|
| 摘要: |
| 提出了一种基于经验模式分解(EMD)的近似熵和支持向量机(SVM)的电力故障类型识别的新方法.利用EMD良好的局域化特性和近似熵来量化故障特征.再与SVM结合进行故障类型识别.首先,对故障线路的三相电压信号进行EMD分解得到若干个能反映故障信息的本征模式分量(IMF);其次,选取三相电压的前4个IMF的近似熵值作为信号的特征向量.最后将构造的特征向量输入到SVM分类器进行故障类型识别.仿真表明,该方法能有效地提取故障特征,不同的故障类型,其三相近似熵变化明显不同,同一种故障类型,在不同故障位置、过渡电阻和初始相角情况下,其三相近似熵变化规律相似;与传统的BP网络相比,SVM网络具有训练样本少、训练时间短、识别率高的特点. |
| 关键词: 经验模式分解近似熵支持向量机 |
| DOI： |
| 分类号:TM711 |
| 基金项目: |
|
| Fault classification for power transmission line using EMD-approximate entropy and SVM |
|
ZHU Zhihui SUN Yunlian
|
| Abstract: |
| A fault classification method based on EMD(Empirical Mode Decomposition) -approximate entropy and SVM(Support Vector Machines) is proposed for power system.The localization charac -teristics of EMD and approximate entropy are used to quantify the fault,and then the SVM is combined to classify the fault.The three -phase voltage signals of fault line are decomposed into some IMF(Intrinsic Mode Function) by EMD;the approximate entropies of the first four IMF are taken as the eigenvectors;the eigenvectors are input into SVM classifier for fault type recognition.Simulation results show that,the change of three -phase approximate entropy is obviously different for different fault types,while similar for different fault locations,different transient resistances and initial phase angles of same fault type,which proves the proposed method can effectively extract fault features.Compared with BP network,SVM network has shorter training time,smaller samples and higher recognition rate. |
| Key words: empirical mode decomposition,approximate entropy,support vector machine |
