基于LCD-Hilbert谱奇异值和多级支持向量机的配电网故障识别方法
发布时间：2018-10-16 17:35
【摘要】：准确识别故障类型是配电网故障处理的首要任务。提出了一种基于时频矩阵奇异值分解(SVD)和多级支持向量机(SVM)的配电网故障识别方法。利用局部特征尺度分解法(LCD)、Hilbert变换以及带通滤波算法,构造配电网母线电压、主变低压侧进线电流等波形的时频矩阵,对其进行奇异值分解以获取波形奇异谱,提取相应奇异谱的分布参数(如反映奇异值大小的奇异谱均值、反映信号复杂程度的奇异熵等)作为特征向量。将特征向量输入基于多级SVM的分类器以实现故障识别。各类典型工况下的仿真和实验结果表明该识别方法的正确率均90%,可实现对各类不同故障的有效辨识,且具有很强的适应性和实用性。
[Abstract]:Accurate identification of fault types is the primary task of distribution network fault processing. A fault identification method for distribution network based on time-frequency matrix singular value decomposition (SVD) and multistage support vector machine (SVM) is proposed. Based on the local characteristic scale decomposition (LCD), Hilbert) transformation and bandpass filtering algorithm, the time-frequency matrix of the waveform such as bus voltage and low voltage input current of main transformer is constructed, and the singular value decomposition is carried out to obtain the singular spectrum of the waveform. The distribution parameters of the corresponding singular spectrum (such as the singular spectral mean which reflects the size of the singular value, the singular entropy reflecting the complexity of the signal, etc.) are extracted as the eigenvector. The feature vector is input into the classifier based on multilevel SVM to realize fault identification. The simulation and experimental results under various typical working conditions show that the accuracy of the method is 90%, which can effectively identify different kinds of faults, and has strong adaptability and practicability.
【作者单位】： 福州大学电气工程与自动化学院;国网福建省电力有限公司技能培训中心;
【基金】：国家自然科学基金(51677030) 福建省自然科学基金(2016J01218)~~
【分类号】：TM711
本文编号：2275128
[Abstract]:Accurate identification of fault types is the primary task of distribution network fault processing. A fault identification method for distribution network based on time-frequency matrix singular value decomposition (SVD) and multistage support vector machine (SVM) is proposed. Based on the local characteristic scale decomposition (LCD), Hilbert) transformation and bandpass filtering algorithm, the time-frequency matrix of the waveform such as bus voltage and low voltage input current of main transformer is constructed, and the singular value decomposition is carried out to obtain the singular spectrum of the waveform. The distribution parameters of the corresponding singular spectrum (such as the singular spectral mean which reflects the size of the singular value, the singular entropy reflecting the complexity of the signal, etc.) are extracted as the eigenvector. The feature vector is input into the classifier based on multilevel SVM to realize fault identification. The simulation and experimental results under various typical working conditions show that the accuracy of the method is 90%, which can effectively identify different kinds of faults, and has strong adaptability and practicability.
【作者单位】： 福州大学电气工程与自动化学院;国网福建省电力有限公司技能培训中心;
【基金】：国家自然科学基金(51677030) 福建省自然科学基金(2016J01218)~~
【分类号】：TM711
【相似文献】
相关期刊论文 前6条
1 赵闻飙,施颂椒,夏伟;固定阶数的结构奇异值反馈控制器的设计[J];自动化学报;1999年03期
2 赵瑞峰;;最小奇异值灵敏度分析法的研究[J];湖南工业大学学报;2010年03期
3 王昆;彭昱;周玮;孙辉;;基于预估/校正技术的改进奇异值指标[J];继电器;2006年17期
4 徐志友,栾兆文,樊涛,骆济寿;衡量节点电压稳定的奇异值法和稳定指标[J];电力系统自动化;1997年08期
5 胡三高;安宏文;马志勇;柳亦兵;滕伟;;基于小波奇异值分析的汽轮机碰磨特征提取[J];动力工程学报;2013年03期
6 ;[J];;年期
，本文编号：2275128
本文链接：https://www.wllwen.com/kejilunwen/dianlidianqilunwen/2275128.html
