Electricity output figures show wind turbine performance deteriorates very slowly with age

I wrote a few weeks ago about the surprising assertion from the Renewable Energy Foundation (REF) that the performance of wind farms declines rapidly with age. A study carried out by Professor Gordon Hughes for the REF in 2012 suggested that ‘The normalised load factor for UK onshore wind farms declines from a peak of about 24% at age 1 to 15% at age 10 and 11% at age 15’. To put this in everyday English, Professor Hughes is saying that a 15 year old onshore wind farm will typically produce less than half its initial output of electricity. Few people in the industry would demur from a conclusion that wind farms very gradually lose output but none accepted Hughes’s finding that electricity generation falls at anything like the rate he stated.

If true, his finding would have serious implications, as the REF was keen to point out. To achieve the UK’s targets for wind-generated electricity, we would have to put more turbines on the ground because ageing wind farms would produce much less power than expected. This is an important topic and I thought it needed more examination.

After meeting REF in early 2013, DECC Chief Scientist David MacKay responded to the study, eventually publicly saying that Hughes’ work had serious statistical flaws. REF has recently rebutted Professor MacKay’s comments saying, with some asperity, that his actions are ‘extraordinary’ and impugning his understanding of econometrics.

Few of us have the detailed knowledge of statistics to say whether Hughes’ conclusions follow from the data he has used. I thought it might therefore be helpful if I analysed the individual performance of all the UK’s oldest wind farms. I’ve looked at the data on the output of 14 farms, all established in the period 1991 to 1993. I’ve been particularly helped by the assistance of Peter Edwards, the entrepreneur behind Delabole, the Cornish wind farm that started the UK’s commercial exploitation of wind for the purpose of generating electricity in December 1991.

Hughes’ study contained no assessment of the performance of specific wind farms. All the data was merged into one large statistical series. On the basis of my assessment of actual production data from the earliest farms – all but two of which are still operating with the initial turbines – I want to suggest that the empirical evidence strongly suggests that Professor Hughes greatly exaggerates the rate of performance decline. None of the 14 wind farms shows ageing effects more than a small fraction of the figures he quotes. Investors and the general public can be confident that performance degradation is not a large problem.


I have two sources of data. First – with many thanks to Peter Edwards – I have the yearly output figures from Delabole from 1992 until the farm was ‘repowered’ with new, much larger, turbines in mid 2010 after nearly twenty years of production.

Second, I have the numbers from Ofgem’s database on the output of renewable generators. These numbers only go back to April 2002. (I have no idea how Professor Hughes could possibly have calculated the rates of decline of electricity output of twenty year old turbines when – at most – he only had ten years of figures).

We also have information on the average performance of UK onshore commercial wind turbines. DECC publishes a yearly estimate of the ‘load factor’ of existing wind farms. (The ‘load factor’ is the percentage of maximum yearly output actually achieved). Load factors vary – principally in response to average wind speeds. Professor Hughes’ work suggests that after accounting for wind speed variations load factors fall every year from the moment a new turbine is installed. This is what I wanted to check using real world data.


Chart 1 shows the yearly output from this Cornish wind farm from 1992 to 2009. (The repowering process started in mid 2010 so later output figures are not available).

Peter Edwards commented to me that the reason the 2009 figures appear to show a drop is that the operators of the wind farm (by then it was the utility Good Energy) decided it wasn’t worth replacing a gearbox because the turbines were scheduled to be taken down in less than a year’s time.

But even with the lower level of output in 2009, the average yearly decline was only  about 0.8% of output, not the 5% estimated by Professor Hughes. [1]2009 electricity production from turbines that were then 18 years old was 85% of the first year’s figure. In 2008 – when the turbines were still being actively repaired – Delabole recorded electricity generation of 99.6% of its initial annual output. Rather than output being more than halved, performance had fallen by a few megawatt hours a year.

Chart 1

Delabole 1

I don’t have UK average ‘load factors’ before 2001. Chart 2 shows how Delabole compared to the typical onshore wind farm in the years between 2001 and 2009. On average it was slightly lower, with a more marked difference in 2009 because of the lack of repairs to gearboxes. But the differences are small and there certainly isn’t any obvious sign that the performance was degrading against the UK average.

