Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Demand Forecasting for Electric Vehicle Charging Stations using Multivariate Time-Series Analysis

Saba Sanami, Hesam Mosalli, Yu Yang, Hen-Geul Yeh, Amir G. Aghdam

TL;DR

The paper tackles the problem of day-ahead, 15-minute interval forecasting of EV charging demand, a critical input for infrastructure planning. It introduces a feature-rich multivariate LSTM with an attention mechanism and SHAP-based explanations to both improve accuracy and provide interpretability. Tested on campus charging data from CSU Long Beach, the approach demonstrates superior predictive performance over univariate and non-attentive baselines and yields insight into how weather, calendar, and temporal factors shape demand. The work offers practical benefits for infrastructure optimization, dynamic pricing, and resource management in EV charging networks, with potential for further expansion to additional drivers and station collaboration.

Abstract

As the number of electric vehicles (EVs) continues to grow, the demand for charging stations is also increasing, leading to challenges such as long wait times and insufficient infrastructure. High-precision forecasting of EV charging demand is crucial for efficient station management, to address some of these challenges. This paper presents an approach to predict the charging demand at 15-minute intervals for the day ahead using a multivariate long short-term memory (LSTM) network with an attention mechanism. Additionally, the model leverages explainable AI techniques to evaluate the influence of various factors on the predictions, including weather conditions, day of the week, month, and any holiday. SHapley Additive exPlanations (SHAP) are used to quantify the contribution of each feature to the final forecast, providing deeper insights into how these factors affect prediction accuracy. As a result, the framework offers enhanced decision-making for infrastructure planning. The efficacy of the proposed method is demonstrated by simulations using the test data collected from the EV charging stations at California State University, Long Beach.

Demand Forecasting for Electric Vehicle Charging Stations using Multivariate Time-Series Analysis

TL;DR

The paper tackles the problem of day-ahead, 15-minute interval forecasting of EV charging demand, a critical input for infrastructure planning. It introduces a feature-rich multivariate LSTM with an attention mechanism and SHAP-based explanations to both improve accuracy and provide interpretability. Tested on campus charging data from CSU Long Beach, the approach demonstrates superior predictive performance over univariate and non-attentive baselines and yields insight into how weather, calendar, and temporal factors shape demand. The work offers practical benefits for infrastructure optimization, dynamic pricing, and resource management in EV charging networks, with potential for further expansion to additional drivers and station collaboration.

Abstract

As the number of electric vehicles (EVs) continues to grow, the demand for charging stations is also increasing, leading to challenges such as long wait times and insufficient infrastructure. High-precision forecasting of EV charging demand is crucial for efficient station management, to address some of these challenges. This paper presents an approach to predict the charging demand at 15-minute intervals for the day ahead using a multivariate long short-term memory (LSTM) network with an attention mechanism. Additionally, the model leverages explainable AI techniques to evaluate the influence of various factors on the predictions, including weather conditions, day of the week, month, and any holiday. SHapley Additive exPlanations (SHAP) are used to quantify the contribution of each feature to the final forecast, providing deeper insights into how these factors affect prediction accuracy. As a result, the framework offers enhanced decision-making for infrastructure planning. The efficacy of the proposed method is demonstrated by simulations using the test data collected from the EV charging stations at California State University, Long Beach.

Paper Structure

This paper contains 12 sections, 8 equations, 8 figures, 3 tables.

Table of Contents

  1. INTRODUCTION
  2. Problem Statement
  3. Methodology
  4. LSTM
  5. Attention mechanism
  6. SHAP for feature importance
  7. Experimental testing of the method
  8. Data collection
  9. Data preprocessing
  10. Test results
  11. Explainable AI
  12. Conclusions

Figures (8)

  • Figure 1: Daily patterns of EV charging requests across different months
  • Figure 2: Weekly patterns of EV charging requests over one year
  • Figure 3: Multivariate data for selected features over one week
  • Figure 4: Actual and predicted charging requests
  • Figure 5: Average attention weights on an hourly basis
  • ...and 3 more figures