Predicting household energy consumption in an aging society

Social aging will significantly affect energy consumption in the residential sector, with a critical influence on the demand side of the energy sector, especially for megacities. Despite previous studies on the total change in energy demand due to aging, limited research has focused on high-frequency effects. Using a machine learning model, this study examines how social aging will affect household hourly consumption patterns, covering both the change in total consumption and the hourly distribution. We first unsupervised cluster household high-frequency consumption patterns in Shanghai from 2016 to 2018 into 12 groups, and then trained a finite mixture model to analyze the correlation between clusters and household features. We further used the well-tuned model based on the out-of-sample result to simulate the consumption patterns under scenarios of social aging and income growth. The simulation results demonstrate that in addition to increasing overall energy consumption, an aging society will also change the hourly consumption pattern, leading to a larger gap between peak and non-peak periods.