东篱科研大数据发现系统（DRDS）

位置：成果数据库 > 期刊 > 期刊详情页

电力电子装置周期性故障波形小波矩阵分类法

ISSN号：1672-7207
期刊名称：《中南大学学报：自然科学版》
时间：0
分类：TM464[电气工程—电器]
作者机构：[1]中南大学物理科学与技术学院,湖南长沙410083, [2]中南大学信息科学与工程学院,湖南长沙410083, [3]河南周口市供电局,河南周口466001
相关基金：国家自然科学基金重点资助项目（60634020）;湖南省自然科学基金资助项目（06FD007）;博士后科学基金资助项目（20060400885）

关键词：平稳小波变换, 奇异值分解, 逆变器故障, 分类, stationary wavelet transform, singular value decomposition, fault of inverter, classification

中文摘要：

提出一种基于平稳小波变换和奇异值分解的电力电子装置周期性故障波形的分类方法。该方法利用平稳小波变换的冗余性和奇异值的稳健性。其步骤为：对周期性故障波形进行平稳小波变换,将信号分解到多个小波子空间;将平稳小波变换后的小波系数矩阵奇异值进行分解,即采用K-L变换对子信号进行特征压缩,并以奇异值向量作为特征向量;按照向量的空间距离对故障波形进行分类实现故障的分类诊断。以精确的半导体器件模型建立的PSPICE逆变器故障波形为例,分别用该方法和小波子带能量法对逆变器的IGBT开关故障进行分类。研究结果表明,与采用小波子带能量法相比,采用所提方法能够精确地对22种逆变器断路故障进行诊断,且受小波分解层数的影响较小,分类边界较清晰,其类间距与类内距之比是小波子带能量法的2.5倍,抗噪性能好,正确识别率高5%。

英文摘要：

Wavelet matrix classification, which takes advantage of redundancy of the stationary wavelet transform and steadiness of singular value decomposition, was proposed in order to classify the periodic fault wave of power electron equipment. The procedures are as follows： Periodic fault wave is decomposed into sub-space by the stationary wavelet transform, singular value is obtained by singular value decomposition of wavelet coefficient matrix, and the essential of singular value decomposition is the decompressed by K-L transform. Classification was completed by calculating the space distance of two singular values. The fault wave, which was from the precise PSPICE model of inverter, was classified by the proposed method and the wavelet sub-band energy method. The results show that, compared with wavelet sub-band energy method, the proposed method can completely diagnose 22 on-off faults of inverter, the wavelet level has little effect on the proposed method, the classification borderline of proposed method is more clear, the distance of different kinds of proposed method is 2.5 times larger than wavelet sub-band energy method, and the correct recognition rate of proposed method is 5% higher than wavelet sub-band energy method.