Skip to content

Latest commit

 

History

History
238 lines (191 loc) · 9.84 KB

README.md

File metadata and controls

238 lines (191 loc) · 9.84 KB
Raw

ChroniX2Grid - The Extensive PowerGrid Time-serie Generator

Chronix2Grid is a python package, providing a command-line application as well, that allows to generate synthetic but realistic consumption, renewable production, electricity loss (dissipation) and economic dispatched productions chronics given a power grid. Reference data that you provide will serve to calibrate the parameters so that the synthetic data reproduce some realistic criteria (KPIs) from the reference data.

See the documentation at https://chronix2grid.readthedocs.io/en/latest/

Chronix2Grid at a glance

Chronix2grid_inputs

Chronix2grid_outputs

Installation

Requirements:

  • Python >= 3.6

(Optional, recommended) Step 1: Create a virtual environment

pip3 install -U virtualenv
python3 -m virtualenv venv_chronix2grid

Step 2 (first option): Install from pypi

source venv_chronix2grid/bin/activate
pip install Chronix2Grid

Step 2 (second option): Install from source

source venv_chronix2grid/bin/activate
git clone https://github.com/mjothy/ChroniX2Grid.git
cd ChroniX2Grid/
pip install -U .

Additional install required for dispatch

A backend for dispatch has been implemented with PyPSA

pip install pypsa==0.17.0

You might need to install cbc solver as well that pypsa will call: https://projects.coin-or.org/Cbc

Getting Started

Four notebooks are provided to get you started with this package:

  • getting_started_cli.ipynb guide you through the use of chronix2grid cli
  • getting_started_api.ipynb is a more detailed notebook that presents chronix2grid api and showcases several ways to analyze the chronics produced
  • running_chronics_grid2op.ipynb is a notebook giving an example of how generated chronics can further be used to create power grid scenarios and run them analyze the chronics produced
  • RenewableNinja_Solar_WInd_KPI_format.ipynb is an example on how to retrieve Solar and Wind reference chronics from the renewable ninja api. There will be use to compare the KPI's with the generated chronics.

Additionally, a data starting kit is provided to run an example in the folder

getting_started/example

The output folder corresponds to the following run of chronix2grid :

chronix2grid --mode LRTK --ignore-warnings --weeks 8 --n_scenarios 1 --start-date 2012-01-01 --by-n-weeks 4

NB:

  • Default input data from package will be used if no --input-data and --case is provided
  • Output will be written in the working directory in a folder output/

To provide --input-data and --output-folder folders and specify --case grid case, you would run this kind of command (make sure to change ChroniX2Grid_path with you own path) :

chronix2grid --mode RLTK --output-folder ChroniX2Grid_path/getting_started/example/output --input-folder ChroniX2Grid_path/getting_started/example/input --ignore-warnings --weeks 8 --case case118_l2rpn_wcci --n_scenarios 1 --start-date 2012-01-01 --by-n-weeks 4

KPIs generation with existing chronics as in grid2op env

If you already generated your data such as from a grid2op environment, you can generate KPIs afterwards under this cli given that a paramsKPI.json exists in the input-folder/kpi/grid_case

chronix2grid --mode K --output-folder pathTo/data_grid2op/l2rpn_wcci_2022/chronics --input-folder ChroniX2Grid_path/input_data --ignore-warnings --weeks 4 --case case118_l2rpn_wcci_2022 --n_scenarios 1

In your "chronics" folder, this will create a "kpi" folder with subfolders of KPIs per chronic

WARNING

In order to limit the size of the output and the running time for this example, chronics are only generated for 8 weeks. This implicates that some kpis that defined on a whole year will not be exploitable for this example.

