PhD Forum Abstract: Integrating data-driven and human-centric approaches for scalable and flexible building management

Buildings are pivotal for grid flexibility under rising renewable penetration. However, unlocking these potential faces scalability bottlenecks and risks occupant comfort. This work proposes a scalable, human-centric framework using ubiquitous smart-meter data to assess and harness building demand flexibility. We develop non-intrusive, data-parsimonious, physics-informed machine learning methods for probabilistic occupancy inference, occupant-driven load flexibility quantification, and sensor-free indoor thermal modeling. These foundations culminate in a fast, multi-objective Pareto optimization framework that comprehensively evaluates energy-comfort trade-offs for flexible HVAC operation. This research bridges building energy dynamics, human behavior, and grid operations, paving the way for resilient, transparent, and occupant-engaged energy transitions.