基于循环神经网络的配网电压异常数据检测方法  

Detection method of abnormal voltage data in distribution network based on recurrent neural network

作　　者：柯子桓 罗楚楠 黎少凡 KE Zihuan;LUO Chunan;LI Shaofan(China Southern Power Grid Co.,Ltd.,Guangzhou 510700,China;Guangdong Power Grid Energy Development Co.,Ltd.,Guangzhou 510160,China)

机构地区：[1]中国南方电网有限责任公司,广东广州510700 [2]广东电网能源发展有限公司,广东广州510160

出　　处：《电子设计工程》2024年第1期106-110,共5页Electronic Design Engineering

摘　　要：配网电压异常数据检测过程易受到动态数据的影响,导致数据检测精准度较低。为了解决该问题,提出了基于循环神经网络的配网电压异常数据检测方法。在分析循环神经网络结构的基础上,以电压误差标准值为依据,构建电压异常数据检测模型。使用归一化处理方式训练模型,获取异常数据集。在标注数据后,计算线路两端节点电压,并将其与预设置的偏差进行对比,完成对异常数据的检测。由实验结果可知,该方法检测准确率和召回率最大值分别为0.991和0.90,说明使用该方法检测精准度较高。The detection process of abnormal distribution network voltage data is easily affected by dynamic data,resulting in low accuracy of data detection.In order to solve this problem,this paper proposes a method for detecting abnormal voltage data in distribution network based on recurrent neural network.Based on the analysis of the structure of the circulating neural network,a detection model of abnormal voltage data is established based on the voltage error standard value.The normalized processing method is used to train the model and obtain the abnormal data set.After the data is marked,the voltage of the nodes at both ends of the line is calculated and compared with the preset deviation to complete the detection of abnormal data.According to the experimental results,the maximum detection accuracy and recall rate of this method are 0.991 and 0.90,respectively,indicating that the detection accuracy of this method is high.

关 键 词：循环神经网络 配网电压 异常数据检测 归一化处理 

分 类 号：TN99[电子电信—信号与信息处理]