Answers are grounded in this paper and cite their sources.
The growing interest to improve HVAC energy efficiency has motivated development of advanced supervisory control strategies, including data-driven and machine learning-based approaches. Model predictive control (MPC) is a well-established framework that predicts thermal dynamics and proactively adjusts system and zone operations to reduce energy use while maintaining comfort. Despite strong results in simulations and testbeds, relatively few studies report on MPC performance and deployment challenges in operational buildings. This paper presents a field deployment of an inverse-model-based MPC in a multi-zone VAV AHU commercial office building in Ottawa, Canada, developed using historical sensor data to model zone-level thermal behaviour and forecast disturbances. Compared to baseline rule-based control, the MPC reduced cooling energy by 4.6% and cut early-morning start-up cooling loads by 15.8%, while decreasing total discharge airflow by 52.5%. Heating energy was reduced by 28.7%, with a 54.3% reduction during the later-afternoon period preceding end-of-occupancy. In-situ deployment revealed several challenges, including semantic data integration, unexpected equipment behaviour, and unresolved system faults, which increased deployment lead times and constrained achievable energy savings. The results highlight the importance of robust data infrastructure and practical integration over model complexity.
No takes yet. Share an insight, caveat, or question.
Markus et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: