Underestimating Renewables

The International Energy Agency, or IEA for short, is an autonomous intergovernmental organization based in Paris. They were established in the 1970’s after the OPEC embargo sent oil prices to new highs. Their main job is to guess the future when it comes to energy production. They do this in their annual World Energy Outlook or WEO.

I’ve tended to trust their predictions, since they seem to have a lot of expertise and don’t seem to have a strong axe to grind. They believe global warming is a serious problem and they’ve outlined some plans for what to do.

However, I’m now convinced that they’ve consistently underestimated the growth of renewable energy. Not just a little—a lot.

This is bad news in a way: who can I trust now? But of course it’s mainly good news! I am now more optimistic about the potential of wind and solar power.

To explain what I mean, I’m just going to quote a chunk of this article:

• David Roberts, The International Energy Agency consistently underestimates wind and solar power. Why?, Vox, October 12, 2015.

Here goes:

David Roberts on the IEA

That the IEA has historically underestimated wind and solar is beyond dispute. The latest look at the issue comes from Energy Post editor Karel Beckman, who draws on a recent report from the Energy Watch Group (EWG), an independent Berlin-based think tank. The report analyzes the predictive success of previous WEOs.

Here’s the history of additions to electric generation capacity by renewables excluding big hydro, along with successive WEO projections:

[Chart from Energy Watch Group. Click to enlarge.]

As you can see, IEA keeps bumping up its projections, but never enough to catch up to reality. It’s only now getting close.

It gets even worse when you dig into the details. Here’s the bill of particulars:

• WEO 2010 projected 180 GW of installed solar PV capacity by 2024; that target was met in January 2015.

• Current installed PV capacity exceeds WEO 2010 projections for 2015 by threefold.

• Installed wind capacity in 2010 exceeded WEO 2002 and 2004 projections by 260 and 104 percent respectively.

• WEO 2002 projections for wind energy in 2030 were exceeded in 2010.

Other, independent analysts (like those at Bloomberg New Energy Finance and Citi) have come closer to accurately forecasting renewables. The only forecasts that match IEA’s inaccurate pessimism are those from the likes of BP, Shell, and Exxon Mobil.

Here are IEA’s wind and solar projections broken out, from a 2014 post by the folks at eco-consultancy Ecofys:

[Click to enlarge.]

Back in 2013, energy analyst Adam Whitmore took a look at the IEA’s track record on renewables. He found it abysmal, like everyone else. This year, he returned to the WEO to see if it has improved and found that, well, it hasn’t.

Here he shows the rate of growth in annual installations of renewables, and what the IEA projects for the future:

[Click to enlarge.]

(The dashed lines are the standard WEO projections, what happens if nothing changes. The dotted lines are from the “bridge scenario” in the WEO Special Report on Energy and Climate Change, which is supposed to represent some policy ambition.)

As Whitmore says, it’s possible that the rate of solar PV installations will suddenly plunge by some 40 percent and then enter a long steady-state period, but there’s no reason to think it’s particularly plausible.

For more

Roberts goes on to analyze various possible reasons for the IEA’s consistent underestimates. They’re worth reading, but none of them seems like an obvious smoking gun.

I suppose if I were very careful I would check all the graphs and numbers in Roberts’ article, but I’m inclined to trust them. He’s getting them from various sources; this is a factual issue that can be easily checked, and I haven’t seen anyone arguing the other side.

If you want to check some numbers yourself, you can download these free books:

WEO 2011.

WEO 2010.

WEO 2009.

WEO 2008.

WEO 2007.

WEO 2006.

The following report, mentioned above, goes into more detail about the IEA’s failures:

• Matthieu Metayer, Christian Breyer and Hans-Josef Fell, The projections for the future and quality in the past of the World Energy Outlook for solar PV and other renewable energy technologies, Energy Watch Group, 2015.

They write:

Summing up, the IEA keeps ignoring the exponential growth of new renewable energies such as solar and wind, and does not learn from its past mistakes.

But this leaves me with a question. Who is doing the best job of predicting energy trends? This is where we could really use a well-developed, easily accessed prediction market.

31 Responses to Underestimating Renewables

  1. nad says:

    just to liven up the discussion a bit – I was just trying to imagine the outcomes of a prediction market predicting the pathway of a tornado. :)

  2. The Sage says:

    In practise, there is a large gap between the nominal capacity of a wind or solar installation, and its actual output. For example, a check at time of writing shows the UK’s 13.5GW wind capacity is actually generating 2.8GW. And with sunrise an hour away, solar isn’t doing anything.

