first one is good, but 2 and 3 still suffer a bit from vagueness. Is there anyway we can quantify the difference between extremely broad and simply large?
I'll try to make it clearer when I write the final Power.
Part of the distinction is that in 2 your options interfere with one another more. In a situation where Strength is useful, Speed likely would also be useful. So only being able to use one of the two is a big deal. But in a situation were Strength is useful Glamours is likely to be useless. So being able to use only one of the two matters less.
Which means that a 2-cost set could include Speed and Strength, but not Glamours and Strength unless it was rather small.
What about if you have it so that the refresh cost is the amount of categories that you are choosing from.
Not a good idea, unfortunately.
The categories have no mechanical weight...they don't do anything rules-wise. Changing that adds another variable to the design of each Power for no real reason.
Also, the categories are not remotely balanced against each other. Faerie Magic contains like four Powers, and having two makes the other two redundant. Creature Features contains like a dozen Powers, with some that have multiple variants. Plus there are a lot of homebrew Creature Features.
And finally, that would really limit the usefulness of the Power. For example, it'd make your Zodiac Power idea impossible.