Posts at Knowledge Problem acknowledge the range of results from Part I and Part II in my series; Katzenstein and Apt; and an article by Michael Milligan et al, Wind Power Myths Debunked, but attribute much of the differences to characteristics of the power system to which wind power is added.
However, although results will vary by jurisdiction, the differences I reported are not derived from this consideration but from general issues with respect to wind power integration. Milligan claims low reductions from the theoretical maximum (negligible to 7 per cent), apparently from Gross et al’s literature review, but this does not survive critical assessment.
The work of Katzenstein and Apt is cited in the bibliography to Part I, even though they show that as much as 75–80 per cent of the CO2 emissions reductions presently assumed by policy makers is realized. The reason for its inclusion is that the underlying approach is used in the calculator. The difference is that the calculator takes into account the limitations that they acknowledge in their article, for example:
Even so, according to the Knowledge Problem post, they have been criticized as overstating the need for backup power supplies by Mills et al, and that geographic diversity helps to smooth out variability. In an update to the post attention is drawn to the Milligan article. This article contains often used, and questionable, arguments to support the ability of wind to offset fuel consumption and the resulting emissions despite its high degree of variability. The following addresses some examples of these.
Milligan claim – The greater the number of wind turbines the lesser the variability. Milligan demonstrates this with two samples of data from a wind plant having several interconnection points. The sample sizes are one of 200 turbines and one of 15 turbines. The comparison is graphically displayed in Milligan’s figure 3, which shows that the “relative” variability decreases as the number of wind turbines sampled increases from 15 to 200. The reason for the inclusion of the term “relative” is that Milligan “normalizes” the results by dividing the output at one second intervals by the mean of each sample. Why show the “relative” variability, which distorts the scale, when it is the absolute variability that is important? The following calculations demonstrate this:
The mean of the 200-turbine sample is about 13 times that of the 15-turbine sample (200/15).
From the graph in Milligan’s figure 3, the variability from the mean is about 20 percent for the 200-turbine sample and 40 per cent for the 15-turbine sample.
The increase in variability with the larger sample is over 6 times greater than the smaller sample ((0.20 x 13) / (0.4 x 1)).
Figure 3 should have displayed the absolute values for comparison purposes, which would have shown increased variability with the larger number of turbines. This is to be expected because wind output is stochastic in nature and the turbines in this wind plant will be strongly positively correlated. Further, it is not clear what the inclusion of the standard deviation information and the ratio of this to the mean in figure 3 add to the comparison.
Milligan’s analysis also leaves open questions about the results of using other 15-turbine samples and much more extensive timescales.
Milligan’s conclusion that, as a result of this analysis, aggregation reduces wind variability for small-scale and large-scale geographical aggregation and all timescales is therefore questionable.
The claim of the benefits of geographic diversity does not stand up to other illustrations by Oswald, Apt, and Adams. The Adams paper shows a high degree of correlation between geographically dispersed wind plants in Ontario and the Nordic region. Hugh Sharman of Incoteco (ApS) Denmark, a Danish energy consulting firm states, “We have seen how large wind carpets, composed of many small units, can act like a single, virtual, ‘out of control’ power station.”
Milligan introduces Nordic system’s ability to balance net variability and generation response, because of existing interconnections. These connections were established to bring the relatively extensive hydro generation from this region to northern European countries. They also facilitate the export northward of Danish wind power, which would otherwise overwhelm the Danish electricity system. Adams shows a strong, positive correlation of wind output in the Nordic region even at distances of 700 km, and which remains positive at 2,000 km plus. No negative correlation is shown.
The Milligan article depends substantially on the questionable argument of the benefits of geographic diversity to support many of its conclusions.
Milligan claim – There is typically sufficient responsive generation capability already built into most electricity systems, which can handle wind’s variability. No additional reserve capacity is required. The Milligan article cites the Gross paper, which is generally favourable to wind. Nevertheless, the following is acknowledged by Gross, and shown in context:
“Intermittent renewable energy plants can save fossil fuel, but may also increase the amount that conventional plants must vary their output, operating in response to market signals. This change in utilization of generation is a separate issue from the need to establish additional reserves. These effects can be quantified using time series data on intermittent outputs and demand, and the implications for the operation of conventional stations assessed.” (emphasis added)
Gross acknowledges the types of analyses needed, and not yet performed, to determine the impact as recommended in my Part I and II posts. As indicated, the Gross paper is generally favourable to wind, and is included in the bibliography in recognition of the contribution it attempts to make to the subject, and because important statements are made, which might be missed by the casual reader. I have addressed some of the considerations in the Gross paper in my Case Study on Methods of Industrial-scale Wind Power Analysis. (See Appendix A for comments on the Gross study.) My position is that papers representing all views should be read carefully.
