Suzhou Electric Appliance Research Institute
期刊号: CN32-1800/TM| ISSN1007-3175

Article retrieval

文章检索

首页 >> 文章检索 >> 往年索引

基于卷积神经网络的多源局部放电模式识别

来源:电工电气发布时间:2023-10-28 09:28 浏览次数:112

基于卷积神经网络的多源局部放电模式识别

余祉宏1,邵振华2,冯旗1
(1 温州大学 电气与电子工程学院,浙江 温州 325035;
 2 闽江学院 计算机与控制工程学院,福建 福州 350108)
 
    摘 要:为验证开关柜多源局部放电直接分类的可行性,设计了四种典型局部放电模型,采集单局部放电源和双局部放电源信号,并绘制 PRPD 图谱作为数据集,利用卷积神经网络 (CNN) 模型进行模式识别。实验以经典模型的性能作为参考,再对表现较好的模型进行优化,得到最终模型。实验结果表明,优化后的模型准确率均超过98.5%,且训练时长较经典模型明显减少,适用于多源局部放电模式识别。
    关键词: 多源局部放电;PRPD 图谱;卷积神经网络;模式识别
    中图分类号:TM835 ;TM85     文献标识码:A     文章编号:1007-3175(2023)10-0024-08
 
Multi-Source Partial Discharge Pattern Recognition Based on
Convolution Neural Network
 
YU Zhi-hong1, SHAO Zhen-hua2, FENG Qi1
(1 College of Electrical and Electronic Engineering, Wenzhou University, Wenzhou 325035, China;
2 College of Computer and Control Engineering, Minjiang University, Fuzhou 350108, China)
 
    Abstract: In order to verify the feasibility of directly classifying multi-source partial discharge in switchgears, four typical partial discharge models are designed. They collect signals of single and double partial discharge sources, draw PRPD map as the data set, and adopt the Convolution Neural Network(CNN) model to recognize patterns. The experiment, taking the performance of classical model as the reference,optimizes models with better performance to screen the final model. According to the experimental results, the optimized model has the accuracy of more than 98.5% with less training time, which is suitable for the pattern recognition of multi-source partial discharge.
    Key words: multi-source partial discharge; PRPD map; convolution neural network; pattern recognition
 
参考文献
[1] DUAN Lian , HU Jun , ZHAO Gen , et al .Identification of partial discharge defects based on deep learning method[J].IEEE Transactions on Power Delivery,2019,34(4) :1557-1568.
[2] 唐志国, 唐铭泽, 李金忠, 等. 电气设备局部放电模式识别研究综述[J] . 高电压技术,2017,43(7) :2263-2277.
[3] 邓兴宇. 高压开关柜局部放电检测中的抗干扰技术研究[D]. 广州:广东工业大学,2021.
[4] 陶加贵. 组合电器局部放电多信息融合辨识与危害性评估研究[D]. 重庆:重庆大学,2013.
[5] 范路,陆云才,陶风波,等. 人工智能在局部放电检测中的应用(二) :模式识别与状态评估[J]. 绝缘材料,2021,54(7) :10-24.
[6] 黄亮,唐炬,凌超,等. 基于多特征信息融合技术的局部放电模式识别研究[J] . 高电压技术,2015,41(3) :947-955.
[7] 陈敬德,李峰,孙源文,等. 基于 KNN 和 MSR 的局部放电模式识别研究[J] . 电气技术,2018,19(1) :10-14.
[8] 周文潮,周子涵,靳冲. 基于 SVM 的变压器局部放电故障诊断研究[J] . 铁路通信信号工程技术,2022,19(S1) :137-140.
[9] FENG X Y, HU X L, YONG J, et al.Application of Improved BPNN Algorithm in GIS Insulation Defect Type Identification[C]//Journal of Physics Conference Series,2019.
[10] 陈继明,许辰航,李鹏,等. 基于时频分析与分形理论的 GIS 局部放电模式识别特征提取方法[J] .高电压技术,2021,47(1) :287-295.
[11] SUN Shengya, SUN Yuanyuan, XU Gongde, et al.Partial Discharge Pattern Recognition of Transformers Based on the Gray-Level Co-Occurrence Matrix of Optimal Parameters[J].IEEE Access,2021,9 :102422-102432.
[12] FIRUZI K, VAKILIAN M, PHUNG B T, et al.Partial discharges pattern recognition of transformer defect model by LBP & HOG features[J].IEEE Transactions on Power Delivery,2019,34(2) :542-550.
[13] BARRIOS S, BULDAIN D, COMECH M P, et al.Partial discharge classification using deep learning methods—Survey of recent progress[J].Energies,2019,12(13) :2485.
[14] 黄雪莜,熊俊,张宇,等. 基于残差卷积神经网络的开关柜局部放电模式识别[J] . 中国电力,2021,54(2) :44-51.
[15] 陈健宁,周远翔,白正,等. 基于多通道卷积神经网络的油纸绝缘局部放电模式识别方法[J] . 高电压技术,2022,48(5) :1705-1715.
[16] 孙抗,轩旭阳,刘鹏辉,等. 小样本下基于 CNN-DCGAN 的电缆局部放电模式识别方法[J] . 电子科技,2022,35(7) :7-13.
[17] TANG Zhiguo , CAO Zhi . Application of Convolutional Neural Network Transfer Learning in Partial Discharge Pattern Recognition[C]//2020 IEEE International Conference on High Voltage Engineering and Application(ICHVE),2020.
[18] GAO Angran, ZHU Yongli, CAI Weihao, et al.Pattern recognition of partial discharge based on VMD-CWD spectrum and optimized CNN with cross-layer feature fusion[J].IEEE Access,2020,8 :151296-151306.
[19] 朱霄珣,林佳伟,刘宝平,等. 基于 Iradon-CNN 的变压器局部放电状态识别方法[J] . 电子测量技术,2022,45(17) :36-42.
[20] 谢荣斌,杨超,申强,等. TEV 与 HFCT 法测量开关柜局部放电的特性对比[J] . 中国电力,2022,55(3) :37-47.
[21] 吴闽,蒋伟,罗颖婷,等. 基于改进 SSD 的 GIS 多源局放模式识别[J] . 高电压技术,2023,49(2) :812-821.
[22] MANTACH S, ASHRAF A, JANANI H, et al.A Convolutional Neural Network-Based Model for Multi-Source and Single-Source Partial Discharge Pattern Classification Using Only Single-Source Training Set[J].Energies,2021,14(5) :1355.
[23] HE Kaiming, ZHANG Xiangyu, REN Shaoqing,et al.Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016.