论文部分内容阅读
为了实现高压直流(High Voltage Direct Current,HVDC)换相失败的快速诊断,提高交直流互联系统运行的安全与稳定性,提出了一种基于小波包分解(Wavelet Packet Decomposition,WPD)和广义回归神经网络(General Regression Neural Network,GRNN)的高压直流输电系统换相失败故障诊断方法,该方法对直流和交流的故障信号进行小波包分解,获取各频段下的能量系数,计算其小波包能量(wavelet packet energy,WPE)作为故障特征,再通过两种模式下的广义回归神经网络结构分别进行模式识别来判断是否发生换相失败以及引起故障的原因,并且对两种拓扑结构的鲁棒性对比分析。仿真和结果表明,两种拓扑结构均能很好的识别换相失败故障和直流线路故障,进而准确地诊断引起换相失败的原因,但模式一的鲁棒性较好。
In order to quickly diagnose the commutation failure of High Voltage Direct Current (HVDC) and improve the security and stability of the operation of AC / DC interconnection system, a Wavelet Packet Decomposition (WPD) and generalized regression neural (HVDC) system based on GRNN, this method performs wavelet packet decomposition on the fault signals of DC and AC to obtain the energy coefficients of each frequency band and calculates the wavelet packet energy (wavelet packet energy, WPE) as fault features, and then through the two models of generalized regression neural network structure pattern recognition to determine whether commutation failure and the cause of the failure, and the robustness of the two topologies comparative analysis . Simulation and results show that both topologies can identify commutation failure and DC fault, and then diagnose the cause of commutation failure accurately. However, the robustness of the first one is better.