Risk-Aware Multi-Agent Reinforcement Learning for Modern Energy Networks

Speaker: Sarah Keren (Assistant Professor) at The Taub Faculty of Computer Science at Technion)

Title: Risk-Aware Multi-Agent Reinforcement Learning for Modern Energy Networks 

The rapidly changing architecture of electrical networks, coupled with the rising integration of renewable and distributed energy resources, introduces various operational challenges and new sources of uncertainty. These have rendered traditional centralized energy-market paradigms insufficient due to their inability to support the constantly evolving nature of the network and the presence of non-regulated agents that dynamically respond to market signals. In this talk, I will first specify key computational challenges in managing energy networks and markets and then present how multi-agent reinforcement learning can mitigate the associated challenges. Specifically, I will explore our development of risk-aware reinforcement learning strategies that account for the objectives and constraints of both the grid-edge agents and the system operator and our integration of robust optimization techniques to enhance stability and efficiency.