This post is the first in a three part series that critiques the recently published article “Wind Power Always Replaces Fossil Fuels” by Chuck Kleekamp, which provides material for another in the series of my critiques of wind proponents’ claims. Previously analyzed were papers by Milligan, Komanoff and Gross. My understanding is that this author has previously made notable contributions to environmental matters. Let’s see how he does with respect to wind.
To begin, I cannot help commenting on the inclusion of “Always” in the title. The apparent certainty in this term immediately alerts me to a questionable analysis. Perhaps the author meant to be provocative, and was not serious in the use of this word. If so, this does not give due consideration to the importance of the matter.
This leads to another general comment. In a circulation of a draft of these posts to a panel of reviewers, one commented on the nature of Kleekamp’s article as that of not having sufficient knowledge of the subject, but attempting to appear so. He provides descriptions, but makes errors in the process. Cases in point are his (1) example of the Mirant Canal oil-fired plants and (2) description of electricity system markets and activities of the System Operator of New England (ISO NE).
Mirant Canal Plants
As the illustration for conclusions, the author uses information about the Mirant Canal oil-fueled, supercritical, steam turbine plants. He describes the high efficiency performance of these two plants and the low loss of efficiency at notably low loading, which are characteristic of the supercritical type. Based on two points of steady-state operation, he therefore draws two conclusions:
“Therefore to say when wind power comes on line it will not reduce fossil fuel consumption is simply erroneous.”
“The point of this discussion is that no fossil fueled generating unit consumes more fuel than that which is absolutely necessary to produce the dispatched power allowed under the rules of the ISO. When the wind comes up or air conditions [sic] and lights are turned off, less fuel is consumed on a linear basis.” (Emphasis added)
Although the detailed information provided is interesting and may well influence a general audience, it is not relevant to the discussion. Frequent cycling is required to balance wind’s volatile output. It is also required to balance short term demand fluctuations, and fast acting, but less efficient (in the case of fossil fuel plants), generators are required for the continual up and down ramping required in both cases. According to a National Petroleum Council (NPC) report on Electric Generation Efficiency”, supercritical plants are not capable of cycling without reducing their longevity and efficiency. Power-Gen Worldwide makes the same point.
So, the inherent assumption, and not considered in his description of normal demand fluctuations (Kleekamp’s example of when lights or air conditioners are turned off and on), is that there is no increased fossil fuel consumption impact as a result of the frequent cycling of the balancing units, aside from their less efficient operation at lower load levels. These units just “magically” change to operating at the new level in response to continuous short term demand fluctuations. Unfortunately for his argument, there is an impact, and this is part of the necessary “cost” of keeping the electricity system in balance. Add wind, especially large quantities, and this effect is magnified significantly. See the following section for more on this.
Kleekamp thus makes two mistakes. The first is extending his conclusions, derived from the specific oil-fueled plants he focuses on, to all fossil fueled plants, and the second is to ignore the impact of the frequent cycling required, which many fossil fueled plants, especially the Mirant Canal plants, are not suitable for.
Finally, why choose oil-fueled electricity plants to make his point? As reported by the EIA, oil-fueled plants account for only 1% of all U.S. electricity generation, admittedly higher (about 5%) in New England, but still not worthy of general conclusions on the subject.
Perhaps the explanation is Kleekamp’s reliance on this to support his questionable argument that electricity generated from oil is the first displaced by wind, and the even more questionable conclusion of meaningful, increased energy independence.
Electricity Market Considerations
The author’s description of the electricity market appears to focus on the day-ahead market, versus the spot market, which is the one that wind largely participates in. Although there are variations, in general the day-ahead market typically provides about 90% of the electricity generation requirements (not one-quarter as Kleekamp suggests) and the spot market is used for final balancing requirements. Wind plant owners typically do not participate in the day-ahead market because this market has an implicit guarantee of availability, which wind cannot provide. In some cases though, wind does participate but is uniquely “excused” for non-delivery in the event of the unavailability of its fuel, wind.
Kleekamp provides a very simplistic (and incorrect) explanation of the operation of the New England electricity market as managed by the ISO NE. He states the following:
“The ISO dispatches generators in the region from an hourly bid stack that starts from the lowest-priced bids (this includes generators that bid $0, such as [wind], hydro, and nuclear units) and progress to higher-priced bids (i.e., from coal, natural gas, and oil fueled generating units) until there is sufficient generation to meet consumers’ demand for each hour of the next day.” (Emphasis added)
Although, based on a general statement in an ISO NE document, Kleekamp has inappropriately altered it by adding wind, admittedly contained in square brackets indicating an edit. Note this ISO NE statement refers to the day ahead electricity market operation and dispatchability. His edit alters the high-level generality of the statement that addresses the general operation of the electricity system, naturally starting with baseload generation sources and progressing further from there.
A source of Kleekam’s error is in attributing dispatchability to wind. True, by mandate wind is used whenever it is available, with an exception that will be described later, but this does not mean that it is dispatched as such. In this context dispatchable means sources of electricity that can be called on at the request of power grid operators (ISOs), that is, it can be turned on or off upon demand. The author persists in his interpretation in the subsequent discussion with a commenter.
