Abstract
Oil-pressboard insulation defect is the main cause of insulation failure of oil-immersed transformer. The traditional pulse current method for partial discharge detection of oil paper insulation has low frequency and narrow frequency band, which is difficult to fully reflect the partial discharge and affect the estimation of insulation state. In this paper, an ultra-wideband partial discharge detection method with an effective measurement frequency band of 30 kHz to 50 MHz is designed and used. The method has a wide effective measurement frequency band, which can reflect partial discharge information more accurately and is more suitable for the identification of partial discharge types. The experimental platform of partial discharge under three typical defect models of transformer oil-paper insulation is built in the laboratory, and the partial discharge signals are collected simultaneously by the ultra-wideband partial discharge detection method and the traditional pulse current method respectively, and their detection data are compared and studied. In order to visually compare the application effects of the two partial discharge detection methods, a restricted Boltzmann machine (RBM) combined with a gray wolf optimized support vector machine (GWO-SVM) algorithm model is designed for partial discharge type identification based on the two types of detection data. The results show that: under the ultra-wideband detection method, the phase distribution of partial discharge amplitude and discharge number under three typical defect models have their own characteristics and are different from the traditional pulse current method; the feature parameter values extracted based on the two methods respectively have large differences;With the classification methods cited in this paper and theselection of feature quantities, the accuracy of partial discharge type identification based on the ultra-wideband method reaches 97.22%, which has a better classification and identification effect compared with the traditional pulse current method.
Translated title of the contribution | Partial Discharge Detection Method in Oil-pressboard Insulation Based on Ultra-wideband Method |
---|---|
Original language | Chinese (Traditional) |
Pages (from-to) | 4770-4779 |
Number of pages | 10 |
Journal | Gaodianya Jishu/High Voltage Engineering |
Volume | 48 |
Issue number | 12 |
DOIs | |
Publication status | Published - 31 Dec 2022 |
Externally published | Yes |