基于零序电流小波变换系数均方根值的故障选线ANN方法

An ANN method of faulted feeder detection based on continuous wavelet transform coefficients' RMS values of zero sequence current

配电网单相接地故障产生的高频信号可以用于接地故障选线.利用馈线零序电流特征频带(SFB),提出一种连续小波变换(CWT)系数的均方根(Root Mean Square,RMS)值与人工神经网络(ANN)相结合的故障选线方法.通过对各条馈线故障后5ms时零序电流进行CWT变换,剔除工频信号,并根据能量和最大原则选出故障特征频带.将各条馈线特征频带上CWT系数的均方根值作为ANN选线的输入样本属性,故障馈线编号作为输出样本属性,构造智能选线网络.该方法不需要提出明确的故障选线判据,利用ANN非线性拟合和记忆功能进行故障选线.大量的实验仿真数据表明,该方法选线结果准确可靠.

High-frequency transient signals generated by single-phase grounding fault can be ap- plied to the faulted feeder detection for distribution systems. This paper combined root mean square （RMS） value of zero-sequence current continuous wavelet transform （CWT） and ANN to complete the faulted feeder detection. Analyzing the 5 ms data of each feeder zero-sequence cur- rent, the characteristic frequency bands were determined based on the maximum energy principle. By using the RMS value of wavelet coefficients in the selected frequency band, smart faulted feed- er detection network was constructed. The fault detection method took the advantage of ANN＇s nonlinear fitting and memory functions instead of putting forward a clear detection criterion.Large amount of simulation and experimental results showed that the proposed method is accurate and reliable.