The development of district-level cooling energy profiles is important for the design, sizing, and
operation of district cooling plants. This task is often challenging due to the lack of building stock
data, input uncertainties, modeling strategies, and the time and space resolutions needed from the
cooling demand profiles. The work presented in this thesis attempts to address these challenges by
leveraging a data-driven methodology to produce monthly cooling demand profiles at the district
level for a hot and arid climate region. The methodology used in this work is comprised of two
main phases. The first conducts a comparative analysis to identify the optimal building archetype
development methodology suited for the studied climate zone. The second phase utilizes the
obtained results to develop an archetype library for the district’s building stock. This library is then
used to generate synthetic data for the training and testing of machine learning models’ districtscale energy prediction. The results obtained in this work indicate that for the building archetypes
studied, ML models can accurately predict profiles generated by physics-based models all while
reducing the computational time involved. The results also indicate that the uncertainty in
predicted cooling demand profiles remains considerable in the absence of realistic input
distributions. Ultimately, this study contributes to the current body of work by introducing a BA
library for high-rise buildings in hot and climate areas and by proposing an entire pipeline for data
generation, ML model training and testing, and data aggregation.