    In light of such considerations, the IEA’s cautious estimate of the impact of renewables seems a sensible thing to do.

    • John Baez says:

      It’s true that the nominal capacity and the actual output are completely different things. People who don’t know this can sometimes be fooled into thinking wind and solar are much better than they actually are.

      The IEA is not so dumb. But note, the graphs and discussion here are comparing the IEA’s predictions of various quantities to the actual historical record of the same quantities. They are not comparing apples and oranges. The IEA has been bad at predicting these quantities.

  3. Peter Bloem says:

    I’m no expert on any of this, but I’ve always felt that banking on the continuation of a perceived exponential growth is very dangerous. Usually the exponential regime only represents a transition between stable states.

    Of course, the end of this exponential growth may be when the whole of the Sahara is covered in solar panels, but there may also be an earlier ceiling that will stop the growth within a few years.

    Perhaps that is the reason that the IEA is so cautious?

    • Well, Seba relies on the many instances of exponential growth based upon semiconductor technology as the rationale for extrapolating that to zero Carbon energy. True, that applies more to solar PV and electric cars than it does to wind turbines and energy storage, but the latter two are increasingly being dominated by great computing, good algorithms, and clever alternative, as inspections of IEEE Transactions on Smart Grid, Power Systems, and Power Electronics for the last year or two will indicate. Even the Control Theory people find this a rich area of research, because, for instance, the tuning of power output from a field of wind turbines is a combination of great wind forecasting and adaptively tapering and adjusting individual turbines in the field to conditions. According to his publications, Professor Greg Plett at Colorado State University is using smart controls and state-space models for optimizing batteries (see http://mocha-java.uccs.edu/HCBRTL/).

      So, I’d say there’s more here than simple extrapolation, but, true, there could be obstacles in the way.

      I think the model is that decentralized generation of power and electrical transport, perhaps “juiced” by self-driving cars and the Uber model, will cause people to simply (and wastefully) abandon fossil fuel assets and infrastructure, even if they were relatively recently built.

  4. Three pillars of zeroing Carbon emissions: (1) electricity generation, (2) heating, and (3) transport. To the degree “(2)” and “(3)” can be made to depend upon “(1)”, that’s great. There is substantial and (very) private infrastructure which relies directly on fossils for “(2)”, although the efficiencies to be had (e.g., air heat pumps) for converting are great. Capital investment is needed. “(3)” relies upon electrification of the transport fleet and probably another economic disruption, like self-driving cars. Stanford’s Tony Seba has dramatically argued for these.

    Another group, led by Stanford’s Mark Jacobson, has looked at what it would take to provide electrical energy for all purposes.

    If global zero Carbon energy growth is government by a Moore’s-type law implying exponential growth, projections even a few years out are sensitive to small errors in the growth coefficient. I do not know if shortfalls in the IEA’s projections could be explained by that. The other question to answer with respect to their projections is why do they consistently underestimate if that’s the case?

    I am heartened by all this discussion for the same reasons as John, but also because of mathematical law. In particular, we know exponential growth in consumption is not sustainable, per the dramatic illustration of David Suzuki. But the same argument goes for any exponential process. Accordingly, if zero Carbon energy has (hypothetically) a two year doubling time, the percentage of market captured just 10 years away from total domination is 3%, because 2^{-2 \cdot 5} = 0.03125.

  5. Rob Mac. says:

    To be fair to iea part of the reason is the low ration of solar/wind to conventional. In early adoption stage we can go after the low hanging fruit. This appears exponential bit surly follows a resource limited growth pattern like logistic curve.

    Did you read “the worst hard time”? It is about the dust bowl. I sometimes think the dust bowl is a reason to be hopeful. Using only ford tractors we turned the great Western grasslands into a desert in a few short years. It is a testament to our short sightedness. But also a testament to our ability to make massive rapid changes to our environment quickly.

    Subsidies for wheat during ww1 caused the dust bowl. Government conservation subsidies fixed it. Someday we may look up and realize we put up so many solar panels we are heating up the planet because we increased the earths albedo or something crazy like that. Let’s hope.

  6. Jenny Meyer says:

    They aren’t very good at predictions, are they? I suspect they’re starting with incomplete and/or faulty assumptions, and the people making these predictions are working within the mindset of the energy industry — which seems to be all about getting the most energy at the least cost. In particular, infrastructure: energy companies are notorious for neglecting their own infrastructure — as the state of California is discovering right now with the natural gas leak there — because they don’t want to waste one penny of the value of the old infrastructure by replacing it before it fails. This is kinda like “better to ask forgiveness than seek permission” writ large. I don’t think this approach is a good plan for a small planet.

