Scholarlink

<p>Accurate knee point detection in lithium-ion batteries is critical for ensuring electric vehicle safety and preventing sudden performance degradation. However, existing methods struggle under nonlinear degradation patterns and diverse chemistries. Accurate knee point prediction facilitates timely maintenance, optimized battery usage, and reduced operational risks before reaching end of life (EOL). This study introduces a novel hybrid machine learning (ML) model, XGBoost-random forest (XG-RF), which leverages data from the initial 500 battery cycles of NMC cells (2.5 Ah) and LFP cells (3.2 Ah), which were analysed at 0&#xa0;°C, 25&#xa0;°C, and 35&#xa0;°C enhanced predictive accuracy is evaluated using Mean Absolute Deviation (MAD) and Mean Absolute Error (MAE). Several recent studies emphasize the significance of early-life information (&lt; 500 cycles) for lifespan estimation and degradation phase detection. Our approach aligns with this trend by leveraging only early-cycle data for knee prediction, reducing training time and enabling proactive diagnostics. The XG-RF model outperforms conventional methods such as support vector machine (SVM), AdaBoost, and neural networks (NN), achieving mean absolute error (MAE) values of 0.13 to 0.15 and mean absolute deviation (MAD) values of 0.11 to 0.14 respectively. Knee point for NMC and LFP cells at 0&#xa0;°C, 25&#xa0;°C, and 35&#xa0;°C, compared against existing models including SVM, AdaBoost, and Bacon-Watts. In comparison, traditional methods like tangent ratio and maximum distance exhibited higher root mean square error (RMSE) values ranging from 4.15 to 12.90, underscoring their limitations in handling non-linear degradation patterns. The XG-RF model, combined with the slope change method, enables early and precise detection of critical transitions, offering proactive degradation assessments that enhance battery safety and extend operational lifespan. The findings set a benchmark for accurate knee point prediction, advancing research in energy sustainability and automotive safety while addressing challenges in real-world applications.</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

Advanced knee point prediction and health monitoring technique for improved electric vehicle battery safety

  • 期刊论文

  • 发表日期：

    2026年5月20日

查看全文
  • Namrata Mohanty,
  • Neeraj Kumar Goyal,
  • V. N. Achutha Naikan

摘要

Accurate knee point detection in lithium-ion batteries is critical for ensuring electric vehicle safety and preventing sudden performance degradation. However, existing methods struggle under nonlinear degradation patterns and diverse chemistries. Accurate knee point prediction facilitates timely maintenance, optimized battery usage, and reduced operational risks before reaching end of life (EOL). This study introduces a novel hybrid machine learning (ML) model, XGBoost-random forest (XG-RF), which leverages data from the initial 500 battery cycles of NMC cells (2.5 Ah) and LFP cells (3.2 Ah), which were analysed at 0 °C, 25 °C, and 35 °C enhanced predictive accuracy is evaluated using Mean Absolute Deviation (MAD) and Mean Absolute Error (MAE). Several recent studies emphasize the significance of early-life information (< 500 cycles) for lifespan estimation and degradation phase detection. Our approach aligns with this trend by leveraging only early-cycle data for knee prediction, reducing training time and enabling proactive diagnostics. The XG-RF model outperforms conventional methods such as support vector machine (SVM), AdaBoost, and neural networks (NN), achieving mean absolute error (MAE) values of 0.13 to 0.15 and mean absolute deviation (MAD) values of 0.11 to 0.14 respectively. Knee point for NMC and LFP cells at 0 °C, 25 °C, and 35 °C, compared against existing models including SVM, AdaBoost, and Bacon-Watts. In comparison, traditional methods like tangent ratio and maximum distance exhibited higher root mean square error (RMSE) values ranging from 4.15 to 12.90, underscoring their limitations in handling non-linear degradation patterns. The XG-RF model, combined with the slope change method, enables early and precise detection of critical transitions, offering proactive degradation assessments that enhance battery safety and extend operational lifespan. The findings set a benchmark for accurate knee point prediction, advancing research in energy sustainability and automotive safety while addressing challenges in real-world applications.