Agent-based modeling framework for predicting regional electricity consumption considering occupant behavior shift and exogenous policy impact

• This model addresses spatial–temporal challenges for predicting urban electricity consumption in diverse and mature city. • Agent-based approach evaluates shifts in electricity usage, accounting for cluster transitions and socio-economic factors • Integrating Geographically Weighted Regression and Decision Trees mitigates heterogeneity in predicting citywide energy consumption patterns. • Findings reveal decision probabilities and impacts on the occupants’ electricity usage behavior, urban energy dynamics, and energy policy formulation. This study introduces a model to predict regional electricity consumption in urban areas with mixed land use patterns and high energy intensity. It employs an agent-based approach to tackle the temporal-spatial and heterogeneity distribution challenges brought about by rapid urbanization. The model evaluates potential shifts in electricity usage driven by occupants’ behavior, considering the urban environment, demographics, and energy policies. Using the GAMA platform, it forecasts electricity consumption for 907 agents in Taipei City, Taiwan. Integrating 5 clusters of 5 independent variables, the model utilizes Geographically Weighted Regression (GWR) and Decision Trees (DT) for analysis. Findings indicate a higher shift likelihood among residential occupants (Cluster 1 and Cluster 4) than commercial occupants (Cluster 2 and Cluster 3), especially in areas with diverse clusters. Consequently, the projected electricity consumption in 2030 is only 2 kWh/household lower than in 2020, underscoring the challenge of achieving carbon neutrality through changes in occupants’ electricity usage behavior in a mature city. Future research aims to enhance accuracy by delving into cluster shift mechanisms and socio-economic backgrounds. Despite the hurdles, the model contributes to understanding urban energy dynamics and informs policy formulation for energy-efficient practices.