Nonlinear Behavior Discrimination of Vibration for Loose Fault of Isolation Switches in Distribution Networks
After the mechanical switch looseness fault occurs in the flexible high-voltage DC distribution network,a large number of submodule capacitors of MMC quickly discharge towards the fault point,causing a rapid increase in fault current and a sharp drop in system voltage,affecting the safety of the power grid.However,the relevant signal characteristics will be affected by external exci-tation of variable frequency current,leading to distortion in loose behavior detection.Research has found that when the base of the iso-lation switch is loose and defective,the frequency domain vibration spectrum will exhibit nonlinear changes.Based on this feature,a vibration nonlinear behavior discrimination method for loose faults of isolation switches in distribution networks is proposed.Taking the GW16 type isolation switch commonly used in distribution networks as an example,multiple feedback self mixing interference vibra-tion measurement technology is used to collect vibration signals such as the relative position of the dynamic and static contacts,contact pressure of the dynamic and static contacts,circuit resistance,and balance spring force value of the distribution network isolation switch in a loose state;Decompose the switch signal using the EEMD method,calculate the envelope of the IMF component,and based on this,obtain the energy entropy of the IMF component,which is used as the waveform nonlinear feature of the switch loose-ning signal;Establish an extreme learning machine to input the nonlinear characteristics of the waveform of the switch loosening signal into the machine,and achieve switch loosening fault diagnosis and detection.The experimental results show that the proposed method can collect high-precision vibration characteristic signals of switch looseness behavior and extract nonlinear waveform features,which has the advantages of high looseness fault diagnosis accuracy and detection accuracy.
distribution networkmechanical switchvibration waveform characteristicsnon linearityloose faultextreme learning machine
万方数据
维普
