A great deal of the electricity in the United States goes to waste.
Much is lost in the initial generation of electricity. And much is lost through the use of inefficient devices, like incandescent light bulbs that heat up a filament to produce light.
But power is also lost in between, on the grid, as it is carried along hundreds of miles of wires, repeatedly shifted between different voltages, and converted from AC to DC and back, all in the split second between the time it enters the grid and the time it powers your computer.
How much power is lost on the grid?
The consensus among experts in the field is that most electricity is lost on the two ends, in generation and use, and not that much in between. The Department of Energy estimates that, of 37.7 quads (quadrillion BTUs) of “energy consumed to generate electricity,” 23.24 quads (about 62 percent) is wasted as “conversion losses.” After that, only 0.84 quads (roughly 2.2 percent) is lost or “unaccounted for” in transmission and distribution (T&D).*
Now a research and development lab-cum-start-up out of North Carolina’s research triangle has begun commercializing a technology it says can measure and manage electricity with a level of accuracy and precision far beyond any existing technology, using a cutting-edge application of real-time computing.
3DFS has two core messages, both of which promise to shake up the status quo, not only in the power sector but in our general understanding of electricity.
First: A close-up, real-time view of electricity reveals that there is a lot more waste happening on the grid than current estimates capture. 3DFS contends that the waste DOE clusters under the nebulous term “conversion losses” is in fact spread out across the grid, in generation, transmission, distribution, and consumption. That’s what their measurements have shown.
In a sense, that’s good news. Some amount of losses are inevitable in the conversion of fuel to steam. But electrical losses on the grid are preventable. Which bring us to …
Second: Waste on the grid is the result of poor power quality, which can be ameliorated through digital control. Real-time measurement makes that possible. 3DFS technology, which the company conceives of as an “operating system for electricity,” can not only track what’s happening on the electricity sine wave from nanosecond to nanosecond, it can correct the sine wave from microsecond to microsecond, perfectly adapting it to the load it serves, eliminating waste.
The company has an unorthodox plan to grow slowly and organically. But if it can scale its technology up to wide use, across the grid, 3DFS says, it could potentially double the energy efficiency of the electricity system, getting twice the energy services out of the same amount of generation. That would radically hasten both electrification and decarbonization, proving a weapon against climate change at least as potent as renewable energy itself.
Along the way, it could vastly enhance our understanding of electricity.
Interlude: your bullshit detector
By now, many BS detectors will be ringing at full volume. I get it. This sounds like magic beans.
And this is a field full of magic beans. Manifestos promising revolutionary energy solutions (if only The Establishment weren’t suppressing them) abound. I get those emails myself.
3DFS tech directly challenges a lot of conventional wisdom in the field and involves a technology and a level of data that are genuinely new, such that basically nobody beyond 3DFS has experience with them. It makes it difficult to consult outside experts — the best I got was, “Well, I don’t hear anything that sounds impossible…” — and quite rightly inspires a heightened level of skepticism. One professor of electrical engineering, when I mentioned that 3DFS believes DOE numbers on energy waste are mistaken, became enraged and literally hung up on me.
But 3DFS is not a fly-by-night operation, it’s a research lab that’s been around for 15 years and employs well-known and experienced engineers. And as I learned by talking to people who have used it, the product is not vaporware. It exists, and it works.
And 3DFS is not secretive with its data. It is happy to show anyone who asks what real-time visibility into electricity looks like. Its claims are testable and it is eager to have them tested.
It took a great deal for me to get past my own skepticism. Reporting this story, I was repeatedly reminded of science-fiction author Arthur C. Clarke’s famous dictum that “any sufficiently advanced technology is indistinguishable from magic.” But I’m convinced that this thing is not magic. It is real, and it’s a big deal.
Digital measurement and management of electricity is a huge deal — and very complicated
If it proves out, the implications of what 3DFS calls “software-defined electricity” (SDE) could be very big. To begin with, recovering some or most of the lost electricity on the grid would amount to finding a huge new source of zero-carbon power — a powerful resource in the fight against climate change.
And that’s just the beginning. SDE promises to improve efficiency on both ends as well, making generators more efficient and boosting the performance of every electrical device (including storage, like batteries) in every environment, and reducing consumption and waste heat.
