局部放电UHF脉冲的时频特征提取与聚类分析

发布时间：2018-09-10 11:14

【摘要】：制作了4种人工缺陷模型模拟典型的局部放电源,并进行局部放电试验采集UHF脉冲信号。引入S变换(ST)对局部放电的UHF脉冲进行时频分析,探索不同放电源脉冲的聚类分离。算法首先对UHF脉冲进行S变换,并采用非负矩阵分解(NMF)对S变换幅值矩阵进行分解得到频域基向量和时域位置向量,从中提取尖锐度、导数平方和、信息熵以及稀疏度等特征参量,构造出能充分反映局部放电时频信息的特征空间,最后利用模糊C均值算法对提取的特征向量进行聚类得到放电源脉冲的聚类结果。对试验数据的分析结果表明,提取的ST时频特征能够有效实现不同局部放电源脉冲的聚类,当NMF参数r=2时,10维时频特征能够取得最高为90.33%的聚类正确率;与常用的Wigner-Ville分布(WVD)相比,ST具有更好的聚类效果;当存在复杂的多重信号折反射时,本文提出的时频特征聚类结果较差,需要进行进一步的研究。
[Abstract]:Four kinds of artificial defect models were made to simulate the typical local discharge power supply, and the partial discharge test was carried out to collect the UHF pulse signal. S-transform (ST) is introduced to analyze the UHF pulse of partial discharge, and the cluster separation of different discharge pulse is explored. At first, the UHF pulse is transformed by S transform, and the S transform amplitude matrix is decomposed by non-negative matrix factorization (NMF) to obtain the frequency domain basis vector and the time domain position vector, from which the sharpness and derivative square sum are extracted. Characteristic parameters such as entropy and sparsity are used to construct the feature space which can fully reflect the time-frequency information of partial discharge. Finally, the cluster results of discharge pulse are obtained by clustering the extracted eigenvector using fuzzy C-means algorithm. The analysis of the experimental data shows that the extracted ST time-frequency features can effectively realize the clustering of different local discharge pulses, and the maximum clustering accuracy is 90.33% when the NMF parameter r = 2 ~ 2. Compared with the usual Wigner-Ville distribution (WVD), the clustering effect of Wigner-Ville St is better than that of (WVD), and the clustering results of time-frequency feature proposed in this paper are poor when there are complex multi-signal refraction, which needs further study.
【作者单位】：中国电力科学研究院;重庆大学输配电装备及系统安全与新技术国家重点实验室;
【基金】：国家创新研究群体基金(51021005) 国家自然科学基金(51277187)资助项目
【分类号】：TM855

【参考文献】