PyPSA Model of the European Energy System
Switch branches/tags
Clone or download
Type Name Latest commit message Commit time
Failed to load latest commit information.
resources Add resources/powerplants.csv Aug 3, 2018
results Add empty results directory Oct 12, 2017
scripts add a variable to select the clustering algorithm in busmap_for_count… Nov 26, 2018
.gitattributes initial commit Oct 11, 2017
.gitignore Add cutouts and data to ignored directories Aug 13, 2018
LICENSE.txt Add README, LICENSE Jul 24, 2018 README: Update installation instructions Oct 28, 2018
Snakefile Snakefile: Comment out group settings Oct 26, 2018
borg-it Add borg-it Feb 19, 2018
cluster.yaml Update Snakefile and cluster.yaml Oct 22, 2018
config.yaml {make,plot}_summary: Add rules to flexibly create summaries and plot … Oct 26, 2018
environment.yaml add_electricity: Adapt to reverse compatible argument change in times… Sep 3, 2018
matplotlibrc Add matplotlibrc to automatically use Agg backend Sep 19, 2018

PyPSA-Eur: An Open Optimisation Model of the European Transmission System

PyPSA-Eur is an open model dataset of the European power system at the transmission network level that covers the full ENTSO-E area.

PyPSA-Eur Grid Model

The model is described and partially validated in the paper PyPSA-Eur: An Open Optimisation Model of the European Transmission System, 2018, arXiv:1806.01613.

This repository contains the scripts and some of the data required to automatically build the dataset from openly-available sources.

Already-built versions of the model can be found in the accompanying Zenodo repository.

The model is designed to be imported into the open toolbox PyPSA for operational studies as well as generation and transmission expansion planning studies.

The dataset consists of:

  • A grid model based on a modified GridKit extraction of the ENTSO-E Transmission System Map. The grid model contains 6001 lines (alternating current lines at and above 220kV voltage level and all high voltage direct current lines) and 3657 substations.
  • The open power plant database powerplantmatching.
  • Electrical demand time series from the OPSD project.
  • Renewable time series based on ERA5 and SARAH, assembled using the atlite tool.
  • Geographical potentials for wind and solar generators based on land use (CORINE) and excluding nature reserves (Natura2000) are computed with the vresutils library.

Building the model with the scripts in this repository uses up to 20GB of memory. Computing optimal investment and operation scenarios requires a strong interior-point solver compatible with the modelling library PYOMO like Gurobi or CPLEX with up to 100GB of memory (for the 356-bus approximation).

This project is maintained by the Energy System Modelling group at the Institute for Automation and Applied Informatics at the Karlsruhe Institute of Technology. It is currently funded by the Helmholtz Association. Previous versions were developed by the Renewable Energy Group at FIAS to carry out simulations for the CoNDyNet project, financed by the German Federal Ministry for Education and Research (BMBF) as part of the Stromnetze Research Initiative.


The steps are demonstrated as shell commands, where the path before the % sign denotes the directory in which the commands following the % should be entered.

Clone the repository using git (to a directory without any spaces in the path)

/some/other/path % cd /some/path/without/spaces
/some/path/without/spaces % git clone

Python dependencies

The python package requirements are curated in the conda environment.yaml file. The environment can be installed and activated using

.../pypsa-eur % conda env create -f environment.yaml
.../pypsa-eur % conda activate pypsa-eur   # or source activate pypsa-eur on older linux installations

Note that activation is local to the currently open shell! After opening a new terminal window, one needs to reissue the second command!