There is considerable evidence from the German Energy Agency (dena) and E.ON Netz that the installed capacity of wind power is approximately 90 per cent duplicated by other generation capacity, and this duplicate capacity is in excess of that needed to meet peak demand plus reserves. A presentation by Hoppe-Kilpper, Managing Director of deENet, Energie mit System (a consortium of 90 research institutions and service providers in Germany) graphically demonstrates this on slide 13.
Milligan claim – The reserves needed to balance variations in net load (the effect of the combination of demand and wind) are less than the sum of reserves needed to balance variations in the load alone or the wind alone. This is based on the fact that wind power does not correlate with the variability of load. Presumably Milligan means that the correlation is close to zero, versus the unlikely expectation of significant negative correlation. Zero correlation produces a random result, with reinforcement just as likely as opposition, not a generally offsetting one. In this case, the conclusion should be that the variations will be greater with wind present.
Milligan claim – Grid operators in some countries are gaining experience with higher penetrations of wind and with the variability of wind power. Denmark is cited as a good example of handling high penetrations, but no mention is made of the fact that it does so by exporting most of the wind production to Norway and Sweden where it is absorbed by their relatively large hydro generation facilities.
Milligan claim – The impact of forecast errors for individual wind plants is not much of a concern. The aggregate forecast error of all the wind plants is what drives the errors in committing and scheduling generation. Again Milligan relies on geographic diversity, and, as usually claimed by wind proponents, that good forecasting is beneficial. For example, sometimes average deviations from forecast over time are quoted and used as the basis for the need for limited, additional reserves. The reality is that it is the real-time performance of wind power that the electricity system has to deal with, not an average deviation for forecasts over time. Even if the forecast was 100 per cent accurate does not change the real-time impact on wind shadowing/backup capacity. So wind forecasting is not an important or relevant factor. Amongst considerable questionable treatment of the subject, Gross also has this to say:
“However fuel saving may be partially offset by a range of efficiency impacts:
· More frequent changes in the output of load following plant and/or greater use of flexible plant to manage predicted variations. This may decrease the efficiency of thermal plant and cause more fuel to be burnt. Frequent start up and shut down of certain types of plant can use a lot of fuel to ‘warm’ plant, without generating any electricity. The way such changes are provided for is also affected by the accuracy with which fluctuations can be forecast. In general terms better forecasting results in fewer losses, since the most efficient changes can be planned. However improved forecasting does not eliminate these costs, since the need to manage predicted fluctuations will still lead to the effects described above.” (emphasis added)
The question remains: what is the amount of impact? Gross’s assumption of a “partially offset” result is not substantiated without the called-for detailed studies.
Milligan claim – The UKERC determined that the “efficiency penalty” was negligible to 7% for wind penetrations of up to 20%. This also refers to deductions in CO2 emissions from the theoretical, and appears to be from the Gross et al paper listed in the bibliography. First, no jurisdiction absorbs this level of penetration domestically in energy terms. Second, the four studies mentioned appear to be those in Gross’s Table 3.8 that use the Gross-defined C2 and C5 metrics. Of these, three pre-date the turn of the century, when less than 20 per cent of the wind capacity world-wide was installed and two of these are dated 1981 and 1983, when wind capacity, and experience, were minimal.
Milligan claim – Wind power costs compare favourably to nuclear and coal. No attempt is made here to look at all the pricing representations used. However, it is interesting to note two important considerations, in connection with capital cost per unit of energy produced. A very aggressive capacity factor of 40 per cent is assumed for wind and no mention is made of plant life considerations, which differ substantially. This can be as little as 10-15 years for wind turbines. The net effect of is that wind is understated by a factor of about 3 times. The other pricing claims should be looked at carefully.
