Emissions savings from wind power

Recently, economist Colm McCarthy noted that:

Wind generators can be relied on to produce power only about one hour in three over a year, and those productive hours are unpredictable. So conventional capacity has to be kept in reserve for the periods when the wind does not blow. These stations will be utilised less than optimally and this is a hidden cost of wind generation.

In addition to inefficient use of capital, critics have argued that wind generation has a potential cost in terms of CO2 emissions. When the wind is blowing, priority is given to wind generation over conventional capacity. However an idling thermal plant is like a car crawling along in traffic – not doing very much but still burning fuel. This may cause thermal plant to burn more fuel per unit energy generated than would otherwise be the case.

Is there any direct evidence of reduced CO2 savings when wind generation is high? Surprisingly, the answer to this question is yes.


The scatterplot shows the relationship between total instantaneous CO2 emissions and instantaneous wind generation using data from the Irish grid operator Eirgrid. The data cover the period from 1-Nov-2010 to 30-Aug-2011 at 15 minute intervals (~ 29,000 data points). The blue line is a local regression (loess) fit*.

As expected, wind generation does reduce CO2 emissions. A linear regression fit suggests an emissions saving ~ -0.38tCO2/MWh.  However, the real world relationship between wind generation and emissions is clearly non-linear. At wind generation ~600MW, fuel savings begin to slow. Above ~800MW, they cease altogether. Above 1000MW, emissions increase again.

Heat rate curve

Carbon intensity is CO2 emitted per unit energy generated. To see why emissions savings decrease as wind generation increases, we need to look at the carbon intensity of thermal generation. Thermal generation is extracted from the Eirgrid data as the difference between demand(MW) and wind generation(MW) (assumes no power is dumped). The graph shows the carbon intensity of thermal generation (tCO2/MWh) versus thermal generation (MW) for the period Nov 2010 – Aug 2011.

There is an optimum point on the curve around 3000MW. 3000MW is close to the average electricity demand. In the absence of wind generation, thermal generation fluctuates in line with demand around the the optimum point, a design feature which ensures maximum efficiency . Unfortunately, high wind generation forces thermal plant to operate far to the left of the optimal point on the “heat rate curve”.

*The loess fit has span parameter 1.0. A non-parametric regression curve is shown in grey.

Another plot …

Wind penetration is defined as instantaneous wind generation as a % of instantaneous demand.

This graph of thermal carbon intensity(tCO2/MWh) vs wind penetration(%) tells the same story.



  1. This was predictedin the 2004 Eirgrid report, but delibrately ignored by the Administration:


    Nice analysis on the above also available at:


    Finally this wind energy programme was initiated by illegal means and is now the subject of a compliance case at the United Nations Economic Commission for Europe (UNECE):


  2. Interesting analysis, but I don’t think you can look at the impact of wind as above without looking at the other constituents of generation. Hydro is more or less negligible, peat always runs, despite it’s high co2 emissions. As a rough generalisation, you can divide the generation stack into Coal (Moneypoint/Ballylumford, CO2 emissions somewhere about .95 t/MWh), >50% efficient CCGTs (About 8 of them, emissions are ~.35-.4 t/MWh. All of them are about 400MW, which explains the optimal point of around 3GW I think.), and older less efficient plants (non CCGT gas, Oil, etc).

    Coal is often running for system reasons, and less flexible, so isn’t displaced by Wind, the new CCGTs will tend to be. Hence at low generation levels CO2 emissions are higher, as proportionately more Coal and Peat. At very high generation levels, older less efficient plants are needed, so CO2 increases again. The optimal in the middle is where new CCGTs make up the highest proportion of generation. Wind will push low CO2 CCGTs off the system, as CCGTs have the flexibility (and less costly) to swtich off and on agin (or lower output and increase) as needed.

  3. @Pat Swords

    Thank you for the interesting links. Your references use wind penetration as a useful variable, so I added a plot of thermal carbon intensity vs wind penetration.

  4. interesting analysis, but I wonder why you used linear regression for what you acknowledge is not a linear function. There are certainly other factors (independent variables) that determine co2 intensity, such as total load and load variability (determining which plants are loaded and how.)Then of course there’s the need for local voltage and frequency support that drives dispatch away from economic order.

  5. @Tom Tanton

    You are right, non-linearity is the point of the post. The blue lines are non-linear regression fits (LOESS with smoothing parameter = 1).

    @Brendan @Tom Tanton

    I agree that CO2 intensity at a particular time depends on the outputs and heat rate curves of the various individual stations and that this in turn depends on demand, expected demand, even factors like ambient temperature etc etc. However, it would be wrong to say that you cannot answer any simple questions without modeling all of this in detail.

