In particular the radiating spokes model seems too limited. If I'm not mistaken, the parameters here are any shape that would fit inside the circle, unless it needs to be attached in the middle. Even so, radiating shapes are just one 'family' of shapes that meet these parameters.

I'd have a go at programming the random shape generator, although I don't have any experience with wind simulations.

My recurring thought as I watched : is it possible to ensure a greater variety of shapes at the outset? They all seemed to be too 'full' to be likely blade shapes. For instance, it looked like an evolved Lenze blade was nowhere in sight. Widening the starting parameters helps to prevent 'lock-in' later on to less desirable shapes.

Most of all though, thanks for making this, as I've yet to find the time and determination to follow through with it ;)