Establishing energy usage profiles of urban residential households: A case study of the Beijing-Tianjin-Hebei region

With improving living standards, energy consumption in urban residential households has continuously increased. In 2020, the energy consumption during the operational phase of buildings accounted for 21.3% of the total energy consumption in China, with the urban residential energy consumption accounting for 38.7% of the energy consumption during the operational phase of buildings. Energy usage in urban residential households is overly complex and differs considerably among households. For sustainable energy saving and emission reduction, it is important to monitor energy conservation in urban residential households by accurately analyzing and identifying user characteristics of different households. Therefore, we develop an urban residential household energy usage profile model using a multidimensional perspective of user profiling theory.

Herein, we focused on urban residential households in the Beijing-Tianjin-Hebei region by establishing a labeling system for household energy usage profiling in six dimensions: household attributes, building features, household appliances, energy usage behaviors, energy consumption, and use of renewable energy; this labeling system included 18 indicators. A total of 351 valid household energy usage datasets were collected through surveys and semistructured interviews. A comprehensive search method was used to calculate the silhouette coefficients of the 18 indicators using different numerical combinations. Backward feature selection was used to filter the indicators based on their average silhouette coefficients. This process was terminated when the insignificance of indicators led to inconsistent results across different indicator sets. Consequently, the silhouette coefficient of the household energy dataset clustering was>0.5. The remaining indicators represented the optimal subset of the household energy usage profile indicators. Finally, the optimal number of clusters k was determined using the elbow method principle. The k-means algorithm was applied to cluster analysis of urban residential households. The t-distributed stochastic neighbor embedding (TSNE) dimensionality reduction method reduced multidimensional data to two dimensions and visualized the distribution of different urban residential households, demonstrating the scientific approach used in this study.

(1) The optimal subset of energy usage profile indicators for urban residential households in the Beijing-Tianjin-Hebei region includes eight indicators: household population, building area, house use pattern, number of appliances, air conditioning behavior, annual electricity consumption, annual gas consumption, and number of solar energy equipment. (2) The optimal number of clusters for the household energy dataset is four. (3) Using the optimal subset, the k-means clustering algorithm classifies urban residential households in the Beijing-Tianjin-Hebei region into energy utilization quality pursuit, energy-saving potential, energy utilization regularity, and energy conservation and environmental protection types.

By analyzing the characteristics of the four types of urban residential household energy usage profiles in the Beijing-Tianjin-Hebei region, we examine the energy-saving potential of households with energy utilization quality pursuit, energy-saving potential, and energy utilization regularity types and propose energy-saving recommendations. We provide new insights for relevant departments to understand the energy usage characteristics of urban residential households in the Beijing-Tianjin-Hebei region and develop accurate energy conservation strategies based on household types.