search

IntelligentSystemsLab / ST-EVCDP

A real-world dataset for EV-related research, e.g., spatiotemporal prediction and urban energy management.
MIT License
60 stars 15 forks source link
data-mining eletric-vehicle graph-learning spatio-temporal-prediction

readme

Spatio-temporal EVCDP (Shenzhen)

This is a real-world dataset for spatio-temporal electric vehicle (EV) charging demand prediction. If it is helpful to your research, please cite our paper:

Qu, H., Kuang, H., Li, J., & You, L. (2023). A physics-informed and attention-based graph learning approach for regional electric vehicle charging demand prediction. IEEE Transactions on Intellgent Transportation Systems. Paper in IEEE Explore Paper in arXiv

@ARTICLE{10539613,
  author={Qu, Haohao and Kuang, Haoxuan and Wang, Qiuxuan and Li, Jun and You, Linlin},
  journal={IEEE Transactions on Intelligent Transportation Systems}, 
  title={A Physics-Informed and Attention-Based Graph Learning Approach for Regional Electric Vehicle Charging Demand Prediction}, 
  year={2024},
  pages={1-14},
  doi={10.1109/TITS.2024.3401850}}

Author: Haohao Qu (haohao.qu@connect.polyu.hk)

Updates

Data Description

ST-EVCDP

The data used in this study is drawn from a publicly available mobile application, which provides the real-time availability of charging piles (i.e., idle or not). Within Shenzhen, China, a total of 18,061 public charging piles are covered during the studied period from 19 June to 18 July 2022 (30 days) with a minimum interval of 5 minutes and 8640 timestamps. As shown in Figure 1, the city is constructed into a graph-structure data with 247 nodes (traffic zones) and 1006 edges (adjacent relationships).

avatar Figure 1. Spatial distribution of the 18,061 public EV charging piles in ST-EVCDP.

Besides, the pricing schemes for the studied charging piles are also collected. Among the 247 traffic zones, 57 of them (enclosed in red lines) deploy time-based pricing schemes, while others use fixed ones. More statistical details are illustrated in the following table.

avatar

ST-EVCDP-v2

Expanding on the foundation of ST-EVCDP, we have gathered an extensive dataset called ST-EVCDP-v2, specifically tailored for EV-related research. This dataset covers a timeframe of one year, spanning from September 2022 to September 2023, which includes comprehensive information such as coordinates, charging occupancy, duration, volume, and price for a total of 1,682 public charging stations with 24,798 public charging piles. Notably, it provides detailed information on charging stations, with a granularity that allows analysis at the charging station level. And its temporal interval is one hour. You can download the data from Google Drive Link.

avatar Figure 2. Spatial distribution of the 24,798 public EV charging piles in ST-EVCDP-v2.

Files

ST-EVCDP

ST-EVCDP-v2

Enviroment Requirement

pip install -r requirements.txt

An simple example to run Spatio-temporal Prediction on the dataset

We developed a physics-informed and attention-based approach for spatio-temporal EV charging demand prediction, named PAG. Expect that, some representative methods are included, e.g., LSTM, and GCN-LSTM, GAT-LSTM. You can train and test the proposed model through the following procedures:

  1. Choose your model in line 45 of main.py or use the default model (PAG-) by skipping this procedure.
  2. Run main.py via Pycharm, etc. or change your ROOT_PATH and command:
cd [path] && python main.py

Extend:

More updates will be posed in the near future! Thank you for your interest.