基于XGBoost-CNN的多端有源配电网故障检测  

Multi-terminal active distribution network fault detection based on XGBoost-CNN

作　　者：郭雪丽[1] 张健壮 龚正国 王莹 张赐源 王盼宝[2] GUO Xueli;ZHANG Jianzhuang;GONG Zhengguo;WANG Ying;ZHANG Ciyuan;WANG Panbao(Institute of Economics and Technology,StateGrid Nanyang Power Supply Company,Nanyang 473000,China;School of Electrical Engineering and Automation,Harbin Institute of Technology,Harbin150001,China)

机构地区：[1]国网南阳供电公司经济技术研究所,河南南阳473000 [2]哈尔滨工业大学电气工程及自动化学院,黑龙江哈尔滨150001

出　　处：《安徽大学学报（自然科学版）》2024年第6期63-69,共7页Journal of Anhui University(Natural Science Edition)

基　　金：国家自然科学基金面上项目(52377173);国网河南省电力公司科技项目(B7178023K138)。

摘　　要：随着配电网中的分布式电源及新能源用户数量的增多,传统配电网已发展为具有多个电源端点的结构.为防止在多端有源配电网支路故障时其他分布式电源向检修区域持续输电而出现反送电事故,提出改进的极端梯度提升(extreme gradient boosting,简称XGBoost)算法,构建基于改进XGBoost算法及卷积神经网络(convolutional neural networks,简称CNN)的多端有源配电网故障检测模型.提取多端有源配电网正常和故障状态的各频段电压峰值、相间电压差和6次谐波分量,且将其作为模型的输入,使用CNN网络对特征数据进行处理.仿真实验结果表明:相对于其他3种模型,该文模型有更好的检测性能、有更强的鲁棒性.该文模型能准确且有效隔离配电网故障区域、预防反送电事故发生.With the increasing number of distributed power sources and new energy users in the distribution grid,the traditional distribution grid has developed into a multi-source structure.In order to prevent the occurrence of reverse power accidents when other distributed power sources continued to supply power to the maintenance area in the branch fault of the multi-source active distribution grid,an improved extreme gradient boosting(XGBoost)algorithm was proposed,and a multi-source active distribution grid fault detection model based on the improved XGBoost algorithm and convolutional neural networks(CNN)was constructed.The peak values of each frequency band voltage,phase-to-phase voltage difference,and 6th harmonic component in the normal and fault states of the multi-source active distribution grid were extracted as the model's input,and the feature data was processed by the CNNnetwork.The simulation experiment results showed that compared with the other three models,the model in this paper had better detection performance and stronger robustness.The model in this paper could effectively and accurately isolate the fault area in the distribution grid and prevent reverse power accidents from occurring.

关 键 词：多端有源配电网 故障定位 故障检测 集成学习 神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TM732[自动化与计算机技术—控制科学与工程]