Engineers turned wind turbines slightly away from the wind and unlocked a major power boost, but one seemingly minor mistake could erase the entire gain
AI-madeWake steering has long looked like one of wind energy’s most promising free gains: deliberately point a turbine slightly off the wind, redirect its shadow away from the machines behind it, and lift a farm’s total output by double digits — no new hardware required.
New high-fidelity simulations suggest the technique works, but only within a surprisingly tight margin. According to the research, a bias of roughly ±3–4° in the assumed wind direction — about the angular width of a hand held at arm’s length — can be enough to erase a predicted 23% power gain and push the farm into a net loss. Real wind farms routinely operate with direction uncertainties well beyond that threshold.
What is wake steering — and why does it matter?
Wind turbines operating downwind of others pay a real performance penalty. The upstream machine extracts energy from the air, leaving a slower, more turbulent wake that cuts power output for every turbine behind it and accelerates blade fatigue. In a tightly packed offshore farm, those losses compound row by row.
Three yaw configurations were tested: an unyawed baseline, the lead turbine alone yawed 20° anticlockwise, and the first three turbines all yawed 20° anticlockwise.
Wake steering tries to sidestep the problem. By rotating a turbine’s nacelle a few degrees off the true wind direction — a maneuver called yaw misalignment — engineers can deflect the wake sideways, steering it clear of the rotors downstream. The yawed turbine takes a small power hit, but the machines behind it gain more than enough to offset it. At least in theory.
Under ideal conditions, the numbers look compelling. Simulations and field trials have shown farm-level power gains exceeding 20%, with zero capital cost — a combination that has made wake steering one of the most actively pursued wind farm control strategies of the past decade. The catch is precision: because the yawed turbine deliberately sacrifices some output, downstream gains need to be accurately targeted.
A high-fidelity simulation built to stress-test the real world
To probe how wrong things can go, researchers at the Technical University of Denmark (DTU) ran large-eddy simulations (LES) of four turbines arranged in a row. LES resolves turbulent flow structures directly rather than approximating them, making it more computationally demanding than standard engineering models — and more realistic.
The simulated atmosphere represented offshore conditions in northern Europe, with wind shear and a mild wind veer of 4.5° across the rotor area. Turbulence intensity at hub height was 3.8%, and wind speed was below rated. Three yaw configurations were tested: an unyawed baseline, the lead turbine alone yawed 20° anticlockwise, and the first three turbines all yawed 20° anticlockwise. The team simulated six different mean inflow directions by rotating farm coordinates while keeping everything else identical.
Peak gains are real — but the window is narrow
The results confirmed that wake steering works — within a specific range. With a wind direction uncertainty standard deviation of 2.3°, yawing all three upstream turbines produced a peak farm power gain of 23% when the mean wind direction sat at roughly 2° off the turbine row axis. Yawing only the lead turbine delivered a more modest 7.5% peak gain over a similarly narrow band.
The range of wind directions over which the three-turbine strategy produced any net gain spanned only about 8.5°. Outside that corridor, the same yaw angles that boosted power became a liability. Atmospheric veer alone introduced an 8% asymmetry in baseline farm power between mirror-image wind directions — a reminder of how sensitive these systems are to directional details that simpler models often ignore.
When uncertainty grows, the gains shrink — fast
The more consequential finding emerged when the team applied Gaussian convolutions to simulate higher wind direction uncertainty. Reported values from real sites typically range from 2.5° to 6°. Stepping up to a standard deviation of 4.5° cut the three-turbine peak gain from 23% to just 7.5%, while the one-turbine strategy’s peak collapsed from 7.5% to 2% — a margin barely distinguishable from measurement noise.
A mean wind direction bias of just ±3–4° — the kind of error that arises from assuming a single wind direction across a large farm, or from sensor drift — was enough to erase all predicted gains and push the farm into a net loss. Spatial variability in mean wind direction of roughly ±2° has been documented on scales of around 10 km, and ±4° on scales of around 50 km, according to New Europe Wind Atlas data cited in the study. The asymmetry matters too: power losses outside the beneficial window decreased far less than peak gains shrank as uncertainty grew.
Load reductions offer a silver lining — with caveats
One finding complicates the narrative of wake steering as purely a power tool. Partial wake conditions — where only part of a downstream rotor sits inside an upstream turbine’s wake — produced the highest flapwise blade fatigue loads in the simulation. Applying wake steering in those scenarios reduced fatigue loads on downstream turbines even when total farm power decreased. At a wind direction offset of 8°, yawing three turbines caused a net power loss yet still delivered a meaningful reduction in flapwise damage equivalent loads.
This suggests wake steering could serve a secondary role as a load-management tool, deployed to extend turbine lifetimes rather than maximize output. The researchers note, however, that only flapwise blade root loads were assessed — edgewise bending and tower loads were not evaluated.
What this means for wind farm operators
The practical implications are clear: wind direction measurement quality may be the binding constraint on whether wake steering delivers its promised value. Using a single farm-wide wind direction estimate introduces exactly the kind of mean bias the study identifies as problematic. Local measurements at each turbine — using nacelle-mounted sensors or nearby lidars — are more appropriate, though those too carry calibration challenges when the nacelle is yawed.
Setting a minimum predicted-gain threshold before activating wake steering is one practical safeguard the study points toward. Approaches such as axial derating or “helix” active wake mixing may also prove more robust to wind direction uncertainty, though higher-fidelity validation is still needed.
You can check the full study here: Hodgson, E. L. and Andersen, S. J.: Wake steering under inflow wind direction uncertainty: an LES study, Wind Energ. Sci., 11, 2173–2190, https://doi.org/10.5194/wes-11-2173-2026, 2026.
Carlos is an engineer with strong expertise in technical and industrial topics. He previously worked at international companies such as Siemens and is multilingual.