Chart 2

Delabole load factors


The oldest UK wind farms

If Hughes is right, then the oldest turbines should be very much less productive than the average UK figures. Of course wind farms established in the early 1990s might have been placed in particularly wind locations which might push their outputs upward. Balancing this, newer wind turbines could be expected to be better designed, and able to turn more of the energy from wind into useful electricity.

Chart 3 shows that the 14 oldest wind farms have load factors slightly below the UK average for the years 2001 to 2011. But there is no evidence of any widening of the differences. And, most importantly, the absolute level of output of these geriatric turbines is very much higher than Professor Hughes said. He wrote that turbines in their fifteenth year of operation should typically produce 11% load factors. In fact, these elderly wind farms – all of which were over eighteen years old in 2011 –  had average load factors of well over twice Hughes’ predicted output. They seem to have suffered more than expected in the historically highly unusual low wind speed year of 2010. (I suspect this is a consequence of better engineering for low air flows in newer turbine designs). But otherwise performance shows no relative decline from a decade ago.

Chart 3

load factors for pre 1994 and all wind farms



One last request. Anybody in active communication with Professor Hughes might want to ask him two questions. First, can he show us any individual wind farms that demonstrate the rate of deterioration his forecasts suggest? There were about 380 onshore wind farms recorded in 2012. The oldest 14 show nothing like the signs of ageing that Hughes grimly forecasts. Do any others? Are there any examples of farms whose wind-speed adjusted output has actually fallen 5% a year as he predicts?

Second, given that the outputs from wind farms are only publicly available from 2002, how is the Professor able to estimate exactly what the rate of decline in output of a twenty year wind farm is likely to have been? Because of Peter Edwards’ generosity in releasing Delabole figures to me, I can show that the decline of that single farm’s output is nothing like Hughes’s statistical forecasts. How did the Professor get to his numbers when he only had – at most – ten year’s data available for all the rest of the UK’s fleet of turbines?


(Please write to me if you are interested in seeing the data I used).

[1] This is estimated using simple linear regression.

[2] These calculations exclude Delabole and Goonhilly wind farms for the years after mid 2010, when both were repowered with new turbines. The other farms have unchanged configurations. The load factor I have used for the UK as a whole is also on an ‘unchanged configuration’ basis.

  1. Martin Normanton’s avatar

    Given that the reverse named REF exists purely to oppose windfarms cos they are ugly, why do we take anything they say seriously? Has this chap Hughes ever done any credible research?
    A similar case is Fred Singer in the USA,a bona fide academic who in the 90s challenged the idea that passive smoking caused cancer, and now gives very well paid speeches claiming that increased CO2 does not cause temperatures to rise.

  2. Paul D’s avatar

    ‘They seem to have suffered more than expected in the historically highly unusual low wind speed year of 2010.’
    I see no reason to assume that the wind speeds are constant every year, any more than the sunshine or rainfall records are constant. I wonder if Peter Edwards keeps the wind speed data for Delabole, as well as the generation numbers? If you put the wind speed data onto the generation graph and rescale them so they are close, then what does the correlation look like?

    Obviously outages of turbines, voluntary or compelled, need to be taken into account when producing a statement about deterioration of wind farm output. Is this the well known ‘bathtub curve’ for engineering equipment where manufacturing faults are exposed in the early stages and a gradual rise in ageing faults takes place towards the end of life.

    What I do not get is any sense of why a gradual deterioration of all wind turbines takes place, as seems to be implied in this debate. If all the turbines are degrading progressively, year by year, then what is the engineering mechanism for this degradation and why can maintenance not stop it?

  3. Michael Knowles’s avatar

    Chris – you have done a commendable exercise but the sample is rather small at 14 wind farms. Delabole of course has had some new turbines to replace the ones installed in 1991. Presumably Good Energy financed that out of its ongoing RO payments and were not given any additional grants?

    20 year is not a very long life is it? Old Nuclear and Coal fired plants are operating over 40 years and new nuclear plants like the over-priced EDF Hinkley C EPR will last for probably 60 years.

