基于同步提取变换和卷积神经网络的有源配电网单相接地故障选线方法  

A Single-phase Grounding Fault Line Selection Method for Active Distribution Network Based on Synchroextracting Transform and Convolutional Neural Network

作　　者：翟乐庆 刘益青[1] 魏元健 徐枫 张宸恺 ZHAI Leqing;LIU Yiqing;WEI Yuanjian;XU Feng;ZHANG Chenkai(School of Electrical Engineering,University of Jinan,Jinan 250022,China)

机构地区：[1]济南大学自动化与电气工程学院,山东济南250022

出　　处：《山东电力技术》2025年第2期65-77,共13页Shandong Electric Power

基　　金：山东省自然科学基金项目(ZR2022ME097)。

摘　　要：分布式电源接入配电网导致单相接地故障时故障电流的幅值和相位发生改变,而现有时频分析法分辨率低导致故障特性区分度不高,因此基于时频分析法和卷积神经网络(convolutional neural networks,CNN)的故障选线方法准确率仍较低。提出一种基于同步提取变换(synchroextracting transform,SET)和CNN的有源配电网单相接地故障选线方法。首先分析分布式电源影响配电网单相接地故障电流特征的机理,选用不受分布式电源影响的零序电流作为选线依据,并将其处理成同步提取变换时频图。然后分析SET和CNN用于有源配电网故障选线的可行性,阐述所提方法的完整实现流程,设计评价指标,开展验证实验和对比实验。实验结果表明,在高阻故障以及噪声干扰等情况下,所提的SET-CNN选线方法相较于现有方法具有更高的选线准确率,选线准确率能提高3.09%和4.12%。Connecting distributed generations to the distribution network leads to changes in the amplitude and phase of the fault current during single-phase ground faults.The existing time-frequency analysis methods have low resolution,resulting in low discriminability of fault characteristics.Therefore,the accuracy of fault line selection method based on time-frequency analysis method and convolutional neural network is still low.A grounding fault line selection method for active distribution networks based on synchroextracting transform(SET)and convolutional neural network(CNN)is proposed.Firstly,the mechanism by which distributed generations affect the characteristics of single-phase grounding fault current in distribution networks is theoretically analyzed.The zero-sequence current,which is not affected by distributed generation,is selected as the basis for line selection and it is processed into a time-frequency image by synchroextracting transform.The feasibility of using synchroextracting transform and convolutional neural networks for fault line selection in active distribution networks is analyzed.A complete flowchart of the proposed method is designed;the evaluation indicators and verification experiments are described;the verification experiments and comparative experiments are implemented.The experimental results show that under conditions of high resistance faults and noise interference,the proposed line selection method has a higher accuracy rate than existing methods,with an increase in accuracy of 3.09%and 4.12%respectively.

关 键 词：有源配电网 单相接地故障选线 同步提取变换 卷积神经网络 

分 类 号：TM77[电气工程—电力系统及自动化]