Scholarlink

<p>With the rapid growth of Urban Air Mobility (UAM), electric vertical takeoff and landing (eVTOL) aircraft are emerging as a key solution for short-range transportation, placing stringent requirements on the reliability of lithium-ion batteries. Accurate state-of-health (SOH) prediction under complex flight missions is therefore essential. However, most existing approaches rely on constant-current (CC) discharge data and fail to capture the highly dynamic degradation patterns of eVTOL batteries. To address this gap, this study makes two key contributions. First, a phase-segmented feature extraction framework is developed to reflect the distinct discharge behaviors across different flight phases, and its rationality is validated through comparative analysis with CC datasets. Combined with Random Forest (RF) importance and mutual information (MI) analysis, nine representative health indicators are identified. Second, a hybrid optimization framework is proposed that embeds Levy flight and Brownian motion into a genetic algorithm (GA) with a complexity-regularized fitness function, enabling efficient hyperparameter tuning and enhanced model generalization. Experimental results on a publicly available eVTOL dataset show that the proposed approach improves <i>R</i><sup>2</sup> by 4.6% and reduces prediction errors by over 30% compared with RF and Long Short-Term Memory (LSTM) models optimized by standard GA. These results demonstrate the effectiveness of the proposed strategy for accurate and reliable SOH prediction in eVTOL battery management systems.</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

Phase-segmented feature extraction and hybrid optimization framework for state-of-health prediction of lithium-ion batteries in electric vertical takeoff and landing aircraft

  • 期刊论文

  • 发表日期：

    2026年3月14日

查看全文
  • Zhangang Yang,
  • Chengzheng Lai,
  • Yanan Zhang

摘要

With the rapid growth of Urban Air Mobility (UAM), electric vertical takeoff and landing (eVTOL) aircraft are emerging as a key solution for short-range transportation, placing stringent requirements on the reliability of lithium-ion batteries. Accurate state-of-health (SOH) prediction under complex flight missions is therefore essential. However, most existing approaches rely on constant-current (CC) discharge data and fail to capture the highly dynamic degradation patterns of eVTOL batteries. To address this gap, this study makes two key contributions. First, a phase-segmented feature extraction framework is developed to reflect the distinct discharge behaviors across different flight phases, and its rationality is validated through comparative analysis with CC datasets. Combined with Random Forest (RF) importance and mutual information (MI) analysis, nine representative health indicators are identified. Second, a hybrid optimization framework is proposed that embeds Levy flight and Brownian motion into a genetic algorithm (GA) with a complexity-regularized fitness function, enabling efficient hyperparameter tuning and enhanced model generalization. Experimental results on a publicly available eVTOL dataset show that the proposed approach improves R2 by 4.6% and reduces prediction errors by over 30% compared with RF and Long Short-Term Memory (LSTM) models optimized by standard GA. These results demonstrate the effectiveness of the proposed strategy for accurate and reliable SOH prediction in eVTOL battery management systems.