Grid2Op

Grid2Op is a platform, built with modularity in mind, that allows to perform powergrid operation. And that's what it stands for: Grid To Operate. Grid2Op acts as a replacement of pypownet as a library used for the Learning To Run Power Network L2RPN.

This framework allows to perform most kind of powergrid operations, from modifying the setpoint of generators, to load shedding, performing maintenance operations or modifying the topology of a powergrid to solve security issues.

Official documentation: the official documentation is available at https://grid2op.readthedocs.io/.

• 1 Installation
  • 1.1 Setup a Virtualenv (optional)
  • 1.2 Install from source
  • 1.3 Install from PyPI
  • 1.4 Install for contributors
  • 1.5 Docker
• 2 Main features of Grid2Op
• 3 Getting Started
  • 0 Basic features
  • 1 BaseObservation Agents
  • 2 BaseAction Grid Manipulation
  • 3 Training An BaseAgent
  • 4 Study Your BaseAgent
• 4 Citing
• 5 Documentation
• 6 Contribute
• 7 Test and known issues
• 8 License information

Installation

Requirements:

• Python >= 3.6

Setup a Virtualenv (optional)

Create a virtual environment

cd my-project-folder
pip3 install -U virtualenv
python3 -m virtualenv venv_grid2op

Enter virtual environment

source venv_grid2op/bin/activate

Install from source

git clone https://github.com/rte-france/Grid2Op.git
cd Grid2Op
pip3 install -U .
cd ..

Install from PyPI

pip3 install grid2op

Install for contributors

git clone https://github.com/rte-france/Grid2Op.git
cd Grid2Op
pip3 install -e .
pip3 install -e .[optional]
pip3 install -e .[docs]

Docker

Grid2Op docker containers are available on dockerhub.

To install the latest Grid2Op container locally, use the following:

docker pull bdonnot/grid2op:latest

Main features of Grid2Op

Core functionalities

Built with modulartiy in mind, Grid2Op acts as a replacement of pypownet as a library used for the Learning To Run Power Network L2RPN.

Its main features are:

• emulates the behavior of a powergrid of any size at any format (provided that a backend is properly implemented)
• allows for grid modifications (active and reactive load values, generator voltages setpoints and active production)
• allows for maintenance operations and powergrid topological changes
• can adopt any powergrid modeling, especially Alternating Current (AC) and Direct Current (DC) approximation to when performing the compitations
• supports changes of powerflow solvers, actions, observations to better suit any need in performing power system operations modeling
• has an RL-focused interface, compatible with OpenAI-gym: same interface for the Environment class.
• parameters, game rules or type of actions are perfectly parametrizable
• can adapt to any kind of input data, in various format (might require the rewriting of a class)

Powerflow solver

Grid2Op relies on an open source powerflow solver (PandaPower), but is also compatible with other Backend. If you have at your disposal another powerflow solver, the documentation of grid2op/Backend can help you integrate it into a proper "Backend" and have Grid2Op using this powerflow instead of PandaPower.

Getting Started

Some Jupyter notebook are provided as tutorials for the Grid2Op package. They are located in the getting_started directories.

These notebooks will help you in understanding how this framework is used and cover the most interesting part of this framework:

• 00_Introduction and 00_SmallExample describe what is adressed by the grid2op framework (with a tiny introductions to both power systems and reinforcement learning) and give and introductory example to a small powergrid manipulation.
• 01_Grid2opFramework covers the basics of the Grid2Op framework. It also covers how to create a valid environment and how to use the Runner class to assess how well an agent is performing rapidly.
• 02_Observation details how to create an "expert agent" that will take pre defined actions based on the observation it gets from the environment. This Notebook also covers the functioning of the BaseObservation class.
• 03_Action demonstrates how to use the BaseAction class and how to manipulate the powergrid.
• 04_TrainingAnAgent shows how to get started with reinforcement learning with the grid2op environment. It shows the basic on how to train a "PPO" model operating the grid relying on "stable baselines 3" PPO implementation.
• 05_StudyYourAgent shows how to study an BaseAgent, for example the methods to reload a saved experiment, or to plot the powergrid given an observation for example. This is an introductory notebook. More user friendly graphical interface should come soon.
• 06_Redispatching_Curtailment explains what is the "redispatching" and curtailment from the point of view of a company who's in charge of keeping the powergrid safe (aka a Transmission System Operator) and how to manipulate this concept in grid2op. Redispatching (and curtailment) allows you to perform continuous actions on the powergrid problem.
• 07_MultiEnv details how grid2op natively support a single agent interacting with multiple environments at the same time. This is particularly handy to train "asynchronous" agent in the Reinforcement Learning community for example.
• 08_PlottingCapabilities shows you the different ways with which you can represent (visually) the grid your agent interact with. A renderer is available like in many open AI gym environment. But you also have the possibility to post process an agent and make some movies out of it, and we also developed a Graphical User Interface (GUI) called "grid2viz" that allows to perform in depth study of your agent's behaviour on different scenarios and even to compare it with baselines.
• 09_EnvironmentModifications elaborates on the maintenance, hazards and attacks. All three of these represents external events that can disconnect some powerlines. This notebook covers how to spot when such things happened and what can be done when the maintenance or the attack is over.
• 10_StorageUnits details the usage and behaviour of the storage units in grid2op.
• 11_IntegrationWithExistingRLFrameworks explains how to use grid2op with other reinforcement learning framework.

