Interesting idea!
So to figure out if you are even in the right ball park i'd just calculate the moments of intertia for the rocket and the wheel to start with.
Ignore air, friction and everything else, calculate the moment of intertia for the rocket, and the moment of intertia of your wheel, then see how many rotations of your wheel it will take to rotate the rocket 180 degrees. If you get something like 100,000 rotations of the wheel; well, there is a good chance all you will be doing is flying a quickly spinning wheel. If the rotations is in the thousands or less then there exists a much better chance for you to be able to do some control actions on your rocket.
Next you can calculate how much kinetic energy you have in your wheel. I would have the starting rpm on your wheel be set halfway of its total energy level, if it can spin 15k rpm then its center point energy level would be somewhere around 13k rpm (remember v is squared). That way you have an equal level of energy available in either direction for your control.
These should get you at least give you a good enough idea on the feasibility of the system to see if it is worth looking deeper.
From here you can start to get some better resolution on the system. To get some numbers on the applied torque from weather cocking, take a look at a similarly sized rocket that you know weather cocked. Ideally you should have a video of it. Take a look how much the angle changed, and how long it took for it to make that change. From there with some guesstimations on speed to get the aerodynamic forces, and the inertia you can get the torque that it applied. This should give you a average torque value that is in the ball park to see how much you will have to compensate for.
Alternatively go peek around through some old NASA technical papers. There has been a lot of study into passively guided rockets, so there is probably some paper floating around on weather cocking that can give you some sort of formula. I know there is one for fin flutter, that gave an empirical formula for calculating the likelihood of flutter taking place.
If you want to get a much better idea than that you will probably have to build a simulation for your system. That way you can plug is some reasonably good calculations for the aerodynamic factors, inertia, lever arms, as well as the rotational elements of your gyro. Then this can give you a good enough evaluation of your system to move on to putting it into hardware.
Or if that sounds like not much fun, take the guesstimates that you had and just go right to hardware.