Question-and-Answer Resource for the Building Energy Modeling Community
 | 1 | initial version |
Just leave it a single zone. It depends how carefully you want model solar and daylighting in the atrium and any spaces with interior windows that open on to it. For example, if you want to model solar distribution on the floor of the atrium, then the IR transparent material is not helpful and should be avoided. I would avoid the approach of splitting the atrium up into horizontal zones. By leaving the zone wide open you can better model daylighting control sensors at the occupant level in the usual way. The solar distribution model will be much more accurate with solar energy penetrating deep into the atrium as it should rather than being stopped in an upper sub-zone. This is because Material:InfraredTranparent does not transmit solar into the zone below it.
I wouldn't recommend air walls for horizontal sub-zones but they may be desirable for connecting to adjacent zones when there are open spaces spreading away from the atrium. Use a single, large high-transmission interior window filling the opening between the zones. Only the window model can move solar and daylight from one zone to another. And also mix air between the zones using AirflowNetwork or more simply at some assumed air change rate using ZoneMixing objects.
The AirflowNetwork will be simpler with a single zone and can focus on the connections with the outside and adjacent zones.
If you expect a significant temperature gradient in the atrium, then prescribe it using RoomAir:TemperaturePattern:NondimensionalHeight, or similar. This is also how one could make use of results from a CFD model of the atrium, if you have one. As other answers have said, what is missing here is a decent/fast general model that can predict the temperature variations as the simulation runs. But if you are willing to make broad assumptions, in advance, about the resulting temperature variations inside a zone, the implications can be modeled. Obviously this screams out for an approach that brackets the problem by running several models with a range of temperature gradients and return air temperature offsets.