It promises to hasten electrification of the economy, radically reduce infrastructure costs, and open up new lines of science and inquiry.
Here’s the problem, though: It is a devil to explain. 3DFS won one of Popular Mechanics’ “breakthrough awards” in 2017 and there have been write-ups here and there, but nothing yet has captured the full potential.
It doesn’t help that 3DFS technology measures and manipulates electricity at a level most of us never think about, acting on characteristics of power most people aren’t even aware exist.
So we’re going to walk through it together, step-by-step: why the device exists, what it does, its implications, and what the company plans to do (and not do) with it.
By the end, it should be clear that if 3DFS can shepherd its technology through the temptations and culs de sac modern tech capitalism, it could change the entire landscape of energy.
Electricity in its natural state is chaotic
So we’re wasting electricity on the grid. Where does it go?
In a nutshell, the electricity gets converted to heat or vibration instead of useful energy services (or “cold beer and hot showers,” in the lingo).
This low-level waste happens from the moment electricity is generated at a power plant to the moment it enters your iPhone. The generators that create it, the transmission lines that carry it to transformers and distribution systems, which carry it to buildings and electrical panels, which carry it to devices, which convert it to services — all along the way, everything heats up, hums, and vibrates. That’s electricity being thrown off as heat and kinetic energy.
Electrical engineers are very familiar with “transients” and other macro distortions that travel on the grid. But the closer you peer, the more chaotic things become. In its “natural” state, electricity is turbulent, like water flowing down a mountainside. The electrons tumble and splash, and all that splashing amounts to waste. Every device that requests electricity from the grid sticks its little cup into that same turbulent flow, with all its little spikes and surges. More splashing.
And splashing’s not the only problem. To perhaps strain the metaphor a bit, each electrical device is designed to work with water at a very precise flow and pressure. When it sticks its cup into the chaotic gusher, it is constantly having the cup over- or under-filled. It is receiving poor-quality water, never quite performing optimally.
All the electrical devices hooked into the grid are receiving varying levels of poor quality electricity, constantly being over- or under-powered, creating waste, backfeed into the grid, and unreliable performance.
Why have we allowed this constant low-level inefficiency to continue for as long as there’s been electricity, despite electricity becoming ever more central to our lives?
The answer is simple: We couldn’t see the waste. We didn’t have the data.
The mismatch of electricity with loads happens at the microsecond level
What does “poor power quality” mean, anyway? Most people know power, maybe voltage and frequency, that’s about it.
But power turns out to be pretty complex. The technology 3DFS developed in its Pittsboro, North Carolina, research facility measures 26 parameters of electricity, including voltage, phase angle, phase imbalance, active power, reactive power, harmonics, power factor, and more.
All of those aspects of power matter to loads. And the exact needs of loads using electricity change from microsecond to microsecond.
A microsecond is about a millionth of a second. The US grid runs on a frequency of 60 hertz, which means electricity is transmitted at a frequency of 60 cycles per second. Each cycle, each little sine wave, is one-sixtieth of a second. Each contains roughly 16,667 microseconds.
So, imagine an electrical panel connected to a variety of machines using electricity. “Each one of those loads has a circuit board with resistors and capacitors and inductors and coils and all sorts of components,” says Chris Doerfler, co-founder of 3DFS. “If you slow the electricity down to the microsecond, and then look at all of these different circuit boards, they all have different needs, in the moment. Every microsecond a different component within the circuit board needs different electricity in a different way,” with a particular balance of inductance and capacitance.
Each load expects perfectly synchronized electricity and never quite gets it. The waste, the constant mismatch of power supply and demand, is happening at the subcycle level, continuously.
We’re still rounding and approximating measurements of electricity
The problem is, we’re still not measuring electricity digitally, continuously, using real data about real electrons passing through wires. We’re still using the same analog method we’ve been using since the 1890s, when electrical meters came into wide use to track the electricity going into buildings.**
Meters employed, and still employ, the root mean squared (RMS) method of measuring electricity, which involves taking periodic measurements of current and averaging those measurements to estimate how much power is traveling past.
Some 130 years later, the number and frequency of the periodic measurements have grown. We can now measure dozens of points within a cycle.
