Update pythonocc to enable Mac ARM support

[reyery]

Could potential fix #3120

[ShiZhongming]

We paused merging this PR for now until @reyery figure it out what to do with it.

[reyery]

Not sure if this can be merged since the results for the thermal networks is very different due to changes to wntr

@jimenofonseca maybe you can take a look at the differences and see if it makes sense. I attached the outputs in the zip file
Archive.7z.zip

[reyery]

@MatNif maybe you have some ideas as well. Apparently one of the changes in wntr is that it is using EPANET version 2.00.12 vs 2.2.0, not sure if that matters.

Pressure losses of edges seems to differ by 10^-3

https://wntr.readthedocs.io/en/latest/whatsnew.html#v0-3-0-november-2-2020

[reyery]

seems like one of the differences has something to do with reservoirs, probably due to the "headloss" differences from the simulation results

[MatNif]

When I try to start up the dashboard from a Windows PC on this branch I am running into a bit of an issue already. It looks like the libpysal.io library is not installed by default. I'm not sure if the warnings-handling in utilities/dbf.py that was there before the changes made in this branch was meant to handle these errors.

[MatNif]

Then, as I run the code of the network-layout script, I run into the following error:

which might have something to do with the fact that the 'nearest'-method is not available for all versions of the geopandas library:

[reyery]

I have tried recompiling wntr for the version that we are using (v0.2.3), to avoid the difference in results for the latest version. Everything seems to be working now, and there should not be differences when running it anymore.

@MatNif the EPANET version used in wntr would still be v2.00.12 vs v2.2.0, not sure what the differences are but I guess we can investigate this next time. So no further actions are required from your end.

@ShiZhongming It now works for Apple Silicon Macs natively, it should be a little faster now as well, since there is no need for a translation layer anymore. I have updated a lot of the python dependencies as well, so testing it would be much more complicated, as you would need to create a new conda environment from the new environment file.

First run a workflow using the current environment that you have, in the master branch.

Then checkout to this PR branch, run the following commands in the CEA git repo folder:

mamba env create -n cea-arm -f environment.yml
conda activate cea-arm
pip install -e .

Run the same workflow using this new environment. Compare if the results are the same (especially for thermal network results)

Let me know if you encounter any issues. Thanks!

[ShiZhongming]

@reyery Thanks for the PR again.
I have just tested it as instructed using two scenarios: one in Zurich and one in Singapore.

For thermal-network, for both scenarios using master and this PR, the results were identical.
For solar-potentials, in the Singapore scenario, the results were identical, while in the Zurich scenario, the results are mostly slightly different. I have attached the two .csv here.

Zurich-PR:
PV_total_buildings.csv

Zurich-Master:
PV_total_buildings.csv

[reyery]

Interestingly there are differences in the area of PV, but I don't think it is that significant. Would it be possible to try a larger scenario and see if affects more buildings.

[ShiZhongming]

Hi @reyery
I have just tested this PR using a much larger scenario (>200 buildings) in Zurich. The results are identical.
I think we can merge this! :D