Add more options for temporal aggregation

[lisazeyen]

This PR should add two more options to aggregate the network temporally:
(1) aggregate timeseries to representative snapshots with e.g. 25sn as sector_opts would results in taking every 25th snapshot
(2) using segmentation method from tsam similar to PyPSA-Eur, except that all time-dependent data is considered (to include e.g. also the COP of heat pumps) and that the options is added to overwrite the pypsa time-series with time profiles generated by tsam. For example 60SEG in sector_opts would results in 60 representative snapshots of the original timeseries.

[lisazeyen]

I am actually pretty sure that the options to aggregate to x representative snapshots with xsn was tat some point in the code. @fneum to you have any experience with overwriting the timeseries of pypsa with the ones generated by the tsam package? I just did a quick check but they didn't seem to be very similar. Therefore, I am currently setting the option to overwrite them as False and haven't moved the option to the config. We could also remove this part if it is not helpful.

[fneum]

Nice PR!

Stylewise, just minor comments, in particular using n.iterate_components().

Results with segmented time series (here 2190 snapshots ~ 4H resolution) look good. The peak and trough shaving is an expected inaccuracy of the temporal aggregation. I didn't check whether there are further options to tweak the segmentation clustering in tsam.

solar:

wind:

heat demand:

electricity demand:

One thing I'm unsure about is the temporal aggregation of e_min_pu. We use this to set the required SOC of electric vehicles in the mornings. The higher the temporal aggregation is, the more this constraint gets washed out. In any case, I don't think it should play into the temporal clustering.

Another thing to be careful about is resource consumption. It took quite a while for tsam to do the temporal clustering.

[lisazeyen]

Thanks a lot for the review @fneum !