Reinforcement Learning Environment for Cyber-Resilient Power Distribution System

Recently, numerous data-driven approaches to control an electric grid using machine learning techniques have been investigated. Reinforcement (RL)-based provide a credible alternative conventional, optimization-based solvers especially when there is uncertainty in the environment, such as renewable generation or cyber system performance. Efficiently training agent, however, requires interactions with environment learn best policies. There are RL environments for power systems, and, similarly, communication systems. Most simulators based UNIX while Windows operating system. Hence of cyber-physical, mixed-domain has challenging. Existing co-simulation methods efficient, but resource and time intensive generate large-scale data sets agents. Hence, this work focuses on development validation OpenDSS leveraging discrete event simulator Python package, SimPy system, which agnostic. Further, we present results agents cyber-physical network reconfiguration Volt-Var problem distribution feeder.