Airborne wind energy: control that must scale from kite to plant

Dynamics change with size

What happens. A control law that flies a small kite well does not simply carry over to a larger one. Mass, line length, line sag and the time the wing needs to react all change with size. Gains tuned on the small system make the large one slow, or let it oscillate in gusts.

What it costs. Test flights spent on retuning instead of power, damaged wings when a manoeuvre fails, and a schedule that slips with every new size.

Why the usual fix fails. Scaling the old gains by a rule of thumb works for one step and not for the next. Testing everything on the large system is slow and risky, and flight days depend on the wind.

How a fixed-price remote diagnosis checks it. From your flight logs (line force, wing attitude, line length, wind estimates, control commands) and your model I look at which dynamics change with size and which controller parts depend on them. The fixed-price remote diagnosis tells you whether a controller that takes the size from the model, instead of from retuning, is realistic. I have worked on flight control and trajectory optimisation for airborne wind energy (kites) at SkySails.

Background: Trajectory optimisation and setpoint generation

From prototype to commercial unit needs robust control across wind

What happens. A prototype flies well on good days with an engineer watching. A commercial unit has to fly in all the wind it meets, start and land without help, and run for weeks. The step from one to the other shows how many cases the control never saw.

What it costs. Downtime and lost energy when the system stays on the ground in marginal wind, manual intervention that defeats the business case, and unclear reasons for faults that appear once in a hundred flights.

Why the usual fix fails. Adding rules for each failure produces a controller nobody can check. Waiting for more flight hours is slow, and the interesting cases are rare. A team that wrote the first controller is busy keeping the prototype in the air.

How a fixed-price remote diagnosis checks it. From flight data of the prototype I look at how much the control margin shrinks in strong, gusty or weak wind, and which cases end in manual action. The fixed-price remote diagnosis tells you whether robust control with automated operation is within reach, and what flight-data analysis you need for it.

Background: Learning-augmented adaptive predictive control