bloom Energy Server, delivers resilient, always-on electric power that is clean, cost-effective, and ideal for microgrid applications. Bloom’s customers include many Fortune 100 companies and leaders in manufacturing, data centers, healthcare, retail, higher education, utilities, and other industries. For more information, visit www.bloomenergy.com.

Supplement 9

Taking the nuclear option Data centers need a steady source of power, with no greenhouse emissions. Could nuclear power be the answer?

Peter Judge Global Editor ata centers need to have a steady supply of electricity that comes from a sustainable source, which doesn’t pump CO2 or other greenhouse gases into the

atmosphere. A small number of organizations are starting to think that nuclear power could fit the bill. Nuclear power has an image problem. It’s tinged with its military origins, there’s a very vocal campaign against it, and nuclear projects all seem to be too costly, too late, or - in cases like Chernobyl - too dangerous. But countries like France rely on nuclear electricity, and are lobbying to have it classified as a clean technology, because it delivers steady base load electricity, without making greenhouse gas emissions. Environmentalist George Monbiot has become a supporter of nuclear power, arguing that its health risks are “tiny by comparison” with those of coal. The Fukushima accident in Japan was an unprecedented nuclear disaster, but it caused no noticeable increase in cancer deaths, even amongst workers clearing the site. Meanwhile, many are killed by pollution from coal-fired power stations - around 250,00 in China each year for instance - and that’s before we consider the greenhouse effect. “The nuclear industry takes accountability for its waste,” says Alastair Evans, director of corporate affairs at aircraft engine maker Rolls-Royce, a company which aims to take a lead in small nuclear reactors. “The fossil fuel industry doesn’t do that. If the fossil fuel industry had managed their waste in as responsible a way as the nuclear industry, then we wouldn't be having COP26 and climate conferences to try to solve the problem we're in now.”

"If the fossil fuel industry had managed their waste in as responsible a way as the nuclear industry, then we wouldn't be having COP26" Nuclear but better For all its green credentials, today’s nuclear industry is all too often bad business, with projects that take far too long, get stuck in planning and licensing, and go over budget. The UK has a new reactor being built by EDF at Hinkley Point, but it is terminally over-budget and late. Its output will cost €115 per MWh, which is double that of renewables. To make matters worse, its delay blights the grid and causes more emissions, because utilities have to burn more gas to make up for its nonappearance. Rolls-Royce is one of a number of companies worldwide that say we can avoid this with small modular reactors (SMRs), which can be built in factories and delivered where they are wanted. Standard units can be pre-approved, and other approvals are easier because they can be done in parallel, says Evans. “You don’t have to go back to government for a once in a generation decision like Hinkley Point,” he says. Nuclear you can buy They’re also easier to finance. At 470MW, Rolls-Royce’s SMRs will be a fraction of the size of Hinkley’s 2.3GW output, but cost less than a tenth the price, at around €2 billion. [Note to the reader. Nuclear reactors normally quote figures their thermal output in MWt and use MWe to refer to the amount that can be converted and delivered as electrical power. In this article, we will only refer to the electrical output, and quote it in

“A user, say a data center, books a slot for a unit, and it rolls off the production line the same way you'd order an aeroplane engine” 10 DCD Supplement • datacenterdynamics.com

MW for simplicity.] According to Rolls-Royce's site, the SMR "takes advantage of factory-built modularisation techniques to drastically reduce the amount of on-site construction and can deliver a low-cost nuclear solution that is competitive with renewable alternatives". “Like wind farms, the cost of a nuclear plant is all up-front,” says Evans. “And they give steady power for six years.” A decommissioned nuclear plant in Trawsfynydd in Wales is being considered for the first of Roll-Royce’s SMRs, and the company has spoken publicly of its ambition to build 16 in the UK. It’s reckoned that the Trawsynydd site could support two SMRs and already has all the cables and other infrastructure needed. At the time of writing, there’s no official government policy on this, however, UK Prime Minister Boris Johnson told the Conservative Party conference in September that nuclear power was necessary to decarbonize the UK electricity grid by 2035, as there is a lot of gas-fueled generation to phase out. Rolls-Royce’s SMR program had some £200 million from the UK government, and a similar amount of funding from industrial partners in a consortium which includes Cavendish Nuclear, a subsidiary of Babcock International, along with Assystem, Atkins, BAM Nuttall, Laing O’Rourke, National Nuclear Laboratory (NNL), Jacobs, The Welding Institute (TWI), and Nuclear AMRC. Having achieved its matched funding, the Rolls-Royce SMR business will submit a design to the UK Generic Design Assessment (GDA) process which approves new nuclear installations, and will also start identifying sites for the factories it will need to build the reactor components. Where and when the SMRs themselves will land is not yet clear.