    For example, now you know that on average the grid was emitting just as much CO2 at 1200MW as it was at 800MW wind generation over the past year.

  6. thank you for the explanation. I note another factor that may be missing…800 MW of wind at a consistent 800 MW has vastly different impact than 800 MW which changes from 600-1000. While the data says “instantaneous” it’s really hourly averaged. Gustiness of wind plays a major role.
    but thank you for pushing the analysis forward.

  7. Here is the URL of a study that is easy to understand.

    A new report by Dr. Fred Udo, a Dutch engineer, analyzes the CO2 emissions of the Irish electric grid, managed by EirGrid, which posts on its website 1/4-hour operations data of total electricity demand, wind energy and CO2 emissions. Analysis of the data proves wind energy reduces the CO2 emissions by just a few percent. 

    The study proves 20% wind energy on the grid reduces CO2/kWh by 4%, 28% reduces it by 1%, 34% reduces it by 6%, and 30% reduces it by 3%, not anywhere near to what is claimed by wind energy proponents.

    Note: The above variations of the CO2 percentages are largely due to the heat rates, Btu/kWh, of the combination of CCGTs and OCGTs selected by the grid operator during wind energy balancing. See website. 


    Note: Paste this URL in the left field of your browser window to access the site.

    The following is a direct quote from the site of EirGrid:

    “EirGrid, with the support of the Sustainable Energy Authority of Ireland, has developed together the following methodology for calculating CO2 Emissions.

    The rate of carbon emissions is calculated in real time by using the generators MW output, the individual heat rate curves for each power station and the calorific values for each type of fuel used.

    The heat rate curves are used to determine the efficiency at which a generator burns fuel at any given time.

    The fuel calorific values are then used to calculate the rate of carbon emissions for the fuel being burned by the generator“

    Note: The heat rate degradation due to ramping down the fossil-fired plants with wind energy surges and ramping up with wind energy ebbs is not accounted for in the above calculation method; i.e., the CO2 emissions posted on the EirGrid site are understated. 

    This means the above CO2 reductions will likely disappear, or become increases.

  8. It only took a few minutes to unravel the statistical trick being used in the Irish data here and in similar work by Fred Udo. This appears to be a classic case of a lurking (or confounding) variable being used to misleadingly present correlation as causality; a comparable example is arguing that cigarette lighters cause lung disease since people who buy them tend to develop lung disease. In this case, the lurking variable that is the actual causal factor appears to be cold weather and its impact on heating demand, data that is available but that (for reasons we can only speculate) was not used in these correlational analyses.

    What tipped me off was part 3, Figure 3 of Mr. Udo’s text, where he called out an event in Ireland around June 9-12, 2011, when the carbon intensity of Ireland’s electricity production surged. I was curious as to what might have caused that event so, on a hunch, I pulled weather records for Ireland. Sure enough, there was an abnormally cold spell when temperatures fell into the 30’s and 40’s F, 10 to 20 degrees below normal for that time of year. Aha! Cold temperatures drive heating demand, forcing Ireland’s numerous fossil-fired combined heat and power (CHP) plants to fire up and run at a high level of heat production (and subsequently more emissions per megawatt-hour, MWh, of electricity, since CHP plants relative to the rest of the fleet are not optimized for electricity production, and CHP plants being run to produce maximum heat are not being operated in a way that is optimized for electricity production; moreover, it appears that the emissions associated with heat production are rolled into the data that Mr. Udo is using, so a CHP plant producing only or mostly heat and little or no electricity under cold conditions like these would score at infinite emissions/MWh). A smaller possible factor is that higher demand for electric heating drives higher merit order, less-efficient fossil plants to operate to meet the abnormally high electric demand.

    As one would expect, cold spells and home heating demand often correlate with high wind speeds, which is how Mr. Udo was able to draw his false conclusion that wind was the causal factor. Sure enough, a closer examination of the spikes in emissions/MWh in his data show that all are associated with cold spells, and only some are associated with an increase in wind output. It doesn’t take a statistician to tell you which is the causal factor in that relationship. Had Mr. Udo himself been more interested in finding the actual causal relationship at play here, he might have noted that the correlations between wind output and emissions intensity varied widely from month to month (as one would expect for weather-driven seasonal changes in electric demand), usually a strong indication that another variable may be the actual causal factor.

