摘要
针对传统GIS设备定期定点检修时设备拆卸复杂、故障定位难等缺点,提出了一种利用GIS分合闸过程中产生的驱动电流特征值搭建BP神经网络,并引入遗传算法对神经网络的权值和神经元的阈值进行初始化,以实现GIS分合闸故障自动分类的方法。试验结果表明,进行样本学习后的BP神经网络,在进行故障判别时具备明显的方向性,能够较好地表明线圈电流信号的特征值和分合闸故障之间的非线性关系,具有较高的预测精度和较好的分类准度。
Aiming at the disadvantages such as complex disassembly of equipment and difficult location of faults during regular spot inspection of traditional GIS equipment,a BP neural network based on the characteristic value of drive current generated during tripping and closing of GIS was proposed and the weight of neural network and the threshold of neurons were initialized by introducing genetic algorithm in order to achieve automatic classification of GIS trip/close faults.The experimental results show that the BP neural network after sample learning has obvious directionality in fault identification,which can well show the nonlinear relationship between the characteristic value of coil current signal and trip/close faults with higher precision of prediction and better accuracy of classification.
作者
付光晶
于跃
张峰
张士文
Fu Guangjing;Yu Yue;Zhang Feng;Zhang Shiwen(School of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)
出处
《电气自动化》
2018年第2期50-54,共5页
Electrical Automation
关键词
GIS
分合闸
故障分类
BP神经网络
遗传算法
GIS
trip/close
fault classification
BP neural network
genetic algorithm
作者简介
付光晶(1985-),女,江苏人,硕士生,从事电工理论与新技术,嵌入式技术应用等研究。;于跃(1992-),男,天津人,硕士生,从事嵌入式技术应用研究。;张峰(1968-),男,江苏人,博士,教授,博士生导师,从事电工理论与新技术、轨道交通设备检测理论分析等研究。;张士文(1976-),男,黑龙江人,硕士,讲师,从事电工理论与新技术、计算机控制技术等方面的教学与研究工作。
