小波包与改进BP神经网络的配电网故障选线.PDFVIP

小波包与改进BP神经网络的配电网故障选线摇 赵摇 峰,等 小波包与改进 BP神经网络的配电网故障选线 Fault Line Selection Based onWavelet Packet and Improved BP Neural Network for Power Distribution Grid 赵摇 峰摇 尹德昌 (兰州交通大学自动化与电气工程学院,甘肃 兰州摇 730070) 摘摇 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。 关键词: 配电网摇 小波包摇 BP神经网络摇 LM算法摇 故障选线 中图分类号: TM773摇 摇 摇 摇 文献标志码: A Abstract: Aiming at the problem of the low accuracy in fault line selection in power distribution grid by using information about single fault feature,thecomprehensivefaultlineselectionmethodthroughcombiningthewaveletpacketandimprovedBPneuralnetworkisproposed. With this method,thetransientzerosequencecurrentofeachlineisdecomposedbyusingwaveletpacket,thewaveletenergyofthefeaturefrequency band is extracted;then three of the fault characteristic values,including wavelet packet energy,zero sequence active power,and the 5th harmonic of each line,areusedforfaultlineselectionthroughtheBPneuralnetworkthatisimprovedby LevenbergMarquardt(LM)algorithm. Through Matlab simulation,itisindicatedthatcomparingwithtraditionalBPneuralnetwork,theimprovedBPnetworkpossessesbetternetwork performance and line selection accuracy. Keywords: Power distribution grid摇 Wavelet packet摇 BP neural network摇 LM algorithm摇 Fault line selection 线精度和可靠性,较好地解决了配电网故障选线问题。 0摇 引言 1摇 小波包算法与故障特征量提取 在中低压配电网中,单相接地故障率约占配电网 [1] 总故障率的80% 。 当系统发生单相接地时,由于接 1.1摇 小波包算法 地电流小、馈线数量多等原因,导致其选线问题一直未 小波包理论的出现,为分析具有随机性、局部性及 能得到很好的解决。 对此人们做了大量研究,提出了