Grid Scale Supplement

Learning from submarines Other countries have been making similar steps, with Nuscale in the US getting government support for a small-scale reactor program (see box). In France, President Macron has pledged to fund EDF developing SMRs for international use. The reactors are simplified versions of the large projects - most are pressurized water reactors (PWRs), and they call in experience from sea-going reactors, such as those in nuclear submarines. SMRs will be the size of a couple of football pitches. They’d be put together inside a temporary warehouse building which would then be removed, leaving the power station in situe. These reactors have of necessity been made to be small, portable, and reliable. Nuclear sub reactors range up to around 150MW. French submarines are powered by a 48MW unit which needs no refueling for 30 years. Russia’s SMR program is based heavily on nuclear submarine and icebreaker units - the state energy producer Rosatom has put together a 70MW floating unit, which

can be towed into position offshore where needed. In the UK, the first few of Rolls-Royce’s units will be paid for by the government. Subsequent ones will be available commercially, as the project has been “derisked,” and it’s easy to get debt and equity to build more. Ready to invest in At that stage, Rolls-Royce would float off the SMR business as a standalone company, financed by equity investors, and start taking orders for new nuclear plants. The company hopes to get factories established in the next few years to start making the SMRs. And that’s where data centers and other industrial customers will come in: “A user, say a data center, books a slot for a unit, and it rolls off the production line the same way you'd order an aeroplane engine.” Why would people invest? This kind of nuclear could be much more viable than the giant projects. Rolls-Royce expects its SMRs to produce electricity at around €50 per MWh. That makes it as cheap

Beyond the SMR Alongside the SMR, there’s an other UK project involving Atkins and Cavendish, to build an Advanced Modular Reactor (AMR), called U-Battery. This is also aiming to deliver reactors built in factories, with a capacity of around 10MW, later in the 2030s. U-Battery has demonstrated a mock-up of its reactor vessel and heat exchangers, as a milestone towards delivery of an actual system. AMRs would be a next generation of nuclear plant based on different technology. Some of those under consideration include high temperature gas reactors (HTGR), sodium-cooled fast reactors (SFR), lead-cooled fast reactors (LFR), molten salt reactors (MSR), supercritical water-cooled reactors (SCWR) and gas-cooled fast reactors (GFR).

Grid Scale Supplement 11

as offshore wind power - but with the very important benefit that the power is delivered continuously. If the whole of society is decarbonizing, then our need for electricity will expand. Heating and transport must be switched to electricity, and that means more electricity must be generated. And beyond that, the industry’s dependence on energy has been revealed graphically by the effects of the current hike in gas prices. By providing long-term, guaranteed low carbon power which can support baseload without the variability of solar and wind power, SMRs could help support the decarbonization of industries. A large amount of baseload electricity could also help foster other energy storage systems. For instance, hydrogen could be used as a portable fuel, but to be green it would have to be made by electrolysis using renewable electricity. Benefits for data centers “We are keen to present the off-grid

"We are keen to present the off-grid application of stable secure green power to any and all carbonintensive industries" application of stable secure green power to any and all carbon-intensive industries,” says Evans. By paying the costs of an SMR located nearby, a steelworks could switch to green electricity, and escape from increases in the cost of gas - while also reducing its emissions drastically. Data centers should find this an easy market to participate in. Large operators like Google and Facebook are well used to making power purchase agreements (PPAs) for wind farms or solar plants: Rolls-Royce believes they could take a PPA for a portion of an SMR project’s output. For a data center operator, a PPA for a portion of an SMR might seem a distinct upgrade on a PPA for a wind farm.

While the operator has paid for green electricity to match its consumption, the wind farm would deliver it at particular times, instead of when the data center needs it, so the renewable energy would be more of an offset for the electricity used by the data center, which would be made by whatever mix is actually on the grid at that time. Data centers that buy a PPA for nuclear energy, on the other hand, would be able to just plug straight into the power station. At that point, going nuclear provides the best of both worlds: as well as emitting no greenhouse gases, the data center would also be independent - free of the fear of blackouts on the grid.