    I should also point out that, contrary to Mr. Udo’s claims, the method Irish utility system operator EirGrid uses to calculate emissions savings from wind is accurate. The plant-specific heat rate curve that they are using would account for all of the impacts wind energy would have on the efficiency of the fossil fleets under all operating conditions.

    There’s a large, well-established body of empirical evidence showing that as states like Colorado and Texas have added wind energy to their grids, their carbon dioxide emissions decreased by even more than had been expected. That data was used to directly refute an erroneous study by consulting firm Bentek even before the study came out 18 months ago, and the Bentek study has since been widely dismissed as the fossil fuel industry attack piece that it is, except for continued efforts by the fossil fuel industry to continue spreading its misinformation.

    There are also dozens of power system studies conducted by utilities, government agencies, and independent grid operators showing that adding wind energy to the grid results in larger emissions savings than the 1:1 offset commonly expected, largely because additional wind energy forces inflexible coal plants to be taken offline for extended periods of time and the share of their output not directly replaced by wind is replaced with more flexible and cleaner natural gas generation. I won’t waste time repeating the full discrediting of the Bentek study that has already taken place, but you can read through that with the numerous links below:


    Michael Goggin,
    American Wind Energy Association

  9. Do you have links OTHER than to AWEA press releases?
    More importantly, I think you miss the one large factor that likely dominates, and that is fluctuation or the wind output (whether high or low) for that is what drives thermal plants into off-design operation. +/- 10% fluctuation is much more important if the wind is blowing at, say, 40mph than at 20. That effect of course is magnified with greater penetration.
    The plant specific heat rate curves used in EirGrid assume steady state (at various levels of opstate) not transient operation.

  10. Further to my question @10, the Inhaber curve charts “emissions avoided per MWh of wind generation” versus “wind energy penetration”. At 20% wind energy penetration, the Inhabler equation says that an extra 1 MWh of wind generation would avoid only 4% of the emissions from the generators displaced by the wind generation.

    There are, of course, many caveats and uncertainties, it is highly specific grid dependent and the equation is “schematic”. But given all this, my question is: does the EirGrid data support the general thrust of the Inhaber equation?

  11. @Peter Lang

    Thanks for the suggestion.

    I don’t think formula used by Inhaber has enough freedom to fit the Eirgrid data very well. To describe both the low and wind penetration behavior at the same time, an additional parameter is needed. The problem is that the vast majority of the data points are at low wind penetration.

    A non-linear regression fit using A + B*exp(-C*x) gave a reasonable fit (very similar to the more sophisticated loess method). The fit parameters are:

    A = 0.17 (tC/MWh)
    B = 0.34 (tC/MWh)
    C = 0.017

    This translates into CO2 intensities:

    @zero wind penetration = 0.51tC/MWh
    @100% wind penetration = 0.24tC/MWh

  12. Joe,

    Thank you for that reply. However, I am not quite sure how to use it because I am interested in the ‘CO2 emissions avoided per MWh of wind generation’. Your figures are not for the “CO2 avoided”.

    From the data you have can you calculate what would be the CO2 emissions avoided by wind generation at 20% wind energy penetration (or at 10% or any value of wind energy penetration between 1% and 20%)?

  13. @Joe, I believe the additional parameter you’re looking for is te wind variability (aka gustiness) although that is unlikely to be in data sets. While met data includes whether the wind is gusty (and gusty is officially defined) it does not quantify the gustiness.

  14. (For benefit of other readers, and to clarify my question, below is an edited copy of an email I sent to Joe)

    Hi Joe,

    Thank you for your reply on your web site.

    I didn’t understand the opening sentence “I don’t think formula used by Inhaber has enough freedom to fit the Eirgrid data very well. ”

    I was hoping you might be able to say for the EirGrid data you analysed what is the CO2 emissions avoided by wind generation at a particular value of wind energy penetration, e.g. 10%, 14%, 20% or whatever value you can determine from your data. Then we could see how that value compares with the Inhaber equation at the same value for wind energy penetration..

    To help understand what I am getting at and how the calculations are done see the following papers discussing cost and quantity of CO2 emissions avoided by different electricity generation technologies.

    CO2 avoidance cost with wind energy in Australia and carbon price implications

    Cost and Quantity of Greenhouse Gas Emissions Avoided by Wind Generation

    Emission Cuts Realities

    Other papers on this subject can be accessed from this page (see those authored by Peter Lang):

  15. (a) I can only gasp at the comment by Michael Goggin of the AWEA. Ireland simply doesn’t have fossil fuelled Combined Heat and Power (CHP) plants for domestic and commercial heating, except for what I believe is a small scale gas engine in the civic offices in Dublin, a similar system at Dublin Airport. There are a few industrial units, such as the power plant at Aughnish Alumina, which is connected to their Bayer based alumina process or the Ballyraggert Dairy Plant.

