There are no realistic timeseries for production and consumption of individual actors

[j-ti]

• consumption: random timeseries
• production: characteristic timeseries (PV curve)
  • timeseries x scale + noise:

first idea: (1) prediction is moving average of realized (2) realized is predicted + noise

• https://github.com/oemof/feedinlib
• https://loadprofilegenerator.de
• https://github.com/tillboeckmann/stromlastprofile_GHD/tree/main/data_out/results_lastgaenge
• http://opendata.ffe.de/dataset/normalized-industrial-electrical-load-profiles-germany/
• https://pvspeicher.htw-berlin.de/wp-content/uploads/Repr%C3%A4sentative-elektrische-Lastprofile-f%C3%BCr-Wohngeb%C3%A4ude-in-Deutschland-auf-1-sek%C3%BCndiger-Datenbasis.pdf
• https://pvspeicher.htw-berlin.de/veroeffentlichungen/daten/lastprofile/

[j-ti]

• d3a uses: https://loadprofilegenerator.de
• also: leonard in netz_elog did something similar?
  • rather ask bashini to create timeseries
  • or ask stefan for existing csv pv timeseries

[j-ti]

Frieda found this for specific load profiles of industry etc.: https://github.com/tillboeckmann/stromlastprofile_GHD/tree/main/data_out/results_lastgaenge

[DanielL35]

• load profiles for private consumers from https://loadprofilegenerator.de, this is a windows software that lets you create a customized household and generate a load curve, how can we integrate it?
• load profiles for businesses: "büroähnliche Betriebe (WZ64-71), Handel (WZ47), Beherbergung (WZ55), Krankenhäuser (WZ86) und Schulen (WZ85) nach Destatis (2008)", available in csv from https://github.com/tillboeckmann/stromlastprofile_GHD/tree/main/data_out/results_lastgaenge
• how can we randomize these time series?
• How do we distinguish between types of actors (consumers, industry etc.) ?
• I'm talking to Bashini next week on Wednesday about load curves

[j-ti]

* load profiles for private consumers from
  https://loadprofilegenerator.de, this is a windows software that lets you create a customized household and generate a load curve, how can we integrate it?

• There could be a database of different households in a csv file from which we sample
  • load profiles for businesses: "büroähnliche Betriebe (WZ64-71), Handel (WZ47), Beherbergung (WZ55), Krankenhäuser (WZ86) und Schulen (WZ85) nach Destatis (2008)", available in csv from https://github.com/tillboeckmann/stromlastprofile_GHD/tree/main/data_out/results_lastgaenge
  • how can we randomize these time series?
• to start with we could just add a noise signal, we just want different actors to be a bit more realistic and may analyse which actor type has which (dis-)advantage within the market
  • How do we distinguish between types of actors (consumers, industry etc.) ?
• take a look at scenario.py: def create_random2(num_nodes, num_actors):
  • I'm talking to Bashini next week on Wednesday about load curves
• ++

[DanielL35]

* load profiles for private consumers from
  https://loadprofilegenerator.de, this is a windows software that lets you create a customized household and generate a load curve, how can we integrate it?

* There could be a database of different households in a csv file from which we sample

• Load Profile Generator already has results for 64 different types of households (family, couple, single, etc.) which we could prpbably integrate. Each data set is one year in 1 min resolution https://www.loadprofilegenerator.de/results2/

[DanielL35]

meeting with Bashini:

• feedinlib contains "implementations of photovoltaic models to calculate electricity generation from a PV installation based on given solar radiation" https://github.com/oemof/feedinlib there is also pvlib and windpowerlib
• installation is tricky but should be easy to use
• only parameter we need is "area of PV". Maybe we can generate a set of time series for different areas for one year and sample from them in the simulation tool

[j-ti]

http://opendata.ffe.de/dataset/normalized-industrial-electrical-load-profiles-germany/

[DanielL35]

http://opendata.ffe.de/dataset/normalized-industrial-electrical-load-profiles-germany/

I saved the data to the "data" folder in the repo. The data.py code loads the JSON and converts it to a DataFrame. We can choose between internal_id[1] = 0: Industry (total) internal_id[1] = 1: Iron & steel industry internal_id[1] = 2: Chemical and Petrochemical industry internal_id[1] = 3: Non-ferrous metal industry internal_id[1] = 4: Non-metallic Minerals internal_id[1] = 5: Transport Equipment internal_id[1] = 6: Machinery internal_id[1] = 7: Mining and Quarrying internal_id[1] = 8: Food and Tabacco internal_id[1] = 9: Paper, Pulp and Print internal_id[1] = 10: Wood and Wood Products internal_id[1] = 11: Construction internal_id[1] = 12: Textile and Leather internal_id[1] = 13: Non-specified (Industry)

The time series is in 60 min resolution (household is 1 min resolution, commercial builidngs is 15 min resolution)

[j-ti]

• dataset csv file should contain columns {Datetime, }, series name could be a abbrev. like HH1 or Ind1, etc.
• loads, pv or other component types should be in different csv files in different folders