Transformer partial discharge pattern recognition based on incremental learning

被引：0

作者：

Xiao J.-P. ^{[1
]}

Zhu Y.-L. ^{[1
]}

Zhang Y. ^{[1
]}

Pan X.-P. ^{[1
]}

机构：

[1] Department of Electrical Engineering, North China Electric Power University, Baoding

来源：

Dianji yu Kongzhi Xuebao/Electric Machines and Control | 2023年 / 27卷 / 02期

关键词：

attention map; convolutional neural network; incremental learning; learning without forgetting; partial discharge; pattern recognition;

D O I：

10.15938/j.emc.2023.02.002

中图分类号：

学科分类号：

摘要：

In view of the variety of transformer partial discharge types and the existence of partial discharge data in the form of data stream, it is impossible to obtain a complete sample set at one time. In this paper, the method based on incremental learning is used to study the partial discharge pattern recognition of transformer. Firstly, the time-frequency analysis of the PD signal was carried out, and the marginal spectrum image of the PD signal was obtained as the input of the network model. Secondly, the data set was constructed, and the actual situation was simulated to divide different types of partial discharge into different tasks. Finally, using GradCAM + + to generate attention map and construct attention loss, the non-forgetting learning algorithm was improved, so as to realize the automatic update of network model failure mode database. Experiments show that the method proposed is closer to the actual situation and different from the static model of previous classifiers. It can retain the knowledge of previous tasks as much as possible without changing the model structure, and gradually expand the knowledge of network model. Compared with other incremental learning algorithms, this method further alleviates the problem of catastrophic forgetting and achieves the intelligence of partial discharge pattern recognition to a certain extent. © 2023 Editorial Department of Electric Machines and Control. All rights reserved.

引用

页码：9 / 16

页数：7

共 24 条

[1]

(2018)

[2]

YANG Dayong, GE Qi, DONG Yongchao, Et al., Research on operation state pattern recognition of PV station based on the principle of K-means cluste-ring [J], Power System Protection and Control, 44, 14, (2016)

[3]

(2018)

[4]

ZHOU Wangping, WANG Rong, XU Shenrong, Et al., Gear fault diagnosis based on VMD energy entropy and random forest [J], Machinery Design & Manufacture, 3, 14, (2021)

[5]

AN Qingguo, SHI Zhewen, MA Shifeng, Et al., Fault diagnosis of WOA-SVM transformer based on RF feature optimization [ J], High Voltage Apparatus, 58, 2, (2022)

[6]

(2020)

[7]

MAO Zhengyu, LI Fangli, YE Yuming, Et al., Pattern recognition of cable partial discharge type basedon BP neural network algorithm, Mechatronics Information, 27, 3, (2019)

[8]

GONG Maofa, WEI Jingyu, JIANG Wen, Et al., HHT and PNN based fault identification of transformer PD[J], Modern Electronics Technique, 41, 9, (2018)

[9]

CHI Minghe, XIA Ruochun, LUO Qinglin, Et al., Partial discharge characteristics and defect type identification of typical defects in oil-pressboard insulation, Electric Machines and Control, 26, 2, (2022)

[10]

ZHANG Congcong, WANG Gang, GAO Dong, Et al., Partial discharge pattern recognition based on convolutional neural network [J], Advanced Technology of Electrical Engineering and Energy, 40, 3, (2021)

← 1 2 3 →