Home // ENERGY 2024, The Fourteenth International Conference on Smart Grids, Green Communications and IT Energy-aware Technologies // View article

On Sound Experiment Execution with Learning Agents in Cyber-Physical Energy Systems

Authors:
Eric MSP Veith
Stephan Balduin
Arlena Wellßow
Torben Logemann

Keywords: agent systems; learning agents; reinforcement learning; complex co-simulation; cyber-physical systems; modelling and simulation

Abstract:
Autonomous and learning systems, such as Multi-Agent Systems (MASs) or learning agents based on Deep Reinforcment Learning (DRL), have firmly estblished themselves as a foundation for approaches to create resilient and efficient Cyber-Physical Energy Systems (CPESs). A substantial amount of research into different aspects of these systems is backed by simulation. However, the presentation of the simulation setup, experiment design, and experiment results evaluation often lacks crucial information, making it hard to reproduce or compare to other researcher’s results. In this paper, we present the experiment design tooling of arsenAI, a part of the palaestrAI software ecosystem. We describe the work in progress on the experiment definition and mechanisms in place to aid in sound and reproducible experimentation with learning agents in co-simulated CPESs.

Pages: 14 to 19

Copyright: Copyright (c) IARIA, 2024

Publication date: March 10, 2024

Published in: conference

ISSN: 2308-412X

ISBN: 978-1-68558-139-8

Location: Athens, Greece

Dates: from March 10, 2024 to March 14, 2024