Abstract
Indoor temperature forecasting is an area of interest and importance as it contributes to improving HVAC systems control, thus reducing wasted energy and improving health and comfort. Acquiring high-quality transitory regime data for training Machine Learning models is challenging due to the scarcity of publicly available dataset. Additionally, such a dataset acquisition incurs high costs from repeated heating and cooling buildings in ranges of temperatures that go beyond normal operation thresholds. In response, we propose an open-source dataset called ‘RICO Dataset’. It is acquired in a dedicated and controlled physical test-building, alleviating potential issues encountered by digital simulation and modelling. It contains 305, four hours long 80-features rich multivariate transitory time series data from sensors in both internal and external environments sampled at a rate of one per minute.