Transfer Learning with TD3 for Adaptive HVAC Control in Diverse Building Environments

Research output: Chapter in Book or Conference Publication/ProceedingConference Publicationpeer-review

1 Citation (Scopus)

Abstract

This paper studies the application of the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm on two heterogeneous transfer scenarios. Transfer learning has shown to be effective in addressing challenges faced in RL for HVAC control by leveraging knowledge acquired during the development of an agent for one building to tackle a problem related to another building. However, buildings exhibit significant variability in size, construction materials, and geographical location; thus, simply transferring neural networks would be a challenge because of the need to adapt to diverse building characteristics. In this research, we extend prior work and investigate the efficacy of transfer learning with the TD3 algorithm. We use this algorithm to optimise HVAC control systems across different building environments. Our experimental results demonstrate the competitive performance of our transfer learning methods compared to rule-based control and training from scratch. Our transfer learning methods see up to 2–3% improvement in performance when compared to these agents. Overall, this study highlights the potential of transfer learning with the TD3 algorithm to enhance adaptive HVAC control systems in diverse building environments.

Original languageEnglish
Title of host publicationHighlights in Practical Applications of Agents, Multi-Agent Systems, and Digital Twins
Subtitle of host publicationThe PAAMS Collection - International Workshops of PAAMS 2024, Proceedings
EditorsAlfonso González-Briones, Vicente Julian Inglada, Alia El Bolock, Cedric Marco-Detchart, Jaume Jordan, Karl Mason, Fernando Lopes, Nada Sharaf
PublisherSpringer Science and Business Media Deutschland GmbH
Pages256-267
Number of pages12
ISBN (Print)9783031730573
DOIs
Publication statusPublished - 2025
EventInternational Workshops on Practical Applications of Agents and Multi-Agent Systems, PAAMS 2024 - Salamanca, Spain
Duration: 26 Jun 202428 Jun 2024

Publication series

NameCommunications in Computer and Information Science
Volume2149 CCIS
ISSN (Print)1865-0929
ISSN (Electronic)1865-0937

Conference

ConferenceInternational Workshops on Practical Applications of Agents and Multi-Agent Systems, PAAMS 2024
Country/TerritorySpain
CitySalamanca
Period26/06/2428/06/24

Keywords

  • Continuous HVAC control
  • Reinforcement learning
  • Transfer learning

Fingerprint

Dive into the research topics of 'Transfer Learning with TD3 for Adaptive HVAC Control in Diverse Building Environments'. Together they form a unique fingerprint.

Cite this