Issue702 single zone commercial adrenalin update

[dhblum]

This is for #702, following from PR #703 to finalize. This is also for #635, #432, and #733.

[dhblum]

@HWalnum This PR is almost finished. I didn't see the updates you made to the releasenotes.md previously but I added them and also marked the changes in the KPI results for all the scenarios, which I think is good to track if they change significantly version to version. Can you look at the KPI changes I marked and double check they are what you'd expect or experienced based on your changes? https://github.com/ibpsa/project1-boptest/pull/734/files#diff-bf0fb780cd5da59cf3fcbb2aeac981b23584b3089c661658f30b386ff4440229.

[HWalnum]

Hi @dhblum. For the peak_heat scenario looks reasonable. The changes in the typical_heat scenario seems a bit drastic. I think I might have not updated the days after switching back to the original weather data. However, I tried to have a look with the find_days algorithm. Looking at the result I got from that, I wonder if the median of the daily max is reasonable for commercial buildings with much lower consumption in weekends. This makes the whole algorithm a bit random.

In the Adrenalin competition we selected the days based more on visual inspection, because the algorithm ending putting one of the scenarios during a holiday, which gave different occupancy profiles.

[dhblum]

@HWalnum Hm ok on the day selection algorithm, thanks for reporting. It certainly isn't as bullet-proof as we thought in the beginning, various issues have come up over time. Perhaps there can be another option in the script to id and remove weekends. At least it's documentable in terms of the options chosen to produce the days. Or, we just say that the given days are the time period for a given scenario because the test case developer thought it was reasonable for the intent of the scenario.

[HWalnum]

I definetly see the benefit of having a documentable procedure, but it is challenging to make it fit all cases. For typical heat day, it could be an idea to look at weekly averages. However, it will not remove the risk of including special holdiays etc. However, the last is mainly a problem for cases like this, where real measured occupancy is used.

For this particular case, I would suggest to go for 308 as typical_heat_day. This is based on median of weekly averages, and it should not coincide with any holidays.