Following a recent and unfortunately misleading article in the Sunday Telegraph (08.12.13) by Christopher Booker there has been a certain amount of renewed comment on Professor Gordon Hughes’ work concerning the economic lifetime of wind turbines, work that used empirical performance data gathered by REF. This comment includes blog postings by Bishop Hill and Roger Helmer MEP, to which Professor David Mackay, Chief Scientific Advisor to the Department of Energy and Climate Change has responded personally with online comments, by releasing a draft paper criticising Professor Hughes’ work, and also with an extended blog posting on his own site.
The following statement comments on the extent and significance of the disagreement with Professor Mackay. For the avoidance of doubt, readers should be aware that Professor Hughes does not accept Professor Mackay’s criticisms of the statistical methods employed in his analysis, criticisms which he regards as ill-informed, and unconstructive.
In December, 2012, REF published Professor Hughes’ preliminary research, The Performance of Wind Farms in the United Kingdom and Denmark (2012) which showed that after allowing for variations in wind speed and site characteristics the average load factor of wind farms declines substantially as they get older, probably due to wear and tear. This decline in performance would mean that it is rarely economic to operate wind farms for more than 12 to 15 years, after which they would have to be replaced with new machines.
As a result of this research, we concluded that the structure of contracts offered to wind generators under the proposed Electricity Market Reform (EMR) should be modified since few wind farms will operate for more than 12–15 years. Subsequent to this work, the Government did in fact limit EMR contracts to 15 years, though it is not clear that this was a tacit admission of the accuracy and significance of Hughes’ analysis.
Discussions with Professor Mackay of DECC
Following publication in December 2012, Professor David Mackay, Chief Scientific Advisor to the Department of Energy and Climate Change (DECC) corresponded with Professor Hughes and REF concerning the results of the study. Professor Mackay’s initial position was that the results were wrong and that the input data could not be interpreted as showing a decline in wind farm performance with age. However, in the course of discussions, it became clear that he agreed that there was in fact a decline in performance over time, though he disputed the rate detected by Professor Hughes, and raised certain methodological issues. We were shown a draft paper (Draft 4, dated 11.02.13) summarising Professor Mackay’s views, to which Professor Hughes responded, and which was discussed at a meeting in REF’s offices on the 21.02.13.
At this meeting the question of whether the methods used by Professor Hughes were, in the technical statistical sense, ‘identifiable’, was resolved, when Professor Mackay appeared to agree that they were so. This point, and several others, are summarised in a letter sent by John Constable, director of REF, to Professor Mackay on 26.02.13.
A later draft of Professor Mackay’s paper is now publicly available dated May 2013, and he has confirmed this material in his blog posting on his Sustainable Energy - without the hot air site. For reasons that are not clear he has reverted to the earlier position regarding the identifiability of the methods, though in other respects his position is much as it was in February 2013, when he wrote to REF: “I agree that when it comes to the graphs, we are mainly disagreeing about the size of an apparent performance decline. And to be quantitative about it, we’re comparing (in my language) your 5.1% per year decline (or more, in the case of figure 10), with my “roughly 2% per year”.
Professor Hughes’ responded to both these questions, identifiability of the methods, and the rate of the decline, in a document dated and sent to Professor Mackay on 14.02.13.
Professor Hughes explains that his analysis goes beyond that attempted by Professor Mackay, and that this fact explains the differences the conclusions reached:
Clearly there are differences in our estimates of the rate of decline but there is no doubt that the performance of wind farms does decline with age. What is left is the task of assessing the average rate of decline and how this varies with other characteristics of wind farms. My analysis goes beyond Professor Mackay’s in two respects.
First, there seems to be a clear indication in my analysis that the rate of decline in performance accelerates as wind farms get older, so that models which impose a constant rate of decline perform less well than ones which do not include this restriction.
Second, my analysis indicates that larger wind farms – and, probably, larger wind turbines – experience a more rapid decline in performance than smaller ones. This is tied up with the issue of capacity weighting. Professor Mackay’s analysis can shed no light on this issue because there were only 3 wind farms with a capacity > 30 MW in existence in 2004 whereas there were 27 that commenced operations after 2004 including all wind farms of > 60 MW capacity. Large wind farms tend to rely upon large wind turbines (with a capacity of > 2 MW) which only came into widespread use after 2002. Work that I have carried out since my REF paper was written suggests that wind farms with large wind turbines have a distinctive pattern of performance. They have higher load factors than wind farms with smaller turbines up to age 4 and then their performance starts to decline much more rapidly so that by age 8 they are no better than the average and the rate of decline suggests that they will be significantly worse by age 10 and thereafter.
