After our recent success stories with winning this years Learning to Run a Power Grid Competition and Austrian state award for innovation we are now super excited to announce that our accompanying paper got accepted for presentation at the NeurIPS 2022 Reinforcement Learning for Real Life Workshop.

‍

Power Grid Congestion Management via Topology Optimization with AlphaZero

‍Matthias Dorfer*, Anton R. Fuxjäger*, Kristián Kozák*, Patrick M. Blies, Marcel Wasserer. (*equal contribution)

‍

“The transition to renewable sources like wind or solar leads to increased volatility in power generation,making grid management increasingly difficult and costly. We propose an AlphaZero-based topology optimization agentthat integrates well with existing grids and workflows, such as day-ahead planning,and achieves up to 60% reduction in required redispatching, decreasing costs and emissions.”

‍

We have been investing a lot of work and resources into our research and developments and are very proud to now get the chance to present at one of the top venues for ML research among such renowned organizations and researchers.

Thanks once more to Antoine Marot and his team at RTE France for enabling this line of research with their power grid simulations tools.We are extremely excited to see where this is goingand hope that our work will supports the decarbonization of the energy sectorand contribute to the green energy transition.

‍

Our paper: https://arxiv.org/abs/2211.05612

Additional resources (code, poster, etc.): https://github.com/enlite-ai/maze-l2rpn-2022-submission