Household electricity consumption in Harbin: a motivation-environment-behavior framework with a residential functional spatial accessibility index
Residential electricity consumption varies across urban neighborhoods, but existing forecasting models often provide limited explanation for these spatial differences. This study proposes a Motivation-Environment-Behavior framework to organize the analysis of household electricity consumption. The framework links occupancy motivation, geographical environment, and electricity-related behavior. To represent the geographical environment, this study develops the Residential Functional Spatial Accessibility Index (RFSAI). RFSAI measures functional accessibility between residential buildings and surrounding urban facilities by combining spatial proximity with facility importance weights derived from household survey data. Using Harbin, China, as the study area, this research integrates 1,823 complete household questionnaire records with residential building data, urban facility data, RFSAI values, vacancy information, and residential electricity consumption data. A backpropagation neural network model was constructed to predict residential electricity consumption at the building level. The model was evaluated with an independent test set. The mean absolute percentage error on the independent test set was 9.4%. Partial correlation analysis showed that RFSAI was associated with residential electricity consumption after selected socioeconomic variables were controlled. Feature ablation experiments showed that the removal of RFSAI variables increased the prediction error by more than 6% points. Replacing RFSAI variables with random noise did not restore model performance. The results indicate that the MEB-based feature structure and RFSAI variables provide useful information for predicting and interpreting spatial differences in residential electricity consumption in the Harbin dataset. The findings support the proposed analytical pathway at the level of prediction and association. They do not provide formal causal identification. This study provides a measurable way to include residential facility accessibility in household electricity consumption modeling and provides empirical evidence for urban residential energy analysis in the Harbin case.