Identification, characterization, and stochastic simulation of residential electric vehicle (EV) charging behavior
Document Type
Article
Publication Date
8-1-2026
Abstract
The widespread integration of electric vehicles (EVs) into the residential sector has introduced significant stochasticity to grid operations. Synthesizing realistic occupant charging behaviors is critical for accurately assessing localized grid impacts; however, existing approaches often fail to bridge the gap between micro-level physical charging states and macro-level stochastic human routines. This paper proposes a hybrid data-driven framework dedicated to the generative simulation of individual occupant EV charging behaviors. Methodologically, we address this gap through a two-stage approach. First, a Gaussian Hidden Markov Model (GHMM) autonomously segments raw telemetry data to accurately distinguish active charging phases from standby idle noise. Second, a hierarchical stochastic model parametrizes the underlying occupant behavior by decoupling the dimensions of daily frequency, start time, and duration to fully capture temporal heterogeneity. During experimental evaluation, clustering analysis demonstrated that these extracted distributional parameters naturally segment the dataset into highly representative occupant profiles: Routine Overnight (RO), Routine Evening (RE), and Sporadic (SP) users. This natural segmentation rigorously validates the framework’s ability to encapsulate complex human habits into interpretable statistical traits. Leveraging this parameterized model, a scenario analysis of a synthetic community comprising 100 households reveals distinct grid stress patterns driven by these distinct occupant routines. While Routine Evening users dominate the traditional evening peak, synchronized Routine Overnight behaviors trigger a sharp rebound peak in the early morning, introducing new bottlenecks during nominally off-peak hours. Ultimately, these results demonstrate that the proposed framework accurately reproduces essential behavioral distributions and macroscopic load characteristics, providing a robust, generative basis for risk-informed grid impact assessments and privacy-aware scenario synthesis.
Publication Source (Journal or Book title)
Energy and Buildings
Recommended Citation
Wang, Z., Feng, F., & Pang, Z. (2026). Identification, characterization, and stochastic simulation of residential electric vehicle (EV) charging behavior. Energy and Buildings, 364 https://doi.org/10.1016/j.enbuild.2026.117641