The US view In the US, Nuscale is the front runner for small reactors, working on a water-cooled design. The company’s Nuscale Power Module (NPM) is a 77MW integral pressurized water reactor (PWR), with 12 of these modules combined in a flagship plant design, that has a total gross output of 924MW. “This is just short of traditional gigawatt nuclear plants, which provide around 1,000MW,” according to Ryan Dean, senior public affairs specialist at Nuscale. “We also offer smaller power plant solutions in four-module (308MW) and six-module (462MW) sizes.” The project started at the University of Oregon with Department of Energy funding from 2000 to 2003, then went commercial when the funding was cut. After problems with its first major backer, the company is now funded by engineering firm Fluor, and expects to produce a working reactor soon. Dean says the reactor will be “an ideal solution for decarbonizing energy intensive industries. The level of plant safety and resiliency is appealing to hospitals, government installations, and digital data storage facilities that serve as mission critical infrastructure and need a limited amount of reliable electricity," he says.

Nuscale claims to offer 154MW at 99.95 percent reliability or 77MW at 99.98 percent reliability - both over the 60-year lifetime of the plant, and it’s designed to work in island mode as part of a microgrid. It’s also the first nuclear plant design capable of so-called “black start”, ie it can be switched on without any external grid power, according to Dean. The station is earthquake proof and EMP resistant, and will keep itself cool in the event of losing power. Dean also points to other benefits, including using the plants for desalination, with each module producing around 77 million gallons (290 million liters) of drinking water per day. They are good for load following, helping grid capacity deal with the intermittency of wind, solar, and hydro generation. All nuclear reactors produce a lot of waste heat, and Nuscale is looking to find ways to use the waste process heat from its reactors for industrial applications such as the generation of hydrogen for fuel cells. “We are progressing towards the commercialization of our first project,” says Dean. “By the end of this decade, a Nuscale small modular reactor (SMR) power plant will become part of the Carbon Free Power Project (CFPP), an

12 DCD Supplement • datacenterdynamics.com

initiative spearheaded by the public power consortium Utah Associated Municipal Power Systems (UAMPS). “In August 2020, we made history as the first and only small modular reactor to receive design approval from the US Nuclear Regulatory Commission (NRC),” he continues. “Nuscale is actively pursuing projects around the world. We have memorandums of understanding for the deployment of Nuscale SMRs in 11 countries.” Meanwhile, the US has at least one nuclear-powered data center in the works. Talen Energy operates a 2.7GW nuclear power plant at Susquehanna in Pennsylvania, and in September 2021 it broke ground on a project to build a data center on the site that could grow to 300MW. Unlike the large scale projects we’ve looked at here, the project won’t be contributing to decarbonizing the grid, however. It’s more of a scheme to soak up surplus energy for profit and detoxify the crypto market. Talen has said the Susquehanna Hyperscale Campus in Berwick, Luzerne County, will be home to cryptocurrency miners, simply using energy to generate speculative assets, rather than any effort to decarbonize existing industry.

Grid Scale Supplement

Energy storage by air Could the answer to a stable grid be literally all around us? Peter Judge Global Editor

low up a balloon and then release it. It flies around the room. What you’ve just demonstrated could give us a stable electric grid, cut emissions at data centers, and avoid tons of greenhouse gas entering the atmosphere. You see, we want to decarbonize the electric grid, by switching off coal plants and replacing them with renewable sources like solar and wind. UN SecretaryGeneral António Guterres says there must be no more coal power plants built after this year. But there’s a big problem. One that is easy to state - but really, really hard to solve. Renewables don’t work all the time. Fossil fuels do. Solar and wind power only flow when the sun shines and the wind blows. But our demand for electricity continues regardless, so we have to have power sources, like coal and gas-fired fossil fuel generation, which can be switched on as

needed. To wean ourselves off fossil generation, we need ways to store energy created at the peaks of renewable production, so it’s available on a cold, dark night. We can’t all do pumped hydro We’re already using big energy storage systems, like pumped hydroelectric storage, based on hydroelectric stations fed from large lakes in mountains. Run the power station in reverse, and it pumps water up to the lake, storing energy. The UK’s Dynorwig station in Wales can store up to 9GWh of energy. and deliver it at a rate of up to 1.8GW. But you can’t build a Dynorwig anywhere. So other options are being proposed, ranging from high-tech new

battery technologies, through using electrolysis to create green hydrogen, to the really low-tech option of cranes that simply lift giant concrete blocks into the air to store potential energy. Compressed air is also on that list. Blow up a balloon and you compress air inside it, storing energy. Allow it to expand, and the energy is released. Physicists explain this with Boyle’s Law. The idea has been used to run trains in mines without a risk of explosions, and more recently has been scaled up to utility levels. “Storelectric’s CAES plants are usually designed as stand-alone plants with standalone profits based on buying electricity from the grid and selling it back to the grid,” says Mark Howitt, CTO at Storelectric,

“Oil and gas companies can really start leveraging their assets to beautifully get into clean energy”

