Abstract
Identification of grey-box models to describe the heat dynamics of buildings is treated extensively in literature, However, to identify these models, data from controlled heating experiments is needed, which is not feasible in many cases. In this work, the aim is to overcome this threshold by using datasets from simulations of detailed white-box models to identify the grey-box models. Given that the white-box models are validated and calibrated against real measurement data, the need for real measurement data is bypassed, thus enabling the use of grey-box models identified from simulation data in real - world predictive controllers. The results show that a threestate grey-box model is able to capture the heat dynamics of the white-box model over a time period of ~1 week in the summer months.