基于LOF和SVM的智能配电网故障辨识方法

Fault identification based on LOF and SVM for smart distribution network

针对现有智能配电网保护方法存在保护装置整定复杂、协调性差以及易误动等问题,提出一种基于局部异常因子(LOF)检测的配电网保护算法,并对配电网在故障定位后不能进行有效的故障类型辨识这一问题,提出LOF和支持向量机(SVM)相结合的智能配电网故障类型判别方法。根据各节点LOF值的大小实现智能配电网的故障检测与定位;然后对故障处的三相电压进行小波变换,以三相电压的小波奇异熵值建立故障特征样本库,利用反映接地故障信息的零序电压低频能量对故障进行预分类,并以此为基础建立SVM故障类型判别预测模型。该算法可对智能配电网的故障进行有效的检测与定位,并能对故障区域的不同故障类型进行合理分类。

Aiming at the complicated settings,poor coordination and misoperation of protective equipmentsof existing protection methods,a protection method based on LOF（Local Outlier Factor） detection is proposedfor smart distribution network. A hybrid fault identification method combining L0F and SVM（Support VectorMachine） is proposed to effectively identify the fault type after fault is located. The fault of smartdistribution network is detected and located according to the L0F value of each node. The three-phasevoltages at the fault point are analyzed by wavelet transform to obtain the wavelet singularity entropy forconstructing the fault characteristic sample base. The low-frequency energy of zero-sequence voltage,whichreflects the information of grounding fault,is used to pre-classify the fault,based on which,an SVMprediction model is constructed for identifying the fault type. The proposed method can effectively detectand locate the fault of smart distribution network ,as well as reasonably classify its type.