Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Using Machine Learning to Generatively Simulate EV Charging in Urban Areas

Marek Miltner, Jakub Zíka, Daniel Vašata, Artem Bryksa, Magda Friedjungová, Ondřej Štogl, Ram Rajagopal, Oldřich Starý

TL;DR

The paper addresses predicting EV charging profiles in urban areas with limited data to assist DSOs in planning infrastructure. It introduces a generative neural network that learns $K$ latent archetypal load profiles conditioned on spatiotemporal inputs, with $K$ set to $4$ to capture distinct charging behaviors. Results reveal four interpretable archetypes (e.g., daytime public charging, morning and evening peaks, irregular multi-peak demand) and show that the local basic administrative unit type strongly shapes load curves. This approach enables scenario analysis and informed infrastructure expansion while acknowledging data limitations and the need for broader geographic validation.

Abstract

This study addresses the challenge of predicting electric vehicle (EV) charging profiles in urban locations with limited data. Utilizing a neural network architecture, we aim to uncover latent charging profiles influenced by spatio-temporal factors. Our model focuses on peak power demand and daily load shapes, providing insights into charging behavior. Our results indicate significant impacts from the type of Basic Administrative Units on predicted load curves, which contributes to the understanding and optimization of EV charging infrastructure in urban settings and allows Distribution System Operators (DSO) to more efficiently plan EV charging infrastructure expansion.

Towards Using Machine Learning to Generatively Simulate EV Charging in Urban Areas

TL;DR

The paper addresses predicting EV charging profiles in urban areas with limited data to assist DSOs in planning infrastructure. It introduces a generative neural network that learns latent archetypal load profiles conditioned on spatiotemporal inputs, with set to to capture distinct charging behaviors. Results reveal four interpretable archetypes (e.g., daytime public charging, morning and evening peaks, irregular multi-peak demand) and show that the local basic administrative unit type strongly shapes load curves. This approach enables scenario analysis and informed infrastructure expansion while acknowledging data limitations and the need for broader geographic validation.

Abstract

This study addresses the challenge of predicting electric vehicle (EV) charging profiles in urban locations with limited data. Utilizing a neural network architecture, we aim to uncover latent charging profiles influenced by spatio-temporal factors. Our model focuses on peak power demand and daily load shapes, providing insights into charging behavior. Our results indicate significant impacts from the type of Basic Administrative Units on predicted load curves, which contributes to the understanding and optimization of EV charging infrastructure in urban settings and allows Distribution System Operators (DSO) to more efficiently plan EV charging infrastructure expansion.

Paper Structure

This paper contains 16 sections, 1 equation, 10 figures, 2 tables.

Table of Contents

  1. Introduction and motivation
  2. Generative modeling of public charging demand
  3. Model inputs
  4. Model architecture
  5. Results
  6. Conclusion and further work
  7. Prague EV public charging landscape
  8. Prague area classification per basic administrative unit
  9. Charging demand localization in Prague
  10. Public charging load curve analysis
  11. Load-curve analysis based on basic administrative unit type
  12. Average load curves per month and weekday
  13. Weekday analysis
  14. Easter holiday analysis
  15. COVID lockdown total load impact analysis
  16. ...and 1 more sections

Figures (10)

  • Figure 1: These 4 plots showcase trained latent load curve shapes within the neural network architecture. Note that the latent profiles are probability distributions, and thus the sum of the area they define is equal to 1.
  • Figure 2: ZSJ categories found in Prague proper
  • Figure 3: Heatmap of public charging point locations per ZSJ in Prague based on the available data.
  • Figure 4: Number of chargers per ZSJ category across Prague
  • Figure 5: Temporal relative share development analysis for charging instances and installed chargers in Prague, classified per ZSJ type. Note that the red-bordered gray fill area in the upper chart of charging instances corresponds to the interpolated region of unavailable data, and the lower chart of the number of installed chargers includes chargers put into operation before the span of the chart timeline, December 2019.
  • ...and 5 more figures