Data and analytics for heating energy consumption of residential buildings: The case of a severe cold climate region of China

Building energy conservation has been a globally important issue not only because of the greenhouse gas emissions that occur during the construction and operation of a building, but also due to the increasing emphasis on indoor thermal quality, which is directly related to the quality of life, especially in northern China. The behaviors of heating companies and building users are considered to play important roles in the reduction of building energy consumption. However, it is necessary to understand the relationship between the heating fee, which represents the heating policy, and the energy performance, which considers both the actual energy consumption and thermal ability of a building. With this goal in sight, this study focuses on residential buildings in a climate zone that experiences quite low temperatures, and the heating energy consumption of 40 residential buildings from five typical regions in Inner Mongolia were monitored and analyzed. The thermal comfort, occupant behavior, and heating energy consumption of typical households and buildings in these regions were quantified. In view of the monitored results, building models were built and validated, with the results treated as training samples in clustering and classification procedures. For the sake of the further formulation of a heating energy cost allocation policy, the K-means clustering algorithm and discriminant analysis (DA) were used in this work to evaluate and group residential buildings according to their heating energy consumption and thermal-physical properties of the envelopes. The results of this study indicated the existence of regional heating energy consumption differences. Moreover, the main influential factors for grouping residential buildings and the number of clusters were suggested based on the K-means clustering algorithm and DA method. Meanwhile, the classification accuracy rate obtained from the DA method not only verified the clustering results, but also confirmed the further applicability of the combination of these two methods.