基于改进变分模态分解和Hilbert变换的变压器局部放电信号特征提取及分类

发布时间：2019-06-25 18:33

【摘要】：针对现有局部放电(PD)信号特征提取方法存在的不足,提出一种基于变分模态分解(VMD)和Hilbert变换(Hilbert-VMD)的特征提取方法,并提出一种双阈值筛选法来确定VMD算法中的分解模态数。首先,根据PD信号功率谱,采用双阈值筛选法确定VMD算法中的分解模态数;其次,采用VMD算法对PD信号进行分解,得到数个有限带宽的固有模态分量(BLIMFs);然后,对各模态分量进行Hilbert变换并线性叠加后得到PD信号的Hilbert时频谱,并计算各模态分量的边际谱;最后,根据各模态分量的边际谱提取PD信号频域内的特征量,并用支持向量机(SVM)对所提取的特征量进行分类。实验结果表明,对试验环境下和现场实测两种环境下的PD信号,采用该文方法提取得到的特征量均具有较高的正确识别率,充分说明该特征提取方法可以有效提取PD信号特征。对于噪声较大的实测信号,采用该方法得到的正确识别率并未明显降低,说明该方法具有较好的噪声鲁棒性。此外,该文所提Hilbert-VMD方法也为PD信号提供了一种新的时频分析方法。
[Abstract]:In view of the shortcomings of the existing partial discharge (PD) signal feature extraction methods, a feature extraction method based on variant modal decomposition (VMD) and Hilbert transform (Hilbert-VMD) is proposed, and a double threshold screening method is proposed to determine the number of decomposition modes in VMD algorithm. Firstly, according to the power spectrum of PD signal, the decomposition modal number of VMD algorithm is determined by double threshold screening method. Secondly, the PD signal is decomposed by VMD algorithm, and several natural modal components (BLIMFs); with limited bandwidth are obtained. Then, the Hilbert time spectrum of PD signal is obtained by Hilbert transformation and linear superposition of each modal component, and the edge spectrum of each modal component is calculated. Finally, according to the marginal spectrum of each modal component, the feature quantity in the frequency domain of PD signal is extracted, and the extracted feature quantity is classified by support vector machine (SVM). The experimental results show that the feature quantity extracted by this method has a high correct recognition rate for the PD signal in the experimental environment and the field measurement environment, which fully shows that the feature extraction method can effectively extract the features of the PD signal. For the measured signals with large noise, the correct recognition rate obtained by this method has not been significantly reduced, which shows that the method has good noise robustness. In addition, the Hilbert-VMD method proposed in this paper also provides a new time-frequency analysis method for PD signals.
【作者单位】：新能源电力系统国家重点实验室(华北电力大学);
【基金】：国家自然科学基金(51677072) 中央高校基本科研业务费专项资金(2016XS101)资助项目
【分类号】：TM855

【相似文献】