Analysis of Model-Agnostic Meta-Reinforcement Learning on Automated HVAC Control

Filiz U., Hekimoglu M. B., Alioğlu A., Ulusoy I.

33rd IEEE Conference on Signal Processing and Communications Applications, SIU 2025, İstanbul, Turkey, 25 - 28 June 2025, (Full Text) identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/siu66497.2025.11112293
  • City: İstanbul
  • Country: Turkey
  • Keywords: EnergyPlus, HVAC, MAML-DDQN, Model-Agnostic Meta-Learning, Reinforcement Learning, Sinergym
  • Middle East Technical University Affiliated: Yes

Abstract

This paper introduces a Model-Agnostic Meta-Reinforcement Learning framework for HVAC automation, integrating Model Agnostic Meta-Learning with Double Deep Q-Networks to improve adaptability across varying environmental conditions. The proposed approach is evaluated using Sinergym, an EnergyPlus-integrated RL Simulation framework, and benchmarked against conventional RL-based HVAC controllers. Results demonstrate that Model-Agnostic Meta-Learning integrated Double Deep Q-Network achieves a 7% reduction in overall power consumption while dynamically adapting to climate variations. These findings highlight the potential of Model Agnostic Meta-Learning in optimizing HVAC control strategies.