智能配电网中的故障检测与自愈系统研究  

Research on Fault Detection and Self-Healing System in Intelligent Distribution Network

作　　者：肖志峰 XIAO Zhifeng(Tongyu County Power Supply Company of State Grid Jilin Electric Power Co.,Ltd.,Baicheng,Jilin Province,137200 China)

机构地区：[1]国网吉林省电力有限公司通榆县供电公司,吉林白城137200

出　　处：《科技资讯》2024年第22期119-121,共3页Science & Technology Information

摘　　要：传统电力系统的接地故障检测和自我修复手段往往因反应迟缓和精度不高而受到挑战。首先,构建了一个单相接地故障识别框架,核心在于聚焦零序电流这一关键指标,对零序电流信号进行了多维度的处理,将其转化为时域、频域和小波域的特性向量,并借助随机森林算法赋予这些特征不同的权重。其次,利用轻量级梯度提升机(Light Gradient Boosting Machine,GBM)算法对这些经过分类的特征向量进行深度学习训练,从而实现精确的故障预测。该算法能有效抵御过渡电阻和初始相位角变化对预测结果的影响,准确性表现达到98.9%,远超其他比较算法。The traditional grounding fault detection and self repair methods in power systems are often challenged due to slow response and low accuracy.Firstly,this paper constructs a framework for identifying one-phase ground⁃ing faults,with a focus on the key indicator of zero sequence current,which processes the zero sequence current sig⁃nal in multiple dimensions,converts it into characteristic vectors in the time domain,frequency domain,and wave⁃let domain,and assigns different weights to these features using Random Forest Algorithm.Then,the LightGBM al⁃gorithm is used to perform deep learning training on these classified feature vectors,in order to achieve accurate fault prediction.This algorithm can effectively resist the influence of transition resistance and initial phase angle changes on the prediction results,with an accuracy performance of 98.9%,far exceeding other comparative algo⁃rithms.

关 键 词：广域信息 小波分析 随机森林 故障检测 

分 类 号：TM76[电气工程—电力系统及自动化]