    I don’t know the answer to your question, but I suspect people looking at the energy sector as just one component in a long-term investment strategy are going to have a clearer picture.

  7. Samuel says:

    Non-Hydro renewables in 2014: 317 of 13500 Mtoe, or around 2% of total. In other news, you need to be cautious with an insignificant energy source which will have an exponential growth at the very beginning as any new energy entering the picture will. You don’t forecast on an exponential curve when it’s going from 0.1% to 1% in a few years, you wait for the market to mature… I’m pretty sure that when this growth will subside like it is already in countries like Germany, the IEA projections will be deemed too optimistic, but then of course the renewable bandwagon will not complain of “overestimating renewables”

    • Well, while their views might still be considered optimistic, both McKinsey & Co, and Bloomberg have projections which see zero carbon energies dominating well before the end of the 21st century, and emissions reducing to the agricultural production floor. McKinsey, in particular, puts a lot on energy reduction, because of behavior change and automated interventions inserted in between drivers and their vehicles, or users of energy in business and homes and their heating and cooling plants. Their emphasis on energy savings and emissions reduction comes from savings being multiplied by the network effects of providing energy to end users (see any Sankey diagram for electrical or heat energy delivery to homes), and greater efficiencies overall.

      It’s possible, therefore, that projected shortfalls in energy supply might be due to misjudgments of energy needs as much as availability.

  8. I did some quick fact checking on the stats for solar and wind and they don’t match the claims in the article.

    I downloaded a copy of WEO 2010. In their “New Policies Scenario” they have (fig 10.3, p 306) the following projected increases from 2008 – 2035:

    Wind – 2600 TWh (approx, this is from their graph)
    Solar PV – 618 TWh (from the text, p.307)
    Solar CSP – 339 TWh (” “)

    In BP Statistical Review of World Energy from June 2015 we have total world consumption of renewables:

    Wind – 706 TWh ( = 27% of the projected addition to 2035)
    Solar (total) – 186 TWh ( = 19% of the projected addition to 2035)

    And estimated increase from 2008 to 2014:

    Wind: 706 – 219 TWh = 487 ( = 19% of the projected addition to 2035)
    Solar (total): 186 – 11 TWh = 175 ( = 18% of the projected addition to 2035)

    So at least part of the article seems way off.

    Projections made in 2010 for the next 25 years for wind and solar have not already been exceeded.

    Note: they give a breakdown of their “New Policies Scenario” which is between “Current Policies” and “450 Scenario”. From the graph it looks in total to be about 20-25% above “Current Policies”.

    • John Baez says:

      Thanks for checking some numbers! Let’s see if we can figure out what the problem is here.

      You mention ‘consumption of renewables’, measured in terawatt-hours—which I assume means actual production of energy by wind and solar plants. But this is completely different than ‘capacity’, which is the theoretical amount of energy that could be produced by these plants if it were windy and sunny all day (and night?).

      All the graphs and numbers in David Roberts’ article seem to concern ‘capacity’. If so, there’s no reason they should match ‘consumption’. They should be much larger.

      A bit more precisely, when you say

      I downloaded a copy of WEO 2010. In their “New Policies Scenario” they have (fig 10.3, p 306) the following projected increases from 2008 – 2035:

      Wind – 2600 TWh (approx, this is from their graph)
      Solar PV – 618 TWh (from the text, p.307)
      Solar CSP – 339 TWh (” “)

      you don’t say whether these figures are for capacity or consumption. Later you mention consumption.

      • John,

        (And see my updated comment below specifically on 2015 predictions).

        IEA Outlook figures are generation in TWh.
        BP Statistics figures are consumption in TWh.

        There can be a discrepancy between these two but not much, and the errors in calculating the values will be more.

        When you see figures in TWh you expect this is actual energy rather than capacity.
        When you see figures in GW (or TW or PW) you expect this is nameplate. Nameplate x capacity factor x number of hours in the year = energy generation.

        Of course, you have to check.

        The energy produced / consumed is the best figure to use because you can install 1GW nameplate of wind turbines in Oklahoma and get 3.5 TWh of annual output but if you install it in Germany you might get 1.6 TWh.

        This problem of quoting nameplate values is worse for solar. Because first you have varying efficiencies – e.g. 13% – 20% for the type of solar PV panel, then you have to remove losses with heat (values are quoted as nameplate at 25’C) and losses from inverters, and then you have to work out where they will be installed. So if you install 1GW of solar panels you don’t get 8.76 TWh even in very sunny countries.