But averaging periodic measurements is never going to give you real-time data about what is happening at the subcycle level. To correct the electricity, you need continuous subcycle information about it, and we just haven’t had the computing power necessary to absorb and analyze that much data.
“Today, power quality loss measurements are rounded and approximated in every industry, in every instrument, and in every tool,” Doerfler says. “Smart meters are a few hundred bucks because they do not have processors inside them. This is how you know electricity is measured in an analog way; the lack of processing power.”
3DFS measures electricity digitally, in real time
This is the first half of 3DFS’s breakthrough: It can measure electricity continuously. Specifically, it measures 26 separate parameters, in 24-bit resolution, in real time. Over the course of a single one-sixtieth-of-a-second cycle, 3DFS tech gathers and analyzes over a million points of data.
“We have a perfect digital replica of the analog signal,” Doerfler says, “without any flaws or errors, zero noise, within a few nanoseconds after its produced.” (A nanosecond is one billionth of a second.)
That data runs through a series of analytic and predictive algorithms 3DFS has been working on for more than 10 years, which extract usable information and then discard 99 percent of the data. The result is actionable analysis of power quality in real time.
You might wonder how 3DFS is able to handle these all these terabytes of data that have stymied previous computing methods. That brings us to another interlude.
Interlude: 3DFS real-time computing
3DFS is able to gather and analyze data so fast through a new method of real-time computing. It’s not something it can own or can patent, just something its engineers have learned to use over a decade of R&D. They call it “task-oriented optimal computing” (TOOC).
Arguably, TOOC is the real story here, since its applications are, if anything, as broad as the applications for electricity. At one point Doerfler casually mentioned that the Library of Congress is having trouble digitizing its collection — and that his team could do it in a year.
But we’re not going to go down that rabbit hole, because this is a story about clean energy, and it’s just getting interesting.
Using real-time data, 3DFS can clean up electricity
So 3DFS tech is gathering and analyzing enormous amounts of data in real time. What does it do with the analysis?
Using nanosecond-level data, it makes microsecond-level predictions about how to correct the signal, “noise canceling” along all of the 26 parameters it measures, yielding perfectly synchronized electricity.
It does this with its flash energy storage system (FESS). Using the real-time analytics, the FESS can inject or extract microamps of electricity from the three-phase signal, radically boosting power quality.
“We’re gathering data and making corrections before existing [systems] can even acquire their data,” Doerfler boasts.
SDE offers an immediate boost in power quality, but it also gets better over time, because the system learns, using artificial intelligence.
Learning about loads to better meet their needs
Once 3DFS tech is attached to, say, a data-center electrical panel (installation is non-intrusive, with no interruption in power, and it takes about a half-hour), it begins analyzing and correcting the electricity passing through it. But it also uses the artificial intelligence algorithms 3DFS has developed to learn. And, over time, it can create a perfectly accurate digital profile for each load attached to the panel.
Doerfler compares it to a game of tug-of-war — eventually, each side of the rope reveals a signature mix of tug strength, length, and frequency.
Just how sensitive is the system to differences in loads? Down to the individual circuit board component level.
“You could have 100 televisions coming off the exact same assembly lines, with the exact same components, the exact same robots putting them together,” Doerfler says, “but here’s the thing: Each one of those components has variances in tolerances.” It can come down to the thickness of individual wires, or the tension at which wires were coiled. All of these differences matter at the microsecond level.
3DFS tech can use those differences to learn. Over time, “the metadata, the actual operation of that circuit, becomes usable data,” Doerfler says, allowing the system to anticipate its needs.
The ability of the 3DFS system to learn the consumption profile of all the loads to which it is attached gives it the ability to continuously perfect the electricity supplied to them.
Not only does this make electricity use more efficient, it gives the system real-time visibility into each load’s performance, which allows it to immediately detect any changes. These changes might be faults or degradation (anticipating them reduces maintenance costs), or they might indicate that the system has been hacked.
“Not only can we identify when an inverter is hacked,” Doerfler says, SDE “can autonomously correct the problem. It’s not only cybersecurity awareness, it’s prevention.”
