多港口无人车关键部件剩余使用寿命联邦预测OA
Federated Prediction of Remaining Useful Life for Critical Components of Autonomous Vehicles in Multiple Ports
针对港口间因数据孤岛与隐私问题导致的设备剩余使用寿命预测困难,提出一种基于长短时记忆网络(long short-term memory,LSTM)的联邦学习方法.首先,在各港口本地部署自编码器提取关键特征并构建联邦 LSTM 模型;其次,设计了特征相似度驱动的自适应客户端分组与加权聚合策略,通过对具有相似退化模式的客户端分组,并在组内基于验证集性能进行模型聚合更新,以提升非独立同分布环境下的全局模型性能.在真实港口电池数据集上的实验结果表明,所提方法能有效缓解模型漂移,显著提升剩余使用寿命预测准确率.
To address the challenge of remaining useful life prediction caused by data silos and privacy concerns among ports,a federated learning method based on long short-term memory(LSTM)was pro-posed.Firstly,autoencoders were locally deployed at each port to extract key features and constructed a federated LSTM model.Secondly,a feature similarity-driven adaptive client grouping and weighted aggre-gation strategy was developed.In this strategy,clients with similar degradation patterns were grouped,and model aggregation was performed within each group based on validation performance to enhance glob-al performance with non-independent and identically distributed conditions.Experiments conducted on a real-world port battery dataset demonstrated that the proposed method effectively mitigated model drift and significantly improved remaining useful life prediction accuracy.
马超;张呈呈;关志博;张恺琪;黄海
哈尔滨理工大学 计算机科学与技术学院 黑龙江 哈尔滨 150006哈尔滨理工大学 计算机科学与技术学院 黑龙江 哈尔滨 150006哈尔滨理工大学 计算机科学与技术学院 黑龙江 哈尔滨 150006哈尔滨理工大学 计算机科学与技术学院 黑龙江 哈尔滨 150006哈尔滨理工大学 计算机科学与技术学院 黑龙江 哈尔滨 150006
信息技术与安全科学
剩余使用寿命预测联邦学习长短时记忆网络自编码器电池
remaining useful life predictionfederated learningLSTMautoencoderbattery
《郑州大学学报(理学版)》 2026 (6)
1-8,8
黑龙江省重点研发项目(2022ZX01A36)
评论