Wake is a fleet problem, not a turbine problem

A turbine operating in the wake of another sees a slower, more turbulent flow, and produces accordingly. Depending on spacing, atmospheric stability, and direction, the deficit can run from a few percent to well over ten in the affected sectors. None of this is a defect. It is a consequence of the layout, and it was in the energy yield assessment before the site was built.

The failure mode this creates for a performance system is not a missed detection but a permanent false one. Score every machine against the site's best performers and the downstream rows are flagged forever. The findings are statistically sound, they will reappear every month, and there is no intervention that closes them - which is precisely the pattern that teaches an operations team to ignore a system. A finding nobody can close does more damage than no finding at all.

The correct treatment is conditional. A turbine's expectation has to depend on wind direction, because that is what determines whether it is in clean flow or behind two rows of machines. Once direction is part of the model, a wake deficit is expected in the sectors where it occurs, and what remains detectable is a change: a machine that is now four percent below its own wake-adjusted expectation in the sectors where it was previously matching it.

That conditional structure also makes a second class of finding possible. If a wake deficit is larger than the layout implies, the interesting question is why - and the usual answers are an upstream machine yawed a few degrees off, or a control configuration that is more aggressive than the design assumed. Those are findings about a different turbine than the one showing the deficit, which is a category conventional per-asset reporting has no way to express.

The honest limit is that wake modeling from operational data alone is coarse. Atmospheric stability drives a large part of the variation in wake recovery and is not recorded by SCADA; nor is turbulence intensity, on most machines. A model conditioned on direction and speed captures the first-order structure and does not reproduce a computational fluid dynamics study, and it should not be presented as though it does.

The consequence for the operator is a reporting decision rather than a modeling one. Structural deficits should be separated from recoverable ones on the face of a finding, because they lead to different actions: a recoverable loss becomes a work order, and a structural one becomes an input to a repowering or layout decision. Merging them produces a total that is larger, more impressive, and unusable.

Worth doing on your own data: pick the most heavily waked row on a site and plot its production by direction sector against its own history rather than against the front row. If the deficit is stable across years, it is layout. If it has grown, something changed, and that is a question with an answer.