It depends on how much of the processor you want to dedicate to the wind. A wind map would be a nice option, similar to a lightmap. If your flag ended up behind a building where no 'wind light' was shown then the current wind velocity would be blocked, You could set up random wind, the end result at any point a function of the value of the windmap at that location.
In order to create a dynamic wind system, you'll need to do raycasting. Personally I'd combine the windmap and raycasting systems, that way you don't have to raycast if the wind is already blocked by a static structure. You will have to raycast from the wind source across the screen in its direction. If something has the ability to catch wind but doesn't recieve a raycast then its wind is blocked. The problem is unless you raycast a lot then you could miss objects and end up getting no wind when you should. And, if you raycast a lot then you'll bog down your game. If you try something absurd like trying to curve wind around dynamic objects then you shouldn't even bother starting your game.
At the end of the day, I think that wind is nice, but unless you're making a weather simulation you shouldn't spend too many resources on it. It's not going to improve the gameplay and very likely won't even get noticed. Games need to be an efficient approximation of something, not a realistic representation.
Come see the WIP!