I’m always reviewing what I’ve written and if new evidence comes to light I’ll update accordingly. A group of us have been discussing my results about how wind turbines lose efficiency due to wear and tear. Several of the group had some reservations about my numbers being so large, on the order of 1.5 – 2% per year decline in Capacity Factors. They surprisingly large to me as well, but I will generally go wherever the numbers take me.
One of the group, Paul-Frederik Bach, ran a lengthier analysis of Denmark’s wind turbines, running 10 “cohorts” through the years since their installation, with each cohort made up of the turbines placed in service during a particular year. He didn’t correct for the wind speed during each year, but over that long of a period (18 years total) you’d think the wind speeds would average out enough so that his analysis would be reasonably accurate. In place of my 1.5%, he averaged 0.38%.
None of us doubt there’s a decline, but how big is it? And how could our numbers, both figured using reasonable (although different) techniques, come to such different results? The answer, at least for Denmark, may lie in the period I had to use to do my (well, originally, John Harrison’s) technique. One of my requirements is that the total number of turbines doesn’t change appreciably during the period I’m analyzing. In Denmark’s case, only the 5 years 2004 through 2008 qualified. I went back to 2000, calculating the average wind speed through 2011, using Copenhagen’s daily average data. This is the result:
Note that the years I considered show up as 5 through 9 above and note their trend – sharply increasing. You’d expect the Capacity Factors to increase along with the cube of the increasing wind speed, but they didn’t. And so when I normalized the CF’s they showed a decline which I consequently attributed to wear and tear. Bach’s numbers are over a longer horizon, when the wind, if anything, was in a downward trend.
So why might my numbers be too large? Well, in a windier year there may be a significantly larger number of hours when the wind turbines are maxed out, and their output is no longer increasing with increasing wind speed. Typically turbines are in this condition about 10-15% of the time, but in a higher average wind speed year that percentage may become enough larger that my technique starts attributing a turbine’s declines due to normal operating limitations to declines due to wear and tear.
So we know for sure that my numbers are off? No, we don’t. And that they’ve held up in widely different environments, with widely different average Capacity Factors, would be a tick in their favor. How could we figure this out for sure? The only way I see would be to work on much finer time increments, like hourly. Mars Hill and Ontario were monthly, Denmark was annually. Unfortunately, that kind of data is quite difficult to obtain, except maybe in Ontario. And even there it would take quite a bit of effort, plus Environment Canada tends to shut down overnight in remote locations (Ontario considers “remote” pretty much anywhere but Toronto!).
So this story isn’t finished yet. After all, this is how science works. I’ll continue analyzing projects as I can, and as always go where the numbers take me.