Kleekamp then incorrectly claims, based on fuel costs and price alone, that wind replaces oil, natural gas and coal in that order, drawing on the above (edited) quote. Closer investigation reveals different information. The ISO NE responded to a question from a reviewer about what can expected to be backed down in a wind event, and the response was (quote from the reviewer, not from the ISO NE):
At least 500 variables are run in the dispatch model every 5-10 minutes, and there is no way they [ISO NE] could tell us what the system will do. They said, we [ISO NE] can only look at what resources are easiest to back down, and given we have a good supply of pumped storage in the region, and a good supply of gas, as well as a good supply of biomass (which can be backed down some) there is a reasonable chance wind will back down other renewables, or our cleanest source of generation. (Emphasis added)
On the subject of exceptions to taking wind as available, there are times when wind production is not used, known as wind “curtailment”. This usually occurs when wind penetration exceeds a few percent. It is a widespread practice in Germany and is notable in Texas and the UK. Do not be distracted by questionable comments by wind proponents in the referenced articles.
Also not mentioned by the author, is the price paid to wind plant operators for delivery of electricity. Regardless of the price in the auction process, wind is paid at its agreed premium price in any long term Power Purchase Agreement (PPA) it may have. Who pays for the difference? The electricity customers and tax payers in the jurisdiction where the wind plants are situated pay for this.
Kleekamp’s claim that Cape Wind Project PPA will save $4.6 billion based on a study by Charles River Associates has been refuted here.
Also, the argument that long term, fixed price PPAs protect against future price volatility does not withstand scrutiny, especially considering the very high prices in wind plant PPAs (about 20 cents/kWh with escalators for Cape Wind, or many times the going rate in a 20 year contract with escalators). Would you take a 20 year term mortgage on your house at 20% (plus escalators) to protect against future interest rate volatility? Some of his references to prices date from just after the 2005 very high gas price spike.
Whoever the wind production is ultimately sold to will pay the auction price and receive the associated cost benefit. This is what is happening in Denmark for most of the wind produced there. In this connection, it is also notable that in Germany there are cases where some fortunate customers are actually paid to take wind generation, often at night. Readers should be careful not to take this as evidence that wind contributes to lower electricity rates. Although such lower prices may be enjoyed by a few fortunate local customers occasionally, or more frequently by those in other jurisdictions, the “piper must be paid” by all who have the wind plants in their country, state or province.
Is Wind Production the Same as Normal Demand Variation?
The answer is no. This is another area where error is demonstrated by Kleekamp.
There are two distinct changes in normal user demand. The first is the regular daily increase to one (sometimes two) peaks and subsequent reduction as night approaches. This is relatively predictable and is responded to with intermediate generation sources that typically cycle once per day in response. The peaks experienced each day are met by responsive peaking generation units. The author seems to say that these are rarely used “…except for unusually cold winter days or extremely hot summer days…”, which is not correct, and the possible source of this error is explained below. Secondly, there are the short term variations in demand that are balanced by responsive online “spinning” reserves, which provide for this and other contingencies. Again these short term variations are fairly predictable, especially with respect to size.
In addition to spinning reserves, there are other reserves, on a different availability basis, designed to meet (1) extreme weather conditions (the possible source of the error above), (2) replacing spinning reserves committed to meet an unscheduled outage of a single major plant as well as (3) scheduled maintenance.
Because the vagaries of wind production must be responded to with other generation means, it is often considered to be negative demand, versus production. When the random wind production is netted against normal demand, two things occur. The first is that the sum of two such events produces a resulting random event with larger ranges of fluctuations. Further, the wind variations alone have amplitudes that are not predictable and can be larger than the normal demand short term fluctuations. Do not be mislead by the use of statistical analyses over long periods of time, which are a form of averaging that smoothes out this effect.
The net of this is that wind introduces a larger, less predictable net demand that must be continuously met with other generation sources by the system operator, as already indicated. At low wind penetrations, say in the range of 1-2% (in energy terms, that is watt-hours), this might be “manageable” in some cases without undue perturbation or changes. As wind penetration increases this cannot be easily masked and becomes a significant problem, as experienced elsewhere, especially in Denmark and Germany. These two countries, in combination, appear to be able to handle wind penetration of about 7%, albeit with some difficulty, and only due to access to the regulation capabilities of the other Nordic countries, the large hydro power in Norway and Sweden. This actual experience likely represents the upper limit for wind. It is also quite unique because of the availability of very large hydro resources.
The impact of this consideration, and other “schemes” being discussed to mitigate the inclusion of intermittent renewables in electrical systems, are covered in an assessment by the North American Electric Reliability Corporation (NERC). An overview is available here.
Nowhere has actual experience shown that there will be no major impact from wind until it reaches about 20% (in energy not capacity terms). Denmark does produce wind electricity in this range, but must export the majority of it, which is otherwise not manageable. Readers are particularly directed to Part III of this series for more information.
EnerNex Simulation Study
This is one of the studies cited by Kleekamp and is discussed very briefly here, based on an admittedly quick review. It is very comprehensive (Volume 1, 91 pages and Volume 2, 135 pages), involving simulation and considerable statistical averaging over long time periods, of a type which usually misses important points. It was published in 2006 and appears to be based on a previous, also very extensive, EnerNex study (145 pages), published in 2004. Although involving long time series of data for short time intervals, the approach is useful primarily for long-term capacity planning, not short term impacts. Admittedly impressive in scope, this study must be viewed in the light of some of the following statements taken from the Executive Summary (shown in italics), which reveal important considerations as described for each.
