Abstract

The decentralization of energy markets, driven by electrification and renewable energy adoption, necessitates a shift from a “follow-the-supply” to a “follow-the-demand” model, requiring grid operators to digitalize and modernize infrastructure to better accommodate distributed energy production and balance demand. In this context, energy aggregators play a crucial role by enabling building clusters to function as unified entities, thereby optimizing interactions with day-ahead coordination and intra-day markets.
This thesis investigates two key aspects of demand-side management: the role of energy aggregators in shaping residential load profiles and the development of optimal aggregation strategies. These aspects have been scarcely investigated, especially by exploiting measured data. To achieve these goals, a hierarchical control methodology is proposed for energy aggregators to coordinate individual homes within clusters. Leveraging data from smart thermostats and power meters, the methodology integrates predictive modelling and control strategies at a building cluster scale. To this aim, buildings are modelled as reduced-order grey-box networks, capturing thermal dynamics in a simplified yet accurate manner. Machine learning techniques are employed to support the creation of these data-driven models, ensuring robustness and adaptability. Advanced control techniques, such as the economic Model Predictive Control, evaluate energy flexibility by comparing performance to reference demand profiles, ensuring adherence to technical constraints while minimizing the economic expense. A Monte Carlo estimation technique is used to account for variability and heterogeneity within portfolios, facilitating probabilistic decision-making to address uncertainties and diverse operational conditions. Clustering techniques are then used to propose a flexibility-informed benchmarking procedure, supporting the creation of control archetypes and diverse strategies.
The proposed methodology is validated through three case studies and measured dataset of varying populations and resolution: (1) the Experimental House for Building Energetics in Shawinigan, Quebec, a fully instrumented, unoccupied research house to test real implementation of demand-side management; (2) a virtual community of 30 houses in Trois-Rivières, Quebec, equipped with smart thermostats and sensors; and (3) a dataset of approximately 100,000 monitored homes provided by a North American thermostat company enables the creation of energy aggregation strategies and prediction of their aggregated impact on the grid. These case studies demonstrate the effectiveness of energy aggregators in optimizing household costs and support transactive power grid management.
By addressing the dual objectives of cost minimization for customers and operational efficiency for grid operators, this research contributes to the design and implementation of energy aggregators as key enablers of the “follow-the-demand” energy paradigm. This thesis provides actionable insights into energy flexibility and portfolio management, paving the way for the scalable deployment of sustainable and efficient energy systems.