Wind turbine by AI

What you get: shaft power, Cp, torque, tip-speed ratio, power curve versus wind speed, blade geometry — computed by Heliciel.

→ About the Heliciel MCP connector
→ Install in Claude
→ Get an MCP pass
→ Wind turbine blade tutorial
→ Wind turbine power and wind speed
→ Version française

Wind turbine power calculation with ChatGPT or Claude

With the Heliciel MCP connector, your AI assistant computes the power harvestable by a wind turbine from diameter, blade count and wind speed, plots its power curve, optimizes its RPM and exports the blades. The power coefficient Cp, tip-speed ratio and torque are those of Heliciel's BEM (Blade Element Momentum) computation, not the Betz formula with a "typical" efficiency.

The problem: Cp depends on the blade, not only on the wind

Everyone knows P = ½ ρ A V³ Cp; what nobody knows offhand is the Cp of your rotor at your RPM. It results from the blade geometry (twist, chords, airfoils), the tip-speed ratio and the polars at the local Reynolds numbers. A language model alone answers "Cp ≈ 0.35 to 0.45": an interval, not a computation. Heliciel computes the Cp of the actual blade by blade element theory applied to energy-capturing rotors.

Real dialogue: 3 blades, 4 m, 8 m/s wind

Real session in ChatGPT web (from the demo video).

"Load a wind turbine, 3 blades, 4 m diameter, 8 m/s wind — what power can I harvest?"

ChatGPT answer with Heliciel MCP for a 3-blade 4 m wind turbine at 8 m/s: Cp 0.426, shaft power 1528.94 W, torque 81.11 N·m, tip-speed ratio 4.712, 180 rpm inherited from the model

Heliciel loads the turbine_eolienne model and returns: tip-speed ratio 4.712, Cp 0.426, shaft power 1,528.94 W, torque 81.11 N·m. 0 errors, 0 warnings.

What the AI flagged rather than invented. The 180 rpm was not in the request: it comes from the loaded model, and the answer says so explicitly ("inherited from the loaded Heliciel model"). You know which data is yours and which is an inheritance to confirm.

"Sweep the wind from 3 to 15 m/s."

The assistant first asks the method question: off-design (geometry and RPM fixed, what a real turbine undergoes) or redesign (optimal blade at every speed, what an ideal turbine could reach). The user answers "Off-design".

Heliciel chart shown in ChatGPT: power, torque and Cp curves of the wind turbine versus wind speed, banner Computed by Heliciel not AI-generated

The chart is plotted by Heliciel and passed as-is ("Computed by Heliciel — not AI-generated").

Off-design sweep table from 3 to 15 m/s: Cp and shaft power per wind speed, maximum Cp 0.428 at 7 m/s, maximum power 3.31 kW at 11 m/s

13 points from 3 to 15 m/s, geometry and 180 rpm fixed.

Reading the sweep: maximum Cp 0.428 at 7 m/s; maximum power 3.31 kW at 11 m/s (Cp 0.355); beyond that, the fixed-RPM blade progressively stalls (Cp 0.241 at 12 m/s, 0.091 at 15 m/s). At 3 m/s, −30.6 W: the rotor at imposed RPM runs as a motor, it does not harvest. The assistant explains it instead of smoothing the curve.

And with RPM optimization

On the same rotor, the request "Optimize efficiency" (method asked to the user before the call) returned during the validation campaign a Cp of 0.575 at 250 rpm, against 0.426 at 180 rpm: the tip-speed ratio was too low for this blade. Exactly the kind of question a rule of thumb does not settle.

Rotor export

STL export of the 4 m 3-blade wind turbine rotor from ChatGPT: eolienne_3pales_4m.stl, 148.8 MB, temporary download link

"Export the propeller as STL": complete rotor with hub, 148.8 MB, link valid 60 minutes. The long export returns milestones; the assistant checks completion before giving the link.

What you can ask next

Design the blade for a given site (rated wind, diameter, generator RPM) rather than loading a model: Heliciel computes the optimal twist and chords.
Compare 2 and 3 blades, or two diameters, at the same wind.
RPM × wind map to choose the control law of a variable-speed generator.
Hydrokinetic turbine: same approach in water, with cavitation check (hydrokinetic turbines).
Export the blades as STL/IGES and the sections as JSON for manufacturing.

Limits specific to wind turbines

Steady, ideal BEM: no atmospheric turbulence, wind shear or dynamic stall.
No blade structural computation or extreme loads (storm); no dB(A) acoustics.
The power returned is at the shaft: generator and drivetrain efficiency to be deducted.

Connect

Get a Heliciel MCP pass on mecaflux.com.
In Claude: connectors directory. Elsewhere: MCP connector at https://mcp.heliciel.com/mcp (per-client details).
Authorize with the pass key, then describe your turbine and your wind.

Frequently asked questions

Why is the power negative at 3 m/s?
Because the RPM is fixed at 180 rpm: at low wind the tip-speed ratio becomes too high, the blade drags more than it harvests. A real turbine would slow down. Ask for a sweep at constant tip-speed ratio to see the other behaviour.

Does a Cp of 0.575 exceed the Betz limit?
No: the Betz limit is 0.593. But 0.575 is an ideal theoretical BEM Cp (smooth surfaces, no drivetrain losses); a real machine will be below. Heliciel states it and the connector reports it.

Can I reproduce these figures?
Yes: same model, same wind, same RPM give the same result.