In conclusion, the Milligan article is not a satisfactory treatment of the subject.
Bibliography
This contains entries in addition to the bibliographies in Parts I and II.
Adams, Tom and Cadieux François, Wind Power in Ontario: Quantifying the Benefits of Geographic Diversity, 2009.
Apt, Jay, The Spectrum of Power from Wind Turbines, 2007.
Center for Politske Studier, Wind Energy – The Case of Denmark, 2009.
Gross, R. et al, The Costs and Impacts of Intermittency, 2006. This paper is generally favourable to wind and is cited by Milligan.
Hoppe-Kilpper, Martin, Systems Studies and Best Practices – Germany – Results from the dena Grid Study. See slide 13.
Oswald, James et al (See Part I bibliography)
Sharman, Hugh, Planning for Intermittency: the Importance of Evidence from Germany and Denmark, (emphasis is Sharman’s), UK ERC Workshop – Imperial College, 2005,
The following is an addendum to the Part I bibliography:
E.ON Wind Report 2004
E.ON Wind Report 2005 (English)
Kent:
Thank you for a thoughtful and professional dissection of the Milligan paper.
Re costs, several sources have shown his contention to be false. Even the somewhat sympathetic EIA shows a significant difference between wind and nuclear when looking at levelized costs.
In reviewing Germany’s extensive wind plants, E.ON Netz, manager of one of that country’s largest grid systems, showed that at no point in 2003 did wind energy exceed 80% of its rated capacity. For more than half the days in 2004, the sum of wind plant output to the grid was lower than 11% of its rated capacity.
That same E.ON Netz report revealed sizable weather forecasting errors for wind energy—ranging from -370 MW to + 477 MW on 6250 MW of installed capacity, and that, during individual hours, the error was as large as +/-2900 MW. Furthermore, when reviewing the results of a timed study of eight-hour-ahead wind production forecasts over 6650 MW of installed wind capacity from 7,000 widely dispersed wind turbines, E.ON Netz found that, although forecasting errors of more than 1000 MW were “fairly rare,” about one-third of the time they exceeded 500 MW. This casts doubt on persistent claims that better forecasting of wind availability will allow grid operators sufficient knowledge to accurately plan the amount of backup generation necessary to safely compensate for the widest wind fluctuations. Such evidence reinforces the axiom that wind patterns are inherently random. Moreover, the size of the E.ON Netz sample demonstrates that spreading large numbers of wind turbines across the landscape—diversifying the aggregate wind supply–would not vastly improve the stochastic likelihood of more accurate forecasting, as many wind developers maintain.
Subsequent analysis by Britain’s Jim Oswalt confirms this finding.
Kent Hawkins in this commentary effectively demolishes the “methods” used by wind boosters to bolster self-serving claims that wind volatility can be functionally tamed through widely-spaced, coordinated wind projects. Stochastics don’t work very well when applied to making such wind projects that much more predictable. Utter randomness can only spawn more randomness. What a wind project does at one time, at one place, typically has little bearing on what another wind project might do a hundred miles away. Or what the same project might do at a different time. And the history of one day–or one hour–has no bearing on how the wind might behave at any future time.
With about 35,000 industrial wind machines now in operation across North America, most of them widely-scattered, there should be plenty of opportunity to investigate whether stochastically refined wind behavior can achieve a more meaningful capacity credit beyond what a stopped clock can achieve on an hourly basis (8.3%).
Hawkins’ call for long-term, finely grained chronological load dispatch analyses should be an imperative for those interested in the application of disinterested science–or those simply interested in making an industry accountable for its claims.
But wait…! The wind industry’s actual performance data is enshrouded behind proprietary confidentiality. How, uh, convenient.
[…] 2009/12/04: Kent Hawkins responds to some of these issues in another post up today at Master Resource. Comments […]
To Jon Boone’s points, I have found that in the PJM system there is a rule called the 10%/5% rule which allows for some variability versus schedule. This information can be accessed at https://admin.acrobat.com/_a16103949/p51803184/ – slide ten.
[…] Hawkins has a further post at Master Resource examining the effects of wind power on overall emissions produced in a power […]
Any author who uses the E.ON 2004 report as a source disqualifies himself from informed analysis. The authors of Myths are experienced power systems researchers with professional credential and I expect that the article was peer reviewed by IEEE.
