Potential Fault Detection Algorithm of Power Grid Based on Collective Anomaly Mining
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
维普
万方数据
针对城市电网故障早期阶段的隐蔽性、潜伏性特点,通过构建"集体离群点-故障模式"的度量规则,对电网系统中全等级异常电流波动信号进行层次聚类分析,将区域性潜在故障检测问题转换为挖掘故障信号数据中的集体离群点问题.为提高检测效率,设计了 一种基于不动点迭代法的层次聚类改进算法(Fixed point iteration based k-medoids,FPK-medoids),利用不动点较强的局部搜索能力提高聚类收敛速度.在测试数据集和实例数据集上进行实验,结果表明改进算法的收敛性能优于传统算法,检测模型能够精准识别电网中的区域性潜在故障.
In view of the latent and concealed characteristics of potential faults in the early stage of urban power grid,this paper proposed a novel detection approach based on hierarchical clustering.According to the measure-ment rule of"collective anomaly-fault pattern",through clustering analysis on all levels of abnormal current fluc-tuation signals in the power system,the problem of regional potential fault detection is transformed into the detec-tion of collective anomaly.Besides,an improved multi-layered clustering algorithm based on fixed point iteration(FPK-medoids)is designed to enhance the detection efficiency.The experimental results show that the conver-gence performance of the improved algorithm is better than the traditional algorithm,and the detection model can identify the regional potential faults in the early stage.
维普
万方数据
