SSA(BBO)-optimized neural networks for remaining useful life estimation and health monitoring of lithium-ion batteries
Wenlong Tao、Chao Lu、Y Jin
1. SSA(BBO)-optimized neural networks for remaining useful life estimation and health monitoring of lithium-ion batteries
基本信息
- 标题:SSA(BBO)-optimized neural networks for remaining useful life estimation and health monitoring of lithium-ion batteries
- DOI/URL:https://doi.org/10.1038/s41598-026-62566-4
- 期刊/会议/预印本平台:Scientific Reports
- 发表日期:2026-07-18(Nature 页面显示 Received 2026-05-05,Accepted 2026-07-13,Published 2026-07-18)
- 作者和机构:Wenlong Tao(College of Automotive Technology, Zhejiang Agricultural Business College)、Chao Lu(Department of Logistics Management, School of Economics and Management, Zhejiang Shuren University)、Yichao Jin(College of Automotive Technology, Zhejiang Agricultural Business College)
- 通讯作者:Chao Lu(Nature 页面明确 “Correspondence to Chao Lu”,Chaoluhangzhou@163.com)
- 脚本评分:83
- 期刊指标核验:未核验;脚本期刊层级为 broad open-access journal / Q1;本报告未额外核验 JCR/JIF/CiteScore 数值。
- 权威作者加分:无;score_reasons 未包含 authority first/corresponding author。
纳入理由和命中证据
- 纳入理由:题名和摘要均明确为 lithium-ion batteries 的 remaining useful life (RUL) estimation / health monitoring,属于电池寿命预测核心范围。
- 命中证据:页面摘要写明 “Accurate prediction of the remaining useful life (RUL) of lithium-ion batteries” 和 “DNN acts as the RUL predictor”。
- 与普通材料/控制论文区别:输出目标是 RUL,验证指标为循环误差和 R²,而不是材料性能表征或调度控制。
创新点、主要内容、模型架构/方法流程
- 创新点:将 DNN 与 Biogeography-Based Optimization(BBO)、Improved Sparrow Search Algorithm(ISSA)结合,并引入线性 SVM 作为 discriminator,把 min-max 惩罚用于抑制物理上不合理的预测。
- 方法流程:真实/增强工况特征输入 DNN 生成 RUL;SVM 判别器构造竞争/对抗学习;ISSA/BBO 用于全局参数优化;与 Autoencoder-DNN、LSTM、CNN-LSTM 等方法比较。
- 输入特征与输出目标:输入包括 variable charge/discharge C-rates、load profiles、ambient temperature、depth of discharge (DoD)、driving cycle statistics;输出为 RUL(循环数)。
数据来源/数据结构/电池类型/循环协议
- 电池体系:锂离子电池。
- 数据来源:NASA-based datasets augmented with realistic variable operating profiles。
- 数据结构与工况:可访问页面确认包含变 C-rate、负载、环境温度、DoD、驾驶循环统计;具体电池数量、原始 NASA cell 编号、循环协议细节未能从页面摘要确认。
- 跨工况验证:作者声称增强真实变工况以提升泛化;是否严格按电池/工况外推划分,页面摘要未说明。
结果指标、Baseline/对比方法
- 指标:RMSE 2.25±0.12 cycles,MAE 1.68±0.09 cycles,R² 0.995±0.002(10 次独立运行均值±标准差)。
- Baseline:Autoencoder-DNN、LSTM、CNN-LSTM;摘要称 RMSE 相比标准方法降低 58–72%;ISSA 也显著优于 BBO variant。
泄漏风险/局限性/复现性
- 泄漏风险:中。指标非常高,且基于 NASA 数据增强;如果随机切分循环片段或增强数据与原始电池泄漏到测试集,可能虚高。页面摘要未确认按电池 split、跨 cell/跨 protocol holdout 或外部验证。
- 局限性:全文是 Nature early access unedited manuscript;可访问内容未确认代码、超参、随机种子、完整数据划分、增强工况生成方式。
- 复现性:信息不足到中。NASA 数据可获得,但增强配置和训练细节需下载/阅读全文或补充材料确认。
对哥哥研究/落地的启发和后续动作
- 启发:可把“真实工况统计 + RUL”作为哥哥现有 RUL pipeline 的扩展输入,重点复核 ISSA-DNN-SVM 是否真的超过强 baseline,尤其是按电池/工况外推时是否仍稳健。
- 后续动作:加入文献库;尝试获取完整 PDF/补充材料;抽取其 NASA preprocessing、增强方法、split;必要时复现一个无泄漏 baseline。
- 跟进结论:强烈跟进。
查看关键词分类依据
- RUL prediction · 标题包含「remaining useful life」
SSA(BBO)-optimized neural networks for remaining useful life estimation and health monitoring of lithium-ion batteries - LSTM · 摘要包含「lstm」
ods (including the baseline Autoencoder-DNN, LSTM, and CNN-LSTM). The framework also significantly outperforms the BBO-optimized - CNN · 摘要包含「cnn」
ding the baseline Autoencoder-DNN, LSTM, and CNN-LSTM). The framework also significantly outperforms the BBO-optimized variant, - Autoencoder · 摘要包含「autoencoder」
to standard methods (including the baseline Autoencoder-DNN, LSTM, and CNN-LSTM). The framework also significantly outperforms the BBO- - Generic neural network / deep learning · 标题包含「neural network」
SSA(BBO)-optimized neural networks for remaining useful life estimation and health monitoring of lithium-ion batt - SVR / SVM · 摘要包含「support vector」
sarial learning mechanism utilizing a linear Support Vector Machine (SVM) as a discriminator. Unlike conventional approaches that primarily - Bayesian · 摘要包含「uncertainty」
s input features. This significantly reduces uncertainty and enhances the model's generalizability across diverse operating conditions. - Hybrid model · 摘要包含「hybrid」
orage systems. This study introduces a novel hybrid deep learning framework that integrates a deep neural network (DNN) with advanc - NASA · 摘要包含「nasa」
cally implausible predictions. Validation on NASA-based datasets augmented with realistic variable operating profiles demonstrate - Li-ion / LIB · 标题包含「lithium-ion」
ful life estimation and health monitoring of lithium-ion batteries