微型电脑应用2025，Vol.41Issue(2)：42-46,5.
基于不平衡样本的中压电缆分布式故障诊断方法
Distributed Fault Diagnosis Method for Medium Voltage Cables Based on Unbalanced Samples
摘要
Abstract
Due to the problem of data imbalance in high and medium voltage cables,it is difficult to meet the requirements of deep network training.Therefore,a distributed fault diagnosis method for medium voltage cables based on imbalanced samples is proposed.This method adopts a bidirectional long short-term memory network as the local model training method,and uses federal learning to aggregate and update the model in the central server aggregation.Distributed learning by multiple partici-pants to alleviate model training pressure and improve training efficiency.Simultaneously,introducing a data sharing strategy to distribute cloud data to local models,reducing data imbalance.The results show that this method achieves collaborative training of participants under independent and unbalanced data,and the accuracy of cable fault diagnosis can reach 96％.关键词
分布式故障诊断/联邦学习/双向长短期记忆网络/数据孤岛Key words
distributed fault diagnosis/federal learning/bidirectional long short-term memory network/data islanding分类
信息技术与安全科学引用本文复制引用
任广振,曹俊平,周铭权,陈荣鑫,陈森杰..基于不平衡样本的中压电缆分布式故障诊断方法[J].微型电脑应用,2025,41(2):42-46,5.基金项目
国网浙江省电力有限公司科技项目(5211DS20007P) （5211DS20007P）
