Drop duplicate hydrogen ports in add_export.py

[yerbol-akhmetov]

Changes proposed in this Pull Request

Good day. This PR aims to minor a bug related to duplicate entries of H2 export ports which prevents addition of links that connect H2 export bus and H2 buses. This issue happens particularly for US, where multiple ports are available. The PR is linked to issue #1231. Dropping the duplicates solve the issue.

Checklist

• I consent to the release of this PR's code under the AGPLv3 license and non-code contributions under CC0-1.0 and CC-BY-4.0.
• I tested my contribution locally and it seems to work fine.
• A note for the release notes doc/release_notes.rst is amended in the format of previous release notes, including reference to the requested PR.

[GbotemiB]

I have also tested this and it works properly. Before this, the US model was not feasible, but with this solution the model is now feasible.

[ekatef]

Thank you @yerbol-akhmetov! Great to hear that the fix works, and a great investigation of the reasons #1231.

Just to be sure, I wonder if we could drop some available shipping capacity when dropping duplicated buses in the current implementation? From the way how the hydrogen_buses_ports are being formed, it appears that the duplicates correspond to the same GADM index:

pypsa-earth/scripts/add_export.py

Lines 63 to 66 in 5d02b29

	ports = ports.set_index("gadm_{}".format(gadm_level))
	# Select the hydrogen buses based on nodes with ports
	hydrogen_buses_ports = n.buses.loc[ports.index + " H2"]

May it be the case that duplicates in hydrogen_buses_ports are different ports but located in the same GADM region?

[davide-f]

Thanks a lot @yerbol-akhmetov for this great proposal; I'd propose just a minor revision that may make the code more robust.

If I get the issue right, the problem in the current workflow is that the column "gadm_{}".format(gadm_level) used as index may contain duplicates.
For your case, however, drop_duplicates works.
The function drop_duplicates() by default drops only duplicated rows, if rows have the same index but different content they are not dropped.

To ensure robustness and keep the proposed formulation, a possible proposal would be to move the drop_duplicates few lines up with something like:

    gcol = "gadm_{}".format(gadm_level)
    ports_sel = ports.loc[~ports[gcol].duplicated(keep='first')].set_index(gcol)

May that work? Feel free to revise and improve as you feel best

I'd also recommend to add a comment like "TODO: revise if ports quantity and property by shape become relevant" just to keep track of this approximation

[yerbol-akhmetov]

Thanks @davide-f, @ekatef. I have applied the latest changes that were suggested and tested it locally. It worked as intended. Also, the comment about TODO was added too.

[ekatef]

Ahh, true!! The check for duplicates relate to the whole lines, not only indices. Agree then that data loss is highly unlikely.

I think the implementation looks great, and if @davide-f doesn't have further comments, happy to merge 😄

[davide-f]

Merged, many thanks @yerbol-akhmetov :D