Design and implementation of an environment for Learning to Run a Power Network (L2RPN)

TL;DR

This paper presents a software environment for simulating power grid management, enabling reinforcement learning agents to automate grid control and facilitating benchmarking and competitions in this domain.

Contribution

Developed a flexible, open-source simulation environment for power grid control, supporting reinforcement learning and benchmarking efforts.

Findings

Environment successfully simulates power grid dynamics.

Supports reinforcement learning agents for grid control.

Facilitates organized benchmarks and competitions.

Abstract

This report summarizes work performed as part of an internship at INRIA, in partial requirement for the completion of a master degree in math and informatics. The goal of the internship was to develop a software environment to simulate electricity transmission in a power grid and actions performed by operators to maintain this grid in security. Our environment lends itself to automate the control of the power grid with reinforcement learning agents, assisting human operators. It is amenable to organizing benchmarks, including a challenge in machine learning planned by INRIA and RTE for 2019. Our framework, built on top of open-source libraries, is available at https://github.com/MarvinLer/pypownet. In this report we present intermediary results and its usage in the context of a reinforcement learning game.