Try submiiting your paper to IEEE for peer review, rather than an anti-wind site.
Don’t believe what’s written here folks.
Jack Cadogan
The nature of the remarks by John Cadogan is disappointing. I leave it to informed reviewers to assess my analysis of the Milligan article. One cannot hide behind peer review of any sort.
Mr. Codagan–the authors of Myths are AWEA lobbyists. Their work has not been peer-reviewed. I agree that we ALL need to dissect analyses from both/either side. The encouragement of THAT is precisely the purpose of Mr. Hawkins paper in my view. Until and unless you’ve run grid simulations with data that includes REAL heat rate penalties and actual wind measurements along with load profiles and with different wind penetrations, don’t believe the shams that wind will always or even sometimes reduce CO2 or fossil fuel use.
“Science is the disinterested search for the objective truth about the material world.” – Richard Dawkins
“Theories crumble, but good observations never fade.”—Harlow Shapley
These quotes wafted into mind after reading John Cadogan’s criticism of Kent Hawkins’ thesis about using a comprehensive system accounting for the relevant thermal variables involved in “integrating” wind technology’s highly variable intermittent flutter.
There are at least two issues inherent in Cadogan’s comment. He’s perfectly correct about the desirability of getting Hawkins’ work into a first rate energy journal, which I trust will happen soon. On the issue of the substance of Hawkins’ ideas as he exposes the slipshod methods of Milligan, et al, Hawkins’ argument would be correct even if he presented it on papyrus to the Washington Press Corps.
The energy pundits from the National Renewable Energy Lab are hardly engaged in dispassionate work, for their jobs and career futures are in thrall to the success of technologies like industrial wind. Any magician knows how easily it is to fool people who have a stake in the outcome of an illusion.
And any scientist, in contrast to pundits, understands the necessity of eliminating bias in the conduct of experiment. Science insists upon conclusions that account for all the evidence, not selective pieces which fit the convenience of a particular point of view. A given experiment, however honestly and intelligently conducted, can yield the “wrong” answer for a variety of reasons. This is why experiments must be checked by other scientists, using other instruments, other conditions, even other ideas.
Which brings me back to Harlow Shapley’s quip about the overriding importance of good observation–as the final arbiter of successful hypotheses. The Myth authors are very good at using Authority, statistical calumny (as Hawkins demonstrates), and abstract “projections” to bolster their arguments. But they assiduously avoid measuring real-world behavior at meaningful time increments to account for all factors, in ways consistent with the dictates of the First Law of Thermodynamics.
Who is really fooling whom? The idea that a wholly random, unreliable, always fluctuating, non-dispatchable energy source controllable only through curtailment can be incorporated without significant thermal consequences into a system driven by the quest for utter reliability and control is, in my view, delusional.
Perhaps Mr. Cardogan might agree that genuine real-time measurement is required, in the form of sustained chronological load dispatch analyses at 15-minutes intervals, closely examining the heat rate penalties involved as wind sporadically displaces existing power sources, and its continuous ebb and flow is in turn balanced by reliable thermal generators. If so, perhaps he’ll join me in urging an independent, dispassionate evaluation of wind performance–and not simply accept a government report that tailors both methods and conclusions to fit the needs of its client. And then submits them for review by those who also have a stake in the outcome. At the very least, he might consider insisting that wind subsidies be indexed to actual measured reductions of CO2 throughout a grid system.
Admittedly, since wind performance data is typically proprietarily “confidential,” actual measurement might be difficult to do. People like Shapley would be appalled, for how can one get good observations when the data is intentionally hidden from view? Which makes it easier for NREL pundits to pull a Dragon in My Garage stunt as they attempt to bamboozle the public into believing their pigs can fly.
The assault on science by those who “Believe in the Wind” (do recall the Vestas ad campaign) is profound. Good for Kent Hawkins for asking people to think about industrial wind technology in a meaningful scientific way, even though they might only want to clap their hands to help the industry survive its life and death struggle against unbelief.
[…] You might think that wind energy proponents would have their own studies to show that geographical spread is workable. You’d be wrong. The most notable effort I’ve seen is from Milligan, but when you analyze his evidence it falls apart. […]
[…] Part 3 […]