Home // ENERGY 2023, The Thirteenth International Conference on Smart Grids, Green Communications and IT Energy-aware Technologies // View article

An Architecture for Reliable Learning Agents in Power Grids

Authors:
Eric Veith

Keywords: agent systems; reinforcement learning; trustworthy AI; resilience; power grid

Abstract:
Agent systems have become almost ubiquitous in smart grid research. Research can be roughly divided into carefully designed (multi-) agent systems that can perform known tasks with guarantees, and learning agents based on technologies, such as Deep Reinforcment Learning (DRL), that promise real resilience by learning to counter the unknown unknowns. However, the latter cannot give guarantees regarding their behavior, while the former are limited to the set of problems known at design time. This paper presents a hybrid architecture that enables a learning agent to give guarantees about its behavior, making it suitable for usage in Critical National Infrastructures (CNIs), such as the power grid.

Pages: 13 to 16

Copyright: Copyright (c) IARIA, 2023

Publication date: March 13, 2023

Published in: conference

ISSN: 2308-412X

ISBN: 978-1-68558-054-4

Location: Barcelona, Spain

Dates: from March 13, 2023 to March 17, 2023