A Reinforcement Learning-based Volt-VAR Control Dataset

This repository contains a suite of Volt-VAR control (VVC) benchmarks for conducting research on sample efficient, safe, and robust RL-based VVC algorithms. It includes the IEEE 13, 123, and 8500-node test feeders wrapped as a gym-like environment along with baseline algorithm implementations to reproduce the results of [1]

Installation

1. Download .zip of this repository
2. Install the packages

TODO

Usage

Run baseline algorithms

Set environment and algorithm parameters in the config variable in main.py, then run the program

python main.py

Implement custom off-policy RL algorithms

Create implement the update(), act_deterministic, and act_probabilistic functions in the /algos/template.py file

References

[1] Y. Gao and N. Yu, “A reinforcement learning-based volt-VAR control dataset and testing environment,” arXiv.org, 20-Apr-2022. [Online]. Available: https://arxiv.org/abs/2204.09500.

Citing this benchmark

To cite this benchmark, please cite the following paper:

@misc{gao2022dataset,
  doi = {10.48550/ARXIV.2204.09500},
  url = {https://arxiv.org/abs/2204.09500},
  author = {Gao, Yuanqi and Yu, Nanpeng},
  title = {A Reinforcement Learning-based Volt-VAR Control Dataset and Testing Environment},
  publisher = {arXiv},
  year = {2022},
}