Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Capturing Multivariate Dependencies of EV Charging Events: From Parametric Copulas to Neural Density Estimation

Martin Výboh, Gabriela Grmanová

Abstract

Accurate event-based modeling of electric vehicle (EV) charging is essential for grid reliability and smart-charging design. While traditional statistical methods capture marginal distributions, they often fail to model the complex, non-linear dependencies between charging variables, specifically arrival times, durations, and energy demand. This paper addresses this gap by introducing the first application of Vine copulas and Copula Density Neural Estimation framework (CODINE) to the EV domain. We evaluate these high-capacity dependence models across three diverse real-world datasets. Our results demonstrate that by explicitly focusing on modeling the joint dependence structure, Vine copulas and CODINE outperform established parametric families and remain highly competitive against state-of-the-art benchmarks like conditional Gaussian Mixture Model Networks. We show that these methods offer superior preservation of tail behaviors and correlation structures, providing a robust framework for synthetic charging event generation in varied infrastructure contexts.

Capturing Multivariate Dependencies of EV Charging Events: From Parametric Copulas to Neural Density Estimation

Abstract

Accurate event-based modeling of electric vehicle (EV) charging is essential for grid reliability and smart-charging design. While traditional statistical methods capture marginal distributions, they often fail to model the complex, non-linear dependencies between charging variables, specifically arrival times, durations, and energy demand. This paper addresses this gap by introducing the first application of Vine copulas and Copula Density Neural Estimation framework (CODINE) to the EV domain. We evaluate these high-capacity dependence models across three diverse real-world datasets. Our results demonstrate that by explicitly focusing on modeling the joint dependence structure, Vine copulas and CODINE outperform established parametric families and remain highly competitive against state-of-the-art benchmarks like conditional Gaussian Mixture Model Networks. We show that these methods offer superior preservation of tail behaviors and correlation structures, providing a robust framework for synthetic charging event generation in varied infrastructure contexts.

Paper Structure

This paper contains 17 sections, 4 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Our Contribution
  4. Data and Experimental Setup
  5. Methodology
  6. Model Architectures
  7. CODINE
  8. GMMNet
  9. Parametric Copulas
  10. Training Procedure
  11. Evaluation Framework
  12. Density and Distributional Fidelity
  13. Dependence Structure and Tail Behavior
  14. Generalization
  15. Grid Impact
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Bivariate density contours for Arrival Time vs. Duration (Trondheim dataset). The data exhibits a multimodal distribution that the parametric copula fails to capture, whereas the neural-based CODINE recovers the multimodal dependency structure.
  • Figure 2: Comparison of bivariate density contours (Duration vs. Energy) for the Dundee dataset. The plots visualize the z-space, highlighting CODINE's superior ability to capture non-linear dependence and tail structures compared to the GMMNet baseline.
  • Figure 3: Average daily load curve for the Dundee dataset. While the Vine underestimates the load magnitude, the neural-based methods exhibit higher fidelity in replicating the average load, with CODINE having lower $\text{MAE}_{\text{Load}}$.