Research on Partial Discharge Information Monitoring Method based on UHF Signal Power Frequency Characteristics

According to the power frequency characteristics of UHF signal, the partial discharge information monitoring method of high voltage switchgear is studied. Analyze the principle of different discharge modes, build a monitoring system to obtain various characterization parameters of partial discharge, and use the deep learning algorithm model to classify the partial discharge information of different spectra. The experimental results show that the average accuracy of different neural network models is different. The recognition accuracy of resnet50 model without SVM classification algorithm is as high as 96.8%, while the recognition accuracy of vggl6 model with SVM classification algorithm is 97.8%. Use the monitoring system to obtain various discharge data, and plot the prpd ΔU and ΔT isogram, using the method of multi information fusion, gives different weights to different spectra. The experimental results show that the classification accuracy of the fused algorithm model in all kinds of partial discharge information reaches 99.35%, which is far higher than the recognition effect of a single spectrum, providing a solution for the monitoring of related problems.