Scholarlink

<p>Under cold conditions, the driving range of electric vehicles decreases significantly, and inaccuracies in the displayed remaining driving range (<i>RDR</i>) exacerbate range anxiety. This study proposes a knowledge-enhanced hierarchical framework that breaks down the <i>RDR</i> estimation problem into the prediction of energy consumption rate and effective energy coefficient. Both modules employ deep learning as their core models, using data sourced from a cloud-based big data platform with a focus on cold regions in Northeast China. To address real-world driving scenarios, the energy consumption rate module uses a switching mechanism: a base model, using region-specific collaborative features as inputs, is applied in the early stages of trips, while a sequential neural network is used in the later stages. The effective energy coefficient module incorporates battery degradation and environmental factors, correcting discrepancies in nominal battery energy under low-temperature and aging conditions. The model’s performance is validated using real-world data from 8 electric vehicles under cold conditions, demonstrating a 15–20% improvement in prediction accuracy over traditional methods, thereby enhancing <i>RDR</i> accuracy and reliability.</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

A Knowledge-Enhanced Modular Method for Predicting Electric Vehicle Remaining Driving Range under Cold Conditions Utilizing Cloud-Based Big Data

  • 期刊论文

  • 发表日期：

    2025年7月14日

查看全文
  • Yunfeng Hu,
  • Hong Liu,
  • Yao Sun,
  • Xun Gong,
  • Fengxin Zhao,
  • Zhen Cheng,
  • Chong Zhang,
  • Ke Xu

摘要

Under cold conditions, the driving range of electric vehicles decreases significantly, and inaccuracies in the displayed remaining driving range (RDR) exacerbate range anxiety. This study proposes a knowledge-enhanced hierarchical framework that breaks down the RDR estimation problem into the prediction of energy consumption rate and effective energy coefficient. Both modules employ deep learning as their core models, using data sourced from a cloud-based big data platform with a focus on cold regions in Northeast China. To address real-world driving scenarios, the energy consumption rate module uses a switching mechanism: a base model, using region-specific collaborative features as inputs, is applied in the early stages of trips, while a sequential neural network is used in the later stages. The effective energy coefficient module incorporates battery degradation and environmental factors, correcting discrepancies in nominal battery energy under low-temperature and aging conditions. The model’s performance is validated using real-world data from 8 electric vehicles under cold conditions, demonstrating a 15–20% improvement in prediction accuracy over traditional methods, thereby enhancing RDR accuracy and reliability.