IoT Thrust Seminar | Data-Driven Optimization for Human-Centric and Cost-Effective Building Operations 

Buildings account for 30% of global energy consumption while hosting occupants for 90% of their time, underscoring the urgent need for intelligent, adaptive control strategies that balance comfort and efficiency. Traditional building control systems rely on static comfort models that fail to capture real-time occupant preferences, leading to inefficient energy use. In this talk, I will present a novel data-driven framework that integrates online preference learning with closed-loop performance optimization to enable adaptive and cost-effective building operations. By leveraging machine learning and optimization techniques, this approach dynamically adjusts control parameters—such as heating power limits—based on real-time occupant feedback and system performance. A case study in a residential setting demonstrates electricity bill savings of up to 27% compared to default controllers, showcasing the practical impact of our methods. Furthermore, our work won the 2024 BOPTEST Challenge for Smart Building HVAC Control, setting a new benchmark for intelligent building energy management. Beyond this application, I will discuss how our data-driven methodologies can be extended to broader challenges in energy systems, human-in-the-loop optimization, and autonomous decision-making.