基于关联维数的直流牵引网故障识别  

Fault Identification Algorithm of DC Traction Network Based on Correlation Dimension

作　　者：祝冰心 李夏青[2] 田行军[3] 韩龙涛 

机构地区：[1]北京化工大学信息科学与技术学院,北京100029 [2]北京石油化工学院电气工程系,北京102617 [3]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《北京石油化工学院学报》2014年第1期37-43,共7页Journal of Beijing Institute of Petrochemical Technology

摘　　要：地铁牵引供电系统的发散型振荡电流容易造成直流牵引网继电保护误动,因此欲提高直流牵引网保护系统的可靠性必须寻求更为有效的特征提取方法。分形理论中关联维数的特征提取方法可灵敏地反映出牵引网非线性动态特征信息的变化,并准确识别出牵引网中振荡电流和短路故障电流。在确定的时间序列内,牵引网电流信号经相空间重构和关联维数计算后,定义关联维数为其故障模式识别的特征矢量。实测数据证明,该保护方法不仅具有灵敏性高和概念清晰的优点,而且适合复杂的直流牵引网运行状态信息的诊断。Owing to the malfunction of DC traction network protection system caused by divergent oscillation current easily in Metro traction power supply system, more effective feature extraction methods need to be saught in order to improve the reliability of DC traction network protection system. New feature extraction based on correlation dimension of fractal theory can effectively reflect the changes of nonlinear dynamic characteristics information, and accurately distinguish short-circuit fault current from oscillation current of traction network. In the set time sequence, after the traction network current signal is reconstructed by the phase space method and the correlation dimension is calculated, the correlation dimension is defined as the feature vector of fault pattern recognition. It is proved by the experimental data that the new protection method not only has the advantages of high sensitivity and clear concepts, but also is appropriate for the diagnosis of the complex operational status of DC traction network.

关 键 词：直流牵引网 故障电流 特征提取 分形理论 关联维数 

分 类 号：TM922.3[电气工程—电力电子与电力传动] TM713