Scholarlink

<p>As the electric vehicle market expands, the importance of driving performance evaluation in the electric vehicle development process is increasing. Simulation-based performance estimation is more economical than real vehicle driving tests, as it can predict performance without the need for physical experiment. However, existing simulation models often show discrepancies because they struggle to accurately replicate real-world vehicle characteristics and driving conditions. In this study, a data-based electric vehicle performance prediction technique using Gaussian process regression (GPR) and piecewise affine regression and classification (PARC) is proposed. GPR provides not only predicted values but also the uncertainty of the prediction, enabling reliability evaluation of model predictions, and PARC provides high computational efficiency by dividing the input space into several clusters and applying local linear models to each area. Two regression models, GPR and PARC, trained on actual electric vehicle driving data, are applied to a black-box model based only on data and a gray-box model that integrates data with physical simulation models, and their performances are compared. As a result, the data-driven approach is confirmed to be effective in improving simulation accuracy, suggesting its potential to enhance both the efficiency and reliability of EV simulation-based development processes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Data-Driven Approaches to Energy Efficiency Prediction in Electric Vehicles

  • 期刊论文

  • 发表日期：

    2026年7月13日

查看全文
  • Sujin Ko,
  • Seungyeon Ryu,
  • Jiyun Han,
  • Jungwon Han,
  • Miriam Di Russo,
  • Heeyun Lee

摘要

As the electric vehicle market expands, the importance of driving performance evaluation in the electric vehicle development process is increasing. Simulation-based performance estimation is more economical than real vehicle driving tests, as it can predict performance without the need for physical experiment. However, existing simulation models often show discrepancies because they struggle to accurately replicate real-world vehicle characteristics and driving conditions. In this study, a data-based electric vehicle performance prediction technique using Gaussian process regression (GPR) and piecewise affine regression and classification (PARC) is proposed. GPR provides not only predicted values but also the uncertainty of the prediction, enabling reliability evaluation of model predictions, and PARC provides high computational efficiency by dividing the input space into several clusters and applying local linear models to each area. Two regression models, GPR and PARC, trained on actual electric vehicle driving data, are applied to a black-box model based only on data and a gray-box model that integrates data with physical simulation models, and their performances are compared. As a result, the data-driven approach is confirmed to be effective in improving simulation accuracy, suggesting its potential to enhance both the efficiency and reliability of EV simulation-based development processes.