基于粗集神经网络的配电网故障选线研究

为了克服基于神经网络的故障选线方法训练时间长和网络结构复杂的缺点,提出了基于粗集神经网络的故障选线方法.利用ATP-EMTP做大量的单相接地故障仿真试验,得到大量的各馈线零序电流信号,通过小波变换和傅立叶变换从中提取各种暂态和稳态故障特征.利用粗集理论对故障特征进行预处理,将约简后的故障特征作为神经网络的输入,约简后的样本作为训练样本.完成训练的神经网络模型即可实现故障选线.仿真和现场验证结果表明,该方法训练速度快、误判率低.

Study of fault line detection based on rough set neural network for distribution network

To overcome the shortcomings of the longtime training and the complicated structure in the neural network based single phase to ground fault line detection method for distribution network,a rough set neural network based fault line detection method is presented.A lot of zero sequence current signals of feeders are obtained in many single-phase-to-earth fault experiments by using the ATP-EMTP simulation and all kinds of steady state features and transient fault features are extracted from zero sequence curre...