Electric Vehicle Charging Demand Prediction Model Based on Spatiotemporal Attention Mechanism DOI Creative Commons
Yang Chen, Zeyang Tang, Yanyan Cui

et al.

Energies, Journal Year: 2025, Volume and Issue: 18(3), P. 687 - 687

Published: Feb. 2, 2025

The accurate estimation and prediction of charging demand are crucial for the planning infrastructure, grid layout, efficient operation networks. To address shortcomings existing methods in utilizing spatial interdependencies among urban regions, this paper proposes a forecasting approach that integrates dynamic time warping (DTW) with spatial–temporal attention graph convolutional neural network (ASTGCN). First, method delves into correlations between various regions within target city, establishing intricate coupling relationships them. Subsequently, FastDTW algorithm is employed to construct an adjacency matrix, capturing spatiotemporal correlation different regions. Finally, ASTGCN model applied predict power load each region, which can accurately capture characteristics load. experimental results indicate proposed has more powerful comprehensive ability improve accuracy stability steps.

Language: Английский

Traffic accident risk prediction based on deep learning and spatiotemporal features of vehicle trajectories DOI Creative Commons
Hao Li, Lian-Yi CHEN

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(5), P. e0320656 - e0320656

Published: May 2, 2025

With the acceleration of urbanization and increase in traffic volume, frequent accidents have significantly impacted public safety socio-economic conditions. Traditional methods for predicting often overlook spatiotemporal features complexity networks, leading to insufficient prediction accuracy complex environments. To address this, this paper proposes a deep learning model that combines Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), Graph (GNN) accident risk using vehicle trajectory data. The extracts spatial such as speed, acceleration, lane-changing distance through CNN, captures temporal dependencies trajectories LSTM, effectively models structure with GNN, thereby improving accuracy.The main contributions are follows: First, an innovative combined is proposed, which comprehensively considers road network relationships, accuracy. Second, model’s strong generalization ability across multiple scenarios validated, enhancing traditional methods. Finally, new technical approach provided, offering theoretical support implementation real-time warning systems. Experimental results demonstrate can predict risks various scenarios, providing robust intelligent management safety.

Language: Английский

Citations

0
Electric Vehicle Charging Demand Prediction Model Based on Spatiotemporal Attention Mechanism DOI Creative Commons
Yang Chen, Zeyang Tang, Yanyan Cui

et al.

Energies, Journal Year: 2025, Volume and Issue: 18(3), P. 687 - 687

Published: Feb. 2, 2025

The accurate estimation and prediction of charging demand are crucial for the planning infrastructure, grid layout, efficient operation networks. To address shortcomings existing methods in utilizing spatial interdependencies among urban regions, this paper proposes a forecasting approach that integrates dynamic time warping (DTW) with spatial–temporal attention graph convolutional neural network (ASTGCN). First, method delves into correlations between various regions within target city, establishing intricate coupling relationships them. Subsequently, FastDTW algorithm is employed to construct an adjacency matrix, capturing spatiotemporal correlation different regions. Finally, ASTGCN model applied predict power load each region, which can accurately capture characteristics load. experimental results indicate proposed has more powerful comprehensive ability improve accuracy stability steps.

Language: Английский

Citations

0