Stack effect airflow

[Mathadon]

For #1244

[jelgerjansen]

@Mathadon @kldjonge I ran IDEAS.Examples.TwinHouses.BuildingO5_Exp1_1Port.mo again. The results are already better, but still errors compared to the CONTAM model in the order of 1e-2. Could this be due to the default wind pressure computation?

Furthermore, the newly add unit test for BuildingO5_Exp1_2Port throws an error due to the absence of resDoor (which is needed to calculate comparison_W40): in this wall, useDooOpe is true, due to which useResDoor is false (see InternalWall model implementation)

[Mathadon]

Ah, now I remember something else: the 1 port door configuration contains a medium column, which I have removed in my pull request. This indeed seems wrong and was not included in the contam reference results either. So it should probably be removed before rerunning the reference results. @kldjonge ?

[kldjonge]

True. For this branch/unit test it should not be included. (We can discuss maybe adding density columns in a later stage just for the larger openings, but not for now)

[jelgerjansen]

@Mathadon @kldjonge could you explicitly tell me what I should remove in which models? Then I can rerun the unit tests after applying the changes (and such that we can hopefully move on to the other PRs after that ;) )

[kldjonge]

@jelgerjansen, apparently having branches/pull request with exactly the same name causes issues. At least in the tool I use, I can't load this pull request so being very specific is hard.
If I recall correctly, the column we are talking about is in the internal wall component and was added between the door component and the connectors.

[jelgerjansen]

@Mathadon @kldjonge removing the density columns between the door, i.e. going from this:

to this:

gives the same results as plotted here: #1296 (comment)...

Removing the density columns therefore didn't affect any of the reference results...

[jelgerjansen]

Thank you @kldjonge, this indeed fixes the problem; the reference results are now exactly the same as those of the master branch:

@Mathadon can you comment on the specific implementation of Cs as noted by Klaas above?

Furthermore, before pushing the new changes, I would like to know what to do with the unit test for IDEAS.Examples.TwinHouses.Building05_Exp1_2Port, which threw an error (see #1296 (comment))

[Mathadon]

@jelgerjansen no, I'm not too familiar with that model.

With respect to the error, I'd remove the offending variables in the .mos file for now :)

[jelgerjansen]

Due to the change in calculation of Cs, all previously changed reference results are also different of course. Give me some time to reset the reference results again to the 'original' ones, such that we can compare again with those instead of the reference results with the 'wrong' Cs values.

[Mathadon]

okay, sounds like a good extra check!

[jelgerjansen]

Running the unit tests using the updated Cs calculation and comparing to the 'original' reference results leads to lower deviations between the new and old reference results. An overview of the changes can be found below:

I think we can therefore continue and merge these changes in #1322?

The reference results of TwinHouses are, in contrast to the first unit test that I ran, not changed.

[kldjonge]

I think we can move on to the next branch indeed. Great!

[jelgerjansen]

The changes of this PR are included in #1327. Further development of the stack airflow effect will take place in that PR.