Scholarlink

<p>Accurate estimation of battery state of health (SOH) is essential for prolonging battery lifespan and optimizing battery management systems. This study proposes a hybrid SOH estimation method that integrates data-driven models with equivalent circuit models. Health-related features are extracted from charging-phase measurements and equivalent circuit parameters, and their importance is assessed using the random forest algorithm. Meanwhile, the conventional long short-term memory (LSTM) network is enhanced by incorporating quantum neuron structures and a memory augmentation mechanism. Furthermore, the Kepler Optimization Algorithm (KOA) is employed to automatically tune the key hyperparameters of the proposed QLSTM model, resulting in the KOA-QLSTM framework for battery SOH prediction. Validation using both public datasets and real-world aging data demonstrates that the proposed KOA-QLSTM model achieves excellent predictive performance, with an average root mean square error (RMSE) of less than 1%, indicating high accuracy and robustness.</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

State of health estimation for lithium-ion batteries based on KOA-QLSTM

  • 期刊论文

  • 发表日期：

    2025年10月28日

查看全文
  • Yuanjin Zhang,
  • Huawei Wu,
  • Congjin Ye

摘要

Accurate estimation of battery state of health (SOH) is essential for prolonging battery lifespan and optimizing battery management systems. This study proposes a hybrid SOH estimation method that integrates data-driven models with equivalent circuit models. Health-related features are extracted from charging-phase measurements and equivalent circuit parameters, and their importance is assessed using the random forest algorithm. Meanwhile, the conventional long short-term memory (LSTM) network is enhanced by incorporating quantum neuron structures and a memory augmentation mechanism. Furthermore, the Kepler Optimization Algorithm (KOA) is employed to automatically tune the key hyperparameters of the proposed QLSTM model, resulting in the KOA-QLSTM framework for battery SOH prediction. Validation using both public datasets and real-world aging data demonstrates that the proposed KOA-QLSTM model achieves excellent predictive performance, with an average root mean square error (RMSE) of less than 1%, indicating high accuracy and robustness.