Wind Turbine Power Curves: Underperformance, or the Wrong Yardstick?
Plot any wind farm's SCADA output against the manufacturer's power curve and the turbines will almost always look short. Sometimes they are. Often the comparison is the problem: the warranted curve was measured in different air, at a different turbulence level, with a met mast instead of the anemometer sitting behind the rotor. Thin air on a hot, high site, turbulence, wind shear and a mis-set nacelle transfer function can each open a gap of several per cent with nothing wrong with the turbine. Correct the yardstick first. Whatever gap remains — yaw misalignment, pitch offsets, blade damage, hidden derates — is real, and worth chasing.
The power curve is the most quoted chart in wind. It sits in the turbine supply agreement, the energy yield assessment and the monthly performance report. So when the dots from SCADA fall below the line, the conclusion seems obvious: the turbines are underperforming. Sometimes they are. But the line and the dots were measured in different ways, under different conditions, and the difference between them is not automatically lost energy.
Before you blame the turbine, check the ruler.
Who this matters to
Wind asset managers explaining a gap between budget and actual, performance engineers deciding which turbines to investigate, and owners weighing a power curve warranty claim against a turbine manufacturer.
What a power curve actually promises
A warranted power curve is measured under a defined procedure — IEC 61400-12-1 is the international standard — using wind measured upwind of the turbine, typically on a met mast or with remote sensing. The results are normalised to a reference air density, binned by wind speed, and reported with an uncertainty. The 2022 edition of the standard brings wind shear, veer, turbulence and air density into the measurement explicitly, and treats the curve as specific to the climate it was measured in. Your SCADA comparison is usually nothing like that. It uses the turbine's own nacelle anemometer, behind the rotor, in your site's air, with your site's turbulence. Comparing the two directly is comparing two different measurements and calling the difference a loss.
Four reasons the curve and the SCADA disagree
- •Air density. Below rated wind speed, power is proportional to air density. Power curves are commonly quoted at the standard sea-level density of 1.225 kg/m³. On a hot day at 1,000 metres, air can be around 15% thinner than that, so a healthy turbine produces roughly 15% less power at the same wind speed below rated. Without a density correction, a high or hot site looks permanently short.
- •Turbulence. Power curves are built from ten-minute averages, and wind is not steady within ten minutes. Because power rises with the cube of wind speed, higher turbulence tends to lift output at lower wind speeds and round off the curve as it approaches rated power. The same turbine shows a different curve on a gusty site than on a smooth one.
- •Wind shear and veer. A large rotor spans a wide band of height. When wind speed changes strongly with height, or changes direction across the rotor, a single hub-height reading misrepresents the wind the rotor actually sees.
- •The nacelle anemometer. The anemometer on the nacelle sits in air the rotor has already slowed and disturbed. A nacelle transfer function converts its reading into an estimate of the free wind. If that function is wrong — or stops being right after an anemometer replacement, a controller update or a change in yaw behaviour — every comparison built on it shifts, on every turbine affected.
And the reasons that are real
Once the yardstick is corrected, the gaps that remain are worth money. A turbine yawed a few degrees off the wind captures less of it. A pitch angle offset on one blade or all three changes how the rotor converts wind into torque. Leading-edge erosion, dirt and ice degrade the blades' aerodynamics. And control settings — noise or wildlife curtailment, sector management, a derate left in place after maintenance — can cap output without any fault being raised.
| Cause | What it typically looks like | Measurement effect or real loss? |
|---|---|---|
| Low air density | Whole site short below rated; worse in summer and at altitude | Measurement effect, if uncorrected |
| Turbulence, shear and veer | Curve shape differs from the warranted one, site-wide | Mostly measurement effect |
| Wrong nacelle transfer function | Step change after a part or software change; same turbine model affected together | Measurement effect |
| Yaw misalignment or pitch offset | One or a few turbines short of their neighbours | Real loss |
| Blade erosion, soiling or icing | Gradual or seasonal decline below rated | Real loss |
| Hidden derates and curtailment | Output flat-topped below rated power | Real loss, unless contracted |
One site, one six per cent gap: an illustrative case
Numbers illustrative. A wind farm at moderate elevation reports summer output about 6% below the warranted power curve, and a warranty discussion is about to start. Correcting the comparison to the site's measured air density closes about half the gap. A further 2% traces to one turbine model whose nacelle transfer function was not updated after an anemometer retrofit; every turbine of that model shifted on the same day. What remains is about 1%, and it is not spread evenly: two turbines are well short of their neighbours in the same wind sector, one with a yaw offset and one with a derate left in place after maintenance. The warranty claim is dropped. Two work orders are raised instead — and they are the ones that recover energy.
Underperformance is not an outage
This distinction matters beyond operations. In NERC GADS Wind, event reporting is about equipment unavailability, not production below expectation. NERC's own guidance gives the example of a 100 MW plant expected to produce 75 MW that is producing 55 MW with every turbine running: that is not an event. If 20 MW of turbines are shut down by temperature limits, it is. A plant that confuses the two either over-reports outages or misses ones it should file. Ellume Bridge automates NERC GADS reporting. It classifies each event using the plant's own data, weather and operating context, computes hours and energy figures by arithmetic rather than AI, and routes every event through human review before filing.
A check you can run this month
Correct your comparison for measured air density before reading anything into it. Note the date of every anemometer replacement, controller update and transfer function change, and look for step changes on those dates. Then compare each turbine with its neighbours in the same wind direction rather than with the brochure: a turbine that is short of the warranted curve along with every other turbine is probably a measurement issue; a turbine that is short of its neighbours is probably a real one.
Frequently asked questions
- What air density are power curves quoted at?
- Commonly the standard sea-level density of 1.225 kg/m³, though manufacturers often publish curves for a range of densities. Use the one that matches your site's measured conditions.
- Can nacelle anemometer data prove a power curve warranty claim?
- Usually not on its own. Warranty verification normally follows a defined test procedure such as IEC 61400-12-1. Nacelle data is valuable for monitoring trends and comparing turbines with each other, and for deciding whether a formal test is worth commissioning.
- How do I tell a measurement problem from a real loss?
- Look at the pattern. Measurement effects tend to affect the whole site, or every turbine of one model, and often start on a specific date. Real losses tend to affect individual turbines relative to their neighbours under the same wind.
- Does underperformance count as a NERC GADS Wind event?
- No. Events relate to equipment that is unavailable, not equipment producing less than expected. Lack of wind and efficiency losses do not make an event on their own.