The considerations referred to above indicate that there are likely some important threads running through the study that significantly affect the theoretical results. Sufficiently motivated readers can investigate and decide for themselves.
On a final point, in the use of the term “capacity value”, EnerNex means capacity credit as defined here.
So far Kleekamp has failed to be convincing about the value of wind plants. Part II will address further issues, with a focus on capacity considerations. Part III completes the series covering the costs of nuclear versus wind plants, which disproves his apparent warning about nuclear costs exceeding that of wind.
Good beginning. I’ll look forward to Parts II and III. On the matter of why Kleekamp presumed that wind must displace oil, perhaps he was referring to economic dispatch obligations whereby grid controllers remove the highest cost production in response to perceived demand reductions (which is how wind penetration typically appears). Perhaps the Mirant plant had the highest cost production, although Kleekamp does not state this.
Unfortunately Kent, it seems like you still have a few things to learn from the conversation I’ve been trying to have with you on your other post, here:
http://www.masterresource.org/2010/11/the-calculator-14-results-part-i/
The following statement from your article in particular shows that you’ve got a ways to go:
“When the random wind production is netted against normal demand, two things occur. The first is that the sum of two such events produces a resulting random event with larger ranges of fluctuations. Further, the wind variations alone have amplitudes that are not predictable and can be larger than the normal demand short term fluctuations. Do not be mislead by the use of statistical analyses over long periods of time, which are a form of averaging that smoothes out this effect.”
By using the word “random,” you seem to be trying to incorporate what I have been trying to help you understand, but falling short. The key point is that wind and load are random, i.e. not correlated, over the subhourly intervals that are relevant for grid operators and for determining impacts on emissions. Therefore, when these sources of variability combine on the power system, they combine statistically according to the square root of the sum of their squares, not additively. Since wind variability is so much smaller than load variability (consult any wind integration study, like the EnerNex study, to see that this is the case) the incremental variability added by wind is so small as to be almost inconsequential. As an example, a power system with 100 MW of load variability and 5 MW of wind variability over a 10 minute interval will have an aggregate variability of 100.125 MW (using the square root of sum of squares) when those sources of variability are combined, 40 times smaller than the 105 MW you would have assumed by doing a simple arithmatic rather than a statistical combination.
This has nothing to do with conducting a statistical analysis over a long period of time. Two variables that are found to be random (uncorrelated) are random at any point in that dataset. That is what random means.
Also, in the next paragraph, you claim that wind penetrations beyond 7% are not possible. As a counterfactual, you may want to look at the countries of Germany, Spain, Ireland, and Portugal. All get over 10% of their electricity from wind, at times getting more than 40% of their electricity from wind, and all have done so without any negative impact on grid reliability. Let’s focus on the case of Ireland. Ireland gets well over 10% of its electricity from wind, and the government has plans to obtain 50% or more of the country’s electricity from wind. Peer-reviewed government studies have shown that this can be achieved with no negative consequences. What is most remarkable is that Ireland is a small country, has a very inflexible generating fleet made up mostly of coal plants, is an electrical island with only weak abilities to import and export power, and has fairly little diversity in its wind resources. If Ireland can get to 10% wind, let alone 50% wind, under those extremely adverse conditions, then the sky is the limit in a country like the U.S.
Dear Stat Prof,
You have set up a strawman example with regard to variability. In the countries that you mention heavy levels of wind production, especially in the night hours, have led to market and system instability. You might want to look at this post (http://www.masterresource.org/2010/09/german-wind-high-cost-least-cost/) to see how power pools work in the real world with high levels of wind penetration.
In fact, for all of the countries that you have mentioned the wind variability over short periods of time exceeds the load variability. In your conceptualization the time period is too long for pool management; the real period of control has to be less than 10 seconds.
So, if load variability is 100 MW on a 10,000 MW system and wind penetration is 15% (1,500 MW) with an expected temporal variation of 150 MW, then the total variability of the load + wind = 180 MW and the additional variability over expected load is 80 MW, or an 80% increase in variability. In this case the addition of the wind resources would require 80 MW or more of fast acting engines to offset the variability of the wind from one time slice to another.
AWEA and NREL, those revolving door wind boosters, have been citing the old square root of the sum of the squares routine for many years. Not hard to understand. It is statistical and may or may not conform with reality. Reality would impose itself when those 5MW of installed wind fail to materialize, leaving a shortfall that must be compensated for somehow. It also would impose itself with increasing wind penetration, magnifying the flux that seems limited using the square root of the sum of the squares. To say that wind flux is not additive is at best disingenuous, since it is flux that is added to that of load–from the supply side, no less. As Tom Tanton points out, wind and load are “not un-correlated but, depending on the region, anti-correlated.”
It’s not clear what the limits of wind integration might be as a percentage of wind penetration. That would depend upon a number of factors, including the number and quality of interconnections, the overall generation mix, and the quality of regulation, among other things. Sure, there are times when in parts of Germany, the part that has strong hydro interconnections, wind is generating up to 40% of the supply. But there have also been times when it has tripped the system and caused widespread blackouts throughout Europe.
