Magnified resistance after clustering

[ORIOL RAVENTOS MORERA]

Dear PyPSA users,

I'm clustering an European network using a convenient predefined busmap (but the problem persists if I use K-Means or HAC clustering). Some resulting resistances are highly magnified in the process (around 6 % of the lines). This is no mystery, it happens because of the length factor used to compute the new resistances:

https://github.com/PyPSA/PyPSA/blob/245109666ede7105f381c976291a106af53c2e2f/pypsa/networkclustering.py#L271

Question:

What is the standard way to deal with this issue? How is it done in PyPSA-EUR so that one gets more reasonable values for the resistances? Is there literature on that?

Some thoughts about it:

Taking inspiration on how the clustering of apparent power is dealt with in the simplify_network function:

https://github.com/PyPSA/pypsa-eur/blob/49f1c0f6299401fc99ea2ebb849ff45be68397a4/scripts/simplify_network.py#L116

one could just ignore the resistances of the input network and compute them afterwards as if we were modelling the original network by a standard line type (e.g. 490-AL1/64-ST1A 380.0). This gives a sort of lower bound to the resistances.

Another option is to use an heuristic factor, like the line_length_factor (which is 1 by default and 1.25 according to the article [Hörsch & Brown, The role of spatial scale in join optimisations of generation and transmission for European highly renewable scenarios]), but this is meant to increase the line factor to account for the changing complicated line paths by straight lines between cluster centroids. Hence it will only make things worse.

Any inputs are welcome. Thanks,

Oriol