基于概率神经网络的配电网接地故障选线方法

针对小电流接地系统单相接地故障,提出一种基于概率神经网络的配电网接地故障选线方法。采集每条线路在单相接地故障发生时的零序电流暂态信号(一个周波,故障前1/4个周波和故障后3/4个周波),运用db10小波序列进行5层小波包分解。提取零序电流小波能量并将其输入到PNN网络,实现故障线路自动识别。并给出故障选线小波包分解算法和PNN算法的具体步骤,可直接用计算机编程实现。运用Matlab软件对中性点不接地系统和中性点经消弧线圈接地系统均进行了仿真,验证了该方法的可行性与准确性。

Method for grounding fault line selection in power distribution system based on probabilistic neural network

In connection with single phase grounding fault in small current grounding system, a method was pro- posed for grounding fault line selection in power distribution system based on probabilistic neural network. Collected the transient signals of zero-sequence current（ a cycle, 1/4 cycle before fault and 3/4 cycle after fault） of each lines when there was grounding fault; used dbl0 wavelet sequence to execute 5-layer wavelet packet decomposition; ex- tracted the wavelet energy of zero-sequence current and put it into PNN; finally achieved fault line selection auto- matically. Moreover, the specific steps of wavelet packet decomposition algorithm and PNN algorithm in fault line se- lection were showed out, which could be used to program in computer directly. The simulation in Matlab software had proved the method was feasible and accurate in both isolated neutral system and arc-suppression-coil-ground neutral system.