"L2RPN in a sustainable world" NeurIPS competition is now Open

Track Robustness:

Track Adaptability:

Challenge Overview

After a successful first competition at IJCNN 2019 on a reduced version of our proposed challenge, two new competitions will run this year 2020 at WCCI and NeurIPS.

While first competitions aim at demonstrating the feasibility of applying Reinforcement Learning for controlling electrical flows on a power grid, the NeurIPS competition introduces a realistically-sized grid environment along with two fundamental real-life properties of power grid systems to reconsider while shifting towards a sustainable world: robustness and adaptability

Successive challenges in L2RPN series: maintenance events and redispatching actions are successively introduced, while the grid size increases and new harder game objectives are defined such as robustness and adptability.


2019 L2RPN competition paper published at 2020 PSCC Conference

Timeline for 2020 NeurIPS Competition

July 8th, 2020: Beginning of the competition with the release of public RL environment on 2 Tracks: robustness and adaptability. Participants can start submitting agent models on Codalab platform and obtaining immediate feedback in the leaderboard on validation scenarios.

September 7th, 2020: Potential release of a new baseline to foster competition if several participants are already doing better than this baseline.

October 19th, 2020: Start of the testing days on unseen test scenarios.

October 20th, 2020: End of the competition, beginning of the post-competition process

October 30th, 2020: Announcement of the L2RPN Winners.

December 6th, 2020: Beginning of NeurIPS 2020.

Winners of L2RPN 2020 WCCI Challenge

We saw great submissions until the last day and the competition was exciting until the last minute. The WCCI competition can still be overseen here:

This was a hard problem, so everyone among the more than 50 active participants who tried to solve it by making a submission can be proud for being bold!

Of course, this is a competition, and we still want to reward the 2 winning teams for their accomplishments:

1) Shhong - who remained on top of the leaderboard for most of the competition, congratulations!

2) Zenghsh3 - who became a strong challenger towards the end of the competition, getting close to 1rst rank actually.

Here is now a complete leaderboard for participants doing better than Do Nothing agent (score>0.0) and complying with the rules of the competition:

Final Leaderboard of WCCI competition for best compliant submissions

Kaist Team (shhong), the winning team, explains in the following video its approach to solve the problem

Ziming who ranked 3rd also describes its method in a video


Winners of L2RPN 2019 IJCNN Challenge

The competition was tight for the top 5 teams! Geirina Team eventually ranked 1rst and Learning_RL team ranked 2nd, making them the winners of this first episode of L2RPN Challenge! Congratulations to them, they did remarkable team work to reach this performance!

A comparable competition to IJCNN 2019 can still be overseen here: You can read more about it also in our paper that will be presented at PSCC 2020 Conference Learning to run a power network by training topology controllers

In addition, the winners both explain their respective approach in the following videos:

GEIRINA Presentation for L2RPN on ICJNN 2019.mp4

Both teams kindly released and open-source their code implementation at the following links: