特高压直流一次设备运行状态k-means聚类监测仿真  

K-Means Clustering Monitoring Simulation of Operating Status of Ultra-High Voltage Direct Current Primary Equipment

作　　者：宗万里 黄燕燕 罗志恒[2] ZONG Wan-li;HUANG Yan-yan;LUO Zhi-heng(EHV Branch Company of State Grid Zhejiang Electric Power Co.,Ltd.Hangzhou Zhejiang 311100,China;Huadian Yuntong Power Technical CO.,LTD.,Changsha Hunan 410221,China)

机构地区：[1]国网浙江省电力有限公司超高压分公司,浙江杭州311100 [2]北京华电云通电力技术有限公司,湖南长沙410221

出　　处：《计算机仿真》2024年第11期162-166,共5页Computer Simulation

基　　金：国网浙江省电力有限公司超高压分公司项目(2021610340)。

摘　　要：针对特高压直流一次设备运行状态异常可能导致电网运行不稳定的问题,提出特高压直流一次设备运行状态监测方法,并进行仿真分析。首先构建基于能量均衡的动态分簇算法,控制无线传感器网络采集特高压直流一次设备运行状态数据,利用k-means算法动态分簇,通过竞争函数和能量比平衡各簇能量,并在簇中划分节点为不同等级节点,高效采集一次设备运行状态数据。采用改进的稀疏主成分分析法选择采集到的一次设备运行状态数据关键特征,最后将特征选择后运行状态数据输入至训练完成的BP神经网络中,实现特高压直流一次设备运行状态监测。实验结果表明:所提方法具有优异的数据采集能力,并且运行状态监测中的PR曲线和ROC曲线更为理想,同时AUC值更高。Aiming at the problem that abnormal operation status of UHVDC primary equipment may lead to unstable operation of power grid,a monitoring method for operation status of UHVDC primary equipment is proposed and simulated.First,a dynamic clustering algorithm based on energy balance is constructed to control the wireless sensor network to collect the operating status data of the UHVDC primary equipment.The k-means algorithm is used to dynamically cluster,balance the energy of each cluster through the competition function and energy ratio,and divide the nodes in the cluster into nodes of different levels to efficiently collect the operating status data of the primary equipment.The improved sparse principal component analysis method is used to select the key features of the primary e-quipment operation status data collected,and finally the operation status data after feature selection is input into the BP neural network after training to realize the operation status monitoring of UHVDC primary equipment.The experimental results show that the proposed method has excellent data acquisition capability,and the PR curve and ROC curve in operation state monitoring are more ideal,and the AUC value is higher.

关 键 词：特高压直流 一次设备 运行状态监测 稀疏主成分分析法 BP神经网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TM743[自动化与计算机技术—控制科学与工程]