Tired site or tired turbine: what your own data can tell you before you spend anything
A falling farm total cannot say whether the wind changed or the machines did. The farm's own records, read machine by machine, usually can, and most owners already hold them.
When the wind falls, it falls on every machine on the site at once. A machine that falls behind its neighbours, and stays behind, is telling you something the wind cannot.
That is the whole method in two sentences, and in my experience most owners of old wind farms never use it. They judge the farm by its total: this year against last year. When the number has been falling for a few years, the conclusion arrives on its own. The site is tired.
Sometimes it is. More often, on the old sites I see, the wind is where it always was and something on the site has changed. This piece is about telling the two apart with data most owners already hold, before anyone spends money on a new contract, a new machine or a measurement campaign.
Why the farm’s total cannot answer the question
A year’s generation is several things added together: how hard the wind blew, how many hours each machine was able to run, how well each machine turned the wind it got into power, and whether all of it was reported.
A weak wind year, a machine that has quietly lost performance, a machine that stopped and a gap in the reporting all show up the same way, as a smaller total. Lining up more years of totals does not separate them. It only makes the decline look more like a trend.
That is why the comparison that works is not year against year. It is machine against machine.
Read each machine against its neighbours
Start with the monthly generation of every machine on the farm, for as many years as the records go back. It sits in the turbines’ own SCADA records, and an owner can usually ask the servicer for it. In my experience, few owners ever ask for it in that form.
Then compare each machine with its nearest neighbours of the same model, month by month. Not with the farm average, which the weakest machines drag down, but with the machines beside it.
Neighbours never match exactly. Wakes, terrain and spacing mean some positions on a site always do better than others, and the pattern moves with the wind’s direction through the year. That is expected. What you are looking for is not a difference. It is a change in the difference.
A machine that has always made a little less than the one beside it is telling you about its position. A machine that used to keep pace and now falls behind, month after month, is telling you about itself. A weak wind year moves the whole row together. A worn gearbox, a blade whose leading edge has worn, a rotor that no longer points squarely into the wind: those move one machine.
Compare the same month across different years, so the season is the same. Three patterns matter. A machine whose share of its row’s output has fallen and not come back is a question about that machine. A machine with months missing, or flat at nothing, is a stoppage or a gap in the reporting, and both need explaining. The whole row falling together is the one pattern no comparison between neighbours can explain.
The first two are questions about machines. The third is the one that needs the wind measured.
Generation says that a machine fell behind. The availability figure in the monthly report is meant to say why. Before reading it, ask which figure it is, because every availability number answers two questions.
The first is time or energy. Time-based availability counts the hours a machine was able to run. DNV, in its paper on how availability is defined, notes that it “may not capture the added importance of a turbine’s availability during high wind periods”. Production-based availability counts energy instead: what the machine made, against what it could have made in the wind that actually blew. For an owner it is the more useful number, and it needs something the time-based figure does not, an estimate of that wind.
The second is what is left out. Contractual availability is the figure the service contract defines, and DNV says such figures “typically have certain ‘carve-outs’ or allowances”. It also says contractual values “often cannot be used directly as energy loss factor”. Technical availability takes no allowances, and DNV calls it “a more accurate measure of technology performance”.
So the figure that belongs beside a machine’s generation is the technical, production-based one. If a report gives only a contractual, time-based number, it is answering the contract’s question, not yours. IEC 61400-26-1 defines both time-based and production-based indicators, so asking for the second is not an unusual request.
What NIWE’s 15 per cent line says, and what it does not
NIWE’s screening rule, as CSE reports it, is that any wind machine with a capacity utilisation factor below 15 per cent is technically ready for repowering, and that its CUF “can be at least doubled, or in wind-intensive sites, tripled”.
It is a useful line, and it is easy to misread. It is a rule about machines on land that has the wind. On such land, an old machine below 15 per cent is leaving energy behind. The line does not tell you that your land has the wind. A machine below 15 per cent on a weak site is not leaving anything behind; it is getting what the site has.
So the line tells you where to look, not what you will find. Read machine by machine, it picks out the machines worth a closer look. Read as a farm average, it hides them among the rest.
When is the farm’s own data no longer enough? When the whole row has fallen together, so the neighbours have nothing to say. When the decision on the table costs real money, a new contract or a repowering. And when the only wind figure anyone has comes from the turbines’ own nacelle sensors, which describe the conditions the machine itself sees.
Then the wind needs measuring independently: a mast or a lidar on site, at hub height. A wind resource assessment needs on-site measurement over “a minimum of a one year period”, with the short record then extended to the long term against historical data. It is not cheap, and it is the one thing that settles the question when the machines cannot.
When the site really is tired
Sometimes the measurement comes back and the wind is not there. The original assessment was optimistic, or what stands around the farm has changed, or the site was always marginal.
Then the right answer is not a new machine. I will not recommend repowering a site where the wind does not justify it. A larger turbine on weak land is a larger cost on weak land.
Sometimes the right answer is to sell. It is not an answer anyone selling repowering likes to give. It is still the answer the measurement supports, and an owner is better off hearing it before the money is spent than after.
The objection
The fair objection to all of this is that neighbour comparisons are crude. Wakes, terrain and different machines make neighbours differ, the records are often messy, and only a proper assessment counts as evidence.
All of that is true. The comparison is not a measurement, and it should never be presented as one. What it detects is change: a machine drifting away from neighbours it used to keep pace with. Wakes and terrain do not do that on their own, because they stay where they are from one year to the next, unless something new has been built nearby.
And the comparison costs nothing. The records already exist. Reading them tells you whether the expensive step, the measurement, is worth taking, and where on the farm to point it.
What to do before the next decision
Before the next decision about an old farm, whether to renew the contract, repower or carry on, ask for three things. Every machine’s monthly generation, for as many years as the records go back. The technical, production-based availability beside it. And, if the whole row has fallen together, an independent measurement of the wind.
If one machine falls behind its neighbours, the problem is on the site, not in the sky. If they all fall together, measure. And if the measurement says the wind is not there, believe it.
Pavansara Renewables’s page
Senthil Kumar Ramasamy’s profile