Comparative analysis of algorithmic paradigms for HVAC control across diverse building systems: A systematic review

Abstract Heating, Ventilation and Air Conditioning (HVAC) systems buildings account for over 60% of global building energy consumption, yet their HVAC systems predominantly rely on conventional control strategies with limited energy-saving capabilities. This review systematically analyses over 30 studies to transcend existing limitations by shifting focus from system-specific implementations to four core algorithmic paradigms: rule-based/heuristic, model predictive control (MPC), data-driven/learning-based, and hybrid control. Performance evaluation across Direct Expansion (DX), Fan Coil Unit with Fresh Air (FCU+FA), and Variable Air Volume (VAV) systems reveals critical multi-objective trade-offs among energy efficiency, thermal comfort, indoor air quality (IAQ), and computational demands. Analysis shows that while rule-based methods offer straightforward deployment, they struggle to adapt to dynamic nonlinear environments; MPC excels theoretically but faces real-time deployment barriers due to computational intensity; and data-driven approaches remain constrained by dependencies on high-quality training data and infrastructure readiness. In contrast, hybrid frameworks integrate long-term optimization with real-time adaptability, significantly enhancing comprehensive performance, achieving up to 32.06% energy savings. Future research must overcome persistent challenges including humidity-temperature coupling, local optimization, and over-ventilation, thereby advancing HVAC systems toward carbon neutrality and intelligent operation.