Data dependencies

Not all data dependencies are shipped with the git repository (since git is not suited for handling large changing files). Instead we provide two separate data bundles:

  1. pypsa-eur-data-bundle.tar.xz contains common GIS datasets like NUTS3 shapes, EEZ shapes, CORINE Landcover, Natura 2000 and also electricity specific summary statistics like historic per country yearly totals of hydro generation, GDP and POP on NUTS3 levels and per-country load time-series. It should be extracted in the data subdirectory (so that all files are in the data/bundle subdirectory)
.../pypsa-eur/data % curl -OL ""
.../pypsa-eur/data % tar xJf pypsa-eur-data-bundle.tar.xz
  1. pypsa-eur-cutouts.tar.xz are spatiotemporal subsets of the European weather data from the ECMWF ERA5 reanalysis dataset and the CMSAF SARAH-2 solar surface radiation dataset for the year 2013. They have been prepared by and are for use with the atlite tool. You can either generate them yourself using the build_cutouts snakemake rule or extract them directly in the pypsa-eur directory (extracting the bundle is recommended, since procuring the source weather data files for atlite is not properly documented at the moment):
.../pypsa-eur % curl -OL ""
.../pypsa-eur % tar xJf pypsa-eur-cutouts.tar.xz

Script overview

The model has several configuration options collected in the config.yaml file located in the root directory.

The generation of the model is controlled by the workflow management system Snakemake. In a nutshell, one declares in the Snakefile for each python script in the scripts directory a rule which describes which files the scripts consume and produce. snakemake then runs the scripts in the correct order and is able to track, what parts of the workflow have to be regenerated, when a data file or script is updated. For instance, with the Snakefile of pypsa-eur, an invocation to

snakemake networks/

follows the dependency graph Dependency graph for network elec_s_128

In detail this means it has to run the independent scripts,

  • build_shapes to generate GeoJSON files with country, exclusive economic zones and nuts3 shapes
  • build_cutout to prepare smaller weather data portions from ERA5 for cutout europe-2013-era5 and SARAH for cutout europe-2013-sarah.

With these and the externally extracted ENTSO-E online map topology, it can build the PyPSA basis model

  • base_network stored at networks/ with all buses, HVAC lines and HVDC links, and in
  • build_bus_regions determine the Voronoi cell of each substation.

Then it hands these over to the scripts for generating renewable and hydro feedin data,

  • build_hydro_profile for the hourly hydro energy availability,
  • build_renewable_potentials for the landuse/natura2000 constrained installation potentials for PV and wind,
  • build_renewable_profiles for the PV and wind hourly capacity factors in each Voronoi cell.
  • build_powerplants uses powerplantmatching to determine today's thermal power plant capacities and then locates the closest substation for each powerplant.

The central rule add_electricity then ties all the different data inputs together to a detailed PyPSA model stored in networks/, containing:

  • Today's transmission topology and capacities (optionally including lines which are under construction according to the config settings lines: under_construction and links: under_construction)
  • Today's thermal and hydro generation capacities (for the technologies listed in the config setting electricity: conventional_carriers)
  • Today's load time-series (upsampled according to population and gross domestic product)

It further adds extendable generators and storage_units with zero capacity for

  • wind and pv installations with today's locational, hourly wind and solar pv capacity factors (but no capacities)
  • long-term hydrogen and short-term battery storage units (if listed in electricity: extendable_carriers)
  • additional open-cycle gas turbines (if OCGT is listed in electricity: extendable_carriers)

The additional rules prepare approximations of the full model, in which generation, storage and transmission capacities can be co-optimized

  • simplify_network transforms the transmission grid to a 380 kV-only equivalent network, while
  • cluster_network uses a kmeans based clustering technique to partition the network into a certain number of zones and then reduce the network to a representation with one bus per zone.

The simplification and clustering steps are described in detail in the paper The role of spatial scale in joint optimisations of generation and transmission for European highly renewable scenarios, 2017, arXiv:1705.07617, doi:10.1109/EEM.2017.7982024.


For the use of snakemake, it makes sense to familiarize oneself quickly with its basic tutorial and then read carefully through the section Executing Snakemake, noting the arguments -n, -r, but also --dag, -R and -t.

The dependency graph shown above was generated using

snakemake --dag networks/ | dot -Tpng > dependency-graph-elec_s_128.png


The code in PyPSA-Eur is released as free software under the GPLv3, see LICENSE.