How the Lights Stay On · The Sun Doesn’t Ask Permission

The Sun Doesn’t Ask Permission, part 2

This is what the duck curve looks like. A flat morning back. A sagging midday belly. A sharp evening neck. A pointed sunset head.

The duck is not a small problem. Between roughly 4 PM and 7 PM on a spring evening in California, the grid has to ramp up about 13,000 megawatts of generation — the equivalent of about thirteen large nuclear power plants worth of output — in three hours. Every day. As the sun sets.

You cannot ramp a nuclear plant up that fast. You cannot ramp a coal plant up that fast. Even gas plants, which are the fastest-ramping of the conventional sources, strain at this. The grid has to keep enough natural gas plants warm and ready, all day long, to come online when the duck’s neck rises.

The duck, in other words, is the shape of what intermittency does to a grid that still has to keep the lights on.

There is another problem hidden in the duck. The midday belly — the part where solar is producing more than the state needs — is the source of the strangeness we started this chapter with.

When solar is over-producing, the electricity has to go somewhere. Electricity cannot pile up. You cannot stockpile it in a warehouse. The grid balance from Chapter 2 requires that every kilowatt being produced is being used in the same instant.

So California has three choices when its solar floods the grid.

Send it somewhere else. Sell it to Arizona, Nevada, even down to Mexico if the transmission lines exist. Sometimes the neighbors do not need it either. Then the price goes negative — California pays Arizona to take it. This is genuinely happening on real grids in the United States right now.

Use it for something flexible. Charge batteries. Pump water uphill into a reservoir. Run industrial processes that can shift their schedule, like aluminum smelting or hydrogen production. The trouble is that there are not yet enough of these flexible uses to absorb the excess. There will be more in the future. Right now, not enough.

Throw it away. Tell solar farms to curtail. Tell wind farms to feather their blades and stop spinning. The free electricity is not generated.

Curtailment is the option of last resort. The energy is wasted. The owners of the solar farm lose revenue. Nothing useful happens with the photons that hit those panels. But the grid stays balanced, because that is the iron law.

It is not just California. Texas curtails wind regularly. Hawaii curtails solar. Germany has paid surrounding countries to take excess wind power during stormy nights when the wind is howling and demand is low.

This is a strange thing to read in a book that is, in part, about cleaning up our grid. We have an abundance of free, clean electricity at certain hours — and we are throwing some of it away. The reason is not that anyone wants to. The reason is that we have not yet built enough ways to catch it.

Wind is intermittent in different ways than solar is, and the ways are sometimes worse.

The most awkward fact about wind, especially in places like the central United States, is that it often blows hardest at night and dies down during the day. The wind farms of Texas, Oklahoma, and Iowa generate most of their electricity between midnight and dawn — exactly when nobody wants it. Then the wind slackens during the day, when demand is high. Then it picks up again at night, when demand falls.

This is the opposite of helpful.

Wind has another, more dangerous habit. During cold weather, high-pressure systems sometimes settle over large regions and stay there for days at a time. The air is still, the sky is clear, the temperature is freezing. There is no wind. Across hundreds of miles, wind farms stop turning, all at once, for an entire week.

This is the worst possible time for wind to fail, because cold weather is when people most need electricity for heat. A grid that has come to rely on wind, but does not have a backup plan for the windless weeks, can find itself in a very bad place.

Texas found itself in a very bad place in February 2021. That story is coming.

When you read in the newspaper that a country has installed “300 gigawatts of solar,” you might assume that 300 gigawatts is, well, 300 gigawatts. It is not.

A solar farm rated at 100 megawatts produces 100 megawatts only at the brightest moment of the brightest day of the year, with the panels at perfect temperature and perfectly clean. Over the course of a whole year, the average output of that solar farm is closer to 25 megawatts. The ratio of average to maximum is called the capacity factor. For solar in most of the United States, it is around 25 percent.

For wind, the capacity factor is around 35 to 40 percent on good land, and as high as 50 percent at sea.

For a coal plant, the capacity factor used to be 70 percent or more, though it has fallen in recent years as cheaper sources push coal off the grid.

