点击次数：

影响因子：10.7

DOI码：10.1016/j.est.2026.124251

发表刊物：Journal of Energy Storage

关键字：Lithium-ion batteries; State of health estimation; Temporal; convolutional network; Gaussian process regression; Uncertainty quantification; Decommissioning thresholds

摘要：The intensifying reliability requirements for contemporary energy storage systems necessitate advanced State of Health (SOH) monitoring strategies that simultaneously address battery degradation patterns and environmental sustainability. This research introduces a hybrid Temporal Convolutional Network (TCN)-Bidirectional Long Short-Term Memory (BiLSTM) architecture designed to capture multi-scale temporal features, utilizing Gaussian Process Regression (GPR) to deliver probabilistic SOH estimations with robust uncertainty quantification. Based on the SOH estimation, a circular carbon emission intensity (CEI) metric is introduced to characterize the evolution of battery carbon burden by quantifying carbon emissions per unit of delivered energy. This integrated framework establishes a novel methodology coupling real-time health monitoring with lifecycle carbon assessment, facilitating dynamic retirement decisions that overcome the limitations of traditional, fixed SOH thresholds. Validation using NASA and CALCE battery datasets demonstrates superior precision, with RMSE and MAE values maintained below 0.66% and 0.54%, respectively. Furthermore, the observed CEI evolution effectively identifies the transition from stable aging to accelerated carbon burden growth, providing a more sustainable criterion for battery retirement and overall lifecycle management.

备注：中科院2Top

合写作者：Shiyi Yang,Yiqing Yang,Muyao Wu,Zhongwen Zhu,Yan Ma,Jifang Hu

第一作者：Shuhua Li

论文类型：论文集

论文编号：124251

学科门类：工学

文献类型：J

卷号：180

是否译文：否

发表时间：2026-08-16

收录刊物：SCI、EI

发布期刊链接：https://www.sciencedirect.com/science/article/pii/S2352152X26039150