Abstract

The evolving energy landscape, driven by rising demand, electrification, and renewable energy integration, necessitates a shift from traditional “follow-the-load” model to demand-side management. This transition requires accurate prediction of building energy demand, effective demand response participation, and quantification of energy flexibility.
This thesis develops a methodology for predicting and optimizing building thermal energy demand using data from smart thermostats and monitoring infrastructures. Multi-zone buildings and schedule-based operations are modelled using resistance-capacitance (RC) thermal networks. An automated model order reduction approach identifies dominant thermal zones in multi-zone buildings, while control-oriented RC archetypes capture key dynamics in schedule-based operations. Calibration follows a Model Predictive Control Relevant Identification (MRI) process, ensuring models accurately predict thermal dynamics up to 24 hours ahead.
Weather variability is managed through clustering techniques that identify representative days, reducing computational complexity while enabling scenario-driven analysis. This approach bridges the gap between operational and design studies by integrating energy flexibility considerations early in building and community planning.
A distributed economic Model Predictive Control (e-MPC) framework optimizes thermal load management while maintaining occupant comfort and system constraints. It supports applications at both single-building and community scales, such as virtual power plants. Performance is assessed using energy flexibility Key Performance Indicators (efKPIs) against a reference scenario.
The methodology is validated through three case studies: (1) Residential buildings: 30 detached homes equipped with smart thermostats (data from Hydro-Québec); (2) Institutional building: The Varennes Net-Zero Energy Library, Canada’s first net-zero energy institutional building; (3) Community-scale system: A simulated hybrid photovoltaic-battery microgrid in Varennes serving residential and institutional buildings.
Findings highlight how varying building participation in demand response influences aggregated demand profiles, utility metrics (load shifting, peak shaving), and the sizing of grid-supportive technologies. At the single-building level, insights are provided for optimizing thermal load management across convective, radiant, and mixed heating systems. By integrating data-driven modelling, advanced control, and scalable design, this thesis provides actionable solutions for energy efficiency, flexibility, and resilience, supporting a sustainable energy transition.