基于大数据分析的输变电线路故障预测与诊断  

Fault Prediction and Diagnosis of Transmission and Transformation Lines Based on Big Data Analysis

作　　者：韩文通 HAN Wentong(Shenzhen Power Transmission Engineering Co.,Ltd.,Shenzhen,Guangdong 518000,China)

机构地区：[1]深圳市输变电工程有限公司,广东深圳518000

出　　处：《自动化应用》2025年第6期94-96,共3页Automation Application

摘　　要：针对输变电线路故障预测与诊断问题,提出一种基于大数据分析的新方法。该方法结合双向长短时记忆网络(BiLSTM)与注意力机制,从设备状态参数中提取时空关联特征,实现了对多种故障类型的精准识别。通过构建RTDS与Hadoop集成的实验环境,利用330 kV变电站3年运行数据与模拟故障样本进行验证。实验结果表明,所提方法故障预测准确率达96.2%,预警提前量为1.9 s,漏报率仅为0.3%,在处理海量电力监测数据与识别小概率故障模式方面显著优于传统小样本分析方法。This paper focuses on the problem of fault prediction and diagnosis in power transmission and transformation lines,and proposes a new method based on big data analysis.This method combines the Bidirectional Long Short-Term Memory Network(BiLSTM)with the attention mechanism to extract spatio-temporal correlation features from equipment status parameters,thus achieving accurate identification of multiple fault types.An experimental environment integrating Real-Time Digital Simulator(RTDS)and Hadoop is constructed,and the three-year operation data of a 330 kV substation and simulated fault samples are used for verification.The experimental results show that the proposed method has a fault prediction accuracy of 96.2%,an early warning lead time of 1.9 s,and a false negative rate of only 0.3%.It is significantly superior to traditional smallsample analysis methods in handling massive power monitoring data and identifying low-probability fault patterns.

关 键 词：输变电线路 故障预测 故障诊断 双向长短时记忆网络 注意力机制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP277[自动化与计算机技术—计算机科学与技术]