Exploring Deep Reinforcement Learning Algorithms for Enhanced HVAC Control

Heating, Ventilation, and Air Conditioning (HVAC) systems are one of the major sources of energy consumption in buildings. Typically, HVAC control has relied on reactive controllers, which often lack the ability to adapt to specific building dynamics. In recent years, Deep Reinforcement Learning (DRL) algorithms have emerged as a potential alternative to reactive controllers. However, these solutions are still immature, with a lack of standardisation and difficulties in real-world deployment. This paper presents an empirical evaluation of several state-of-the-art DRL algorithms for HVAC control, highlighting their main strengths and limitations. We emphasize the importance of using standard frameworks for comparative analysis, enabling a more comprehensive assessment of these innovative approaches to HVAC control.