        The IEA WEO primarily predicts energy production (not nameplate installation). But it’s 738 pages and I’m sure there are nameplate predictions in there as well.

        • SOD,

          preface this by saying I read SOD at lunch and enjoy it. Thanks for putting it together.

          re “first you have varying efficiencies…” etc. I don’t think so. That is accounted for in the name plate. efficiency for a panel is basically power out @ STC (Standard Test Conditions) divided by module area. Different manufacturers will sometimes play with the area quoting “apperature efficiency” which is a way of excluding the frame from the area calculation.

          The reason for the standard test conditions is fairly obvious. You need a way to compare panels. Thus standards. Why 25C if panels almost never operate at those temps (and typically operate much hotter? The sun is hot, right?) Well, we humans are adverse to working in 41 C environments. If we could just figure out a way to get iguana’s to work for us, we could get a more accurate estimate of output power. Companies will publish, and in fact compete on module temperature coefficients, how much power is impacted by temperature (hotter is bad, colder is good) a fun read on this this is “Third Generation Photovoltaics by Martin Green. He goes through all the basics of solar cell efficiency physics, and all the thermodynamic limits for different cell strategies. Amazingly because the sun is do hot and the earth so cool, the carnot limit for a solar cell is crazy high 95%. The book is basically about why in practice we can’t get there.

          Of course for solar installs, name plate isn’t power output. Where I live, you can expect about 10% of name plate on an annualized basis. I take lots of vitamin D in the winter. There are lots of factors in addition to the one’s you named, like cloud cover, angle the panels are installed (usually you follow the roof line for a house), growth rate of your neighbors trees, bird poop (yes, there is an actual industry problem referred to as “the bird poop problem, the cells in modules are strung in series, so when one cell gets hit, it pulls the entire module down. )

          Given that, it seem unlikely that the IEA would be better at predicting power output than predicting name plate capacity. This is surprising.


        • There are also engineering responses to some of the problems cited in Rob’s post, such as cloud cover, strings going down, and the Bird Poop Problem. From personal experience with our recently installed 10 kW (nameplate) 29 panel array from SunPower and RevoluSun, some of the best days in our situation are diffuse cloudly days. We have a lot of tree cover at the horizon, and the shading that would otherwise affect the array at some Sun azimuths goes away with cloud cover. Also, this array has SunPower optimizers on it, which allow it to work when parts of panels are shaded. Also, the SunPower panels have white margins around the cells, making each look like an eye. This is a passive bird discouragement. Also, because the panels operate warmly, the usual way poop gets removed — I’ve witnessed this — is for a snowfall. The snow melts off and carries dust and poop along with it.

          In summary, just looking at direct physics of cells and Sun is not sufficient to tell generation when there’s control logic and engineering in-between.

      • John Baez says:

        Thanks for your clarification!

        The IEA WEO primarily predicts energy production (not nameplate installation). But it’s 738 pages and I’m sure there are nameplate predictions in there as well.

        Perhaps it’s worth emphasizing that the eye-catching graphs I got from David Roberts’ article claim to show the IEA’s predictions of ‘capacity’. I suppose I should be careful in guessing what that means, but I would imagine it means something more like ‘nameplate’ than ‘production’ or ‘consumption’:

        [Chart from Energy Watch Group. Click to enlarge.]

        div align=”center”>

        [Chart from Ecofys. Click to enlarge.]

        • In terms of what this all means, Kevin Anderson of University of Manchester has consistently maintained that, while wind and solar are good, they cannot be brought online fast enough to keep within +2C above preindustrial, especially if developing nations are going to be allowed leeway to get their economies ready for zero Carbon emissions.

          Accordingly he has maintained, and I’m increasingly in agreement, is that the alternative is to push energy efficiencies in a big way and, meantime, people who are part of the global 1% who are responsible for 50% of the world’s greenhouse gas emissions should curtail their consumption and transportation to buy zero Carbon energies time. He is particularly scolding of climate scientists and other academicians who certainly know better and yet trot across the planet going to meetings by air. Air transport has a particularly large impact, since the same unit of CO2 emission at altitude has a bigger warming effect than if emitted at sea level. (This was part of the problem with the Permian mass extinction. The large igneous provinces created updrafts that lifted gases high in the troposphere and into the low stratosphere.) This is very hard.

          Anderson makes the case that despite the unpopularity of this approach, to the degree that the economies of the 1% benefited from high per capita greenhouse gas emissions and, morally speaking, if anyone should take a cut in standard of living, it’s them, meaning you and me and us.

