Objects floating in space don’t do anything. They maintain constant velocity (linear and angular). They are acted on by gravity, but the effects of gravity tend to be small, relative to their velocity.
Little objects floating or suspended in liquid have a particular kind of behavior where their motion is dominated by Brownian motion: small, random movements, caused by the impact of individual molecules (which themselves are moving because that’s what “temperature” is, just disorganized molecular motion-- drop it to 0 K and you stop seeing Brownian motion. )
Neither of these would be simulated by wind. Simulating space is just a matter of, two keyframes only, vector handles, linear extrapolation. Simulating small objects in fluid would be a bunch of keyframed randomized transforms, with vector handles, without regular timing. (There may be some way to drive it from noise f-curve modifiers instead, but it doesn’t seem to me that would be quite right.)