Skip to content

Have We Peaked?

The heat is on, and climate change will put significant strain on the US power grid – unless we do something about it.

In the absence of any concrete new policy proposals coming out of the not-so-green White House, we figured we would write about something we know even more about – research! Patrick Baylis, Catie Hausman and I have a new paper out today in the Proceedings of the National Academies of Sciences (PNAS), which is of utmost relevance to this blog. It combines Max’s three favorite topics: climate change, electricity and lots of data.


Humans do not like heat. There is ample evidence that when it’s hot outside, more people die, we are unhappier, air quality decreases, we become more aggressive and violent, less productive and our cognitive ability decreases. In order to offset these effects, we seek cool environments. Lakes, shade and, over the past century, the air conditioned indoor space. Movie theatres these days are so cold, you can just bring a bucket of sugary cream and you’ll have ice cream by the time the trailers are done. Supermarket, shopping malls, server farms, packing plants, and most homes are air conditioned.

In most of the U.S., electricity usage spikes in the summer when it’s really hot outside. Alan Barreca and coauthors have shown that the rollout of air conditioning has led to massive decreases in heat related mortality over the past century.  While we use some electricity to run our heating systems in the winter, in most of the country, air conditioning uses more electricity. And so it’s no surprise that rising temperatures from climate change are expected to lead to increased electricity demand.

The question that arises is just how much. Lucas has a nice paper on residences in Mexico and finds that we expect massive increases there. In the US, the literature has been relatively sparse. There are few papers on the commercial and industrial sectors.

And importantly, producing electricity at peak times is much more expensive than at non-peak times. If we need to expand generating capacity enough to allow for one more large window unit air conditioner to run at the system peak demand, the upfront cost of that additional power plant (or share thereof) in California is roughly $900. That number nearly knocked my air conditioned socks off.

In the PNAS paper released today we impose end of century climate on today’s economy and grid. We highlight that climate change would cause electricity demand to increase disproportionately at times when the grid is already stressed. A lot of the previous research has focused on the impact of climate change on the *typical* day in a year, finding moderate increases in demand over the next 100 years. We show that it’s also important to factor in the impact on the highest-usage days — when we show that electricity demand will increase even more.

Under business-as-usual (= somewhat frightening) RCP8.5 scenario from the climate change literature, we estimate that *average* demand will increase 7.9%. But the increase is substantially larger — 17.6% — on the highest demand (meaning sweltering hot) days of the year (specifically, we look at the 95th percentile of usage). That makes sense, since a bigger percentage of demand is coming from air conditioning on those high-usage days.

The reason this matters is that cost-effective electricity storage is not yet widely available. So we build the grid to meet the highest usage “peak” hour of the year. Over the next 100 years, that grid capacity will need to increase substantially to accommodate climate-change driven demand growth.

What this means is that adaptation to climate change will be costlier than existing models estimating these costs globally estimate. There are a number of adaptation routes — building hundreds more peaker power plants; developing more storage; developing more efficient cooling technologies, or getting customers to change their behavior. But the existing cost calculations for adaptation don’t factor in the need to do these things.

Another implication is that actions to *prevent* climate change are more valuable than what existing models say. If you can lessen climate change, you can avoid some of these adaptation costs. So let’s get to solving this climate change problem, which is neither perpetuated by the Chinese (their plans to solve climate change are more ambitious than what I fear we will do) nor a hoax. It’s real. Let’s get to it.

This blog post was coauthored with Catherine Hausman (Michigan) and Patrick Baylis (Stanford/UBC Vancouver).

Maximilian Auffhammer View All

Maximilian Auffhammer is the George Pardee Professor of International Sustainable Development at the University of California Berkeley. His fields of expertise are environmental and energy economics, with a specific focus on the impacts and regulation of climate change and air pollution.

11 thoughts on “Have We Peaked? Leave a comment

  1. This piece has good many good ideas and intentions.

    However, the models used to describe the temperature projections to 2100 are no longer valid. They do not fit [represent] correctly the warming that has been measured up to the present. They over-warm the earth and atmosphere by factors close to 2.4! Their projections of future temperature increases are similarly much larger than those predicted by the measurements – good news for our electrical grid!

    There is much speculation regarding the model failures. One possibility is that the positive feedback due to increased water vapor from a warmer ocean surface is less than assumed – cloud formation limiting warming being one suggestion.

