基于改进EEMD与GA-BP的谐振接地故障选线方法

针对谐振接地系统发生接地故障，存在暂态信号特征辨识度低，且单一特征作为选线判据易受故障条件影响等问题，提出一种基于改进EEMD与GA-BP神经网络的故障选线方法。首先使用边界局部特征尺度延拓法加集合经验模态分解和多尺度排列熵算法的混合算法（MEEMD）分解暂态电流信号，各项分解指标说明MEEMD能准确区分高频特征分量和基频分量并有效改进端点效应和抑制模态混淆。然后提取重构的高频分量能量、方向以及裕度因子等特征并将其用来训练、测试GA-BP神经网络。结果表明所提出的选线方法有较高的准确率且不受线路类型、接地电阻影响，有较强的鲁棒性和容错性。

Resonant Grounding Fault Line Selection Method Based on Improved EEMD and GA-BP Model

Targeting the problem of low identification accuracy of transient signal characteristics when grounding fault occurs in resonant grounding system， and the single feature criterion is easily affected by fault conditions， the paper proposes a fault line selection method based on improved EEMD and GA-BP neural network. Firstly， the hybrid algorithm of boundary local characteristic scale extension method with ensemble empirical mode decomposition and multi-scale permutation entropy algorithm （MEEMD） is used to decompose transient current signals. The decomposition indices show that MEEMD can accurately distinguish high-frequency characteristic component and fundamental frequency component， effectively improving endpoint effect and suppressing mode confusion. Then the energy， direction and margin factor of the reconstructed high-frequency components are extracted and used to train and test GA-BP neural network. The results show that the proposed method is not affected by line type and grounding resistance， and has high identification accuracy as well as strong robustness and fault tolerance.