This piece, originally published in August, has been lightly updated.
For most of the 100-plus years that the electricity grid has been around, grid managers had control over the supply of power but not the demand for it.
Like the weather, or the tides, electricity consumption could be reasonably well predicted, but it couldn’t be controlled. It was something that just happened, to which grid operators responded by adjusting supply, turning power plants on or off, up or down.
The problem with having no control over demand, though, is that you end up overbuilding supply. You have to build enough power plants to satisfy the highest possible peak in demand (and a reserve on top of that, for extra reliability).
The thing is, peaks in demand are by definition exceptions. Most of the time, lots and lots of power plants, especially the “peaker” plants built purely to satisfy these brief windows of high consumption, just sit around, idle. Overbuilding has been the rule in the power system, at great cost in both money and efficiency.
In recent years, that dynamic has been changing, on both ends.
In supply, the story is familiar: As renewables grow, power is becoming less controllable (less “dispatchable” in the lingo). Wind and solar energy can’t be turned on and off at will — they come and go with the weather and time of day, so they must be accommodated, as demand once was.
But the less-familiar story is that demand is also changing, becoming more dispatchable.
Lots of power consumption is not time-sensitive. The iconic case here is your home water heater. Water retains thermal energy pretty well. You can heat the water in the tank at any time and still have hot water available when you want. So in theory, water-heating energy consumption can be time-shifted, from times of high demand to times of low demand, without any loss of service.
Many power-intensive industrial processes are largely automated, so they could, in theory, be run at any time. Electric vehicles could be charged at different times of the day. Lots of energy use can be moved in time.
The problem has always been coordination; before the internet, it was laborious, slow, and “chunky” to plan and execute shifts in demand. But as more and more appliances, homes, buildings, and industrial facilities are wired to the web, it has become easier to synchronize and coordinate their consumption, to treat them like an aggregate unit. Various forms of “load flexibility” are emerging as a significant force in energy systems.
That shift to more dispatchable demand has important consequences. To the extent consumption can be controlled, the big peaks and spikes of demand can be reduced. That, in turn, reduces the need for overbuilding of power plants, potentially saving billions and reducing unnecessary carbon emissions.
Just how big a role could load flexibility play? Conveniently, the Brattle Group, a research consultancy, recently issued a study on just that question: “The National Potential for Load Flexibility.” Brattle researchers built a model for electricity markets (LoadFLEX), tackled a lot of technical problems (like quantifying the future impacts of still-nascent programs) and estimated the total market potential of current and expanded load flexibility through 2030.
What they found is the load flexibility could readily be expanded, bringing in new technologies and producing new benefits, to satisfy almost 20 percent of US peak demand — potentially eliminating the need to generate one out of every five electrons when demand spikes.
Let’s look at a few of the top-line conclusions.
The current market for load flexibility is growing but slowing
Already, there is a healthy US market for load flexibility. It mostly operates through something called “demand response” (DR).
Dozens or hundreds of loads (power-consuming appliances or buildings) can be linked together into what is, effectively, a single aggregate unit. When a peak in demand arrives, instead of dispatching a power plant, grid operators can dispatch an aggregated unit of demand reduction, “shaving” the peak and reducing the amount of expensive peak power that must be produced.
DR is already the biggest category of distributed energy resources (DERs) in the US.
(By way of comparison, the US currently has 67 gigawatts of installed solar photovoltaic (PV) capacity and close to 265 GW of installed natural gas capacity.)
Though it is developing quickly in other regions, today DR is mostly concentrated in the Upper Midwest.
Meanwhile, the US is seeing rapid adoption of the kinds of technologies that can enable more DR, including behind-the-meter (BTM) energy storage, smart meters that can track and report energy use, electric vehicles, and smart (internet-connected) appliances.
However, market growth for current versions of DR — what Brattle calls “DR 1.0” — is slowing. As you can see below, DR’s average annual growth rate, which peaked in 2011 at 15 percent, is now down to 2 percent.
Brattle cites a number of reasons for this slowdown: rules for market participation are getting more stringent; many utilities have overbuilt peaking capacity, so prices in capacity markets are cheap; and energy prices are swinging around less, leaving fewer opportunities for arbitrage.
Looking out to 2030, Brattle estimates that expanding and improving existing DR 1.0 programs could get another 16 GW of capacity online. Developing new load flexibility programs, taking advantage of new value streams, could add another 40 GW, bringing the total to 115 GW.
That’s nothing to sneeze at, but the intriguing bit is the 83 GW of additional capacity that Brattle says could come from “market transformation.” That would bring the 2030 total up to 198 GW, representing 20 percent of the total system peak.
So what’s this “market transformation” anyway?
Load flexibility 2.0: more options, more benefits
Today’s load flexibility programs are mostly focused on reducing peak demand, during limited windows, for a limited number of hours per year, which is a fairly limited market.
DR 1.0 offers three basic benefits that can be monetized:
- generation capacity avoidance — power plants that aren’t built to provide peak power
- reduced peak-energy costs — expensive peaker power that isn’t purchased
- avoided peak-related transmission and distribution (T&D) upgrades — shaving peaks can help utilities delay or avoid upgrades to the power lines that must handle the peaks.
(You’ll notice that all these benefits are things load flexibility allows utilities not to do; it’s all savings.)