          The alternative, of course, is a miracle of solar and wind deployment, and electric transport. Miracles really oughtn’t be depended upon.

    • I was wrong about the article.
      I found the appendix with a breakdown (p.620).

      Projected solar PV 2015: 58 TWh
      Projected solar CSP 2015: 29 TWh
      Total solar 2015: 87 TWh

      Projected wind 2015: 756 TWh

      Solar actual is about 2x the prediction (the article is roughly correct)
      Wind actual is a little less than predicted (93%)

      Total projected 2015 wind + solar = 843 TWh
      Actual wind + solar = 892 TWh

      Their projection is a staggering 5% below the actual for the combination of the main renewable additions.

      The article says
      Current installed PV capacity exceeds WEO 2010 projections for 2015 by threefold.
      – roughly correct

      Installed wind capacity in 2010 exceeded WEO 2002 and 2004 projections by 260 and 104 percent respectively.
      – didn’t check

      They didn’t comment on the 2015 projections about wind (made in 2010) which turned out to be very accurate.

  9. John Baez says:

    For those who want to look for themselves, here are some free books:

    WEO 2011.

    WEO 2010.

    WEO 2009.

    WEO 2008.

    WEO 2007.

    WEO 2006.

  10. Walter Blackstock says:

    Maybe of interest: a snapshot of the UK National Grid supply and demand.

    • Jon Davies says:

      Thanks. That’s a really useful website. Shame it can’t track the UK solar generation explicitly but I guess that is not possible for domestic solar. Would be interested if the commercial sites have to report in any way.

  11. Perry Metzger says:

    One thing seems to be entirely absent from this discussion: what the price of power is as generated by various different means. Price is what drives such things.

    The exponential rise in renewables is not a phenomenon that should be predicted by some sort of model-free curve fitting as the causes here are understood. Uptake of new generation methods is driven by human greed. Most specifically, it is driven by the fact that as the price of power from renewables drops further and further, and especially as it drops substantially below the cost of power from other sources, we can generally expect that people will prefer to purchase power from the lower cost method of generation.

    Technological cutovers have happened many times in human history, often with surprising speed. If something is cheaper and a near or exact substitute for another good, people buy it in preference to the more expensive good. Therefore, if you want to understand uptake, you have to look at prices, not model-free curve fits. If you know what the price is, and what it is likely to be based on foreseeable technological shifts, you can likely predict uptake.

    You will find, if you look closely, that the price of photovoltaics (which I’m more familiar with than wind) has been dropping quite rapidly over the last decade, and is now already substantially below the price of fossil fuel power in places where fossil fuel power is particularly expensive, such as Hawaii. Not surprisingly, in such places the uptake of PV generation has been extremely rapid. In places where the price of PV power is not yet below that of other generation methods, such as places with relatively low insolation and plentiful nearby sources of fuel, the uptake has been much slower.

    However, as the price trend for the fuels is relatively stable over long periods and has a floor at long term production costs, while the price of photovoltaics seems to be bounded more by efficiencies in manufacturing and installation than by any sort of materials costs, the price of PV still has a long way to drop before it hits a steady state baseline. We can therefore expect rapid continued uptake even in places like the Northeastern United States and Canada because of price drops.

    Equally interesting, uptake has been very rapid in the third world where power grids are often absent or unreliable and where operating costs of small (building or village scale) PV systems are much more affordable.

    • I also made the comment today at my blog posting regarding the Lazard’s LCOE study: “Only thing further I’d say about this and evaluations like it, is that they don’t reflect the benefit of residential solar PV being put in with capital that’s been raised and paid for by entities other than utilities, or society. Sure, these are often subsidized with tax breaks and I see these less as incentives for installing residential solar and more as a reward for people to put their capital at risk to provide a social benefit, whether that’s excess electrons generated for society’s use, or load that’s been shed onto a local generator.”

    • John Baez says:

      Perry wrote:

      The exponential rise in renewables is not a phenomenon that should be predicted by some sort of model-free curve fitting […]

      Agreed. I often argue with some people who try to do that. I believe the IEA, whatever their problems, is not so dumb as to do this.

      Regarding your main point, namely that prices matter, the IEA came out with a report that I want to read and post about here:

      Joint IEA-NEA report details plunge in costs of producing electricity from renewables, 31 August 2015.

You can use Markdown or HTML in your comments. You can also use LaTeX, like this: $latex E = m c^2 $. The word 'latex' comes right after the first dollar sign, with a space after it.

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.