软件导刊2025，Vol.24Issue(5)：40-45,6.DOI:10.11907/rjdk.241190
一种基于TCN的核电蓄电池异常检测方法
An Anomaly Detection Method for Nuclear Power Storage Batteries Based on TCN
摘要
Abstract
In the production and operation of nuclear power plants,lead-acid battery packs play a vital role,which can ensure the normal op-eration of key equipment instruments and control systems when the reactor is shut down normally or in the event of an emergency.Due to the long-term operation of the battery,the reliability will be reduced,but the maintenance cost of replacing the battery pack regularly is high.Therefore,an anomaly detection method based on temporal convolutional network was proposed.Firstly,the long-span feature extraction of historical voltage data was carried out by combining the large-kernel depth convolution structure and the deep point-by-point convolution.Secondly,the cross-layer connection feature of the residual convolutional layer is added to speed up the model training speed and avoid the gradient explosion caused by too deep layers.Experiments show that when the size of the convolutional kernel is 2 048,the model has the best performance,with the accuracy,precision,recall and F1 values of 96.5%,93.5%,80.2%and 86.3%,respectively,which can process the key data of the battery in the long-term window in parallel,accurately identify the abnormal battery for single-cell replacement,so as to re-duce the maintenance cost and improve the economic benefit.关键词
蓄电池/异常检测/机器学习/工业安全Key words
lead-acid batteries/anomaly detection/machine learning/industrial safety分类
信息技术与安全科学引用本文复制引用
李吉生,于艇,岳鹏,周剑秋,徐鹏..一种基于TCN的核电蓄电池异常检测方法[J].软件导刊,2025,24(5):40-45,6.基金项目
江苏省科技厅前瞻性联合研究项目(BY2016005) （BY2016005）
江苏省自然科学基金项目(BK20190374) （BK20190374）
中央军委装备发展部装备预先研究项目(61407200302) （61407200302）
