Abstract

Existing commercial buildings in Québec are responsible for a major share of GHG emissions within the local building sector, yet limited progress has been achieved in their operational transformations over the past decades. Building retrofit is recognized as a key approach to improving energy efficiency while considering the intertwining economic and environmental effects of the applied retrofit measures. Balancing these objectives transforms the problem into a high-dimensional, multi-objective, and challenging task requiring iterative simulations. When conventional physics-based approaches are used, such analyses become computationally prohibitive, with the challenge further intensifying when analyzing multiple buildings or extending the scope to future building performance.
This thesis addresses the highlighted challenge by developing surrogate models that allow the investigation of high-dimensional retrofit analyses. The study establishes a holistic building retrofit framework for post-industrialized buildings within the Montréal region to rigorously identify feasible retrofit measures specific to the studied building typology. It further develops robust surrogate models for energy consumption, embodied carbon, and investment costs that aid in decision-making across a wide array of measures. The methodology is further extended to integrate future building performance through feature extraction methods. Similarly, the study extends the generalizability of the surrogates to a multi-building analysis using a bottom-up approach, which integrates additional training data samples. Given the prevalence of data scarcity in commercial buildings, the methodology integrates building archetypes while identifying the modelling approach that accurately represents the studied building typology.
The findings demonstrate the robustness of the surrogate-based approach in building retrofit analyses, highlighting a significant improvement in computational efficiency. This thesis introduces a comprehensive methodological framework that advances building performance assessment using surrogate-based methods, identifying their limitations and effectiveness in the building retrofit research domain.