A few degrees of yaw

Power available to a rotor falls with the cosine of the angle between the wind and the rotor axis, raised to a power that theory puts at three and field measurement generally puts lower. Take the conservative end. At five degrees of static yaw misalignment, a cosine-squared relationship gives a loss of about 0.8 percent; at ten degrees, about three percent. Those are small numbers per hour and large numbers per year, and they are paid in every hour the machine runs.

What makes static misalignment worth singling out is that nothing in the operating chain reports it. The turbine is not faulted, availability is unaffected, and the controller believes it is pointing correctly, because the offset is usually in the relationship between the wind vane and true wind direction rather than in the yaw system's execution. The machine is doing exactly what it was told, slightly wrong, indefinitely.

It is also unusually cheap to fix. There is no crane, generally no intervention at height, and in many cases the correction is a controller parameter. That combination - a persistent loss with a low-cost correction - is why alignment findings routinely sit at the top of a ranked work list even though the per-hour effect is smaller than a fault. Value at stake is a product of magnitude and duration, and duration here is measured in years.

Detecting it from operational data is a matter of looking at the residual rather than the average. Bin production by the difference between wind direction and nacelle position, and a correctly aligned machine peaks near zero. A misaligned one peaks somewhere else, and the offset is readable directly off the shape. Doing this reliably needs the nacelle position channel, which is one of the reasons it appears in the SCADA record rather than among the optional signals.

The honest complication is that nacelle wind measurement is distorted by the rotor itself, so an apparent offset can be an artifact of where the sensor sits rather than a real misalignment. This is a genuine limit and it is why an alignment finding should be verified before a controller is touched: on most sites the confirmation is a lidar campaign or a spinner anemometer, and where neither is available the correct output is a ranked list of candidates rather than an instruction.

There is a second-order effect worth knowing about. A yawed rotor deflects its wake, which changes what the machines behind it receive. Correcting one turbine can therefore move production around a site rather than only adding to it, and any claim about recovered energy that ignores the wake interaction is an overestimate. It is usually still an improvement; it is rarely the full figure a single-turbine calculation suggests.

The action this suggests is small. Pull twelve months of history for one site, plot production against the wind direction minus nacelle position, and look at where each turbine peaks. Machines whose peak sits several degrees off zero, consistently and across wind speeds, are candidates. It costs an afternoon and it is the cheapest energy most sites have available.