基于图卷积网络的光伏配电网故障诊断方法  

Fault Diagnosis Method of PV Distribution Network Based on GCN

作　　者：曾瑞江 黄缙华 李志勇 ZENG Ruijiang;HUANG Jinhua;LI Zhiyong(Electric Power Research of Guangdong Power Grid Co.Ltd.,Guangzhou 510630,China)

机构地区：[1]广东电网有限责任公司电力科学研究院,广州510630

出　　处：《河南科学》2024年第2期202-208,共7页Henan Science

基　　金：南方电网重点科技项目(036100KK52210053,GDKJXM20210066)。

摘　　要：近年来,机器学习技术在电网的故障诊断领域表现出了更显著的性能优势,然而大部分研究仅基于对已有信号进行数据分析,忽略了网络拓扑的分析,具有一定的局限性.鉴于此,提出了一种基于图卷积网络(Graph Convolutional Network,GCN)的光伏配电网故障检测和识别方法,其利用配电网拓扑信息和测量数据来进行故障识别,在标准测试数据和不良数据注入的情况下该方法均具有更优秀的故障特征提取能力.以PSCAD/EMTDC平台中的一个光伏配电网为例,对各种正常和故障电网事件进行了仿真和模型训练,并与其他几种基于机器学习的故障诊断方法进行了对比.结果表明,该故障诊断方法在故障检测和分类方面比其他算法具有更高的准确率,同时在存在不良数据的情况下也具有较好的鲁棒性.In recent years,machine learning technology has shown more significant performance in fault diagnosis of power grids,but most of the research is simply based on data analysis of existing signals,ignoring the analysis of network topology.In view of this,this paper proposes a fault detection and identification method for photovoltaic distribution network based on Graph Convolutional Network(GCN),which uses distribution network topology information and measurement data to identify faults,and has better fault feature extraction ability in the case of standard test data and bad data injection.Taking a photovoltaic distribution network in the PSCAD/EMTDC platform as an example,various normal and fault grid events are simulated and model trained,and compared with several other machine learning-based fault diagnosis methods.The results show that the fault diagnosis method has higher accuracy than other algorithms in fault detection and classification,and has better robustness in the presence of bad data.

关 键 词：光伏配电网 图卷积网络 故障诊断 拓扑信息 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]