基于LSSVM的小电流接地故障融合选线

A Fault Line Selection Method for Indirectly Grounding Power System Based on LSSVM

小电流接地系统单相接地故障选线是尚未解决的一大难题。传统的单一故障选线方法都有其各自的适用范围,故障选线准确率不高。在此基础上,提出利用最小二乘支持向量机(LSSVM)对多种选线方法进行信息融合。分别利用小波包变换和FFT算法,从零序电流中提取暂态能量分量、稳态基波分量和五次谐波分量,形成故障选线的特征分量。作为LSSVM的输入进行信息融合。从而实现故障选线。仿真结果表明:与标准的SVM相比,该方法学习速度更快、泛化能力更强,且不受接地电阻、故障初始角、故障距离等的影响。选线结果具有较高的精度。

The faulty detection of single phase to earth fault in indirectly grounding power system is a problem that has not been well solved. The applicable scope of traditional fault line selection is limited, their reliability is not satisfied. Therefore,, a various methods of fault line selection are combined using LSSVM is proposed. Transient energy component, steady-state fundamental component and fifth harmonic component extracted from zero-sequence current separately by means of Wavelet packet and FFT are used as fault features to perform fault line selection. Simulation indicates this method features high learning speed, good approximation with SVM, and immunes from the influences of load, fault initial phase angle and fault location. The faulty line selection result by the proposed method is more precise.