Reprise: Duck and Cover

Grid Stability and the "Duck Curve"

K.T.’S NOTE: I’m on vacation this week, but I thought I would re-publish this piece on the duck curve that I wrote earlier this year before I had many subscribers. And as California is in the grip of a heat wave this week, it’s particularly relevant. Enjoy!

Children hide under their desks during a fallout drill, 1950s (Source: Diefenbunker.ca)

Falling Out of Line

The Cold War-era “Duck and Cover” drills taught a generation of American Boomers how to stop worrying and hide under their desks in the event of a Soviet nuclear attack. In retrospect we may laugh at such a foolish attempt to avoid the apocalypse. But human nature being what it is, our public service announcements haven’t fallen so far from the proverbial tree.

These days, instead of warnings of impending nuclear winter, California's hot summer days are punctuated by utility-issued “flex alerts,” signaling extreme strain on the grid, and urgently pleading consumers to curtail home and business electricity usage to avoid power failure. But, unlike the atomic bomb annihilation eventualities, which mercifully never materialized, threats of rolling blackouts and their fallout are very real, with serious consequences.

The main culprit of this grid instability is what is affectionately referred to as the “duck curve,” or the peaks and valleys that characterize the daily electricity demand curve, broadly resembling the outline of a duck:

Source: Energy Information Administration/CAISO

As shown in the graph above, net load, or “demand remaining after subtracting variable renewable generation,” bottoms out during the middle of the day (the duck’s belly), then sharply increases during the evening hours (the duck’s neck and head), before flattening out during the overnight hours. Peak hours (4-10pm) roughly correspond with the demand surge that occurs when people get home from work and turn on lights, HVAC, and appliances. The wild swings from valley to peak cause significant stress on the grid, requiring fast ramp-up of reserve and peaker plant production.

And, as clearly shown here, this problem is only getting worse. As Arshad Mansoor, President and CEO of the Electric Power Research Institute (EPRI) recently put it, “the duck now looks like a canyon.” So why is this happening? Put simply: increased reliance on solar power. As of the end of 2023, California had 43,244 MW of installed solar, representing 27.76% of the state’s electricity generation. And, unlike dispatchable resources such as natural gas, coal, and nuclear, that provide constant baseload power, renewables such as solar and wind are inherently intermittent resources. They must be consumed or curtailed immediately upon generation (the ultimate use it or lose it).

At the peak of solar intensity (12-2pm), solar generation brings a glut of electricity onto the grid, providing clean energy, but also posing challenges to grid management. Excessive generation in turn leads to curtailment, where grid operators intentionally shut down solar plants to prevent over-production. Curtailing electricity is a common, and highly inefficient, effect of solar generation, wasting valuable energy and incurring economic losses. For example, in 2020, the California Independent System Operator (CAISO), curtailed 1.5 million MWh of utility solar (5% of total annual production).

If curtailment is one side of the solar balancing act, the mad dash to ramp up dispatchable supplies in the evening is the other. Because solar generation is plentiful during the day, it displaces traditional baseload sources, leading to significantly reduced output of gas-fired electricity. But as the sun sets, these gas plants must quickly increase production to climb the increasingly vertiginous duck curve and balance the spike in demand. Because the grid is a delicately calibrated instrument, and supply must always meet demand for its proper function, this is a tough ask indeed.

Call and Response

As mandated by California’s net zero by 2045 commitments, installed solar will continue to grow its share in electricity production, further sharpening the mismatch between supply and demand. So what’s the solution? Over the past decade, utilities have developed increasingly sophisticated “demand response” (DR) programs to help balance the grid. Their goals are twofold: 1) to influence customers to use less energy during peak hours and 2) use advanced automated control systems to shed load during times of strain.

The first kind of DR program typically involves behavioral interventions to encourage customers to support grid stability. For example, most California utilities now employ time-of-use (TOU) pricing, charging higher rates for electricity consumed during the critical 4-10pm peak period, and presenting an economic incentive to avoid energy intensive behaviors during those times (e.g. operating washing machines, dryers, and dishwashers). As discussed earlier, “flex alerts” may be issued in extreme circumstances via text messages to implore customers to immediately cut non-essential electricity usage to prevent imminent blackouts.

