基于注意力机制与模糊聚类的电网设备数据特征提取  

Feature extraction of power grid equipment data based on Attention Mechanism and fuzzy clustering

作　　者：易立 毛宇 邓楚杭 赵筑雨 YI Li;MAO Yu;DENG Chuhang;ZHAO Zhuyu(Southern Power Grid Supply Chain(Guizhou)Co.,Ltd.,Guiyang 550002,China;Guizhou Qianchi Information Co.,Ltd.,Guiyang 550002,China)

机构地区：[1]南方电网供应链(贵州)有限公司,贵州贵阳550002 [2]贵州黔驰信息股份有限公司,贵州贵阳550002

出　　处：《电子设计工程》2025年第8期192-196,共5页Electronic Design Engineering

基　　金：贵州省科技计划项目(20162031);贵阳市高新区创新创业项目(QC/XZGY202004022A00)。

摘　　要：为提升电网设备的品质管控水平,文中对电网设备的特征提取方法进行了研究。采用模糊聚类算法对在网运行的电力设备特征按照功能性、安全性以及稳定性等方面进行了划分,建立了层次分明的评估模型。使用GRU网络对设备的品质评分进行预测,为区分不同特征的影响力,引入注意力机制并利用模糊聚类的权重对不同节点进行赋值,提升了神经网络的非线性拟合能力。通过对变压器、柱上开关的设备参数、运行数据和维护记录进行数据采集与清理,建立了模型评估的数据集,并使用所提方法计算了设备的运行品质得分。结果表明,在引入Attention层后,算法的RMSE和MAPE较普通GRU网络分别降低了6.78%、5.36%。In order to improve the quality control level of power grid equipment,the feature extraction method of power grid equipment was studied in this paper.Using fuzzy clustering algorithm,the characteristics of power equipment operating in the grid were divided in terms of functionality,safety,and stability,and a hierarchical evaluation model was established.Using GRU network to predict the quality rating of devices,in order to distinguish the influence of different features,Attention Mechanism is introduced and fuzzy clustering weights are used to assign values to different nodes,improving the nonlinear fitting ability of the neural network.By collecting and cleaning data on equipment parameters,operating data,and maintenance records of transformers and column mounted switches,a dataset for model evaluation was established.The proposed method was used to calculate the operating quality score of the equipment.The results showed that after introducing the Attention layer,the RMSE and MAPE of the algorithm decreased by 6.78%and 5.36%compared to ordinary GRU networks,respectively.

关 键 词：设备评估 特征提取 模糊数学 数据挖掘 聚类分析 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]