Here we will write best practices regarding reducing the spatial complexity.
- most models are made for specific regions
- spatial resolutions fine for getting the data, but not good by means of how the system works --> error! how big is this error?
who is using zonal models? what kind of spatial resolutions?
-
smallest unit: municipalities (heat distribution) --> clustering
-
municipalities: problem: borders change over time --> conversion of data huge amount of work
-
balmorel model: for denmark different spatial resolutions for electricity and heat distributions (low resolution for electricty, high resolution for district heating (e.g. one area for copenhagen, clustering of rural areas)
-
problem: consumption data source does not match to spatial resolution of elec. and heat structure! how to solve problem?
-
estimate heat demand by coefficient of type of building, industry by branch total demand devided by number of employees
-
address of company does not mean that production site is there
-
resolution strongly depends on reserch question --> resolutions need to be flexible and adjustable
-
ludwig hülk published paper (usage of landuse data, group data to create load areas --> oedb, validation was carried out, using GDP, time series not included in the paper, but it is the next step and will be published by RLI; for medium voltage first approach Voronoi algorithm but it was mixed with municipality approach)
-
ffe used verteilnetzgebiete publiction
-
what is the error due to limited spatial resolution?
-
calculations with high resolutions (e.g grid and high industrial load) -> then you see the error
-
error may be decreased by working first on high spatial resolution and then aggregating
scaling up load data according to GDP in a specific region, is that approach sufficient? is there a better approach?
- relating to other errors that exist in your model
- comparing it with random data
- ffe: splitting shares of industry loads and then using different scenarios of shares (publication ?)
- if you cannot find out how wrong you are --> carry out sensitivity analysis
- FIAS (David Schlachtberger): euorpe country resolutions sensitivity analysis of resolution quality of electric grid and time step length (publiction is planned to be submitted by the end of october)
compare open grid data like scigrid and gridkit with entsoe, what are the results?
- good data of locations of power data?
- there are data on oedb
- statistics of entsoe have to be handeled carefully
- marktstammdaten are upcoming but maybe just in 2019
- many information see “open power system data”
voronoi, matching algorithm, etc. --> these methods should be better ordered and documented when published! Maybe create an “open energy methods” section on OEP, with a good overview of methods and maybe categorization of methods?
-
for weather data ask stefan pfenninger and renewables ninja, RLI is working on openFRED but it is not finished
-
PSI: pandarius library python with spatial data, will be added to WIKI
Summary
- makes no sense for everyone to do the modular work, coordinate in advance
- do random tests and see what does it to the results
- own approaches are similar to the approaches of others -->> look it up before starting modelling next time (maybe by OEP factsheets)
- set up google alert when opsd updates something
- we cannot model the future correctly, validation is hard