As for Spain and Portugal, again, because of the amount of installed wind capacity, there are times when wind provides 10% of the supply. But this is not routine and it is always dicey. As wind flops randomly around as function of the cube of the wind speed at a rate typically about four times the intensity of the routine load flux, I’d love to see a chronological load dispatch analysis at 15-minute intervals measuring the heat rate penalties from those Irish and Spanish coal generators as they work to balance all that aggregate flux. No wonder the cost of electricity is so high in those countries–and getting higher. And last time I looked, Spain continues to increase it CO2 emissions, even in the electricity sector, despite-uh-all the wind….
And one must wonder why all those subsidies enabling wind activity aren’t actually indexed to measured reductions in fossil fuel consumption and greenhouse gas emissions.
[…] Part I of this series critiquing an article by Chuck Kleekamp dealt with the more general issues of examples used, one of the major references and electricity markets. There is a lot found to question his analysis. This post focuses on capacity considerations and other miscellaneous issues raised by Kleekamp. Finally, Part III addresses his remarkably inappropriate warning, “If you think wind power is expensive, wait till you have to pay for electricity from a new nuclear plant.” […]
Stats Prof
I believe your statistical approach has been effectively dealt with by Don Hertzmark above.
With respect to your comments on random concepts, you do appear to be “falling short”. Hertzmark also addresses this in terms of the very short time frames involved, which is the essence of my point. The combination of two random events is itself a random event. The range of variation increases and the two do not have to be positively correlated for this to occur. The fact that they are random is sufficient. This is consistent with actual experience with electricity systems embodying wind. However actual experience is not necessary to establish this point.
I believe I detect interesting characteristics in your comments in nature, tone and how they have progressed. You start off as presumably a “statistics professor” dealing with related matters and end with the very remarkable declaration that with wind “…the sky is the limit in a country like the U.S.” Your arguments increasingly sound more like those of the American Wind Energy Association (AWEA), and like wind-promoting organizations, citing favourite sources such as the National Renewable Energy Laboratories (NREL) as published in the IEEE. I must confess that I am beginning to become suspicious about what is going on here, and that you are not a “disinterested” third party, perhaps as you would like believed.
One a final note, you make claims about the ability of other countries to absorb large wind penetrations. I suggest you publish a paper to that effect. I would be pleased to critique it.
That said, inasmuch as you are sincere in making your comments, I thank you for your efforts.
Inasmuch as you are sincere in your beliefs, I respect your views.
Inasmuch as you are wrong-headed, wilfully or otherwise, in your beliefs, I am disappointed.
Therefore, I conclude that we are at an impasse. You may choose to continue with your comments, but my intent is not to respond, except perhaps to acknowledge receipt. I suggest we leave the matter to the moderator (exercising his discretion as to the value involved), but most importantly, to the readers to decide for themselves.
Donald, thanks for the comment. I truly am grateful that you are able to grasp the proper method for statistically combining the variability of wind and load. The example of 100 MW load variability versus 5 MW wind variability that I cited, while hypothetical, is in line with what occurs on power systems with large amounts of wind energy. In Texas, the grid operator calculated that 15,000 MW of wind will produce 6.5 MW, 30 MW, and 328 MW of variability at 1-minute, 5-minute, and hourly intervals respectively. The California grid operator estimated 12,500 MW of wind plus 2,600 MW would similarly produce 3.3 MW, 14.2 MW, and 129 MW of variability at those time intervals. New York also calculated 1.8 MW of 5-minute variability and 52 MW of hourly variability from 3,300 MW of wind. The studies are available here: http://www.uwig.org/opimpactsdocs.html At even these very high penetrations, in any of these areas the variability of load is going to dwarf the amount of variability you see from wind on the subhourly timescale. At hourly intervals it is true that wind variability can start to exceed that of load at high penetrations. Fortunately, hourly variability is much easier for grid operators to accommodate, as grid operators are able to use much slower, non-spinning reserves, which are available with very little efficiency/emissions penalty, and typically at a few percent of the cost of faster-response spinning reserves.
Separately, Kent, I must say that it is truly disappointing to see you go back to burying your head in the sand on these issues. I thought we were making some progress for a bit, but now it’s fully clear that this website exists solely for the purpose of providing a forum for you and your three or four other deniers to reinforce your own mistaken beliefs by circularly referencing each others’ work. It must be fun pretending to be Davids to the Goliaths of real discussions that are being conducted on these topics, at real conferences, in real peer-reviewed articles, and among the real grid operators in the U.S. and Europe who regularly deals with large amounts of wind energy. If you actually listened to these grid operators I think you’ll be in for quite a surprise. If you truly think that entities like NREL, DOE, and IEEE are part of some vast wind energy conspiracy, you may want to consult a psychiatrist. I am truly sorry that you have decided to take the tone of this conversation in this direction, but as it is now clear that no amount of facts are reason are going to convince you to change your ways, I am no longer going to waste my time talking with you. If you ever change your mind and decide to open up to constructive criticism on how to fix the flaws in your emissions calculator, please let me know.
We can’t just make an assumption on everything we think is right. We need evidence and hard facts to back it up.
Statistics Professor,
I have a few comments and questions.
1. Like you, I believe the matter of the emissions avoided by wind farms is an extremely important issue that we need to understand, and the sooner the better. Furthermore, I believe the cost per tonne of CO2-e avoided is the key metric we need.