The idea that the performance of wind farms declines with age is regarded as perfectly normal by academic and other independent engineers. It is true for other electro-mechanical equipment subject to large stresses, so why would wind turbines be any different? There is an important question of whether the rate of decline can be slowed by adopting the best operational and maintenance practices. Since I have no access to data on O&M expenditures or regimes, I can only assess rates of declines in performance under the average O&M regime.
In my presentations and my paper I have consistently emphasised that my work should be the beginning of a research program. I believe that my results establish that there is a strong a priori case for believing that the decline in the performance of wind farms with age is sufficiently large to be a significant factor in the economics of wind generation. Professor Mackay’s results seem to reinforce that conclusion. As more data on the output from wind farms is accumulated it should be possible to strengthen the empirical analysis of the performance of wind generation, especially from age 8 onwards, which is critical to assessing the longevity of and returns to investment in the sector. Technical issues of identification and estimation matter, but they are routine and easily dealt with. What we need to focus on is how we can take better account of factors which may affect performance but which could not be incorporated in my models. An example is variations in wind resources, which both I and others are attempting to examine in more detail.
Professor Hughes’ seminal analysis has highlighted an issue with wind farm lifetimes using advanced but standard statistical techniques designed to deal with data of this sort and we believe that the conclusions drawn are informative. In particular, Professor Mackay’s suggestion that considerations of timing or the imposition of key constraints generate spurious results contradicts a large body of standard approaches to identifying statistical relationships in modern econometrics and statistics. It would appear that Professor Mackay is not deeply familiar with this field. Readers are referred to Professor’s Hughes’ remarks attached below.
This technical debate aside, the key finding of Professor Hughes’ paper stands, namely that wind turbine performance declines over time, which is an original finding not before published. Indeed, as a result of the Hughes paper, Professor Mackay himself now accepts that there is such a decline. Furthermore, it must be noted that his own, lower, estimate of the rate of the decline is still economically significant.
Since we now know that the economic lifetime of a wind farm in the UK is likely to be very significantly less than 25 years, further questions arise in relation to energy policy. For example, planning permission for wind farms is routinely given for 25 years, and a wind farm is entitled to the Renewables Obligation (RO) subsidy for 20 years. Given the likelihood of shorter economic lifetimes, it would seem that both planning permissions and RO subsidy entitlement are needlessly extended. We have already noted that the incoming replacement subsidy mechanism, Feed-in Tariffs with Contracts for Difference gives an entitlement period of 15 years.
Interestingly, Professor Hughes’ research comparing rates of decline in the UK, in Denmark and in various US states has revealed that the rate of decline differs from country to country, suggesting mitigating factors. These factors might include O&M regimes, wind farm layouts, particular turbine models, and the economic incentives of particular subsidising policy regimes. Determining which of these factors are relevant goes to the heart of ensuring that the UK consumer is getting value for money out of the very generous subsidies. The information required to resolve these issues is in the hands of the wind farm operators but is available neither to the Government nor to the subsidising consumer. A transparent debate based on all the available data and the best analytical tools is needed to ensure we are not wasting public money.
Professor Mackay has made considerable efforts, first to persuade us to withdraw Professor Hughes’ paper, and now publicly, and on dubious grounds, to discredit work which is obviously original and draws attention to a previously undiscussed phenomenon, the decline in load factor over time, that was not acknowledged, for example, in the Department of Energy and Climate Change’s own levelised cost estimates for wind power. This is extraordinary behaviour for a Chief Scientific Advisor to government. Rather than shooting the messenger, Professor Mackay might more fruitfully be advising government on how best to ensure that consumer gets better value for their subsidy, and that we present a more economically compelling example of the low carbon economy to the developing world.
This letter was sent to Professor Mackay as a summary note of the meeting between Professor Hughes, Professor Mackay, an un-named civil servant, and Mr Ian Staffell of Imperial College, who came as part of Professor Mackay’s party.
Appendix B: Response of Professor Gordon Hughes to Professor David Mackay’s comments on his paper (14.02.13)
The following paper was written in February 2012 and in response to Draft 4 (11.02.13) of Professor Mackay’s critique of Professor Hughes’ paper. The substance of the arguments remains relevant to Draft 6 (28.05.13) of Professor Mackay’s paper.