    (b) The main significance in Joe’s and Fred Udo’s analysis is that it confirms what Eirgrid engineers were predicted in 2004, with regard to increased inefficiencies on the grid due to increasing wind penetration:


    In fact it now seems likely that the impact of the wind variability is worse than what the Eirgrid engineers predicted in 2004.

    However, the real issue is that the Irish Authorities and the EU persist with the basis that 1 MW of wind energy displaces 1 MW of emissions from conventional plants. This is clearly false. The dissemination of information on the environment, which is not transparent, is a breach of the Aarhus Convention. This matter will be raised at the meeting at UNECE in Geneva on the 21st, including the position of the Irish Authorities on emissions from this renewable project:


  16. @Peter Lang

    I looked at CO2 savings per MWh of wind generation (relative to savings at zero wind penetration) as you suggested.

    Here are the numbers:

    Wind Penetration CO2 Savings
    0% 100%
    10% 90%
    20% 77%
    30% 62%
    40% 47%

    It is close to a linear decline.

  17. Pat, Ireland’s hundreds of CHP plants comprise around 5% of Ireland’s generating fleet, and during cold spells like this when they are all running at full output that would be 10% or more of generation, which is large enough to drive the emissions increase noted, particularly if some of those plants are producing all heat and no power and therefore registering infinity emissions/MWh. As I pointed out, another major factor unrelated to CHP that explains the increase in per MWh emissions is that cold weather drives a sharp increase in electric heating demand, which causes less efficient, more expensive fossil plants to run as the higher demand forces grid operators to move up the supply curve of available generating plants. The EirGrid data clearly shows an increase in electric demand associated with the cold spells that are triggering the spikes in emissions/MWh.

  18. Michael Goggin

    Getting facts rights is important, it is also a professional and legal requirement for a chartered engineer. Ireland doesn’t have hundreds of CHP plants. In fact the use of CHP is completely dominated by a limited number of industrial plants, the largest one being at the Aughinish Alumina plant and smaller ones at the Ballyraggert Dairy and the Guinness Brewery.

    Your statement above is false and unprofessional, which leads me to question what professional qualifications you have and what professional standards you operate to.

  19. Pat,
    Mike has a degree in Social Studies from Harvard. He spent most of his professional life in PR roles.

  20. http://www.seai.ie/Publications/Statistics_Publications/EPSSU_Publications/CHP%20in%20Ireland%202010%20Report.pdf

    Page 1 of the highlights says:

    “In 2008, there were 11 units exporting electricity to the grid. These units exported 1,013 GWh of electricity in 2008,
    In 2008, 6.3% of electricity was from CHP installations.”

    It also says that CHP reduces CO2 emissions, not increases them. Michael Goggin’s explanation seems to assume that increased generation by CHP increases emissions. It seems the reverse is the case.

  21. Willem

    You need to read that SEAI report carefully. Firstly Ireland has a mild climate, this year (2011) the temperature has only exceeded 25C for about 2 hrs at one loaction. Our winters are also generally mild, albeit damp. In general persistent temperatures less than -5C would be uncommon, although the last three winters have been the exception with some pretty extreme cold snaps. Yet even then these were not presistent.

    For CHP to be used for heating of buildings, as Michael Goggin is implying is happening in Ireland, you need to get a run time exceeding at least 4,000 hrs per year. This doesn’t really occur in Ireland, so the economics of using CHP for space heating are marginal. Furthermore, as the summer maximum is so low, there is little or no economic justification in using the same facility, through absorption chilling (tri-generation), to provide cooling / air conditioning.

    The net result is that the uptake of CHP in Ireland for building heating purposes has been desperately low. Indeed the SEAI report highlights that in the services sector there is only 41.8 MWe of CHP preforming this duty. Which is absolutely nothing in the context of the grid demand, which runs roughly from 2,000 to 4,000 MW.

    Finally as the SEAI report points out, the Irish CHP sector is dominated by industry (85.6%) of installed capacity. If we take the Aughinish Alumina plant with its Bayer process, the Guinness brewery or the Ballyraggert dairy, there is a steady 24/7 demand for heat related to the processing conditions, which really is unaffected by whatever the weather is like outside.

    I don’t want to labour the point, but there is a difference between proper engineering based on scientific evidence, which includes taking the responsibility for implementing the same and what is PR. I would also point out that as an engineer one can be held responsible for one’s negligence and professional misconduct – clearly this does not apply with PR (note I refuse to use the word profession in this context).

