Logo: to the web site of Uppsala University

Ambitions to reduce energy use in the building sector are increasing in order to mitigate global climate change. There is an increasing need for research to understand how to reach large-scale energy efficiency targets for buildings, and what the outcome would be if certain measures were implemented. Urban building energy modelling (UBEM) has received increased attention due to its inherent capacities for large-scale modelling of building stocks, and potential of varying data inputs, approaches and applications. In this thesis, based on a comprehensive review of the state of the art, a UBEM development approach utilising open-access data and probabilistic techniques was taken. Cadastral data, energy performance certificates, and socio-economic data were used to identify representative archetype buildings using a data-driven approach. The developed UBEM framework further includes new approaches to probabilistic parameter modelling. The model was applied to Gotland, Sweden, as a case study, to analyse the potential for using UBEM techniques for large, urban-rural regions. This case study also includes an evaluation of the aforementioned techniques, concluding that the developed model accurately captures the inherent diversity in energy use in residential building stocks. The developed model was applied to the historic building stock of Södermalm in Stockholm, Sweden, where renovation strategies were developed based on heritage value classifications. The identified archetypes are representative when compared to measured data, and simulated energy savings are 10-39% depending on the scenario. The savings when simulating a balanced scenario show that EU energy efficiency targets can be achieved for a historical building stock while avoiding measures that risk damaging heritage values. The final results show the possibilities of combining building data, heritage classification, and UBEM to form a reliable basis for analysis of building stocks balancing energy use and heritage values.