Domain convergence test
 

I'm doing a 2D transient analysis of air flow around three relatively small cylinder arranged in a triangle.
I'm having issues in my domain convergence test.
I'm running 150 time steps, each one 0.002 seconds long, and taking the accelerated velocity at the 150th step as the converging value.
I'm increasing the domain size by 20% in each direction for each iteration.
I have run into a problem with the fact that the velocity value is not converging, in fact is is varying up and down as I increase the domain size.
I think it might have to do with the fact that I don't have the mesh properly refined in the boundary layer, or the fact that I'm using openings for the size.
The mesh is about 400,000 nodes, cylinders 0.034m in diameter.
I would appreciate anybody who could spread light on the matte.
Also, Would it be bad practice to use a steady state solution for the convergence test?