基于神经网络的柔性直流混合输电线路故障测距  

Fault-location Method for VSC-HVDC Hybrid Transmission Lines Based on ANN

作　　者：杨明玉[1] 杨立红[1] 

机构地区：[1]电力系统保护与动态安全监控教育部重点实验室(华北电力大学),河北保定071003

出　　处：《现代电力》2013年第5期50-54,共5页Modern Electric Power

摘　　要：柔性直流输电线路故障暂态过程具有更强的固有频率信号,因此本文提取固有频率的幅值和频率作为样本属性,提出一种基于神经网络的柔性直流混合输电线路故障测距算法。对于柔性直流线-缆混合输电线路,故障区域不同,固有频率的成分及幅值也不同,首先根据固有频率频谱中有无"定频"以及主频的大小判断故障发生的区域,然后利用分层分布式神经网络进行故障测距,采用粒子群算法对BP神经网络的权值和阈值进行优化,提高了网络的训练效率,使其收敛速度加快。PSCAD和MATLAB仿真结果表明,该故障测距算法具有较高的可靠性和精确性。The natural frequency signal of VSC-HVDC transmission lines is strong during transient process.So the amplitude and frequency of the natural frequency are chosen as the samples,and a novel fault-location method for VSCHVDC hybrid transmission lines based on artificial neutral network （ANN） is presented.The composition and amplitude of the natural frequency are different when the failure occurred at various locations of VSC-HVDC hybrid transmission lines.Because the fault location can be judged by ＂fixed-frequency＂ in the spectra and the amplitude of the dominant frequency,hierarchical distributed neural network model is used to locate the fault distance.Through the optimization of the weights and bias of BP neural network by using of PSO algorithm,the training efficiency and convergence speed of network are improved.Simulation results by PSCAD and MATLAB show that the proposed method has higher reliability and accuracy.

关 键 词：柔性直流输电 混合输电线路 固有频率 故障测距 BP神经网络 

分 类 号：TM77[电气工程—电力系统及自动化]