2. To gain this understanding I believe we need actual measurements of demand, power output( from wind and from fossil fuel generators) and fuel use. We need these measurements in real time at a short time interval. We need these measurements during times of low wind output, high wind output and strongly fluctuating wind output. Preferably, this study would be done on a relatively isolated grid, with relatively high wind capacity penetration and with no energy storage component. The South Australian grid would be an ideal situation to do such a study. It has a relatively high wind capacity penetration, high proportion of gas generation (this makes fuel measurement easier than with coal generators), no hydro capacity and no energy storage, and only two small transmission interconnectors to the rest of the Australian National Grid. Demand, output from each generator and energy flows over the interconnectors, all at 5-minute interval, are all publically available information (see links provided below). The fuel used in the power plants is not publically available.
3. The Australian National Energy Market (NEM) grid has the largest areal extent of any in the world, so I understand. Wind farms are distributed over the southern part of the grid over an area spanning 1200 km east-west and 800 km north-south. There is strong positive correlation of wind farm output over this area. Furthermore, it is frequently the case that wind power output drops as demand increases and vice versa. This causes the fossil fuel generators to have to ramp faster than if there was no wind capacity in the system.
4. Here are links to charts showing the output of wind farms in the NEM. You can download the data from this site if you want to do more with it. http://www.landscapeguardians.org.au/data/aemo/
– May 2010 (6 days of near zero wind output across the 1200km x 800km area; in fact the total output from all NEM wind farms was negative on 65 5-minute periods)
http://windfarmperformance.info/documents/analysis/monthly/aemo_wind_201005_hhour.pdf
– August 2010 – (seven cycles with total output of all NEM wind farms from as low as zero and up to 85% capacity factor)
http://windfarmperformance.info/documents/analysis/monthly/aemo_wind_201008_hhour.pdf
– Wind farm output for 2010.
http://windfarmperformance.info/documents/analysis/annual/2010/wind_capacity_factor_demand_2010.pdf
5. My first question is whether you know of any studies, at appropriate scale, of the actual CO2 emissions avoided by wind farms – that is, studies which use actual measurements of fuel used (other than the three studies Kent Hawkins used to test his calculator against)?
6. I interpret your previous comments upthread to suggest you believe that the problems of wind power’s intermittency would be reduced as the capacity penetration increased. Many wind power enthusiasts argue for high capacity penetration and seem to believe that capacity penetration up to 20% and even 50% of total installed capacity is viable. Do you agree with them on this? If so, could you comment on what would be the emissions avoided by wind farms, per MWh, at those high levels of penetration, and what would be the cost. I expect the cost would be unacceptable, and the cost of emissions avoided would prove to be a very high cost way to avoid emissions. Do you have an opinion on this?
Those who seek to improve an explanatory idea, and especially those who can, by identifying counter examples, debunk those ideas, should be treasured. On the other hand, those who seek to misinform, confuse, misdirect, particularly those selling snake oil, should be exposed for what they are. Readers should be reminded that, despite the attempt by one commenter here to misdirect the argument, Hawkins’s central thesis remains intact.
Perhaps Hawkins and others will incorporate, with fuller explanation than provided here, the statistical integration formula mentioned above, where demand/wind flux wavelets are averaged to help determine levels of supply needed to meet demand during relevant time periods. But as energy expert Tom Tanton reminds, “modern
computing/storage/processing allows each time period to be assessed (say, using minimum or maximum) without the need to ‘average wavelets.”‘
No one to my knowledge here on Master Resource ever maintained that wind, at least at certain levels of penetration, could not be integrated on most grids. Hawkins’s calculator assumes this. What he is after are the implications of any wind balancing, that is, its effects on overall carbon emissions and fossil fuel consumption, particularly any heat rate penalties imposed on thermal generators that may accrue to wind beyond those attributed to demand flux.
As Peter Lang observes, this has been rendered next to impossible because of laws protecting proprietors of generation fuels, including especially our friends in the wind industry. This is a particularly nettlesome problem for accountability, since public funds are responsible for enabling wind enterprise.
But here’s the kicker for anonymous bloggers: no matter where one looks, one can find no evidence that wind performance, measured after the fact, is the responsible agent for reducing fossil fuel consumption, including coal consumption–anywhere. Here’s a good example from Ontario:p://windfarmrealities.org/?p=721. But one can go to the Pacific Northwest, to Denmark, Germany, Spain, Texas, Colorado, California–and find the situation.
If wind is so easy to integrate, why aren’t areas with copies wind supply showing unequivocal evidence that there are corresponding large scale reductions in coal, gas, or oil consumption NOT ATTRIBUTABLE TO REDUCTIONS IN DEMAND, INCREASES IN CONVENTIONAL GENERATION, OR THE VICISSITUDES OF WEATHER? With the lack of access to do the kinds of measurements at appropriately fine-grained time intervals that I, Hawkins, and Lang, have suggested, all the statistical mumbo jumbo in the world is merely happy talk. Still, the idea put forward by Hawkins and Lang that heat rate penalties induced by wind must increase both emissions and fossil fuel consumption sufficient to offset much, if not all, of any initial reductions wind might cause, seems the most efficiently elegant explanation for why there’s no reductions in fossil fuel consumption or emissions–a la Occam’s Razor.
