Key points are not available for this paper at this time.
Heating, ventilation, and air-conditioning (HVAC) systems are central to building decarbonisation, yet improving indoor environmental quality (IEQ) often increases energy demand. This paper presents a systematic literature review of AI-driven HVAC control optimization for jointly improving energy efficiency and indoor environmental quality across different building types and climatic contexts. Following a PRISMA-based screening process, the review synthesizes 143 recent peer-reviewed studies published between 2020 and 2025. The analysis revealed supervised learning as the dominant ML technique for prediction and surrogate modelling and is frequently embedded in model predictive control (MPC). Reinforcement learning and deep reinforcement learning are increasingly applied to multi-objective supervisory control of setpoints, airflow, dampers and HVAC components because they offer model-free adaptability. Hybrid and ensemble approaches are emerging to improve robustness and trade-offs between energy efficiency and IEQ. Across diverse building typologies and climates, the reviewed studies reported an average energy savings of 40% compared with baseline cases, with several studies achieving savings above 50% while maintaining excellent IEQ. Despite these promising results, real-world validation remains limited as only 8 of 130 studies progressed to field or testbed deployment, highlighting a concerning gap between simulation-based research and practical validation. This review provides a comprehensive, context-sensitive mapping of AI-driven HVAC applications for enhancing both energy efficiency and IEQ in buildings.
Otoo et al. (2026) studied this question.