Koopman mode analysis on thermal data for building energy assessment

ABSTRACT Current approaches to thermal control and energy management in residential and office buildings rely on complex or high-dimensional thermal models. We provide a means to extract features from in-office thermal-data sensors which avoid the use of standard models. We develop these data-driven methods through the use of Koopman operator theory. We validate our resulting algorithms via analysing thermal data from a single thermal zone space. The particular advantage of the method is that it associates the temporal characteristics of control mechanisms with the corresponding spatial zones of influence. The methodology enables identification of spatial heating and cooling control modes directly from the data.