As to that other issue of how much wind penetration a grid can absorb with changing its identity and putting on a lot of weight in the form of new conventional generation, one should consider some thought experiments. Imagine a grid with 1000MW of installed generation–200MW of which is installed marginal reserves (IMR) that only come into play when overall grid security is threatened. Of the remaining 800MW, imagine that 25% is wind. Therefore, this grid would have 600MW of conventional generation to meet all levels of routine demand, and we could outfit this with the national average of 50% coal, 20% nuclear, 20% natural gas, and 7% hydro. Finally assume a baseload demand of 400MW–half of the routine baseload/midload/peakload that fills out the 800MW of demand met by the 800MW of installed grid capacity (remember we’re keeping 200MW as IMR).
What must happen in such a grid when there is 800MW of demand (on those peak demand days) and zero wind? Or only 100MW of fluctuating wind? Surely, these situations, which would be rather routine, would invade the IMR–big time, leaving no IMR to handle emergency outages caused by weather or malfunction.
As a number of Loss of Load Expectation studies have shown, this is a major problem for any grid, not least Germany’s. Grids therefore have an IRM ranging from 10%-35%, depending on the kind of generators that dominate, the historic load behavior, and, the number and quality of interconnections with other grids. The IRM is deeply important for ultimate grid security and shouldn’t be deployed for other, more routine purposes–like standing in for wind flux, as AWEA suggests.
AWEA’s statements that wind can easily be integrated within existing grid reserve margins at small levels of penetration are undoubtedly true, although it taxes the grids operating reserves, causing them to operate more inefficiently. As wind penetrations increase, however, this is untrue, since it then encroaches on the IRM. This really cannot be tolerated. So more capacity reserves, in the form of conventional generation, must then be added–in a continuing regression; the more wind, the more reserves, ultimately reaching a 1:1 replacement circumstance.
In consequence, modern grids must have conventional generation in place, most of it operating more inefficiently, to engage 80-90% of wind’s installed capacity. As wind penetration increases, likely at levels approaching 5% on most grids, additional conventional generators must be added that approach 100% of the installed wind capacity. At levels of wind penetration approaching 10%, it seems statistically essential to add reliable capacity at 110% of wind’s installed capacity.
The “sky is the limit” for wind in America only in the sense that there is a trillion dollars worth of diamonds lying hidden beneath the earth’s surface. Problem is–it would take a trillion and a half dollars to extract them. Se la vie.
I’m in the wind industry and I’m just trying to figure out what’s real. I’m fortunate enough that my skill set can translate to other generation projects and could potentially slide into our nuclear division if I truly thought that wind weren’t worth the effort. As such, I would hope you would give me the benefit of the doubt – not as a disinterested third party but as an honest actor.
I took seven days of BPA’s five minute data and have looked at it in several different ways statistically. I will present all of my findings as [average, stddev, max] (all in MW unless otherwise noted).
Load [6671, 723, 8291]
Wind [1104, 808, 2801]
Load delta [38,37,494] (absolute value of change in load over 5-min period)
Wind delta [21,23,261] (absolute value of change in wind over 5-min period)
Wind penetration [17%,12%,44%]
=SQRT(G9^2+H9^2) [48,38,504] (statistical combo of load delta and wind delta)
Not sure what to do with those numbers yet or what they mean but thought I would throw some data on the fire and see if anyone responds.
“Further, the wind variations alone have amplitudes that are not predictable and can be larger than the normal demand short term fluctuations.”
This can be fixed. I know it can be fixed for the ramp-up because I’ve done it. We installed a SCADA system that clamped ramp-ups on a wind farm to conform to the Scottish grid code in 2004. They gave us the specs they required and we simply didn’t allow the output at the substation to ramp up faster than what they allowed.
Ramp down is a little trickier but still manageable, I think. There are three areas of interest on a wind turbine power curve. Area 1 is from cut in wind speed (approximately 3.5 m/s but depends on the turbine make/model) to “rated power wind speed” (12 m/s). This area can be reasonably approximated by a linear equation. From rated power to cut-out, the power curve is flat – no fluctuations in output when the wind speed is in this range. Above the cut-out speed the turbine turns itself off – called high wind shutdown. While potentially the most dramatic change in output (max to zero), this condition is pretty rare and can be handled more gracefully using curtailment and selective shutdown modes (turn off the turbines one at a time and let them coast to a stop instead of slamming them all off at once.) Above the knee, between rated and cut-off wind speed, the output is flat, thus, nothing is needed.
It is only in the middle of the linear part – away from zero where little power is being produced and away from the knee where the curve flattens out – are large fluctuations even possible. Of the 1877 five minute samples I grabbed from the BPA, there was a ramp-down greater than 1% of the load 48 times. One percent was arbitrary on my part.
jam – comment No 12
One can look at this from a statistical point of view, but this is not the real world in which we live and use electricity. As a simple example, if I understand life expectancy tables for the US, it is about 60 years for people born before 1940. So statistically I am dead, but in the real world “rumours of my death are greatly exaggerated”. This statistic may be useful in life insurance applications and demographic projections for example – a real world in another sense.
So as to the question as to what is “real”, here is the result of taking the range of load fluctuation, wind fluctuation and the combination over the same time period as jam used, which he cooperatively provided to me.