The other type of program, also called automated demand response (ADR), leverages machine learning and “smart” technologies to learn behavioral patterns of equipment and buildings, understand their load profiles, and automatically shed load when needed. While customer consent is generally required to opt in to these programs, the process itself is automatic. For example, customers with smart thermostats may enable utilities to control set temperatures in exchange for financial incentives. However, most of the grid balancing opportunities for ADR programs are currently in the commercial and industrial sectors, where buildings are often unoccupied in the evenings, and lighting and HVAC loads can be shed to help meet the demand in residential homes.

Not Enough

While DR programs have enjoyed a modicum of success, particularly at the level of the individual home or business, it’s become clear that TOU rates and curbing commercial building HVAC loads are not enough. Indeed, despite the deepening of utility DR portfolios over the past years, the duck curve also continues to deepen, and the barrage of new solar and wind projects coming online continues apace. Compounding the problem, the drive to fully electrify homes and commercial buildings will increase overall consumer electricity demand.

This conundrum has left policymakers scratching their heads, looking for new blood to squeeze from the DR stone. California’s new paradigm expands the repertoire of tools to include “Shape, shift, shed, shimmy” (which frankly sounds more like Grandpa’s dance moves after his fourth Old Fashioned, but, I digress..) Of these new interventions, the potential to shift load is identified as most promising, according to a recent DR research report and scoping study issued by Lawrence Berkeley National Lab (LBNL). Shifting load refers to the ability to control and defer equipment use to times when the grid is less congested. Customer behavior nudging programs, such as TOU, already encourage shifting by making it more expensive to operate appliances at peak hours. What’s new is that growth in AI and machine learning have made it possible to dynamically control internet-enabled appliances and lighting through automated smart grid signals. For example, electric heat pump water heaters may receive signals to program the heating element to run during off-peak hours. Research currently underway at LBNL’s CalFlexHub includes pilot studies on products such as residential water heaters, pool pumps, and home energy management systems (enabled through smart speakers/hubs), evaluating their potentials for automatic load shift.

Big Brother, Where Art Thou?

It remains to be seen whether the new smart-grid enabled DR programs will attenuate California’s duck canyon. But, to quote Thomas Sowell, “There are no solutions. There are only trade offs.” Critiques of DR have long included invasion of privacy as a potential risk. As these programs necessarily collect data about household energy usage patterns via advanced metering infrastructure and smart connected devices, it is indeed possible for utilities to gain insight about revealing personal information, as previous studies have discovered. What’s more, this data is vulnerable to hackers and unauthorized sales of information to third parties. As more connected devices come online and feed data about our behaviors and preferences into the hivemind, fears of privacy breach and improper surveillance only accelerate and proliferate.

Moreover, with further pushes toward fully automated demand response, individual choice and sovereignty come into question. In general, TOU rates and current opt-in programs do not coerce particular behaviors, and still preserve overall consumer autonomy. However, it would be a wholly different scenario if you weren’t able to use your washing machine or stove at certain times by order of grid diktat. To be clear, no one has seriously proposed such measures (yet), but the nature of control is to exert more control. Policymakers must be made aware of potential mission creep when they evaluate and approve new programs that grant utilities remote control access to their customers’ personal devices and appliances.

Finally, coming full circle, there is an element of folly in depending on increasingly unrealistic Hail Mary passes to stave off grid collapse and the inevitable anarchic chaos that would ensue. Demand response has not cured the duck curve, and never will as long as solar intermittence is poorly aligned with the natural demand cycle (I wouldn’t hold my breath). Much media coverage of this issue points to battery storage as the Deus ex Machina that will solve our problems and stabilize the grid. But this is a future that doesn’t exist yet, and storage has proved to be a tough nut to crack.

A more realistic approach for the immediate future (in addition to current DR measures) would be to incentivize gas power plants to stay on throughout the day, producing minimal generation, but providing the capacity to ramp up quickly at peak time. As Mansoor, the EPRI CEO stated, “some gas generation must remain online” during the middle of the day, when net load is at rock bottom. In short, we cannot depend on flex alerts and TOU rates to be our salvation in times of need — the integrity of the electric grid is too important to leave to the whims of Fate. The last major flex alert sent out across California during the height of a week-long heatwave in September 2022 did seem to avert disaster. But next time we might not be so lucky. Quack.

Electrically yours,

K.T.