Large-scale HVAC systems in public buildings account for approximately 40% of total building energy consumption. This presentation reports an industrial deployment of an FMI-based Model Predictive Control (MPC) framework with AI-driven adaptive optimization for a university library's central air-conditioning system. A Modelica-based digital twin of the HVAC system was developed using FMI co-simulation architecture (Modelica Association 2026a; b). The digital twin is continuously updated with real-time IoT sensor data, enabling the MPC controller to minimize energy consumption while maintaining thermal comfort. An AI-driven iterative strategy automatically adjusts model parameters and control policies based on operational feedback (Zhang et al. 2026).
Field deployment through integration with the existing Building Automation System (BAS) achieved 15%–20% energy reduction versus conventional control, with 10-minute optimization cycles. The talk will share practical deployment experience, lessons learned from laboratory prototyping to industrial implementation, and open challenges in scaling AI-enhanced MPC for building energy management.