Attaching loads at every bus (not only at 'substation_lv' buses) may improve the spatial load distribution

[cristobal-GC]

When obtaining elec.nc, loads are attached only to buses labeled as 'substation_lv':

pypsa-eur/scripts/add_electricity.py

Line 300 in 3ef6d51

substation_lv_i = n.buses.index[n.buses["substation_lv"]]

I don't understand the rational behind this decision. However, a consequence is that the load within a country is distributed only across certain buses (i.e. some Voronoi cells). This may distort the spatial distribution of the load, making the obtained model less accurate as compared with reality. This seems to be the case in Spain. Below (left) is the load density at each Voronoi cell resulting from the current procedure implemented in PyPSA-Eur (year 2022). Only about 50% of the Voronoi cells were utilized. See on the right figure the resulting load density when the load share is aggregated using NUTS3 regions. It can be seen that the load in central Spain was underestimated (where the actual electricity demand is high), because many buses were not eligible for load attachment.

The load can be shared across all the buses by replacing the code line mentioned above with the following:

substation_lv_i = n.buses.index

With this modification, the resulting load density in the Voronoi cells (left) and at NUTS3 level (right) are as follows:

It can be seen that the load density in central Spain is a bit more realistic. For completeness, below is the actual load density in NUTS3 regions in Spain for the same year (data from the Spanish TSO):

I'm aware that the grid representation is transitioning from ENTSO-E to OSM, perhaps this will solve this issue without further changes. In any case, I just wanted to share this analysis with the community. There may be a reason for attaching loads only to substation_lv buses that I haven't considered (that's why I'm not sending this as a PR), but as far as I understand, the proposed change could improve the spatial load distribution.

[fneum]

The reason we use substations is because these are the grid nodes where the lower voltage layers of the power network attach to. ENTSO-E map has significant distortions around Madrid especially. Perhaps this is better now with OSM? #1079

[cristobal-GC]

ah ok I see. Then it could make sense (from a load sharing point of view) to create Voronoi cells only in substation buses, to ensure that the whole territory of a country is considered in the load sharing. This is the result of this approach for Spain, and it can be seen that it slightly improves also the load share in central Spain.

But of course the counterpart is that there are fewer Voronoi cells, and they are larger, which may have a smoother effect in the reneawble profiles.

I will have a look at this issue with OSM, it may actually solve the issue.

[fneum]

Ah, I see. We used to have Vornoi cells per substation but at some point switched to all unique points.

[cristobal-GC]

Just to give a final feedback, yes, with OSM the results look much better. Now the load is distributed over all the Voronoi cells in elec.nc:

which leads to a better characterisation in central Spain, see after aggregating at NUTS3 level:

Thanks!