|
Abstract
|
In order to improve the fault diagnosis effect of commutation failure and ensure the stable operation of power grid, the fault time domain waveform similarity based commutation failure fault diagnosis technology of HVDC transmission is studied. Different frequency excitation signals are applied to high voltage DC circuits under various fault modes to obtain the time-domain waveform similarity of each frequency fault. The mean square error weight was used to fuse the time-domain waveform similarity of each frequency fault to obtain the time-domain waveform similarity of the whole fault. The adaptive whale optimization algorithm was used to optimize the weight of the probabilistic neural network and substitute it into the probabilistic neural network to establish the fault diagnosis model of commutation failure of HVDC transmission. The fault time domain waveform similarity is input into the model, and the fault diagnosis result of commutation failure is output. Experimental results show that this technique can effectively obtain the time-domain waveform similarity of faults at different frequencies. The technology can accurately diagnose commutation failure at different fault locations. When the transition resistance is different, the AUC value of commutation failure diagnosis is higher, that is, the fault diagnosis accuracy is higher.
|