Why can a wind turbine spin faster but deliver less electricity?

Two very different turns

The rotor goes around its main shaft. Each blade can also turn around its own long axis. That second movement is blade pitch. Turning all three blades together is collective pitch. Nacelle yaw changes the direction the rotor faces.

The generator pushes back

Torque means a turning effect. Air can apply torque to the rotor. A working generator applies a resisting torque as it receives mechanical energy. Two equal opposing torques can hold the speed steady while energy keeps moving through the machine.

Faster does not mean more useful

In the disconnected-load trial, electrical output becomes zero and incoming work increases the rotor’s stored rotational energy. With balanced load, speed stays steady and electricity is delivered. Rpm alone cannot tell you which is more useful.

A finite store can help for a moment

Start the 5° pitch trial with the old load. At this operating point, the air supplies less power. The rotor slows and releases some stored energy. That helps explain why electrical energy can briefly exceed the work arriving from the air. It is not free energy or an invisible battery.

An aerodynamic blade, reconstructed from source data

The NREL-5MW AeroDyn file declares 19 stations. Each gives span, curvature, sweep, twist, chord and a profile ID. The last declared span is 61.4999 m from the root; a later commented numeric line is not an extra station. Eight profiles each provide a separate reference point and 399 perimeter coordinates. We loft those declared sections and add display caps; this is not certified manufacturing CAD.

Pitch is a geometry change

In our source frame, X points downwind, Y sideways and Z upward. Positive pitch rotates the leading edge toward upwind. Each profile’s own reference is subtracted before scaling by chord. Source twist and offsets are retained; collective pitch then rotates the entire blade about its pitch axis. Pitch changes between trials are reconfigurations, not measured actuator motion.

One fixed source wind condition

The ROSCO performance file provides 26 tip-speed-ratio rows, 36 pitch columns and one wind-speed plane: 11.4 m/s. It contains calculated Cp, Ct and Cq values, not field measurements. The main trials use the same density, geometry and wind condition. Startup, no wind, reversal, yawed inflow and complete feathering are outside this model.

Two radius conventions that must not be mixed

The nominal radius is 63 m. With a 2.5° precone, the projected radius is 63 cos(2.5°) m and the reference swept area is about 12,445.257 m². Table TSR uses the nominal radius: ωRnom/U. The coefficient relation at source nodes uses projected TSR: Cp approximately equals TSRtable × cos(2.5°) × Cq. The small remaining difference follows from the published rounding.

Interpolation is a declared approximation

Four surrounding source Cp values are combined with bilinear weights. At TSR 6.25 and pitch 2.5°, those values average to Cp = 0.41332725. We calculate aerodynamic power from that Cp, then torque as power divided by angular speed. Separately interpolating Cp and Cq would not guarantee their identity between grid nodes.

The reference power is not a wake measurement

The coefficient reference is one-half density × projected area × wind speed cubed. Here it is about 11.293 MW. It describes an undisturbed-flow normalization through the selected area, not a direct measurement of flow through the slowed rotor. Unconverted reference power is not all frictional heat. This lookup model does not calculate a wake or local blade pressure.

A gearbox trades torque for speed

The reference ratio is 97:1. Generator speed is 97 times rotor speed; generator torque is the low-speed equivalent load divided by 97. In the source’s ideal 100%-efficient gearbox, their products are equal. The generator then has a separate modeled 94.4% efficiency. Neither assumption describes all real machinery.

The generator also contributes inertia

Rotor inertia is 38,677,040.613 kg·m². Generator inertia is 534.116 kg·m² on its faster shaft. Referred to the low-speed shaft, total inertia is Jr + G²Jg = 43,702,538.057 kg·m². The squared ratio follows from the generator’s rotational energy, not an arbitrary correction.

A ledger that includes the rotor

The rigid-drivetrain equation is Jdω/dt = Qa − QL. Electrical power is ηQLω; modeled generator loss is (1−η)QLω. Stored energy is Jω²/2. We integrate air work, electrical energy and modeled losses separately, then check that their difference equals the change in stored rotation. The numerical residual is exported alongside the values.

Why the 5° trial can give back energy

After one model second at 5° with the original load, incoming aerodynamic work is about 4.007 MJ, electrical energy is 4.594 MJ and modeled loss is 0.273 MJ. Stored rotation decreases by about 0.859 MJ. Rounding aside, the complete ledger balances. The finite reservoir cannot sustain unchanged output indefinitely.

Pitch does not have one effect everywhere

At table TSR 4, Cp increases from 0.212709 at 0° to 0.244105 at 5°. At TSR 6, the same pitch change decreases Cp from 0.434596 to 0.356023. Some source cells are negative: at TSR 6 and 15°, Cp is −0.039710. That sign represents the rotor giving mechanical energy back toward the air in this model; it is preserved.

Stop where the model stops

Our exercise uses 6.9–12.1 rpm, pitch 0–15°, a maximum low-speed load of 4.59 MN·m and electrical output at most 5 MW. These are a selected study envelope and authored load cap, not universal failure thresholds. A boundary event freezes the trial and its ledger. It does not add a hidden brake or continue generating at an artificially clamped speed.

