Power System Vulnerability Identification and Defense Through Deep Reinforcement Learning

Drew Levin Factsheet

Project Overview:

We seek to develop a Deep Reinforcement Learning (DRL) algorithm to reveal operational vulnerabilities and generate defensive strategies for an adversarial power systems scenario. New features of our implementation method will include advances in machine learning which will be applied to networked infrastructure systems. Our novel method will provide new capabilities to identify vulnerabilities and learn defensive strategies. Our work will dramatically advance the resilience of networked infrastructure systems, permitting the discovery of potential cyber-physical vulnerabilities and the development of possible mitigation strategies