Brattle researchers propose expanding the model of load flexibility in two directions. First, they propose new benefits that could be monetized, and second, they propose new technologies and techniques that could provide those benefits (and ought to qualify as load flexibility). An expanded definition, with more options, producing more benefits.
The three new benefits are:
- targeted T&D capacity deferral — instead of just avoiding peak-related T&D upgrades, load flexibility could help avoid any T&D upgrade; it could be targeted to any congested area of the grid, at any time throughout the day
- load shifting and building — as variable renewable energy expands on the grid, there will be times not just of peak demand but of peak wind and solar supply, when there is excess power on the grid that needs to be absorbed; load flexibility means demand can be shifted around to accommodate those fluctuations
- ancillary services — to run smoothly, electricity grids need certain special services like frequency regulation and voltage control (don’t ask); load flexibility, especially the fast, grid-connected, automatic kinds, can provide some of those services.
To capture these new benefits, Brattle proposes a whole host of new types of load flexibility, from managed electric-vehicle charging to ice-based thermal storage. I won’t go through the whole list; you can check the report or squint at this summary chart.
This is a much broader conception of load flexibility than what is covered in most studies, and it’s helpful in flexing the imagination. It’s not difficult to see how, as the internet enables more energy uses to be coordinated, this list could expand.
Just how much energy use really is time sensitive? With enough experimentation and the right incentives, we’ll find out.
Putting expanded load flexibility to work
Brattle says that load flexibility can help address three of the “megatrends” in the energy world today.
The first is the growth in renewable energy, which creates intermittency in supply as well as occasional supply spikes that lead to curtailment (shutting wind and solar off temporarily). Demand response can move quickly (providing ancillary services to smooth out micro-fluctuations), and it can also help soak up excess renewable energy in times of surplus. Both of these will help the grid absorb more renewables.
The second is grid modernization, which refers to the simple fact that most of the US grid is old AF and badly needs costly upgrades. Geographically targeted demand response — basically, lightening the load in areas where the grid is stressed — can help delay or avoid some of those upgrades.
The third is electrification, not only the rise in electric vehicles but the eventual electrification of heating and cooling systems and industrial processes. Shifting everything over to run on clean electricity is going to substantially increase demand (and the need for load flexibility), but it’s also going to provide aggregators and grid managers with scads of new dispatchable loads — water heaters, EVs, clothes washers, heat pumps, what have you.
Like energy efficiency, load flexibility is a win-win-win. It saves money on utilities, it reduces consumer energy bills, and it helps clean up the grid.
All told, Brattle estimates that the net national savings enabled by load flexibility could exceed $15 billion a year by 2030.
That annual value in 2030 will be comprised of 57 percent avoided generation capacity, 29 percent avoided energy costs, 12 percent avoided T&D capacity, and 2 percent (what Brattle calls the “cherry on the sundae”) ancillary services.
That’s all based on national averages — the benefits will vary substantially at the regional level. As an example, Brattle compares the benefits in Minnesota and California. The former (with a bunch of coal plants soon retiring) mostly benefits from avoided peaker plants; the latter (with more renewable energy on its grid) benefits more from ancillary services.
Capturing the full benefits of load flexibility — potentially 20 percent peak reduction — will require action from regulators, policymakers, and utility providers. It will require expanding programs, creating new programs, and running demand response much more frequently.
Wholesale markets need to be made friendlier to load flexibility. Regulatory standards and incentives need to be updated, along with utility resource planning.
In the short-term, the simplest way forward is for utilities to offer customers fixed payments for participation in load flexibility programs. As grid infrastructure gets smarter, more sophisticated methods will develop, like real-time locational pricing, which uses a price signal to incentivize load flexibility rather than prescribing particular technologies.
Brattle predicts that load flexibility programs “will get smart before they get bigger,” meaning that a lot of short-term work will go into simply modernizing and revitalizing existing programs, many of which have been neglected. That, along with new regulatory incentives, will set the stage for expansion later in the decade.
There remains much work and experimentation to be done to see exactly what’s possible. When I asked lead author Ryan Hledik what sorts of developments might confound Brattle’s projections, he cited unexpectedly rapid growth of renewables in regions with limited transmission interconnections — that would substantially skew the market to favor energy and ancillary service benefits.
Another possibility is the rapid growth of EVs in particular areas, which could stress local grids. “If EVs are adopted in geographic clusters at levels significantly higher than we’re seeing today,” Hledik said, “then I would expect the value of geo-targeted T&D deferral to significantly exceed what we’ve quantified.”
There’s also enormous learning to be done about the full potential of smart residential technology. Another Brattle prediction: while industrial demand-response has always exceeded residential, the rise in smart appliances, smart home energy management, smart meters, and EVs (technologies people are purchasing for reasons other than the load-flexibility benefits) means that the residential market will grow faster in coming years.
What this means is that someday relatively soon, your energy use — in your home, your vehicle, and your workplace — will be part of a carefully choreographed symphony, most of which you won’t notice. Your EV will charge at 5 am one night, 1 am the next; at work, it will charge at 10 am one day, 2 pm the next. Your water heater will fire up at 2 pm one day, 2 am the next. The ice used to provide you air conditioning will be frozen at 10 pm or at 7 am.
All around you, energy consumption will be subtly and continuously shifted and adjusted to smooth out variations in energy supply and demand, hold prices low, and allow the continued decarbonization of the grid. It’s pretty cool.