Re­modelling buildings Data­efficient models to evaluate energy efficiency in operative buildings

More than half of the world’s population lives in cities, and the number is expected to rise in the coming decades. Given the current population growth rate, it is necessary to optimise how energy is used in the building stock to ensure sustainable urban development. The building sector is being digitalised due to the appearance of cheaper monitoring devices and an increase in computational power. The access to these data has caused a rise of data­driven models to study how energy is used in buildings. These models aim to disentangle the numerous processes governing energy flow to develop strategies to increase energy efficiency. Despite the advance in data­collecting methods, it is often challenging to use building models that require numerous variables since relying on many sensors increases complexity fast, and potential issues during installation and maintenance may occur. In addition, data is collected by different stakeholders, so access to consolidated databases is scarce. Moreover, extensive monitoring of occupied buildings might raise privacy concerns. Thus, we are in a transition period where on the one hand, we acknowledge the need to take measurements from buildings, but, on the other hand, the available data do not match the expectations. This work explores modelling techniques to develop data­efficient building models that are flexible, computationally light, and easy to interpret. We study traditional building models based on physics principles and propose simpler model structures that hold physical interpretation. In addition, we investigate statistical methods so our models can assimilate complex phenomena that are often impossible to incorporate using traditional heat transfer principles. The results are two­fold: i) we see that the proposed methods are flexible and can enhance current building models to work with limited resources; ii) the modelling methods discussed in this thesis serve as a foundation to tailor specific models to describe concrete processes that occur in occupied buildings. Thus, the outcomes of this thesis are useful for researchers and practitioners that want to study building energy use on a large scale. In particular, we discuss different applications such as characterising the building stock, quantifying building energy flexibility, and simulation tools for urban planning.