摘要
变压器涌流是造成变压器差动保护误动的主要原因之一。本文从变压器不同运行工况下记录的差动电流波形出发,结合小波分析和机器学习智能技术,提出了一种基于逻辑斯蒂回归的涌流识别方法。首先,通过仿真手段批量获取变压器涌流和内部故障波形样本。然后,借助小波理论从中抽取反映波形组成成分复杂度的特征指标,并在此基础上,详细阐述了逻辑斯蒂回归分类器从构建优化到性能评估的完整过程。最后,将分类器应用于仿真实验和现场录波数据,通过与现有文献中的检测方法相比,验证了本文方法的正确性和有效性。本文方法原理简单,计算量小,既可以对当前现场配置保护作动作特性分析,也可以为智能保护的研究提供参考。
Transformer inrush current is one of the main causes for differential protection mal-operation.In this paper,based on the differential current waveforms recorded under different operating conditions of a transformer,an inrush current identification method based on logistic regression is proposed with the combination of intelligent technologies,such as wavelet analysis and machine learning.First,batch waveforms of inrush current and internal fault are obtained through simulations,from which the feature indicators reflecting the waveform complexity are extracted by means of the wavelet theory.On this basis,the whole process of a logistic regression classifier is elaborated,including its construc⁃tion,optimization,and performance evaluation.Finally,this classifier is applied to simulations and on-field recorded data.From the comparison with the existing identification methods in the literature,it is verified that the proposed meth⁃od is correct and effective.This proposed method is of simple principle and with less computation load,and it can not only analyze the characteristics of on-filed configuration protection,but also provide reference for the research on intelli⁃gent protection.
作者
丁晓兵
周红阳
黄佳胤
张弛
白淑华
张利强
DING Xiaobing;ZHOU Hongyang;HUANG Jiayin;ZHANG Chi;BAI Shuhua;ZHANG Liqiang(Department of Electric Power Dispatching and Control,China Southern Power Grid,Guangzhou 510663,China;Beijing Sifang Automation Co.,Ltd,Beijing 100085,China)
出处
《电力系统及其自动化学报》
CSCD
北大核心
2020年第12期77-84,94,共9页
Proceedings of the CSU-EPSA
基金
中国南方电网有限责任公司科技资助项目(ZDKJQQ00000021)。
关键词
变压器保护
涌流
内部故障
逻辑斯蒂回归
小波分析
transformer protection
inrush current
internal fault
logistic regression
wavelet analysis
