Be observant with the reference temperature. Fluent is calculating q_wall correctly, but then normalize HTC and Nusselt with the reference temperature, rather than the bulk temperature (which obviously is very difficult to calculate).

If you are familiar with the above, than I suggest to check you turbulence model. Heat transfer is extremly demanding on the TM. If nothing known I use low-Reynolds number 4 (=Abe-Kondoh-Nagano) k-epsilon model

If I want to add another important parameter, it is the mesh size near the wall. Take care of that!

Take reference temperature on this basis calculations are done

You should calculate reference Temperature. The reference temperature has a substantial effect on the Nusselt number. You can use mass average surface integral for calculating Tref.

Cross check whether your solution is converged or not by monitoring any engineering quantity at any reference location.Is your geometry is extracting energy from the fluid or imparting energy to the fluid?Because your bulk motion of the fluid will have substantial effect on the Nusselt number

Dear Popolla

I am working on similar kind of problem, In my problem fluid is flowing past an unconfined elliptic cylinder. While calculating the Nusselt Number, I use area weighted average method tool box. second thing is that if you are using very high Prandtl number and Reynolds number, then you should have super fine mesh near the cylinder surface to capture the thin thermal and hydraulic boundary layer.