  22. Pat,

    I just read the SEAI report. You are right most of these CHP units are very small; micro turbines, etc. What caused Ireland to take this “distributed” approach? The setup must have high owning and O&M costs.

    In Figure 3 is shown that about 6,000 GWh of fuel is used by the CHP plants in 2008. This fuel caused CO2 emissions. How much of this should be added to EirGrid’s CO2 intensity, gram/kWh?

    If that is a small quantity, then Mike Goggins should be made aware of it.

    He will likely not curb his wild statements, because he is speaking to a HIS audience, not us.

    The heat rate of such micro turbines operating at part load most of the time must be well over 13,000 Btu/kWh, for an efficiency of 3,413/12,000 of about 26%, and an annual electricity production of about 6,000 GWh(t) x 0.26 = 1,560 GWh

    Are there MONTHLY data which show fuel consumption, electricity production and heat production for each unit and how much of that electricity was sold by each unit to the grid? It could all be put on one big spreadsheet.

    It would be useful to have such data for each unit for each month of 2010 and 2011; does it exist?

    SEAI must be gathering such data otherwise it could not write its report. Are there reporting requirements?

  23. Willem

    Ireland has no history of district heating, esentially 20 years ago the only CHP plants operating were the two old coal fired plants at the two sugar plants, which are now long gone. With the promotion of CHP by the EU in the late nineties there was a renewed interest, but there still is a limited number of plants, basically because with our climatic conditions and urban structure it is not very economic to go the district heating route.

    Indeed the uptake has been predominately by industry, with the Aughinish Alumina plant completely dominating this sector, see below:


    This is a must run plant, so it is given first priority on the grid. I’m pretty sure the same arrangement is for the other plants, although they are much smaller in size, the below summarisies most of them:


    They are all pretty modern and in CHP mode are giving over 70% efficiency. Although not all are on full load, for instance the Alumina plant was on 50% capacity due to the economic situation post 2008 and the brewery changed its cleaning process a number of years ago to one using less steam.

    With regard to those in the services sector, which are usually small gas engines (1 MW or less), they are often connect to additional heat sources, such as a swimming pool in a Hotel complex. This enables them to achieve a higher utilisation than what would be obtained solely on space heating, which is the bottom line – it is difficult in our climatic conditions and with the absence of district heating networks to find a useful heat sink.

    With regard to reporting requirements, all sites with more than 20 MW thermal input, and that includes standby capacity, have to report to the Environmental Protection Agency ( http://www.epa.ie ) on their annual greenhouse gas emissions. I would also be pretty sure that additional reporting is required to the Commission for Energy Regulation ( http://www.cer.ie ) to verify that the plant is meeting the efficiency requirements determined previously by Irish legislation (> 70% if I remember), but now by an EU Directive:


  24. The case is much worse for wind. Analysis of Eirgrid’s Adequacy reports over the last few years shows that for every 1MW of wind, 1MW of conventional plant was built. This occurred over a period of falling demand so either Eirgrid expect a huge increase in energy consumption or they are ensuring there is a large enough system to hide the increased penetration of wind in.

  25. @Owen, agreed. Gas capacity increased pretty much in line with wind capacity because wind needs flexible balancing thermal plant. In effect wind capacity targets are also gas capacity targets. The extremely expensive gas/wind baseload system being built has nothing to do with market need. It is being built to meet a bad renewable electricity target.

    A multi-billion euro box-ticking exercise.

  26. A real hottie, she tries to get Kevin to impregnate her, but he.
    By and large when this amusement is played the normal
    way citizens will have to be part of the European Community
    but because of the way the Euro – Millions Lottery has been set up by e – Lottery Syndicate it
    gives individuals from all round the earth the option to engage in and triumph the
    huge Jackpot prize fund. Still, there is no harm
    in actually giving some strategies a shot to up your
    chances of hitting the jackpot and preventing your dreams from turning
    into dust.

  27. He used vacant containers of powdered milk to style versions of Haitian stoves, which he marketed.
    Cash from the perk at Carambola will certainly aid give a water pump so the kids will certainly not need to walk a quarter-mile to safeguard water.

  28. In this modern of hybrid technology situations are more
    simplified this sort of born using a natural a sense music.
    Nevertheless, customers want the most effective beat making
    software that you can buy, right. Then the program should include
    educational tools giving you all the ideas you need to create a strong technical know-how base.


Leave a Comment

Your email address will not be published. Required fields are marked *

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