    DECC Energy Trends Dec 2013 published 9th Dec https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/266403/et_dec_13.pdf shows that Onshore LF average for 3 qtrs was 17% with offshore averageing 32.3% against DECC consultants’ reports for costings and targtes at 38%. Maybe Q4 will bring it up? However at high wind speeds >60 mph they shut down but still get paid for electricity they would have produced and more than the normal RO rate! This is a major design factor in the technology that should not allow the generators to be given special favours.

    The offshore wind industry is still asking for even higher payments than already proposed in the EMR delivery plan £155/MWh reducing to £135/MWh in 2017 all for 15 years.That in my opinion is quite high enough and nearly 4 times the current cost of generation!

    Happy New Year

    Michael Knowles CEng MIMechE , 50+ years experience in energy industry and co-author of
    (1) IMechE Energy Policy Statement 11/02 – UK Electricity Generation – Cost effective management http://www.imeche.org/docs/default-source/public-affairs/IMechE_Electricity_Generation_PS_Feb_2012.pdf?sfvrsn=0 April 2011
    (2) IMechE response to DECC consultation on RO banding levels for 2013 to 2017 http://www.imeche.org/docs/default-source/public-affairs/3236-consultation-ro-banding-response-form.doc?sfvrsn=0 19 January 2012
    (3) IMechE Response to DECC consultation on the draft Electricity Market Reform Delivery Plan September 25th 2013 —not published —

  4. Michael Knowles’s avatar

    Correction onshore average for 1st 3 qtrs of 2013 was24% ! apologies.

  5. Roger Anthony’s avatar

    Everyone with a computer and weather station, can see that every day, month and year is different. Because the wind varies so much from year to year, it is impossible to tell how a windmill is performing and wearing. We know that all mechanical devices wear over time, become more inefficient. Why would the gearbox and bearings on a windmill be different? Indeed the article mentions one mill out of action due to gearbox wear.
    Obviously, you have different people trying one the one hand, to justify the enormous costs of building, placing, connecting, running windmills along with their high built in levels of pollution. Others try to present a different view.
    Keeping in mind the long periods of none production, due to no wind and too much wind no one really has the slightest idea how productive each will be and how much wear it will receive in a given time.

  6. Wookey’s avatar

    As PaulD says, I really think these numbers need to take into account the annual wind numbers at the site, to get a usefully accurate comparison. What you have done is interesting and useful, but without normalisation against the actual windspeed it’s like comparing rainfall from one year to the next; that adds a great deal of noise to the data.

    Michael, you say that offshore farms are paid for non-generation when shut down during high wind periods. ROC payments are per MWh, so they don’t get any when shut down. Are you talking about constraint payments?

  7. Chris Goodall’s avatar

    Dear All,

    Thanks for the comments.

    A theme of the remarks is I should have standardised the output figures by using an index of wind speed. I can see why people write this: to know whether Delabole’s output really fell about 0.8% a year, we’d like to know whether the wind speeds in the first years were higher or lower than the last years of the old turbines.

    Two problems. First, I cannot find average wind speed figures earlier than 2001. Can someone help? Second, wind turbine output is not linearly related to wind speed. Turbines don’t start up until speeds reach 3 m/s or so and then output rises very sharply up to about 14 m/s when it stabilises at an approximately constant level. Then, at very high wind speeds, the turbine blades feather and output falls to zero. So average wind speeds are only somewhat helpful. What we actually need to know is how the amount of power generated at specific wind speeds changes over the years for particular turbines. For example, when the wind speed is 8 m/s does the amount of power fall as the turbine ages?

    I don’t know if we can get the data for this from any individual turbines? It would be very useful to have.

    Anyway, the reason I stuck with comparing the 14 oldest wind farms against the UK average was that the Hughes hypothesis is that there is a large, and apparently predictable, annual fall in output from year 1 onwards. Old turbines ought to both have a *much* lower output if the Hughes equation is correct and that output ought to fall every year (after adjustment for wind speed, of course). Neither of these conclusions seem to be true.


  8. Robert Wilson’s avatar


    I can suggest a slightly round about way of addressing this problem.

    Getting technical. If a wind turbine is functioning properly there should be a 1-1 correspondence between wind speed (m/s) and turbine output (kW). This means that you can derive a wind speed record from individual turbines, i.e. a bar chart of total number of hours each year a turbine experienced a particular wind speed. Of course for typical turbines if the turbine output is its rated capacity then the wind speed would be between 12 and 25 m/s.

