All machines get less efficient as they grow older. Wind turbines are no exception to the rule. A new study shows that a turbine has an average ‘capacity factor’ of 28.5% when new and this falls to about 21% in the nineteenth year of its life. (1) This finding implies shows that the average wind farm loses just less than 1.6% of its expected output for each year that passes. Over a twenty year working life, a turbine will therefore produce about 12% less electricity than predicted by the manufacturers. Some of this decline is due to the turbine being out of action and awaiting maintenance more frequently later in its life. Another reason is simple wear and tear. These results are very different to those obtained by Gordon Hughes and published in late 2012. Hughes said that the rate of decline was very much faster, calculating that typical output of a wind farm halved by the fifteenth year, implying a rate of decline three times the speed of the new study. Hughes didn’t use estimates of actual wind speeds and experts such as DECC Chief Scientist Professor David MacKay have strongly criticised the statistical techniques he employed.
Iain Staffell and Richard Green of Imperial College Business School have produced an elegant and clear paper that is accessible to non-technical readers. Their most significant advance over the work of Gordon Hughes is that they incorporate estimates of the hourly wind speed at each of the several hundred UK wind farms. Since we know how much each type of wind turbine should produce at different wind speeds, Staffell and Green were able to calculate whether the performance deteriorated at time. If a turbine aged ten years produces 15% less power at a specific wind speed than it did when it was new, we can use this figure, along with many thousands more from that turbine, to calculate its rate of degradation.
Staffell and Green show that the 1.6% annual rate of output decline is fairly consistent among turbines of different vintages, and across the UK’s many wind farms, although they do suggest that the newest turbines may be performing better than predicted. Perhaps this latter finding is because of better maintenance in the first years of their lives when manufacturers offer performance guarantees. It’s also important to note that their findings are compatible with the real-life experience of wind farm operators, who were amazed at Hughes’ estimates of performance fall-off.
The wind speed estimates that Staffell and Green use aren’t perfect. Although each large wind turbine in the UK has an anemometer on its nacelle that measures and records wind data, this information isn’t made public. Staffell and Green were therefore forced to use a huge NASA database of wind speeds at low heights above the ground taken from weather stations, balloons, aircraft, ships, buoys and satellites. The resolution of this data is only down to squares of about 50km by 50km. However when the researchers looked at how well the NASA data predicted wind power output across the UK’s wind farms they found a very good fit. Their simulations of wind speeds in 50*50 km squares seem to give excellent predictions of power output from wind turbines inside those areas.
Gordon Hughes’ highly controversial 2012 study didn’t use wind speed data at all. In fact his model allowed wind speeds to rise across the last twenty years and used this increase as an input into the model. (Actually, if anything, UK wind speeds have tended to fall over the last couple of decades - at least until the last three months - so this was a very strange technique to use). The reason his research showed much higher rates of performance degradation is therefore that old wind farms, such as Delabole in Cornwall, appear in his model to be losing power because their output has stayed relatively flat, rather than rising with the higher assumed wind speeds in Hughes’ computer model. Hughes defends his approach by saying that it produces the best statistical fit. Critics have commented that any computer simulation that plugs in an assumed rise in national wind speeds that has not actually occurred is clearly inadequate.
Staffell and Green’s detailed analysis shows that turbine performance takes a dive in the last year or so before ‘repowering’, or the replacement of an old machine with a newer, and often much bigger, version. This is also consistent with the real world experience of wind farm owners who reduce maintenance as the wind turbines approach the point of being taken down. It’s far cheaper to repair old machines on the ground prior to reselling them into the second hand market.
The implications of this new study are important. Surprisingly, the financial models used by investors to plan wind farms seem to generally exclude any figure for performance degradation. The loss of power output in later years raises the cost of electricity derived from the turbines. The increment is small – no more than 9% - but it needs to be factored into the calculations about the true cost of wind power.
This isn’t necessarily a comfortable finding to financial people who had assumed that wind turbines had no perceptible performance decline. But Staffell and Green’s comprehensive and lucid work will for the first time provide the industry – and society at large – with proper estimates of the lifetime power output of a wind farm. And, as Gordon Hughes originally suggested, it will mean that a bigger than expected fleet of wind turbines will be needed to provide the UK’s desired electricity output from this source. If the UK does achieve 30 GW of wind power by 2020 - an increasingly unlikely target as offshore operators rapidly retreat from their projects - this will mean installing an extra 435 MW a year, or four large new farms, to counteract the ageing of the fleet.
(1) A turbine's capacity factor is its actual output as a percentage of its maximum yearly production if the wind were to be blowing strongly all the time.