A benchmarking framework for HVAC optimization via competitive evaluation: insights from the 2nd wccbo

This study presents a method to quantitatively evaluate HVAC optimization methods using a virtual emulator of a Variable Refrigerant Flow (VRF) system. Driven by the growing complexity and diversity of HVAC control strategies, including those employing artificial intelligence (AI), the proposed framework enables performance comparison on a unified scale.A simulation-based competition was conducted, where participants submitted control strategies to minimize energy consumption while maintaining occupant comfort. |Over 2,000 operational results were collected and analyzed. A percentile-based Energy efficiency and Comfort Index (ECI) was defined, and a sigmoid-based formula was introduced to normalize scores for fair comparison.The framework was also tested with reinforcement learning, showing that even AI-based methods can be evaluated using the same criteria. Two remaining challenges – overfitting due to repeated evaluation and limited generalizability to real buildings – are discussed, along with potential solutions such as server-managed evaluation environments and integration with real-world systems.