I'm not sure I can speak to the best way to model the ground to sky shaft, but regarding your other proposed simplifications for an initial analysis, they sound largely fine to me. Here are a few couple of things I'd point out:
- Modeling intermediate floors/ceilings as adiabatic seems reasonable as a first approximation as long as you think the surrounding areas will generally be kept at similar temperatures. For example, if they are all tenant units, this is probably a reasonable assumption (on average). If it's mixed commercial and you have different use cases, this may not be reasonable. I think the same goes for walls in between units.
- Regarding which unit(s) to model, you can probably get away with modeling a subset of the units, just remember that the loads on a corner unit will look different than the loads on a central unit. And they also may look quite different based on solar orientation, depending on window-to-wall ratio, glazing characteristics, etc. So, if you really wanted to capture all of the variability, you would probably want to model eight units: four corners, and four internal units on each building side. I'm not sure how much this actually simplifies things, but it would perhaps save you a little time.
