Scholarlink

<p>Accurately estimating the State of Health (SoH) of lithium-ion batteries is vital for designing safe battery management systems (BMS). This is particularly important in electric vehicles, where battery health directly affects driving range, safety, and overall lifespan. In this study, a parallel hybrid deep learning model that leverages the strengths of both Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks is presented. While LSTMs are well-suited for capturing long-term dependencies in battery degradation, GRUs provide efficiency in modelling shorter-term patterns. By running these architectures in parallel, our framework learns a richer representation of degradation dynamics, improving predictive accuracy. To ensure the model’s robustness and adaptability, it was evaluated on two widely recognised datasets i.e. NASA and CALCE. Further enhanced performance using Bayesian optimisation for hyperparameter tuning and 5-fold cross-validation to minimise the risk of overfitting. Recognising that trust and interpretability are essential in safety-critical applications such as electric mobility, integrated Explainable AI (XAI) techniques were applied. In particular, SHAP (SHapley Additive exPlanations) was used to reveal how different input features influence the model’s predictions, providing transparency and interpretability in the decision-making process. Overall, this work presents a powerful, efficient, and explainable framework for predictive health monitoring of lithium-ion batteries, combining high performance with safety for 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

Explainable parallel GRU–LSTM model for predictive maintenance of lithium-ion batteries

  • 期刊论文

  • 发表日期：

    2026年3月7日

查看全文
  • Rahul Kumar Kamboj,
  • Mukesh Singh,
  • Ashima Singh,
  • Anju Bala

摘要

Accurately estimating the State of Health (SoH) of lithium-ion batteries is vital for designing safe battery management systems (BMS). This is particularly important in electric vehicles, where battery health directly affects driving range, safety, and overall lifespan. In this study, a parallel hybrid deep learning model that leverages the strengths of both Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks is presented. While LSTMs are well-suited for capturing long-term dependencies in battery degradation, GRUs provide efficiency in modelling shorter-term patterns. By running these architectures in parallel, our framework learns a richer representation of degradation dynamics, improving predictive accuracy. To ensure the model’s robustness and adaptability, it was evaluated on two widely recognised datasets i.e. NASA and CALCE. Further enhanced performance using Bayesian optimisation for hyperparameter tuning and 5-fold cross-validation to minimise the risk of overfitting. Recognising that trust and interpretability are essential in safety-critical applications such as electric mobility, integrated Explainable AI (XAI) techniques were applied. In particular, SHAP (SHapley Additive exPlanations) was used to reveal how different input features influence the model’s predictions, providing transparency and interpretability in the decision-making process. Overall, this work presents a powerful, efficient, and explainable framework for predictive health monitoring of lithium-ion batteries, combining high performance with safety for real-world applications.