For a nuclear plant, the capacity factor is over 90 percent. Nuclear plants run almost all the time. They take a few weeks off every year and a half for refueling, and that is about it.

This means when you compare a “1,000-megawatt coal plant” with “1,000 megawatts of solar panels,” you are not comparing equivalent things. The coal plant will produce something like three to four times as much electricity over a year as the solar will, even if the nameplate ratings are equal. To replace a 1,000-megawatt coal plant with solar, you need to install three to four times as much solar — and you need to figure out what to do for the nighttime hours.

This does not make solar bad. Solar is now the cheapest source of new electricity in most of the world. But the comparison “we installed enough solar to replace a coal plant” is almost always misleading. The math is not that simple.

Engineers and policymakers have been working on the intermittency problem for thirty years, and they have a partial toolkit.

Geographic diversity. The wind may not be blowing in Iowa, but it might be blowing in Wyoming, or off the coast of Massachusetts. The sun may not be shining in Seattle, but it is shining in Arizona. If you can move electricity across long distances, you can smooth out a lot of the variability. This is one reason new transmission lines — the long road from Chapter 3 — are so important to the energy transition. It is also one reason permitting paralysis is a real cost.

Demand response. Instead of always shifting generation to meet demand, the grid is starting to shift demand to meet generation. Pay electric car owners to charge at noon, when solar is abundant, instead of at 6 PM. Set water heaters and washing machines to run when the grid is flush. Pay big industrial users to ramp down for an hour when demand spikes. This is the smart grid we will spend a whole chapter on later in this book.

Overbuild and curtail. Build so much solar and wind that even on a bad day, there is still enough. Accept that on good days, you will be throwing energy away. The economics of this strategy are actually getting reasonable, because solar and wind are now so cheap that overbuilding is not as expensive as it sounds.

Storage. Catch the excess midday solar in batteries. Pump water uphill in reservoirs. Compress air into caverns. Heat blocks of sand or molten salt. Then release the stored energy in the evening when the duck’s neck rises.

Storage is the big one. It is the reason the next chapter exists.

Solar and wind are now the cheapest sources of new electricity in many parts of the world. The price of solar panels has fallen by more than 90 percent in fifteen years. The price of wind turbines is a fraction of what it was. In sunny places, the wholesale cost of solar electricity is now lower than the wholesale cost from any other source — fossil or otherwise.

But “cheapest per kilowatt-hour at noon” is not the same as “cheapest grid.” A real grid needs electricity at every hour of every day, in every kind of weather. The cost of the system — solar plus storage plus backup plus transmission plus the grid services that keep frequency stable — is higher than the cost of solar panels alone. How much higher depends on local geography, on existing infrastructure, on the mix of other sources available, on how much intermittency the grid is willing to absorb.

The honest summary, the one most cheerful headlines about clean energy will not give you, is this: the cheapest electricity in the world is now intermittent, and learning to live with intermittent electricity is the central engineering project of the next two decades.

This is a real, unsolved problem. Engineers are working on it every day. It is being solved in pieces, in different places, in different ways. It is not yet solved.

The duck curve, when you understand what it is showing, is not a sign of failure. It is a sign of success. The duck exists because California has so much solar that it has run out of demand to absorb it at midday. That is what winning at solar looks like.

California’s solar and wind capacity grew from 9.7 gigawatts in 2014 to 28.2 gigawatts by the end of 2024 — almost tripling in a decade. The duck is the visible mark of that buildout. The duck is also the bill that comes with it.

The challenge now is what to do with the duck.

What to do with the midday belly, when there is more clean electricity than the grid can use. What to do with the evening neck, when the sun sets and the city needs power. What to do with all those hours in between, when supply and demand stop dancing with each other and start fighting.

The answer, as much as there is one, is storage. Catching the noon sun and releasing it at six. Catching the midnight wind and releasing it at three in the afternoon. Making the grid able to remember.

That is the next chapter.

All 18 chapters of How the Lights Stay On · Library · All courses