电气设备局部放电信号特征提取及分类方法研究

发布时间：2018-07-30 07:50

【摘要】：局部放电与电气设备内部绝缘材料劣化和绝缘体的击穿密切相关。不同类型的局部放电产生的机理和发生的位置各不相同,对设备绝缘的破坏程度也不尽相同。因此,准确,快速地识别局部放电类型对电气设备运行维护人员确定放电位置,合理安排检修计划有着重要意义。本文在分析局部放电特性基础上主要研究了电力设备局部放电信号的特征提取及分类方法,主要工作内容如下:提出了一种基于变分模态分解和多尺度排列熵的特征提取方法。利用变分模态分解算法对实验室条件下采集的4种局部放电信号进行分解,得到数个包含不同频带信息的有限带宽的固有模态分量。求得相应的多尺度排列熵,并将其组合成原始特征量。同时使用最大相关最小冗余准则对原始特征量进行优选降维。最后使用支持向量机分类器实现分类。该方法提取的特征能够有效表征PD信号在不同频带下的不确定性和复杂性,并且鲁棒性强,识别率高。研究并建立了基于变量预测模型的局部放电模式识别算法(VPMCD)。从放电信号中提取37维统计特征量和9维时频特征量,使用VPMCD完成局部放电信号的分类。实验结果表明,VPMCD算法在识别率和计算效率均高于BP神经网络和支持向量机。针对局部放电有效样本过少而导致分类器性能下降的问题,提出一种基于改进的VPMCD方法。该方法利用正交完备基函数建立特征值的预测模型,通过网格搜索策略和移动最小二乘法求解模型参数,提高了模型的精度。在只有较少局部放电样本的情况下,改进的VPMCD算法的识别正确率要高于VPMCD、BP神经网络和SVM算法。
[Abstract]:Partial discharge is closely related to internal insulation material deterioration and insulator breakdown. The mechanism and location of partial discharge are different, and the damage degree of insulation is different. Therefore, accurate and rapid identification of partial discharge types is of great significance for electrical equipment maintenance personnel to determine the discharge location and arrange the maintenance plan reasonably. Based on the analysis of PD characteristics, this paper mainly studies the feature extraction and classification of PD signals in power equipment. The main work is as follows: a feature extraction method based on variational mode decomposition and multi-scale permutation entropy is proposed. The variational mode decomposition (VMD) algorithm is used to decompose the four PD signals collected under laboratory conditions, and several inherent modal components with limited bandwidth are obtained. The corresponding multi-scale permutation entropy is obtained and combined into the original characteristic quantity. At the same time, the maximum correlation minimum redundancy criterion is used to optimize and reduce the dimension of the original feature. Finally, support vector machine classifier is used to realize classification. The features extracted by this method can effectively represent the uncertainty and complexity of PD signals in different frequency bands, and have strong robustness and high recognition rate. The partial discharge pattern recognition algorithm (VPMCD). Based on variable prediction model is studied and established. The 37-dimensional statistical characteristic and 9-dimensional time-frequency characteristic are extracted from the discharge signal, and the partial discharge signal is classified by VPMCD. The experimental results show that the recognition rate and computational efficiency of VPMCD algorithm are higher than those of BP neural network and support vector machine. A modified VPMCD method is proposed to solve the problem that the performance of classifier is degraded due to the small number of effective partial discharge samples. In this method, the prediction model of eigenvalue is established by orthogonal complete basis function, and the precision of the model is improved by solving the model parameters by grid search strategy and moving least square method. The recognition accuracy of the improved VPMCD algorithm is higher than that of the VPMCDBP-BP neural network and the SVM algorithm with only a small number of PD samples.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM855

【参考文献】