摘要
快速准确识别局部放电类型对于保证变压器安全稳定运行具有重要意义。针对局部放电信号模式识别中面临的最优特征参数提取和分类器设计难题,提出一种基于分数阶傅里叶变换(fractional Fourier transform,FrFT)和相关向量机(relevance vector machine,RVM)的局部放电模式识别方法。首先将FrFT引入局部放电信号分析领域,利用FrFT将局部放电信号转换至分数域并对其进行多尺度分析,在扩充信息提取维度的同时,提取可反映不同局部放电信号波形差异的14维特征构成特征向量;然后将特征向量作为输入,建立RVM模型进行最优特征选择和分类判决函数的联合优化,从而实现对不同局部放电信号的分类识别。建立电晕放电、沿面放电和气隙放电试验模型并采集局部放电超声信号开展试验,结果表明所提方法对于每种局部放电信号均能获得较高的识别精度,平均正确识别率相对于常规支持向量机(support vector machine,SVM)分类方法提升超过2.7%。
Rapid and accurate identification of partial discharge(PD)types is of great significance for ensuring the safe and stable operation of transformers.This paper proposes a PD pattern recognition method based on fractional Fourier transform(FrFT)and correlation vector machine(RVM)to address the problem of feature selection and classifier design in PD signal pattern recognition.Firstly,it introduces FrFT into the field of PD signal analysis,which is used to transform PD signals into fractional domains for multi-scale analysis.At the same of expanding information extraction dimensions,a feature vector consisting of 14 features that can describe the waveform differences of PD signals corresponding to different discharge types is extracted.Then,the feature vector is used as the input to establish an RVM classification model for joint optimization of feature selection and classification decision functions,so as to achieve optimal feature selection while obtaining optimal pattern recognition results.Finally,the paper establishes experimental models for corona discharge,surface discharge,and air gap discharge,and collects PD ultrasound signals for testing.The results indicate that the proposed method can achieve high recognition accuracy.Under the same conditions,compared to the conventional support vector machine(SVM),the promotion of average correct recognition rate of this method exceeds over 2.7%.
作者
杨新志
李利华
陈锋
赵国汉
雷秉惠
YANG Xinzhi;LI Lihua;CHENG Feng;ZHAO Guohan;LEI Binghui(China Yangtze Power Co.,Ltd.,Baihetan Hydropower Plant,Liangshan,Sichuan 615000,China)
出处
《广东电力》
北大核心
2024年第6期95-103,共9页
Guangdong Electric Power
基金
国家自然科学基金项目(U1866603)。
关键词
局部放电
模式识别
特征提取
特征选择
分数阶傅里叶变换
partial discharge
pattern recognition
feature extraction
feature selection
fractional Fourier transform(FrFT)