Starting with data centers: the Freudenberg IT experience
As you’ve probably gathered by now, the applications of SDE are innumerable, but as a commercial concern, 3DFS has elected to begin by targeting data centers, the huge warehouses where companies house other people’s data on racks of servers. (The company expects a half-dozen other products in the commercial pipeline by the end of the year.)
It makes sense: Data centers are big power consumers, and poor power quality costs them money. The industry is constantly spending more on new ways of buffering and balancing power — uninterruptible power supplies (UPSs), filters, and banks of capacitors — but they all amount to moving power around and storing it. None of the available solutions clean up the power directly.
So 3DFS has developed a line of products called VectorQ, boxes that attach to the electrical panel in a data center and provide it with SDE. It cleans up power, provides each attached load with precisely the level and quality of power it needs, reduces consumption, reduces the costs of managing waste heat, and extends the working life of the machines.
3DFS has developed a power quality rating (PQR) that it claims is more accurate than existing PQ ratings; it takes into account all 26 parameters of electricity.*** Perfect, lossless electricity would have a PQR of 100 percent. Uncontrolled electricity, 3DFS has found, ranges from 20 to 40 percent. VectorQ gets the number up to about 98 percent.
One of the first data center owners to see the potential of the VectorQ was German-based Freudenberg IT (FIT), a global company that hosts critical software services for a range of Fortune 500 companies like Bridgestone and ABB. Data centers are not its main focus, but it maintains a number of them to host services and spares no expense on data security and reliability.
Michael Heuberger, CEO of FIT‘s American division, ran into the 3DFS engineering team at a tech conference a few years ago, when its product was in the prototype phase. “Sounds great,” he recalls saying. “Maybe try it somewhere else first.”
3DFS came back a year later with a commercial product and Heuberger agreed to let them demonstrate it in the miniature data center his company has built as a testbed for new products.
“Hey, dude, it looks like it’s working!”
“I remember the moment and I get goosebumps,” Heuberger recounted to me. “We hooked it up to our system — it took maybe 30 minutes until it was all installed — and [the 3DFS engineer] said ‘I’ll turn it on now.’ Click.”
Power consumption dropped by 20 percent, server temperature dropped by 20 degrees, and PQR reached the high 90s.
“I went to my UPS displays, was able to see everything he was seeing, and said, ‘Turn it back off,’” Heuberger recalls. “He turned it off, everything went back to crazy, ugly, shitty load distribution.”
“We did this a couple of times,” he laughs. “I figure … he cannot manipulate this, right? I said, ‘Hey, dude, it looks like it’s working!’”
Here’s the VectorQ interface, with SDE turned off. Note the high harmonics and current out of sync:
Here’s the interface with SDE turned on. Note the low harmonics and current in sync. (You can’t see all the tiny PQR ratings in the top right, but they are higher.)
FIT is now running the VectorQ in the test center and plans to install it in data centers worldwide. (The first VectorQs were expensive, around $100,000, with what Heuberger estimates is a seven-year payback. But costs are falling rapidly as manufacturing scales up.)
Heuberger says he was not primarily interested in servers lasting longer or power consumption falling. As a roughly $5 billion, high-end service company, FIT spends huge amounts on labor and security to protect data. Power is not a large cost.
But power quality is another matter. Heuberger was a machine and electrical engineer in a previous life and is intimately familiar with the problems poor power quality causes: “engines kicking in, air conditioners kicking in and backflushing dirty energy, servers rebooting, SAM storage having data loss.” All these day-to-day disturbances are like ghosts in the machine, often inexplicable and unavoidable to data center managers who cannot see or manage power quality at the level that matters.
Eliminating these disturbances, making the machines operate more reliably and with less waste heat, is every data center’s dream. “There’s a big market out there,” Heuberger says, “for relief.”
Another environment where power quality matters: ships
It’s not just the data center market, though. There are all kinds of specialist markets where power quality is at a premium.
Take ships. We don’t normally think of them this way, but every ship of any size is effectively a floating microgrid. And because they tend to be small microgrids, with a few as two or sometimes only one generator, there is a very low tolerance for distortion and harmonics.
When Mike Gaffney, a senior engineer at Navis Energy Management Systems who works on nautical electricity systems, first heard about 3DFS through word of mouth, he recalls, “I looked it up and said, ‘Wow, if it works, this would be awesome!’”
