小波包与改进BP神经网络的配电网故障选线  

Fault Line Selection Based on Wavelet Packet and Improved BP Neural Network for Power Distribution Grid

作　　者：赵峰[1] 尹德昌[1] 

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070

出　　处：《自动化仪表》2013年第9期4-8,共5页Process Automation Instrumentation

摘　　要：针对配电网中基于单一故障特征信息选线准确率低的问题,提出了一种小波包与改进BP神经网络相结合的综合选线方法。该方法首先采用小波包分解各线路的暂态零序电流,并提取特征频带的小波包能量;然后将小波包能量、各线路5次谐波和零序有功功率作为选线系统的故障特征量;最后运用Levenberg Marquardt(LM)算法改进BP神经网络进行故障选线。Matlab仿真表明,与传统BP神经网络相比,采用LM算法改进后的BP神经网络具有更好的网络性能和选线精度。Aiming at the problem of the low accuracy in fault line selection in power distribution grid by using information about single fault feature, the comprehensive fault line selection method through combining the wavelet packet and improved BP neural network is proposed. With this method, the transient zero sequence current of each line is decomposed by using wavelet packet j the wavelet energy of the feature frequency band is extracted; then three of the fault characteristic valuesj including wavelet packet energy, zero sequence active power, and the 5th harmonic of each line, are used for fault line selection through the BP neural network that is improved by Levenberg Marquardt （ LM ） algorithm. Through Matlab simulation, it is indicated that comparing with traditional BP neural network, the improved BP network possesses better network performance and line selection accuracy.

关 键 词：配电网 小波包 BP神经网络 LM算法 故障选线 

分 类 号：TM773[电气工程—电力系统及自动化]