Authors: David Biagioni, Xiangyu Zhang, Dylan Wald, Deepthi Vaidhynathan, Rhoit Chintala, Jennifer King, Ahmed S. Zamzam
Corresponding author: David Biagioni
All authors are with the National Renewable Energy Laboratory (NREL).
PowerGridworld provides users with a lightweight, modular, and customizable framework for creating power-systems-focused, multi-agent Gym environments that readily integrate with existing training frameworks for reinforcement learning (RL). Although many frameworks exist for training multi-agent RL (MARL) policies, none can rapidly prototype and develop the environments themselves, especially in the context of heterogeneous (composite, multidevice) power systems where power flow solutions are required to define grid-level variables and costs. PowerGridworld is an opensource software package that helps to fill this gap. To highlight PowerGridworld’s key features, we include two case studies and demonstrate learning MARL policies using both OpenAI’s multi-agent deep deterministic policy gradient (MADDPG) and RLLib’s proximal policy optimization (PPO) algorithms. In both cases, at least some subset of agents incorporates elements of the power flow solution at each time step as part of their reward (negative cost) structures.
Please refer to our published paper or preprint on arXiv for
more details. Data and run scripts used to generate figures in the paper
are available in the paper directory.
Env setup:
conda create -n gridworld python=3.8 -y
conda activate gridworld
git clone [email protected]:NREL/PowerGridworld.git
cd PowerGridWorld
pip install -e .
pip install -r requirements.txt
Run the pytests to sanity check:
pytest tests/
pytests --nbmake examples/envs
Examples of running various environments and MARL training algorithms can be found in examples.
This work was authored by the National Renewable Energy Laboratory (NREL), operated by Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. This work was supported by the Laboratory Directed Research and Development (LDRD) Program at NREL.
If citing this work, please use the following:
@inproceedings{biagioni2021powergridworld,
author = {Biagioni, David and Zhang, Xiangyu and Wald, Dylan and Vaidhynathan, Deepthi and Chintala, Rohit and King, Jennifer and Zamzam, Ahmed S.},
title = {PowerGridworld: A Framework for Multi-Agent Reinforcement Learning in Power Systems},
year = {2022},
isbn = {9781450393973},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3538637.3539616},
doi = {10.1145/3538637.3539616},
booktitle = {Proceedings of the Thirteenth ACM International Conference on Future Energy Systems},
pages = {565–570},
numpages = {6},
keywords = {deep learning, power systems, OpenAI gym, reinforcement learning, multi-agent systems},
location = {Virtual Event},
series = {e-Energy '22}
}