Well just a couple of ideas that might help
a) get a real wind tunnel,
b) if not a, gravity might be your friend -- terminal velocity
might be around your 200mph velocity regime, so if you
can find somewhere to drop the streamlined thing off
a cliff while it's running a data collection / controls
actuation test that you can gather data from, that'll get
you empirical data that's at least in the realm of being
accurate for your problem. Intact recovery shouldn't be
that infeasible depending on your test site and setup.
c) flight test the thing under its own power
if that's easy; if you're collecting data and have appropriate
exercises programmed in to the actuators, I'd think you'd
be able to gather some useful data before the velocity
decreased too much. Start out with neutral controls and
get increasingly adventurous as altitude increases then
let it parachute (or whatever your design does) home
after the high altitude / high velocity period is over.
d) If you tethered it to a rotating arm or cable it seems
like it shouldn't be too hard to use a conventional motor
to spin it up to reasonably high airspeeds safely.
Rather than using a giant propeller with high speed
blades to push air at high speed down a wind tunnel,
it simplifies things a bit to just use the airspeed
at the tip of a rotating arm itself if you can stand the
centripetal accelleration and the radius of body turn
in the airstream is within the realm of servo control that
you wanted to test anyway. Clearly unrealistic for 500mph,
but clearly do-able for 100mph (45m/s), and somwhere
between the two is your limit.
e) you can use a nozzle (just like .... a rocket, hey!)
to convert high pressure/temperature air to low pressure
high velocity air... So it could conceivably be within the
realm of engineering feasibility to generate low pressure
200mph to 500mph air on the output of a
static combustion chamber with a nozzle, but, of course,
you'd need to have some reasonable balance of
temperature vs. speed vs. pressure (which would need to
remain above atmospheric, of course). Certainly
nothing to attempt for a huge diameter / length of test
chamber, but possibly something that would be very
do-able for a very restricted size.
f) Would it be that bad to just use computer models to
estimate the forces involved, then to use some kind of
self-calibrating servo algorithm using feedback sensors
etc. to implement something like a PID loop or whatever?
How far to turn the fin? Well start turning it a little,
if that's not enough, turn it some more... etc.