Grid Scale Supplement 13

Grid Scale Supplement a company which aims to offer CAES as an energy storage service for utility grids. Howitt says CAES is ideal for data centers, as they can fund the storage system through power purchase agreements, and benefit from a steady clean energy supply. “Storelectric’s fully validated simplifications of existing, well-proven technologies offer 30-60 percent IRR (internal rate of return) stand-alone and 50-70 percent in conjunction with renewables, which also reduces the capital costs and improves the profitability of the renewables,” he says. The company has plans for two CAES plants storing 200MWh and 2.5GWh of energy in underground salt caverns, and an engineering company ready to build the plant, but those plans are on hold right now, as the company looks for funding. Brexit has made it harder for Storelectric to raise money and trade on the continent of Europe but Howitt believes it may have the side-effect of increasing the need for energy storage, because in times of grid stress, European nations rely on interconnectors to provide power where there is a shortage. These connectors will still exist, but they will be available on different terms after Brexit, Howitt warned DCD: “While we were in the European Union, legally the UK had to be treated as a domestic customer. Now we are no longer legally treated as a domestic customer - we're an export destination, and would be less favored than domestic users.” So more energy storage will be needed, he believes. On this and other issues, Howitt runs a very informative blog. CAES’ proven history CAES is well proven, having been in use since 1978 at a 300MW plant at Huntorf, Germany belonging to EN Kraftwerke. Meanwhile, Alabama Electric Corp has run a 110MW plant in McIntosh since 1991. Like Storelectric, those plants pump air into underground salt caverns, but these early systems use the energy to assist a gas turbine, which doesn’t eliminate the use of fossil fuels, and isn’t particularly efficient, operating at around 40 to 50 percent. The next generations of CAES are working to improve on that - and the problem they face is that compressing air makes it hotter. If that heat is lost, efficiency is reduced. “All compressed air storage systems try to separate the air from its thermal energy and store those two separately, such that they may then be recombined later,” says Liam Newcombe, senior vice president of engineering at Energy Internet

"All compressed air storage systems try to separate the air from its thermal energy and store those two separately, such that they may then be recombined" Corporation (EIC), a startup proposing giant CAES systems storing air in spent oil wells. Essentially, when a gas is compressed it generates heat. When it expands it needs to take in heat. If you saved the heat when you compress the gas, you can use it in the expansion process. Because CAES can hold energy for a long time, it’s got potential for large-scale smoothing of the grid, says Newcombe: “You can have very long durations of storage using compressed air in subsurface geological features. You don’t use that to move the California solar peak four hours later, every single day, to hit the demand peak for air conditioning. You use it for periodic deep storage interventions on the grid, that get you from an 80 or 90 percent renewable grid to 100 percent renewable.” CAES systems should be “the ones that you call when the batteries the on solar farms have all hit zero and it's still cloudy, and it's still cold and there's no wind,” he explains. “They’re the times the UK calls on Drax [a coal/biomass generator plant in Yorkshire] to start up the turbines, and we call on all the gas capacity - which we do a couple times a year, for a week or so at a time.” They are far more scalable than batteries, which can charge and discharge cheaply, but have a high cost per unit of energy stored, he explains “If you want to store twice as much energy, you buy twice as many batteries. With compressed air, if you're pushing it into a subsurface feature, the incremental cost of an extra million liters of compressed air is very little, because geological features are just bloody big to start with.” Liquid air Highview Power takes a different approach: it compresses the air much further, until it condenses into a liquid. Its pilot plants in the UK and the US, which include a 50MWscale plant near Manchester, use smallscale above-ground equipment which could suit data center needs. Liquid air systems don’t need underground reservoirs, and can store energy for a long time. The downside is they generate much more heat (on condensation) and cold (on evaporation) so the process means a large heat store or else losing a lot of energy.