    Long term changes in turbine output should show up in long term changes in this proxy wind speed distribution. For example, you could split the wind speeds up into bands of 3-3.5 m/s, …. 11.5-12.5 m/, > 12 m/s. How those bands shift over time might be quite interesting. Turbine performance could be fine up to say 8 m/s, but decline a bit above this, which would show up in a long term comparison. And looking at number of hours each year where output is zero would tell you if the ability of turbines to perform at very low wind speeds declines much.

  9. Paul D’s avatar

    1: The normalised load factor for UK onshore wind farms declines from a peak of about 24% at age 1 to 15% at age 10 and 11% at age 15’.
    It just is not credible that the blades are producing full power and the mechanism is absorbing half of it. The heat dissipation inside the nacelle would be 500kW for a 1mW turbine; that would fry everything. Half the turbines shut down is credible, if faults are not being repaired.

    2: Delabole – Chart 1 shows the yearly output from this Cornish wind farm from 1992 to 2009.
    If this is a continuous degrading mechanism, then how did the farm output rise almost continuously from 2003 to 2008? I do not give credence to self-repairing faults! Was it rising wind speeds or good maintenance, or something else?

    3: In 2008 – when the turbines were still being actively repaired – Delabole recorded electricity generation of 99.6% of its initial annual output
    That does rather make the point: good maintenance can keep output at very close to 100%. If all the turbines had degraded to their worst in 2001, then they were repaired by 2002 because the total output graph goes back to 10k Mwh.

    4: What is the evidence that these total output graphs imply that the load changes are uniformly spread over all the turbines?
    These curves are the aggregate generation performance of a fixed number of turbines. What was the underlying performance of each turbine that made up this aggregate? We do not know.

    Did they all degrade together and then improve together?
    That seems very unlikely to me unless maintenance caused the improvement.

    Was it windspeed acting on all of them?
    Was it some number shut down by protection circuits?
    Were the shutdowns temporary or protracted?
    Do they leave fixing dead turbines for better weather in the spring?

    There could be a lot of human response in these graphs, as opposed to just the turbines and the weather.

  10. Chris Goodall’s avatar


    Thanks for the point about 500 kW of heat dissipation. It is quite a lot for a nacelle to absorb……

    By the way, just in case anybody thinks that Professor Hughes’ work is simply wrong and therefore can be ignored, I’ve just seen a list of the ‘most read’ articles on a US energy web site (PennEnergy). The piece about the REF study came number 1….


  11. Michael Knowles’s avatar

    Chris – Thanks for admitting the Hughes’ article coming 1st in ‘most read articles’ on Penn Energy website.

    To partially answer your question on wind speeds earlier than 2001, does not the Met Office have them on its data base? The output from a wind turbine varies as the cube of the wind speed and I believe turbines cuts out at about 100kM/h or 60mph. Therein lie two big design problems. Added to that the low load factor and energy density of wind turbines being 2.5W per sq m according to Prof David MacKay article in Energy World some time ago, you have the mighty reasons for the difficulty in reducing subsidy costs.

  12. Steve Gilkes’s avatar

    So many errors here:

    Hughes’ statistical inaccuracies arise in the normalisation, which is impossible to repeat from the data he makes available.

    The unnormalised data shows that the only way that Hughes’ results could be achieved is if there had been a steady increase in wind speed year-on-year for a decade to the extent of a doubling of energy. e.g. the simplicity reveals that Hughes’ results are implausible.

    Wind farm operators only receive payments for electricity not supplied when they are asked not to supply the previously contracted power, just like all other types of fuel. If the turbines would have been shut down there are no payments.

    Neither the load factor nor the energy density are preventing onshore wind from being cheaper than nuclear. If we asked the nuclear operators to build offshore because it was unacceptable to build nuclear facilities 30 miles upwind of a city of 500,000, might lead to higher costs.

    By the way, the load factor of machines of the 90′s was lower simply because they tended to have higher ratings per swept area than more modern machines, which have relatively cheaper blades. A bit of work with a power curve shows that this also leads to a higher sensitivity to low wind years.

