Towards smart individual-room heating for residential buildings

摘要

In many homes, residents keep their heating system always turned on although they are out or only occupy certain rooms, and thereby large amounts of energy are wasted. With our work, we aim to build an individual-room heating system that automatically detects occupancy, predicts a schedule based on that, and controls the heaters accordingly. First, we present our technical prototype for individual-room heating control. Second, we show that binary occupancy can be estimated using room climate sensors. We collected room climate data and occupancy data for three rooms over several days. We identified the relevant features and applied a Hidden Markov Model in a supervised and unsupervised way. We achieve a F1-score up to 85 % for both variants in rooms which are occupied for longer periods. Third, we describe how a well-known occupancy prediction approach should be integrated into our heating control for optimal performance.