Assign CTS demands to buildings

[nesnoj]

Similar to #435 for households, CTS demands need to be allocated to buildings.

Idea

• Use the aggregated el. CTS profile grid district level from demand.egon_etrago_electricity_cts, but it's probably better to use this function to obtain the timeseries.
• Disaggregation to census cell: This profile is linearly scaled down to census cell level using the annual el. demand per census cell and scenario, it is located in demand.egon_demandregio_zensus_electricity (sector="service").
• Disaggregation to buildings and amenities:
  • Use only
  • buildings with amenities (openstreetmap.osm_buildings_with_amenities) and
  • amenities (openstreetmap.osm_amenities_not_in_buildings) which have no building assigned. Create synthetic building for each amenity (maybe 5x5m as in HH)
  • The annual demand and cell profile is scaled down to building level by using the building's area
  • Additionally for buildings: use count of amenities (col n_amenities_inside) in a building for weighting the building's demand, if I remember correctly @piaulous' analyses have shown that a linear weighting may overestimate the peak load (and therefore the demand) TODO.
  • In cells w/ el. demand but w/o buildings, create synthetic ones. Different to the residential sector, we have no additional data on number of buildings so the demand is the only we got. A reasonable assumption is necessary (some mean value from other cells?) TODO
  • How to split up demand between buildings w/ amenities and amenities? TODO

[nesnoj]

Sorry, the right table for the amenities is not openstreetmap.osm_amenities_shops_filtered but openstreetmap.osm_amenities_not_in_buildings, I adjusted my post above. Unfortunately, the preprocessing does not come to an end in reasonable time, I will restructure this part and restart the DAG in eGon-data2..

[nailend]

• the CTS demand from Demand Regio is distributed over ~500,000 cells.
• OSM covers only 78,000 cells through amenities_in_buildings
• in addition there are another 80,000 amenities_without_buildings
• so best-case (all amenities_without_buildings in individual, not yet covered cell) -> 160,000 cells
• in addition, there are still amenities in 150,000 cells in which there is no CTS demand after deman_regio...

[nailend]

Disagregate cts:

1. mv_grid level
2. census_cell level
3. buildings level

1 -> 2:

• annual demand share per mv_grid (same for both scenarios)
• use demand regio annual demand at census cell level
• group by mv_grid and calc share
• cts-profiles already scaled? @nesnoj

2 -> 3:

• cells with amenities

  • distribute to building by polygon-area
  • distribute to building with amenity-share of building per cell
    • amenities within buildings
      • n_amenity_inside differs to own calculation (osm_amenity_id per osm_building_id)
      • use own calculation? @nesnoj
    • amenities without buildings
      • create synthetic building

• cells without amenites

  • take random building from buildings_filtered
  • define max annual demand per building? @nesnoj
  • https://www.swo-holstein.de/files/bilder/netze/V_TSS_59_SWO_Richtlinie_zum_Netzanschluss_fuer_Neu-und_Bestandsanschluesse.pdf

[nailend]

preliminary diagramm only! for final methodolgy look at the bottom

[nailend]

First implementation is running in SH. Current issues are:

• missing information on educated-guess for cells without information on amenities or buildings: all over median values is used
• currently using 1 building per cell and 1 amenity per building as first implementation
• there are duplicated cell_ids, resulting from SQL scripts. These are adhoc dropped as I couldn't find the source.

This figure shows an example of the bug. The blue framed cells are duplicated. They are selected for df_buildings_with_amenities as well as df_buildings_without_amenities

Solved by new openstreetmap SQL scripts

[nailend]

For SH		cells	buildings	amenities
df_buildings_with_amenities	2716	6510	9826
df_buildings_without_amenities	8370	8370	8370
df_synthetic_buildings_with_amenities	517	654	654
df_synthetic_buildings_without_amenities	1045	1045	1045
total	12648	16579	19895

[nailend]

[nailend]

This is the final implementation for heat and electricity. Depending on the sector different source for the annual demand are used and the correspondent profiles at MV-HV substation.

If no information is available the median number (2) of amenities per cell is used to generate synthetic amenities. Each amenity gets one building assigned or generated if no is available

[nailend]

For SH cells buildings amenities df_buildings_with_amenities 2716 6510 9826 df_buildings_without_amenities 8370 8370 8370 df_synthetic_buildings_with_amenities 517 654 654 df_synthetic_buildings_without_amenities 1045 1045 1045 total 12648 16579 19895

For DE:		cells	buildings	amenities
df_buildings_with_amenities	97,752	219,463	309,279
df_buildings_without_amenities	353,059	657,482	657,482
df_synthetic_buildings_with_amenities	18,079	22,472	22,472
df_synthetic_buildings_without_amenities	40,856	81,712	81,712
total	509,746	981,129	1,070,945

[nailend]

Resulting Peak load compared to residential ones

		count	mean	std	min	25%	50%	75%	max
cts	eGon100RE	981,136	33,254.7	40,134.6	9.8	11,085.1	21,279.7	4,0392.2	2,842,573.8
	eGon2035	981,136	45,371.3	58,926.0	11.5	14,549.0	28,211.3	5,4108.9	6,785,765.0
residential	eGon100RE	21,360,296	3,173.4	3,378.0	759.0	2,135.2	2,475.5	3,029.6	627,676.6
	eGon2035	21,360,296	3,526.6	3,740.8	907.7	2,388.7	2,735.9	3,335.7	678,399.6

[nailend]

Distribution of peak loads for scenario 2035.

Whole range min/max

Selected range min/ 200 kWh

Selected range min/ 100 kWh

[nailend]

@birgits