    Finally, when we quote 20 years life for wind turbines, one should think of this in terms of the replacement of some piece of plant within the building that is a coal or nuclear station. The point in comparison is that the Delabole wind power station started operating in 1992. After an equipment upgrade in 2010, the expected next refurbishment is in 2030, leading to life of 38 years SO FAR.

    Steve Gilkes, Wind Turbine Designer

  13. Paul D’s avatar

    Yes Chris, it certainly is, …

    … stuck on the top of a long pole sticking out of the sea-bed, you would be able to see the flames for miles and miles! I worked in ship machinery spaces and power stations for quite a few years. If any machinery was dissipating 0.5MW near me, then I would leg it as fast as I could. I might then just survive the resulting explosion! In case anyone thinks that I am joking, I tell the tale of the ship’s LOX (liquid oxygen) plant. That plant dissipated a bit of heat in the wrong place and blew up. The result was that two engineering watchkeepers died and the ship had a hole in it, fortunately well above the water line.

    Your last para is worrying. There is more than a chance that some of our American friends might believe this stuff and think that the Brits know what they are doing.

    OK, let us be serious for a moment. There is insufficient data to make any sense of this. I would want the wind speed data and the plant logs that detail exactly what all these windmills were doing for the last fifteen years; that includes the operational logs and the maintenance logs. I find it very difficult to believe that these windmills do anything else than run at a fixed power level for any given windspeed or are completely shutdown. The flat wind / power curve you talk about must be due to the feathering mechanism working and that mechanism can go wrong.

    I accept completely that the windspeed / power output graph would be far from a straight line (a cube law, according to Michael) for a fixed blade angle. If any machine had a fault in the feathering mechanism, then that is dangerous. If the machine was spinning, then the next gale could take it to destruction under centrifugal forces. If shutdown, could the shaft brake hold the torque with unfeathered blades?


    This is a bit like a heat pump in a house. That produces a square law graph of dT against pump power. Double the dT from (20-10)=10C for ambient at 10C and house internal temperature at 20C to a dT of (20-0)=20C for an ambient temperature of freezing and you require four times the power.

    At -10C outside, the pump power rises to three squared equals nine times that required at 10C. This is a considerable design problem as well because one heat pump cannot sensibly and economically cover the whole power range; you need at least two pumps or one pump plus a fossil fired boiler. I am still trying to figure out if two pumps plus a fossil fired boiler is economically sensible; it certainly is technically but does it produce an adequate return on capital?

    The idea of just one big, heavily subsidised, high temperature, retrofit, air-to-water heat pump is just crazy but that is exactly what the government is pushing at people by means of the draft domestic RHI document that comes up for approval in Parliament this autumn. Will Parliament really approve this junk? I note that DECC are to exclude any heat pumps that can exceed an outlet temperature of 55C from subsidy. I think that 40C would be a better number to get the ball rolling. That would leave air-to-air pumps and maybe the very best air/water-to-UFH pumps as RHI compatible. OK, I changed the subject a bit …

    … but back on the main subject, I see that the i, the small Independent newspaper, reports that :-
    ‘ … The European Commission is to order Britain to end wind farm subsidies. Officials have told ministers that the current level of state support for renewable energy sources must be phased out by the end of the decade.’
    Q: How do you get out of admitting a big mistake and changing your policy?
    A: You get the EU to tell you to change your policy.
    Simples! What’s the problem?

  14. Michael Knowles’s avatar

    Whatever the truth is on degradation of these wind turbines, as stated earlier the DECC’s own energy statistics clearly show that the total offshore turbine capacity needs to be increased by nearly 20% to achive the Government electricity output targets.

    Paul D – re your heat pump comments there is a church – St Mary’s Welwyn – that has put in borehole ground source heat pumps and used their old gas-fired boiler as back up/top up.

  15. Paul D’s avatar


    After having investigated DECC’s heat pump policy for months and the need for wind turbines, backup generators and grid reinforcement to drive them, I have been driven to conclude that the policy is based on two notions :-
    1: Carnot’s Law does not exist and you can therefore pump heat through any temperature difference you like without penalty. Using these Carnot indifferent pumps you can fit big, low COP, high temperature, air-to-water pumps to houses and produce identical electricity bills, as compared to the latest high COP (6.36, the best that I know of) air-to-air heat pumps (aka air conditioners).