    Also, as we have been increasing emissions the percentage taken up by CO2 sinks has also been increasing and remains close to 60% of emissions, leaving only about 42% remaining in the atmosphere.

    The RCP models used are based on CMIP-5 models which do not fit measurements of the Earth’s surface temperature increase or that of the atmosphere. For the latter case please see:

    Click to access HHRG-114-SY-WState-JChristy-20160202.pdf

    For surface temperature history please see [Ref. 2]: the IPCC Climate Change 2014 Report, Summary for Policymakers. Page 3, top panel (a), gives “globally averaged” land and ocean temperature measurements. Over the last 50-60 years when anthropogenic CO2 emissions have been rising most rapidly [ bottom panel (d)], the global average temperature rise has been about 0.115 degrees Celsius per decade. Satellite data analyses, only available since 1979, find 0.114 deg. C per decade for 1979-2014.

    RCP4.5 and RCP8.5 models predict 2100 temperatures about 2.0 and 4.0 C above the average 1986-2006 values respectively [page 11, Figure SPM.6 (a) of ref. 2]. “Business as usual” extrapolation of the measured warming rates would give a temperature increase at 2100 of about 1.2 C – considerably lower than the RCP8.5 prediction of 4.0 C. Relative to the latter the 1.2 C implies much-reduced peak-load requirements.

    Of course, “business-as-usual” emission trends are being reduced. Carbon emissions are projected to peak by 2040-2050 and then decline: The US and EU have been reducing emissions 1-2% a year. China has coal-use peaking near 2022 and emissions peaking in 2030. They have and are adding massive arrays of solar panels, are finishing 20 nuclear plants to add to their 36 operating, and 58 more are planned. Natural gas pipelines to Russia and LNG terminals are being built so coal-burning can in-part be replaced.

    The rest of Asia is a decade behind. But citizen and international pressure will, as in China, eventually prevail.

    Therefore, it seems very unlikely that the 2100 global temperature will increase more than about 1.0 C above 1986-2006 values – good news for our electrical grid!

    Best wishes,

    Paul Brady [UCD]

  2. Max,
    Your results “point to the possibility of climate changing [sic] increasing the demand for storage technology, demand response programs, and alternative pricing schemes.” In other words, adaptation will lower the costs from the $180 billion in the BAU scenario. I should think that improvements in energy efficiency will also continue to moderate demand even in the BAU scenario, i.e. w/o additional demand response programs.
    A complementary exercise would be to estimate the savings in non-electric heating costs as well as the net-emissions effects.

  3. The role of time-variant electric rate design in ameliorating this problem is vastly under-appreciated. We need to introduce strong cost-based time-of-use rates, and reduce or eliminate non-time-based rate components such as demand charges, which act a barriers to customer deployment of energy storage and other peak-reducing technologies. Utilities should be introducing very low “super-off-peak” rates to encourage customers to use renewable resources (solar and wind) at times when they are abundant.

  4. My understanding that the warming effect was the other way around–that nights would warm disproportionately, and that more warming would occur at higher latitudes. I think the historic data has been consistent with that assumption.

  5. Max,
    I think your analysis underestimates the power of technological change. Battery costs fell by HALF in the past two years: And batteries aren’t the cheapest form of storage by a long shot (buildings have been creating ice to shift loads off peak for decades). Hot water storage is also cheap. And of course, the sun is usually up for most of the hottest parts of the day, so ever cheaper solar PVs will reduce peak demand substantially. Will we need some storage? Probably, but your analysis is looking out 100 years, and lots can change in just a few years.

  6. 1. Low GWP refrigerants
    2. Extremely low carbon intensity electricity supply, especially in countries where a percent of the population is currently under served by air-conditioning relative to their climate (e.g., India)
    3. Efficient AC/heat pump
    4. Correct installation practice of the efficient AC/heat pump
    Arguably, switch 4 and 3.

  7. I’ll read your PNAS paper as soon as I can get my hands on it. I’m a little surprised by the emphasis on air-conditioning load. My impression has been that since air-conditioning loads correlate with high insolation, they are not a problem for a grid with high percentage of photovoltaic generation, even without adequate storage. The problem comes as the sun sets and lights start to go on. This causes the famous “duck curve.” The belly of the duck is excess capacity in the middle of the day. Isn’t the belly reduced, not increased, by air-conditioning load?

%d bloggers like this: