Abstract

With global efforts aimed at reaching carbon neutrality by 2050, there is an increased focus on improving the energy efficiency of buildings. The interactions between constructions and their local microclimate significantly influence the built environment and building energy performance. This thesis examines the urban microclimate and its impact on building energy consumption from the individual building level to entire urban areas.
Building energy models (BEMs) are essential for understanding building energy consumption, forecasting building energy, and evaluating energy-saving measures. Meanwhile Urban Building Energy Model (UBEM) is an analytical tool for modeling buildings on city levels and evaluating scenarios for an energy-efficient built environment. However, building planners commonly overestimate cooling loads by relying on Typical Meteorological Year (TMY) data in BEM/UBEM simulations, neglecting local microclimate variations and the neighborhood effects of surrounding buildings. This research developed an integrated platform by coupling BEM/UBEM with an urban microclimate model, allowing local aerodynamic data to be exchanged between the two models at each time step.
Since these BEM/UBEM models usually come with a deal of computation cost and prior knowledge to work with. In recent years, Machine Learning (ML) techniques in specific terms have been proposed for predicting building energy consumption. A synthetic dataset from physics-based simulations can serve as a training and testing data source for the ML model during the design phase. Weather clustering techniques are implemented to enhance computational efficiency and feasibility avoiding the high computational costs of day-by-day simulations. By employing weather clustering to select representative days, the approach reduces database size for training ML-based building prediction models.
The study begins with a comprehensive review of the latest methods for incorporating urban microclimate data into urban building energy models, addressing both methodological approaches and practical issues. Subsequently, the research evaluates the effects of urban microclimate on building energy performance, considering both individual buildings and urban-scale contexts. To address the computational cost associated with BEM/UBEM, an ML-based hourly building energy prediction model was developed, leveraging weather clustering techniques. The conclusion summarizes the key contributions of this thesis and offers recommendations for future research directions.