So he had the company come give a presentation in Norfolk and then went down to the 3DFS facility. Gaffney recounted the tale to me:
I hooked up my meters, they turned it on and off, and sure enough, it worked, as far as I could detect — and I have fairly good metering equipment. So I spent about five or six hours down there trying to understand the technology, and went home, and for about a week or so I pondered over it, and said, ‘Ah, I really don’t get it.’ [laughs] So I called them up and said, I gotta come back! I went down for another five or six hours of schooling and got a little bit of a better handle on it.
(Gaffney is a certified power quality professional, a certified energy manager, and a certified energy auditor. Spare a thought for your author, a layman.)
Gaffney is convinced that poor power quality causes the same sorts of problems on the Navy’s new diesel-electric ships that bedevil data centers — breakers flipping, equipment running hot when it shouldn’t, “phantom alarms” ringing for no reason. These ghosts in the nautical machines are, he believes, the result of harmonics caused by dirty power.
“If we can mitigate that stuff,” he says, “we can increase both the reliability and the energy efficiency of the ship.”
He doesn’t have proof, though, and there’s a shortage of research. That’s why he wants to experiment with SDE on Navy support vessels. He worked with 3DFS on a whitepaper, which they recently presented at the Naval Surface Warfare Center in Bethesda, Maryland. They will know soon whether testing will go forward and hope to have data by July.
“Everything in my gut, from everything I’ve learned and seen, says this is going to work,” Gaffney tells me.
Applications for software-defined electricity are everywhere
It’s not just data centers and ships. SDE will help in every situation where electricity is used.
“At the end of the day, the use of electricity needs to be synchronized,” Doerfler says, “whether it’s a substation, a data center, or an international space station.”
He compares today’s electrical devices to cars operating with “dirty gas, dirty oil, dirty fluids — your car is going to operate hotter, burn more fuel, and die sooner. The exact same is true of electricity: Those distortions, when electricity is not perfectly synchronized, are either overpowering or underpowering [loads].”
That’s what all that heat and humming is, at every single stage of the electricity system. It is a constant low level of waste and wear. Today, every single electricity system in the world is overbuilt, generating more than it consumes, to compensate for this waste. That’s true for grids big and small, right down to a US Marine forward operating base in the Afghan desert that must run up to two 100 kW generators to supply a 100 kW load.
SDE can eliminate that waste and wear. A 100 kW load can be supplied with a 100 kW generator. Electricity infrastructure can be correctly sized.
And SDE can eliminate the waste that comes from electricity conversions, like going from DC to AC, which happens in millions of batteries and solar inverters every day. SDE can accomplish that conversion “in one process, in one moment, and it’s digitally perfect,” Doerfler says. “We use the least amount of energy possible on the conversion, and it’s not in series and mechanical, it’s in parallel and software-oriented.” The same goes for shifting voltages. Electricity management becomes near-lossless.
(Note: While energy losses converting fossil fuels to steam and then electricity are unavoidable, the same is not true of wind power or solar photovoltaics. Theoretically, SDE could ensure that 100 percent of the energy generated by those sources reaches end users, opening the possibility of a nearly lossless electricity system.)
Doerfler says there is no barrier to SDE being installed at the microgrid or distribution-node level (imagine having a digital profile of every load on a distribution node). It would even work at the transmission level. Anywhere electricity flows, it can be digitally measured and synchronized.
3DFS’s ultimate vision is to get the technology small enough to fit on a chip. Each electronic device would have an SDE chip (like its wifi chip) that perfectly synchronizes electricity for its circuit board — a kind of “Intel Inside” for power quality.
For now, the work begins by retrofitting current infrastructure. “Heavy, physical use of power, motors and compressors, are going to immediately reduce their energy consumption maybe 20, 25 percent,” Doerfler says, “in IT loads and computers, it will be 10 to 15 percent.” But he stresses that those are initial savings; as the AI system learns, it gets more efficient. He thinks a fully SDE network can eventually reduce consumption by 30 to 35 percent for most applications, more for heavy industrial processes.
In the end, 3DFS believes that SDE can recover about half of the wasted electricity tucked under “conversion losses” on the DOE chart. That in itself would amount to a revolution.
