基于改进小波分析小电流接地选线方法

针对小波分析在对故障线路与非故障线路暂态量差别不大而产生误判的缺陷，提出基于径向基函数RBF（Radia basis function）神经网络。通过免疫机制改善RBF网络隐含层的聚类形态，对不同故障类型的故障线路与非故障线路的暂态零序电流小波模极大值进行聚类。而RBF网络的训练由遗传算法去执行从而得到最优连接权值。将训练后的RBF网络用于的小电流接地故障选线，仿真结果表明，利用该算法选线具有较高的精确度

Line selecting methods for small current grounding based on improved wavelet analysis

In view of the defect of misjudgment caused by the little differences between the transient components in fault lines and nonfault lines in the wavelet analysis, the paper proposed the neural network based on radial basis function. Through the improvement of the clustering form in RBF network by immune mechanism, the clusters could be formed, which was the maximum value of transient zero sequence current wavelet magnitude in fault lines and non-fault lines in different fault types. The optimized connection weight could be obtained by RBF network training by genetic algorithm. The trained RBF network was used in the line selecting for small current grounding. The simulation results showed that this algorithm had high accuracy.