I think that the length scale for a vertical sinusoidal wall (or any other periodic wall shape) would be no different from the length scale of a vertical straight wall. The length of the wall parallel to the direction of gravity would be the characteristic length scale.
To be clear, any lenght could be used. Now the discussion is only about the proper lenght that makes non only non-dimensionaL but also normalized the variables.
Filippo Maria Denaro Thank you! Very interesting. Then I will use the length of the sinusoidal wall. Because dT/dn is normal to the path along the curves of the sinusoidal wall.