openstudio/energyplus - performance discrepancy among versions
So, I first noted odd behavior in this here post (https://unmethours.com/question/26345...), but thought it might be necessary to make this more prominent and give it another, clearer and more dedicated post.
Basically, I am in the process of upgrading from OpenStudio 1.14.0 & EnergyPlus 8.6.0 to OpenStudio 2.2.0 & EnergyPlus 8.7.0. I am looking at a Proposed Building with Air-Cooled Chillers (ACC), Air Handling Units (AHU) and Variable Air Volume (VAV) terminals, and a Reference Building with Roof-Top Units (RTU) and Fan-Powered Box (FPB) terminals. For each, I opened the 1.14.0 .OSM file in OpenStudio 2.2.0 and then saved it as a 2.2.0 .OSM file from there. No other changes were made!
As it turns out, the 2.2.0 proposed building has a slightly different EUI and greatly different unmet hours from those of the 1.14.0 proposed building, even though they should be identical. The difference is even more stark when one looks at mixed air and outdoor air flows for one of the AHU, as well as setpoints and temperature for one of the VAV. This is shown in the images below. Note how the 2.2.0 version cannot even maintain setpoint, even though the equipment summary (i.e. equipment size) is identical for the 1.14.0 and 2.2.0 versions.
A) Proposed Mixed and Outdoor Air
vs. 
B) Proposed Setpoints and Temperature
vs. 
In contrast, the 2.2.0 reference building has identical performance to that of the 1.14.0 reference building, as I would expect. This is shown in the images below.
C) Reference Mixed and Outdoor Air
vs. 
D) Reference Setpoints and Temperature
vs. 
I find this rather striking and am worried about the accuracy of either 1.14.0/8.6.0 or 2.2.0/8.7.0 I should note that aside from a different type of HVAC system, the proposed building has hundreds of intricate shading surfaces where the reference building has none. I would appreciate any insight and can provide the .OSM files if need be. (Per the post mentioned at the beginning, I had already done an .IDF export file comparison, but found nothing striking).
The release page for EnergyPlus 8.7 shows all the changes made in that release. I would look through the 'Defects Repaired' list to see if anything might be relevant to your model. Perhaps this one, which corrects a problem sizing reheat coils.
Yes, I looked there and found nothing that would ring a bell. Also, none of my coils - whether 1.14.0 or 2.2.0 - size to zero, so I doubt this is even applicable? Also, for good measure, I just now reduced the shading of the proposed building to that of the reference building, but left the ACC-VAV vs. the RTU-FPB system in place. I still get the same type of discrepancy, which is characterized by the fact that the unmet heating hours decrease ever so slightly (three hours), but the unmet cooling ours increase massively (factor of seven) when going from 1.14.0 to 2.2.0.
Again, what strikes me particularly odd is how 1.14.0 and 2.2.0 size identically, but during the simulation, 1.14.0 can handle the load, i.e.maintain setpoint, yet 2.2.0 cannot? Seems to suggest to me that in 2.2.0 either loads during simulation are calculated higher than during sizing, or equipment during simulation is not allowed to run all out?