But the electricity savings are just the beginning. Developing a perfectly accurate digital representation of every load on the grid, with real-time data on its performance, would open a new world of energy analysis and management, new areas of product development, and new avenues of scientific inquiry. We could learn more about electricity — the actual electrons bouncing around in our wires — in the next 10 years than we have in the past 100.
As a final example, take batteries, the great hope for enabling renewables and stabilizing the grid. Their internal chemistries demand precise power levels, which they, like all loads, never get. That’s why they run so hot and degrade so quickly.
Doerfler says SDE can build a profile of a battery “so accurate, we can see dendrites and sulfites growing and can react with corrective action in microseconds.” SDE can supply the battery with perfectly synchronized electricity, eliminating waste heat and extending its life.
What is true for batteries is true for every load. With SDE, electricity can operate with perfect digital accuracy.
Rather than tying the security of our electricity system more and more closely to our information and data networks (and all their familiar vulnerabilities), SDE infuses digital intelligence into the electricity system itself, making it self-monitoring and self-correcting.
Digital measurement and management is the natural next step in electricity’s evolution — like any great new idea — unfathomable before you’ve heard it, kind of obvious afterward.
How to get software-defined electricity to the world without capitalism botching it
The engineers at 3DFS, a research laboratory and a business, believe that SDE could be a universal operating system for electricity, bound eventually to be integrated into every load and every piece of infrastructure.
3DFS probably could have cashed out with real-time computing in any number of directions, but they wanted to make sure it was applied to electricity — the big climate problem/opportunity of our age. And they are wary of SDE getting stuck or misdirected.
“The technology needs to be advanced quicker than capitalism will let it,” Doerfler told me. It is not a perspective one usually hears from start-ups.
He cites the technology trajectory of wifi (one of 3DFS’s engineers was an early wifi pioneer), which was once on track for gigabit wireless in the early 2000s; there would have been no need for fiber at all. Instead, private companies released standards, saturated the market with products using those standards, development slowed, and we still don’t have gigabit wireless.
Or take the single-electron transistor, which “could have detected breast cancer when it was only one cell,” Doerfler says, “but we didn’t take it that direction, we took it to defense, because that’s more important, right? And now General Dynamics owns it, and they don’t give a shit about breast cancer.”
3DFS doesn’t want a similar fate to befall SDE. They don’t want real-time electricity analysis and correction to end up the intellectual property of some tech giant that just uses it to improve data centers. They want it everywhere.
“For us, it’s about maximizing the technological potential, not the dollars,” Doerfler says. “The dollars will come. But the control is more important.”
For now, that means moving slowly. So far, 3DFS has followed something like the pharmaceutical model, doing its own long-term R&D, building up its intellectual property, and slowly translating that into products. It hasn’t taken on any investment or debt and is wary of any partner that demands too much ownership of the company or its IP.
The plan is to start on the load side, “behind the meter,” where customers own their own electricity and can make their own decisions. Stabilizing those loads will also redound to the benefit of the grid, which may eventually convince utilities to adopt the technology at a broader level. (Assuming utilities want a technology that eliminates their justification for building new infrastructure.)
With no real budget to advertise, 3DFS is going “door to door,” Doerfler says. It was recently chosen as one of eight companies in Duke Energy’s Joules Accelerator program, which will give it “immediate and focused access to the relevant decision-makers at Duke Energy and the City of Charlotte to expand their awareness and experience with Software-Defined Electricity.” It also won the audience award at the recent NC Tech Startup Showcase.
“We are sustainable enough to be able to walk this through,” Doerfler says. “We just need to get the knowledge out there and the people will come. “
Relying on organic growth is an unconventional strategy these days, and there’s no guarantee of success. It’s not difficult to imagine 3DFS failing, as any company might.
But now that we know it’s possible, it’s difficult to imagine SDE — measuring and managing electricity in real time — failing. It’s an idea whose time has come. As electricity becomes ever-more-central to our lives, and waste becomes ever-more-unconscionable in light of our environmental situation, we will always need more understanding and control over power.
“The use cases are endless,” Heuberger marveled to me. “They’re freaking endless. It’s amazing.”