The command-line interface

Usage: chronix2grid [OPTIONS]

Options:
  --case TEXT               case folder to base generation on
  --start-date TEXT         Start date to generate chronics
  --weeks INTEGER           Number of weeks to generate
  --by-n-weeks INTEGER      Size of the output chunks in weeks
  --n_scenarios INTEGER     Number of scenarios to generate
  --mode TEXT               Steps to execute : L(K) for loads only (and KPI);
                            R(K) for renewables (and KPI) only; LRT (K) 
                            for load, renewable and thermic generation (and KPI); 
                            LRDT(TK) for load, renewable, loss (dissipation) generation 
                            (and thermic and KPI)   

  --input-folder TEXT       Directory to read input files from.
  --output-folder TEXT      Directory to store output files.
  --seed-for-loads TEXT     Input seed to ensure reproducibility of loads
                            generation

  --seed-for-res TEXT       Input seed to ensure reproducibility of renewables
                            generation

  --seed-for-dispatch TEXT  Input seed to ensure reproducibility of dispatch
  --ignore-warnings         Ignore the warnings related to the existence of
                            data files in the chosen output directory.

  --scenario_name TEXT      subname to add to the generated scenario output
                            folder, as Scenario_subname_i

  --nb_core INTEGER         number of cores to parallelize the number of
                            scenarios

  --help                    Show this message and exit.

Launch mode

4 generation submodules and a KPI module are available

  • L - load generation
  • R - wind and solar production generation
  • D - loss generation a priori, that will be used for dispatch and potentially corrected afterwards
  • T - thermic and hydro production generation thanks to an economic dispatch (simplified optimal power flow simulation)
  • K - KPI generation in order to compare synthetic (generated) chronics to reference (real-life) chronics

The figure below shows how these submodules can be launched together with --mode/-m argument. Note that D and T submodules can't be launched without previous L and R modules, and that KPIs can always been computed

Launch_mode

Configuration

Chronic generation detailed configuration

Detailed configuration is made through thematic json files. For instance you will use in the current implementation of Chronix2grid:

  • params.json for general generation configuration (generation timestep, noise for forecast chronics)
  • params_load.json for load generation
  • params_res.json for solar and wind generation. You will use paramsGAN.json in case you are using the backend with Generative Adversarial Networks
  • params_loss.json for a priori loss generation
  • params_opf.json for dispatch and a posteriori loss simulation

Below is an example of params.json which will launch a 5 minutes time step generation, applying a gaussian noise to forecast chronics with a standard deviation of 0.01

{
  "dt": 5,
  "planned_std": "0.01"
}

Below is an example of params_load.json that provides parameters to correlated generation algorithm.

{
  "Lx": 1000,
  "Ly": 1000,
  "dx_corr": 250,
  "dy_corr": 250,
  "temperature_corr": 400,
  "std_temperature_noise": 0.06
}

You'll find all the details on the parameters used in these json files in the documentation about implemented models

KPI configuration

Some general parameters have to be set in INPUT_FOLDER/kpi/paramsKPI.json

  • comparison: name of benchmark folder to use for KPI reference chronics. For example, benchmark France has been implemented with eco2mix and renewable ninja data
  • timestep: timestep for KPI computation. For example, renewable ninja data requires minimum timestep of 60min
  • night_hours: dictionary to provide night hours for each season of year
  • seasons: dictionary to provide months in each season of year

Note: if you just want to generate the KPIs from your synthetic data without comparison to reference chronics, you can set

"comparison":null

Model interface

All generation submodules (LRDT) have a modular backend. You can develop your own load, renewable, loss and dispatch model using as input:

  • Your json parameters
  • Possible pattern files, or trained neural networks
  • Grid demand and generator characteristics in csv

Citing

@misc{chronix2grid,
    author = {A. Marot, N. Megel, V. Renault, M. Jothy },
    title = {{ChroniX2Grid - The Extensive PowerGrid Time-serie Generator}},
    year = {2020},
    publisher = {GitHub},
    journal = {GitHub repository},
    howpublished = {\url{https://github.com/BDonnot/ChroniX2Grid}},
    }

Running the test suite

To launch the unit test suite:

pipenv run python -m pytest tests/unit_tests/ [--verbose -p no:warnings]

To launch integration tests:

pipenv run python -m pytest tests/integration_tests/ [--verbose -p no:warnings]

To launch the Command Line Interface (CLI) test (only if you installed chronix2grid package from Pypi)

pipenv run python -m pytest tests/cli_tests/

You can also analyse the coverage of the tests with coverage and generate an html report:

pip install coverage
coverage run --source=./chronix2grid -m unittest discover
coverage html

This will generate a htmlcov folder containing a static web site with the analysis. Open index.html in a browser to analyse it.