Reference models and real experiments

The NREL-5MW reference was documented in 2009; our pinned ROSCO input deck contains a computational performance table dated January 2022. Fontanella and colleagues published the separate wind-tunnel study in 2023. Such comparisons help examine control and modeling. They do not make every later model or material automatically better.

What the wind tunnel actually measured

The evidence figures concern a 1:100 IEA 15 MW model, not the NREL 5 MW rotor. Instruments measured generator speed, incoming wind and tower-top loads; aerodynamic loads were obtained through processing. Actual blade-pitch feedback was unavailable, so a commanded pitch set point was used. Distinguish the recorded quantity, its processing and its interpretation.

Utility-scale wind generation

Pitch and generator torque help a turbine operate within its intended range. Real control systems include sensors, actuators, dynamics and protections beyond this manual exercise.

Energy briefly stored in motion

A changing rotor speed changes kinetic energy. This provides a useful way to reason about transient input and output, without assuming a hidden battery or creating energy.

Testing a control idea

Reference models let researchers compare methods. Instrumented scale experiments help reveal differences between predictions and hardware. Scaling and measured quantities must remain explicit.

The main shaft is paused, but each blade turns around its own long axis. What changed?

• Collective blade pitch
• Rotor rpm, even though the shaft did not move
• Nacelle yaw only
• The number of blades

At the initial state, the model disconnects generator load and the rotor speeds up. What happens to electrical output?

• It must increase because rpm increased
• No electricity is delivered; the spinning rotor stores more energy
• All wind energy vanishes
• The model proves real turbines have no mechanical losses

A turbine is producing electrical power at a steady rotor speed. Which torque statement fits the model?

• Aerodynamic torque must be zero because acceleration is zero
• Generator resisting torque must be zero whenever the rotor turns
• Aerodynamic torque balances the generator load referred to the same shaft
• Torque and power are the same quantity with different names

In the 5° trial, electrical energy briefly exceeds aerodynamic work while the rotor slows. Where can the difference come from?

• A hidden battery, even though the model contains none
• The generator creates energy whenever load stays fixed
• A permanent exception to energy conservation
• The slowing rotor gives up some of its stored energy

The lookup map describes one turbine at 11.4 m/s. What can we use it for?

• Estimate values between its listed blade angles and speeds at that wind condition
• Predict every turbine's yearly energy from any wind speed
• Call every table cell an observed wind-tunnel measurement
• Extrapolate freely to a stopped rotor and divide power by zero to obtain startup torque

The reference gearbox makes the generator shaft rotate 97 times as fast. What does that establish in our ideal gearbox model?

• It produces 97 times as much mechanical power
• Generator torque equals the low-speed equivalent load divided by 97
• The generator inertia contributes only its unmodified value to low-speed-shaft inertia
• Every wind turbine must contain this gearbox

What should we conclude from the real wind-tunnel figures?

• Our NREL 5 MW model was directly measured in that room
• A smooth simulation curve is automatically more trustworthy than experimental symbols
• The comparison helps identify differences between predictions and hardware under specified conditions
• Every blank area of the coefficient figure equals zero

A trial reaches the lesson's rpm boundary. What is the honest next step?

• Hold rpm fixed without adding a modeled brake, but keep crediting electrical output
• Report that every real turbine breaks at precisely that speed
• Invent a new power curve to continue the animation smoothly
• Stop at the model boundary, keep the last valid ledger and explain the limit

Rotor aerodynamic forces and mechanical-to-electrical energy conversion.

Public mechanism explanation; not the source of the numerical performance map.

NREL 5 MW geometry, performance table, inertias, gearbox and generator inputs.

Apache 2.0; exact original files and credits are packaged. Table dated January 2022, one wind plane.

Source-code/data reuse terms and notice obligations.

LICENSE and original upstream file credits accompany the download. The visual loft is an adaptation.

Torque/pitch coordination and reference-control research.

Our manual trials do not implement the full ROSCO controller.

Blade sections, offsets, twist and profile-coordinate conventions.

Nineteen declared stations and each profile reference are retained; no trailing numeric comment is inserted.

Documented pitch and blade-coordinate relationships.

Our display uses its own declared right-handed transform; positive feather moves the leading edge upwind.

Projected-radius normalization for aerodynamic coefficients.

Current owner source explains the convention; numerical checks against every table cell independently support the projection factor.

Actual apparatus, experimental operating points and simulation comparisons.

Figures 1, 5, 6; CC BY 4.0. Different IEA 15 MW scale model; original axes/legends retained.

Attribution and adaptation terms for the research figures.

Creator, source, license and unchanged-image status travel with the evidence ZIP.

Historical identity of the 2009 NREL 5 MW reference.

DOI 10.2172/947422. Quantitative inputs here come from acquired owner-repository files, not an inaccessible report PDF.