    2: Joule’s Law of electrical resistance heating does not exist in DECCland either. This produces a situation not unlike ambient temperature superconduction. I know that you can get superconduction at the temperature of liquid Helium and a bit higher now but ambient temperature superconduction is definitely something that I have not heard of.

    With Joule’s Law dismissed, you can generate hundreds of gigawatts of electricity off the Scottish coast, using windmills, and send it south to London, without any problems that need concern a DECC civil servant. Electrical power engineers might not feel quite so relaxed about this re-engineered grid.

    3: In the real world, a COP of 3 means that a pump uses twice as much electricity as a pump with a COP of 6, to supply the same amount of heat in both cases. The deceit that it does not is being supplied by the EST (Energy Saving Trust), on behalf of DECC, to punters who do not know any different. The heat pump industry is also complicit in this. They really ought to watch out for the consumer mis-selling legislation when doing this; there are some big compensation claims coming, I reckon.

    4: The grid currently loses 3% of the electricity and the distribution systems 8%(National Grid’s numbers). Double the grid currents with wind turbines and with Joule’s law reinstated you lose (3+8)*4=44% of the electricity in the cables, wires and transformers without reinforcement. Even DECC ought to be able to see that the wind turbines would not be delivering to London as they should with 44% losses. To double the currents, you need to double the number of wires, cables, pylons and transformers just to hold the losses at their current level.

    5: All of this has been formally reported to the DECC Select Committee, the government via my MP and anyone who has read my posts on this blogsite. Yesterday, the government announced the end of DECC’s mad scheme that ignores the laws of physics. With the wind turbine building program stopped, then the ‘green’ energy to drive millions of big high temperature pumps also stops. That also means that the grid reinforcement nonsense stops.

    Fit a 1kW or two 0.5kW pumps in each of 27 million houses and you can solve 90% of the space heating problem with 10% additional generation, as compared to the DECC proposal. This should also mean the end of the draft domestic RHI policy as currently published.

    6: My own view is that yesterday’s announcement is the start of a whole new scheme, based on Carnot’s Law of heat pumps, Joule’s Law of electrical resistance heating and any other relevant law of physics. Cost saving to the country about £900 billion by 2050.

  16. Michael Knowles’s avatar

    Paul – I couldn’t agree more on the daft generaton of hundreds of gigawatts of electricity off the Scottish coast, using windmills, and sending it south to England not London, without any problems that need concern a DECC civil servant. I wonder what compensation the rest of the UK will require from a devolved Scotland for its 40% renewable programme back from the South? Or maybe they have sufficient back up in Scotland?

    When Torness shuts down there is another problem for Alec Salmond.

  17. Paul D’s avatar

    I thought it worthwhile to put this reference onto the thread :-
    Electricity Capacity Assessment Report 2013 – Ofgem
    27 Jun 2013 – Electricity Capacity Assessment 2013: decision on methodology …. installed capacity of wind power will more than double over the same period …

    Electricity Capacity Assessment Report 2013

    p83 Wind generation

    3.39. The source for wind speed data is NASA‟s Modern Era Retrospective-analysis for Research

    and Applications (MERRA) reanalysis dataset 99. This is a long term (1979-2012) dataset

    built up from analysis of remote sensor (satellite) data. The full dataset is global in

    coverage and contains information on all aspects of climate.

    3.40. For the purposes of this study, a subset of the MERRA data has been downloaded. The

    subset contains wind speeds at 2m, 10m and 50m height, for a grid covering the British

    Isles. The grid is at 0.5 degree longitude by 0.75 degree latitude which corresponds to

    approximately 50km spacing over GB. The model uses this data in combination with the

    capacity, hub height and coordinates of all transmission connected and embedded wind

    in GB.

    3.41. To access the raw data, University of Reading were contracted to build an extraction tool.

    The tool is written in FORTRAN90 and compiled into a form which can be run on a

    standard Windows PC. The detailed description of this tool is online100. The tool can be

    run stand alone to extract the wind speed data for an individual wind farm location over a

    specified period of time (1979-2012). The tool interpolates between local grid points to

    derive the wind speed for the specified location. It also adjusts for hub height using a

    logarithmic relationship.

    p84 Figure 57: Wind power curve in the 2013 and 2012 analysis104


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

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>