• Load fluctuation range: 3,006 MW
• Wind fluctuation range: 2,801 MW
• Combination (load minus wind) fluctuation range: 4,751 MW (158% of load fluctuation range)
A frequency distribution might be interesting, but the reality is that the extremes must be dealt with. Also using further statistical analyses to compare two different wind regimes may be appropriate.
My point is that the combination of wind and load fluctuations results in a greater range of overall variation than either alone. There may be exceptions to this for a number of reasons because of, for example (1) time scales used (2) some consideration about the fit of load and wind events to a normal distribution curve (3) grid topology considerations (in other words there may be a number of populations to look at independently and cannot be usefully viewed in aggregate) and (4) wind penetration. There is more to say on this and the more I look the more complex it gets on both sides of the argument. I plan a future post on it.
Although interesting to some perhaps, and depended upon by many wind proponents, statistical treatment of real time wind impact is a type of averaging that produces a statistical expectation over time. This bears little or no relation to the real time impact. Further the impact of wind variations on fossil fuel consumption and CO2 emissions is the real issue to focus on. This would take actual real time data of plans involved in wind balancing and not estimates based on assumptions and algorithms.
There are uses for statistics in capacity planning and a capacity credit can be determined for wind on this basis, taking into account its stochastic nature. Typically the capacity credit is less than 10% of installed capacity.
I am not well versed in statistics, so any comments/corrections by one who truly is (real statisticians please) would be welcome.
jam – comment No 13
I guess more detail on your SCADA application is necessary to fully assess it. It looks like the result is wind curtailment and this is becoming a normal remedy as wind penetrations increase. It is experienced for example in Germany and Texas. Denmark avoids it by exporting the majority of its wind production. This also explains the inapplicability of your application to ramp down, because now wind must be dependably available when called on, which it isn’t.
I agree wind curtailment is an effective wind integration tool. The question then comes back to the overall value of having wind in the system at all, but this is a very large subject.
With respect to your reference of the electricity production profile of a wind turbine you seem to underplay the normal operating range in the middle of your description from cut in to the knee (You say in this range “…are large fluctuations even possible”). About two-thirds of the available wind conditions occurs in this region. Here the electricity output varies as the cube of the wind speed and thus can provide augmented changes in output. Your incomplete description of down ramping conditions (1% of load is still about 65MW) is not convincing (time frames and only half the problem anyway).
In any event, the wind variations are themselves not the issue as much as is the impact on fossil fuel consumption and emissions of the wind balancing generation source. This is the area where we need the real data (not estimates) for generation plants in a wind balancing role to determine the effectiveness of wind power in reducing fossil fuel use and emissions. I must confess to forgetting to emphasize this myself.
By the way, as you admit to being in the wind industry, perhaps you could tell us the electricity consumption of wind turbines at the three ranges of operation (to cut-in, cut-in to rated power, and from there to cut-out)
“the electricity consumption of wind turbines”
I don’t know it by range of operation, but 3-5 kW is the maximum parasitic load unless heaters are required, then another 2-3 kW could be added.
John,
It would be useful to know the average parasitic load say over a year. Assuming it is the lower of the maximums, the sum of the two you provided is 5 kW. This translates to 1% of average output for a 2 MW wind turbine at 25% capacity factor (5 / 2,000 kW x 0.25 = 0.01), and less for higher capacity factors.
Denmark does publish annual “own use in production” for all generation sources except wind (I wonder why not) and this amounts to about 3%.
I suspect your numbers are low, but I have no other direct information on this. So yours are a welcome addition to the scarcity of information on this topic. You didn’t mention air conditioning load though.
As well, do you know if the turbines have to be turned sometimes when not in use to avoid “settling” or “bending” of shafts. I am told that ships in harbour have this problem and this is how it is dealt with. Remember that the rotor assembly of a 2 MW wind turbine is 30 tons plus at the end of the shaft.
What about parasitic load around the low end of the normal production portion of the turbine output curve, when the nacelle/blade assembly (60 tons) has to be turned to the proper orientation for the current wind conditions, as well as alterations in blade pitch? Does the turbine need help from the grid in starting blade rotation at these low wind levels?
Any further help or information would be appreciated.
Kent:
Perhaps my response was too specific to your question. The parasitic loads of a wind turbine are around 1%. However, looking at your question/response writ large would be to examine the parasitic loads of the entire wind plant. The turbines themselves are pretty spartan. The internal lights and computer monitors are rarely used. There is no television in the break room, no coffee pot pulling 12 amps running 16 hours per day in the base of the turbine, etc. Pulling together all of the parasitics for the plant, including electrical losses through the collection system, the machine shop, the laser printer in the office, etc, I would think that 3% would be a reasonable number. That’s a more valid comparison to another generation source than just the parasitics of a single turbine.
To answer your other questions:
There is no air conditioning load in any turbine that I know of.
Wind turbines will generally be producing power 80+% of the time (i.e. spinning), just not at full power. And when they aren’t producing power they are often left to “float” so that even a very small wind will allow the rotor to move (but not produce any power). Thus, there is no parasitic load associated with turning to avoid settling or bending of shafts.
The turbine does not need help from the grid to start.
