基于K均值聚类算法的谐振接地系统故障区段定位方法

Fault section location method in resonant grounding systems based on K-means clustering algorithm

现有的谐振接地配电网单相接地故障定位方法存在通信依赖过度、特征分析复杂和阈值设置困难等问题,现场运行的适用性较低。本文基于深度融合智能开关在配电网中的应用,研究三相电流变化量的波形特征,提出基于K均值聚类算法的就地选段方法。该方法提取各区段数据特征量,发挥K均值聚类算法无监督学习的优点,使各检测节点只需处理本地故障信号,从而减轻通信压力。利用仿真和现场数据验证该方法的可行性,结果表明,该方法在多种故障工况下都表现出较高的可靠性,并且能够较好地适应现场环境。

The current fault location method for resonant grounding distribution networks faces challenges such as excessive communication dependence,complex feature analysis,and difficulties in setting thresholds,resulting in reduced applicability for on-site operation.By studying the characteristics of three-phase current waveform variations,this paper,based on the deep integration of intelligent switches in distribution networks,introduces a local section selection method using the K-means clustering algorithm.This method extracts fault feature parameters,combining the advantages of unsupervised learning through the K-means clustering algorithm to identify section types.This approach allows each detection node to process only local fault signals,reducing the communication burden.The feasibility of this method is validated using both simulation and on-site data.Experimental results demonstrate that this method exhibits high reliability across various fault conditions and can effectively adapt to real-world environments.