But it could be good for digital operators, a spokesperson told DCD: “At very large-scale Highview Power’s cryogenic energy storage technology can help data center operations. Highview has grid-connected systems that can store power from air for hours to even days.” Highview’s latest announcement is a 50MW plant being constructed in the Atacama desert in Chile, in partnership with local utility Energía Latina S.A.-Enlasa. It’s a good fit, as the location can readily produce a large surplus of solar power. Data center partners Among CAES companies, the one that talks most vocally about partnering with data centers is the US-based Energy Internet Corporation (EIC) - perhaps because its CEO Shankar Ramamurthy knows the territory, having previously run VPS, a startup which aimed to get flexible power provision into data center racks. EIC has a system that uses a “liquid piston,” and is simultaneously working on big projects below ground, along with smaller ones above ground, which will use a liquid air system similar to Highview’s. ”We will have products ranging from one megawatt based on liquid air, potentially scaling up to a gigawatt or more,” Ramamurthy told DCD. The company is using engineering partners. like process firm Lummus, to deliver its designs. EIC plans to show pilot schemes for compressed and liquid air next year, Ramamurthy says. One EIC project is a 5MW zero carbon microgrid, being developed with utility company Pacific Gas and Electric, along with oil and gas firm California Resources Corporation (CRC). According to Ramamurthy, the 5MW project will “demonstrate a full operating system, with subsurface storage and the microgrid distribution, as well as a green interface.” He says this project can be scaled up to a 2GW plant, with 30 to 50 days of energy storage: “Which by the way is TeraWatt hour (TWh) scale, supporting a genuinely zero carbon power strategy.” More pertinently to data centers, EIC is also working on a 1MW liquid air project designed specifically to feed steady green power to data centers. Ramamurthy says this project is with a “large chip company.”

Since the company has previously announced exploratory work with Intel, it seems likely Intel is the partner here. And finally, in Bahrain, there’s a 600kW solution designed to offer steady zero carbon power. That should be up and running by summer of 2022, he says. “We expect commercial products to be available towards the end of next year, at all scales,” he says, and the smaller systems will definitely be offered to data centers as off-grid power storage. Recycling old oil wells For its larger infrastructure projects, EIC uses depleted oil and gas wells to store the compressed air. In the 5MW pilot project, the wells are being provided by the oil company CRC. “CRC is the largest oil and gas company in California,” says Ramamurthy. “They have over 5,000 depleted wells and a number of reservoirs in central California. We have identified a number of candidate reservoirs.” CRC and other oil firms will jump at the chance of recycling old wells, and Ramamurthy believes they will be persuaded into a more active green strategy. "Subsurface reservoirs like depleted oil and gas reservoirs are the best assets. They already have all the pipelines in place, and there is decades of analysis data at a very, very granular level to understand precisely how that reservoir will work” he told DCD. “Oil and gas companies can really start leveraging their assets to beautifully get into clean energy.” Many oil wells are coming to the end of their life, he says: “Oil and gas companies will have to spend billions of dollars in decommissioning. So instead of spending that to make them dead assets. How about actually repurposing them, to provide clean energy? That way, an oil and gas company has created a clean energy future where they can sell a much more valueadded product. “This gives a segue for oil and gas companies to recognize an alternate richer future,” he says. “Instead of passively resisting the change, this would encourage them to accelerate and become the movers and shakers in this transition, and the whole of society would benefit.” Taking in other underground features as well, suitable reservoirs are accessible from 80 percent of the US today, he says, and he echoes Newcombe’s comment on scalability: “For other systems, if you want to store more energy, you're going to have to spend more. In our case, you are actually going to have to spend less.”

"We don't really have the 30 or 40 years that governments are planning for a leisurely transition. The climate may be a lot more unforgiving" The liquid piston Within its systems, EIC attempts to keep the expansion as adiabatic as possible - minimal heat transfers in and out of the system, using water sprays within its compression cylinders to maintain a steady temperature (isothermal compression). It also operates a series of compression cylinders at increasing pressures to avoid a large compression stage. The “liquid piston” the company describes, simply means the expansion happens in a vertical cylinder pushing down on a head of water, rather than hydraulically driving a mechanical piston. “The liquid piston is our invention and it’s one of the elements that enables this to work with really large underground reservoirs,” says Ramamurthy, adding that moving a body of water “make this less compressed air and more pumped hydro. “Anyone who doesn't see the rest of the system, and looks at our hydroelectric plant, will go away with the impression that is pumped hydro,” he says. So it benefits from techniques used

in plants like the UK’s Dynorwig, without needing to find a mountain with a high lake because, instead of pumping water up to a lake, it’s pumping water into a compression stage that drives the compressed air underground. “So this is really a pumped hydro system,” he says. “It's a pumped hydro system on steroids.” EIC uses existing off-the-shelf equipment, which is a benefit, but also a limitation, in that it adds the requirement for multiple compression stages, says Ramamurthy. Much of this will improve in future versions. “We have the potential to be the lowestcost energy,” he says. “That's our mission, to make our energy less expensive than what is available on the fossil fuel grid today.” It’s an urgent task, because “we don't really have the 30 or 40 years that governments are planning for a leisurely transition. The climate may be a lot more unforgiving than the delusionary times that we are allowing for our transition.”

Operations Supplement 15

bloomenergy.com

#BeTheSolution