In December 2012 REF published Professor Gordon Hughes’ paper The Performance of Wind Farms in the United Kingdom and Denmark.
This seminal paper used advanced but standard statistical techniques to estimate the rates of decline in performance over time. Professor Hughes found that the economic lifetime of wind farms in the UK was very much shorter than currently estimated, at between ten and fifteen years, with those in Denmark faring better but still declining significantly.
The study has attracted a great deal of attention, and its fundamental finding is not now contested. Decline in performance is real and should be taken into account by policy makers and investors since it has significant implications both for the subsidy cost of the current renewable energy targets, and for the levelised cost of wind electricity and thus the rate of return to capital.
The study has been the subject of considerable criticism, particularly with regard to the methods employed, and Professor Hughes has responded and defended his approach (for example here in response to Professor MacKay, Chief Scientific Advisor to the Department of Energy and Climate Change and here in response to remarks by a blogger, Mr Chris Goodall.
Other researchers have begun to follow the lead given by Professor Hughes, and Dr Iain Staffell and Professor Richard Green of the Imperial College Business School have recently published a paper in the journal Renewable Energy.
This study uses different methods, and though it confirms the finding that performance declines significantly over time, it reports a lower rate, of about 1.6% per year, with “average load factors declining from 28.5% when new to 21% at age 19.” Staffell & Green note that this has the effect of increasing the levelised cost of electricity from wind by 9%.
REF believes that the debate has moved on from whether wind farm performance declines with age, as that is clearly confirmed, to the technical issues involved in measuring the rate and profile of the decline. The Staffell and Green paper proposes an approach that compares actual output with ideal output based on interpolated wind data. The potential difficulties of this specification are outlined in the Appendix to Professor Hughes’s paper and are not resolved in the new paper. Hence, we are not convinced that the Staffell and Green’s approach is superior to the statistical methods employed by Professor Hughes.
In one respect, we believe that the Staffell and Green analysis is clearly inferior to the Hughes specification. Namely, their model does not allow for the effect of technical progress in the wind industry through improvements in operating and maintenance practices, which can be treated separately from changes in turbine design. This was built into the Hughes specification and few would dispute that such technical progress has occurred in the past.
Work carried out by Professor Hughes – and made available to Dr Staffell and Professor Green – shows that the rate of such technical progress has fallen markedly since 2008. The reason this matters is that Staffell & Green and Hughes have both confirmed that specifications that allow for such technical progress yield higher estimates of the rate of decline in performance with age than those, including the new paper, which do not. In statistics this is known as omitted variable bias.
In summary, REF’s view is that while the new paper makes a useful contribution to public discussion of the performance of wind farms, Staffell and Green’s methods are not statistically superior to those used in the Hughes study. Indeed, Staffell and Green’s specification contains an assumption that results in a significant downward bias in the estimate of the rate at which performance declines with age. We will provide fuller details of these points in due course.
24.02.14