John,
First of all thank you for providing the information that you have done. My questions are about some of the less well known aspects of wind turbine operation, and I am looking to your experience in this industry. Of course if you do not know the answer (speaking for the wind industry), say so. That is fine.
I did not intend to get to the level of television use and coffee pot considerations etc. However, I would point out that where these apply, the relationship of such considerations to the amount of reliable electricity produced has to be taken into account, not any absolute amount. In any event, I accept these as part of the 3% number previously referred to by me.
I note with interest your statements (1) that there is no air conditioning load (requirements) for large scale wind turbines, (2) the “floating”/rotating state of the 30 ton blade assembly in low winds, and (3) that wind turbines do not need help from the grid at cut-in.
I find your comment that wind turbines produce power 80+% of the time surprising. One source based on UK Met Office information shows the following and typical of the relationship between the frequency distribution of wind speed and the wind turbine power curve:
– Wind speeds from zero to cut-in for wind turbine production for 110 days per year. This is a rough basis, presumably based on average wind speed for the day. This accounts for 30% of the year. Very little production should result.
– From cut-in to about one-third of the way up the normal production range another 165 days is accounted for. Production should be low. The cumulative total of this (admittedly rough) calculation is now 275 days or 75% of the year.
– From there wind speeds trail out to past the knee of the production curve to cut-off when the turbines must be stopped.
Of course energy production increases as the cube of wind speed. So energy production does increase substantially even as the frequency of occurrence of these wind speeds falls off, but you are talking about time not energy production. In summary, your 80+% of the time number seems questionable. Perhaps I am missing something and you can explain this further.
In your example:
“Wind speeds from zero to cut-in for wind turbine production for 110 days per year. … This accounts for 30% of the year. ”
Thus, the wind speed is greater than cut-in (and producing power) 70% of the time. Working backwards, I find that this number results from a hub height wind speed of 5.2 m/s. I would never recommend someone building a wind plant in this wind regime.
I’ve generally found that a 6.5 m/s hub height wind speed is a minimum for a viable project in the U.S. Here is the wind probability table for that scenario:
0-3.5 m/s 20.8% No Output
3.5-12 m/s 72.8% Output proportional to wind speed
12-25 m/s 6.5% Rated Output
>25 m/s 0.0% No Output
Thus, capable of producing energy approximately 79% of the time.
One of the better sites I’ve ever been to had a wind profile like this:
0-3.5 m/s 5.6% No Output
3.5-12 m/s 74.3% Output proportional to wind speed
12-25 m/s 20.1% Rated Output
>25 m/s 0.0% No Output
Thus, capable of producing energy approximately 94% of the time.
Here is a very rough table showing hub height wind speed and estimated percent time the wind turbine should be capable of running:
5.0 67.5%
5.5 72.3%
6.0 76.1%
6.5 79.2%
7.0 81.8%
7.5 84.0%
8.0 85.8%
John,
Thank you for this information. It appears that the differences in proportion of the year before the wind turbine cut-in point has to do with the strength of the wind regime as you point out. It’s been a while since I last looked seriously at the wind turbine power versus wind speed curve, and the wind frequency and resulting power output distribution overlays, and it’s been an interesting re-visit. Given your numbers, your statement does now appear reasonable, especially when you take into account that any amount of production counts, which is relevant even at average wind speeds below cut-in. Average still means some periods of sufficient wind for production.
I do have to point out that your comment that output is proportional to wind speed over the normal operating range from cut-in to rated output is not correct. Output here varies approximately as the cube of the wind speed, so the output climbs (and falls) more rapidly than linearly as wind speed increases (decreases). I am looking at a wind turbine production curve at http://www.sd-commission.org.uk/publications/downloads/Wind_Energy-NovRev2005.pdf on page 19, which shows the following relationship between wind speed and output:
Wind speed 7 m/s and output of 500 kW (ratio of 500 to 7 is 71)
Wind speed 9 m/s and output is 1,000 kW (ratio of 1,000 to 9 is 111)
Wind speed 10.5 m/s and output of 1,500 kW (ratio of 1,500 to 10.5 is 143)
If output was in proportion to wind speed, the output at 10.5 m/s would be 500 x 10.5/7 = 750 kW
Or 10.5 x 71 = 746 kW
But in fact the output increased to approximately 500 x {(10.5 m/s)^3 / (7 m/s)^3} = 500 x 3.4 = 1,700 kW
The difference between 1,500 and 1,700 kW is because the relationship is not exactly as the wind speed cubed.
This is important when considering the impact of changes in wind speed on wind output production and volatility.
Kent,
You are of course absolutely correct. I was using “proportional” idiomatically rather than in the strict mathematical sense of the word. The power equation is:
P = 0.5 x rho x A x Cp x V^3 x Ng x Nb
where
rho=air density
A=swept area of the rotor
Cp=coefficient of performance
V=wind velocity
Ng=generator efficiency
Nb=gearbox efficiency
The equation is a form of the kinetic energy equation (1/2)mV2. The extra V term that appears in the power equation comes out of the deriviation of the mass as the mass is that of the air moving through the rotor swept area which obviously depends on the velocity.
[…] – Cost of Wind … LionofJudahthemovie.com- Lion of Judah The Movie- Once Upon a StableParts I and II dealt with most of the issues in a recent paper by Chuck Kleekamp and showed the weaknesses […]