Try them out in your own browser without installing anything with the help of mybinder:

Or thanks to google colab (all links are provided near the notebook description)

Citing

If you use this package in one of your work, please cite:

@misc{grid2op,
    author = {B. Donnot},
    title = {{Grid2op- A testbed platform to model sequential decision making in power systems. }},
    year = {2020},
    publisher = {GitHub},
    journal = {GitHub repository},
    howpublished = {\url{https://github.com/rte-france/grid2op}},
}

Documentation

The official documentation is available at https://grid2op.readthedocs.io/.

Build the documentation locally

A copy of the documentation can be built if the project is installed from source: you will need Sphinx, a Documentation building tool, and a nice-looking custom Sphinx theme similar to the one of readthedocs.io. These can be installed with:

pip3 install -U grid2op[docs]

This installs both the Sphinx package and the custom template.

Then, on systems where make is available (mainly gnu-linux and macos) the documentation can be built with the command:

make html

For windows, or systems where make is not available, the command:

sphinx-build -b html docs documentation

This will create a "documentation" subdirectory and the main entry point of the document will be located at index.html.

It is recommended to build this documentation locally, for convenience. For example, the "getting started" notebooks referenced some pages of the help.

Contributing

We welcome contributions from everyone. They can take the form of pull requests for smaller changed. In case of a major change (or if you have a doubt on what is "a small change"), please open an issue first to discuss what you would like to change.

To contribute to this code, you need to:

1. fork the repository located at https://github.com/rte-france/Grid2Op
2. synch your fork with the "latest developement branch of grid2op". For example, if the latest grid2op release on pypi is 1.6.5 you need to synch your repo with the branch named dev_1.6.6 or dev_1.7.0 (if the branch dev_1.6.6 does not exist). It will be the highest number in the branches dev_* on grid2op official github repository.
3. implement your functionality / code your modifications or anything else
4. make sure to add tests and documentation if applicable
5. once it is developed, synch your repo with the last development branch again (see point 2 above) and make sure to solve any possible conflicts
6. write a pull request and make sure to target the right branch (the "last development branch")

Code in the contribution should pass all the tests, have some dedicated tests for the new feature (if applicable) and documentation (if applicable).

Before implementing any major feature, please write a github issue first.

Tests and known issues

Tests performed currently

Grid2op is currently tested on windows, linux and macos.

The unit tests includes testing, on linux machines the correct integration of grid2op with:

• python 3.7
• python 3.8
• python 3.9
• python 3.10

On all of these cases, we tested grid2op on all available numpy version >= 1.20 (nb available numpy versions depend on python version).

The complete test suit is run on linux with the latest numpy version on python 3.8.

NB grid2op should work on python 3.11 but as numba is not (yet) compatible with this python version, and because numba is used for some tests, we do not test grid2op with python 3.11 yet.

Known issue

Due to the underlying behaviour of the "multiprocessing" package on windows based python versions, the "multiprocessing" of the grid2op "Runner" is not supported on windows. This might change in the future, but it is currently not on our priorities.

Perform tests locally

Provided that Grid2Op is installed from source:

Install additional dependencies

pip3 install -U grid2op[optional]

Launch tests

cd grid2op/tests
python3 -m unittest discover

License information

Copyright 2019-2020 RTE France

RTE: http://www.rte-france.com

This Source Code is subject to the terms of the Mozilla Public License (MPL) v2 also available here