Intermittency and Integration: Duck Curves, Curtailment and Firming

A grid has to match supply and demand every second. For a century, operators did that by turning fuel-burning plants and hydro up and down to follow demand. Wind and solar change the job. Their output follows the weather, so operators now plan around what is left over after the wind and sun do their part. That leftover has a name: net load.

The duck curve

Draw demand across a spring day in a sunny region and it is gentle: lower overnight, rising in the morning, peaking in the early evening. Now subtract solar. Midday net load sinks. Then, as the sun sets just when people come home and cook, net load climbs fast. California's grid operator drew this shape years ago and someone noticed it looked like a duck.

Two problems show up in the duck. In the belly, there may be more solar than the grid can use. On the neck, other resources must ramp up very quickly as solar fades. In California, EIA says solar now supplies almost half of the grid operator's demand between 8 a.m. and 4 p.m. The state's grid had 28.2 GW of wind and solar at the end of 2024, up from 9.7 GW in 2014.

Curtailment

When supply exceeds what the grid can use, store or export, operators tell plants to cut output. That is curtailment. EIA reported the California Independent System Operator (CAISO) curtailed 3.4 million MWh of utility-scale wind and solar in 2024, up 29 percent from 2023. Solar was 93 percent of it, and curtailment peaked in spring, when sun is strong, temperatures mild and air conditioning loads low. Some of it was not oversupply at all but congestion: power that could not get across crowded lines.

Potential solar output: 20 x 8,760 x 0.25 = 43,800 GWh = 43.8 million MWh
Curtailed:              3.0 million MWh
Curtailment rate:       3.0 / 43.8 = 6.8 percent

The toolkit

Capacity accreditation: how grids count reliability

Grid operators measure how much each resource helps in the riskiest hours. The common method is effective load-carrying capability (ELCC). Lazard's 2026 report used grid operators' own published ELCC figures. Solar's ELCC was about 39 percent in MISO, where solar is a small share, and about 14 percent in CAISO, where it is a large share. Once a grid has lots of solar, the risky hours move into the evening, and more solar barely helps. Pairing it with storage raised the CAISO figure to about 41 percent in Lazard's table. Wind ranged from 17 percent in NYISO to 38 percent in PJM across the six markets Lazard studied. Gas plants are also increasingly derated for winter fuel and outage risk, a point Lazard notes it did not model.

What the studies say about cost at high shares

This is where serious people disagree, and each camp has real evidence.

Optimistic: renewables plus storage plus transmission

NREL's 100% Clean Electricity by 2035 study found least-cost systems with wind and solar at 60 to 80 percent of generation, about 2 terawatts of combined capacity, 120 to 350 GW of daily storage, and 100 to 680 GW of seasonal storage or other clean firm capacity. It estimated added power system costs of $330 billion to $740 billion, outweighed in its accounting by avoided health and climate damages, for net societal benefits of $920 billion to $1.2 trillion. Note what that does and does not say: the health and climate benefits are what make the net positive, and the study requires deployment at an "unprecedented scale and rate."

Lazard: still competitive after firming

Lazard's 2026 analysis added a "levelized firming cost" to renewables in six markets (MISO, CAISO, SPP, PJM, ERCOT and NYISO), based on each operator's capacity rules. It concluded that renewables remained broadly cost-competitive with new gas combined cycle plants even after firming. In its table, unsubsidized solar plus firming came to about $115/MWh in MISO, with a $30 firming adder on top of the solar cost. Lazard is also explicit that it did not apply any firming charge to gas.

Firm-resource camp

Sepulveda and colleagues at MIT (Joule, 2018) modeled nearly 1,000 zero-carbon scenarios and found that including firm low-carbon resources (nuclear, geothermal, gas with carbon capture, reservoir hydro) cut costs 10 to 62 percent compared with wind, solar and batteries alone. Costs in wind-solar-battery-only systems climbed steeply in the last stretch toward zero emissions, because batteries must cover rare multi-day lulls.

Skeptical of 100 percent wind, water and solar

Clack and 20 co-authors (PNAS, 2017) critiqued a high-profile 2015 study claiming a cheap all-renewable U.S. grid. They argued it contained modeling errors (including hydro output above physical capacity), relied on unproven storage at scale, assumed a low cost of capital that they argued understated costs, and required deployment far beyond any historical pace. The original authors, led by Mark Jacobson of Stanford, published a reply in the same issue of PNAS rejecting the critique and defending their model, so this remains a live dispute between named researchers.

Deployment speed is the hidden constraint

LBNL's Queued Up 2026 report counted about 2,060 GW of projects waiting to connect at the end of 2025, more than the entire existing U.S. fleet. Only about 13 percent of capacity that applied between 2000 and 2020 was built, and the median project finished in 2025 took more than five years from request to operation. Integration studies assume the lines and plants get built. The queue says that is the hard part.

California context, since Idaho writers lean on it. SB 100 requires renewable and zero-carbon resources to supply 100 percent of California's retail electricity sales by 2045. In 2024, 63.3 percent of the crude oil California refined came from foreign sources, 13.4 percent from Alaska and 23.3 percent from in-state, per the California Energy Commission. Joel Kotkin's May 7, 2025 Los Angeles Times column argues those policies made energy "steadily unaffordable"; that is his argument, and Pierson's California's Green Mirage summarizes it.