I know that as a general rule, EnergyPlus doesn’t like Non-Convex Zones. Unfortunately - I need to analyze heating/cooling loads of a large building with very complex envelope geometry (please, see image below as a rough representation of one of the floors - core + perimeter).
Because the geometry is ready and floors are different on each level, it would take a lot of time to re-draw the model siplyfing it or dividing to numerous non-convex zones. I found this very old topic, where @chris said that using non-convex geometry should not affect the results much if I am not interested in surface temperatures.
I assambled a very simple test model (it is attached to this topic) - it contains 2 rectangular zones, that you can either leave as is (creating two neighboring, convex zones with Air Wall inbetween) or solid union into one - concave L-shaped zone. The change can be modified by marked slider.
Please notice i set SolarDistribution to 1 in Simulation Parameters, so the EnergyPlus component does not show any errors. I wonder why the results are different Am I missing something here? Should I avoid using non-convex zones at all costs?
And, if you still get a warning from E+ about convex surfaces but you are using PixelCounting, you can ignore it as this is a bug in E+ 9.3 that the author of the pixel counting method just fixed for the next E+ release:
I am afraid I was misunderstood - I am not really worried about the E+ warning. If I set SolarDistribution to 1 (Full Exterior) I no longer would get it. I don’t need to check surface temperatures or comfort metrics in my model. I am solely interested in cooling and heating loads (+peak loads). Please correct me if I am wrong - I thought it was fine to use Full Exterior for these calculations.
So, as the PixelCounting is really interesting, I think that is not exactly the source of the differences in results (as i used Full Exterior anyways, instead of Interior, Exterior and Reflections). Unless I misunderstood something and using non-convex zones may affect some other results/metrics.
Unfortunately, I will not be able to do it today - but I will test the model on monday with 1.0 plugin (which looks AMAZING btw.)
What I was surprised about was that summed annual heating load was around 25-30% higher when using one, L-shaped, non-convex zone compared to using 2 convex zones with an air wall (please see last two pictures or gh file I posted or run a model I posted with modyfing the slider marked in red). I couldn’t really find a reason for that (as calculated gains seem similar, as well as ventilation energy and airflow volume) - and was hoping for an answer where such big diffrences may come from.
Sorry. I think I just had a reaction to your first sentence that said Energy plus generally doesn’t like Concave zones and this really isn’t true as of E+ 9.0 now that the PixelCounting method exists.
But I think the difference that you see in peaks here doesn’t have anything to do with convex vs. concave zones and everything to do with the art of zoning in general. By joining all of the zones into one big zone, you effectively spread out the peak cooling load of the perimeter spaces over a larger floor area. So, even though the overall loads on the design day are the same, the sum of the peaks in the subdivided perimeter spaces ends up being greater than the one peak for the overall “we’ll mixed” space.
This is one reason why a lot of engineers that I know like to subdivide zones into as small units as possible when sizing zone equipment (so that they account for the fact that that the peak in a particular perimeter space is covered in the case that the perimeter zone is cut off from the rest of the zones). Granted, I think a lot of times people sizing zone equient go overboard with this subdivision but I guess people feel it’s safer to oversize the equipment slightly to avoid occupant complaints, even though it might unnecessarily raise the cost of the HVAC and increase the energy consumption of the system.
This is all just to say that zoning is an art and there isn’t always one right way to do it. You probably don’t need to subvdivide the zones much if you’re only interested in annual loads but when it comes to sizing stuff with peak loads, you’ll probably find some variation in methods from engineer to engineer.