Partial Discharge Pattern Recognition Based on Convolutional Neural Network and UHF Technology

doi:10.16628/j.cnki.2095-8188.2022.10.005

Abstract

Abstract:

Partial discharge is one of the main factors that lead to the deterioration of the insulation performance of power equipment and eventually lead to insulation failure.The traditional method of detecting partial discharge of equipment is complicated,requires regular manual inspection,and cannot monitor the insulation status of equipment in real time.When partial discharge occurs,high-frequency electromagnetic wave signals will be radiated outward.An online monitoring of partial discharge based on UHF technology and a method for constructing a phase resolved partial discharge (PRPD) map is proposed.This method is used to obtain the amplitude and time information of partial discharge pulses to construct a PRPD map,and the PRPD map is gridded to obtain a grayscale image of 36×30 size.Finally,the optimized convoluntional neural network model is used to identify the four typical discharge types.

Key words: partial discharge, pattern recognition, convolutional neural network, PRPD map

CLC Number:

TM930

SUN Tianlong. Partial Discharge Pattern Recognition Based on Convolutional Neural Network and UHF Technology[J]. LOW VOLTAGE APPARATUS, 2022, 0(10): 32-37.

Figures/Tables 12

References 21

[1]	李方利, 钱帅伟, 毛振宇, 等. 基于多传感器融合技术的电缆附件局部放电检测系统[J]. 电器与能效管理技术, 2020(6):53-60,75.
[2]	程实, 陈道强, 唐世虎, 等. 特高频带电检测技术在GIS设备中的应用研究和验证[J]. 电工技术, 2021(11):177-180.
[3]	张翼, 朱永利. 图信号与图卷积网络相结合的局部放电模式识别方法[J]. 中国电机工程学报, 2021, 41(18):6472-6481.
[4]	梁建英, 孙传铭, 杨刚, 等. 基于声发射技术的典型缺陷局放信号特征分析及识别研究[J]. 电器与能效管理技术, 2020(12):14-21.
[5]	刘炳南, 黄沂平, 方国标. 基于一维卷积神经网络的配电网高阻接地故障识别[J]. 电器与能效管理技术, 2020(9):99-103.
[6]	侯春光, 张润奇, 曹云东, 等. 开关柜局部放电信号抗干扰算法研究[J]. 电器与能效管理技术, 2020(12):22-27.
[7]	朱洪伟, 王小军, 王建生, 等. 开关柜局部放电抗干扰系统开发研究[J]. 电器与能效管理技术, 2018(10):50-53.
[8]	孙曙光, 陆俭国, 王景芹, 等. 基于LabView的局部放电故障类型诊断系统的设计[J]. 低压电器, 2012(6):1-4,60.
[9]	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.
[10]	姚聪伟, 李盈, 庞小峰, 等. 500 kV断路器均压电容器10 kV介损异常现象的研究与探讨[J]. 电力电容器与无功补偿, 2021, 42(6):101-109.
[11]	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.
[12]	朱煜峰, 许永鹏, 陈孝信, 等. 基于卷积神经网络的直流XLPE电缆局部放电模式识别技术[J]. 电工技术学报, 2020, 35(3):659-668.
[13]	侯春光, 孙小涵, 赵长权, 等. 基于频带能量特征提取的10 kV固体绝缘柜表面放电音频信号识别[J]. 电器与能效管理技术, 2020(3):38-42.
[14]	许渊, 刘卫东, 陈维江, 等. 基于高灵敏测量的GIS绝缘子表面微金属颗粒局部放电特性[J]. 高电压技术, 2019, 45(9):2707-2714.
[15]	马婷婷, 杜伯书, 孔晓晓, 等. 开关柜绝缘状态检测与故障诊断[J]. 电器与能效管理技术, 2019(15):71-76.
[16]	SONG H, DAI J J, SHENG G H, et al. GIS partial discharge pattern recognition via deep convolutional neural network under complex data sources[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2018, 25(2):678-685. doi: 10.1109/TDEI.2018.006930
[17]	NGUYEN M T, NGUYEN V H, YUN S J, et al. Recurrent neural network for partial discharge diagnosis in gas-insulated switchgear[J]. Energies, 2018, 11(5):1-13. doi: 10.3390/en11010001
[18]	ALEX K, ILYA S, GEOFFREY E H. Image net classification with deep convolutional neural networks:IEEE Conference on Computer Vision and Pattern Recognition[C]. 2012.
[19]	ROSS G, JEFF D, TREVOR D, et al. Rich feature hierarchies for accurate object detection and semantic segmentation:IEEE Computer Vision and Pattern Recognition[C]2014.
[20]	WANG Y, YAN J, SUN Q F, et al. A mobilenets convolutional neural network for gis partial discharge pattern recognition in the ubiquitous power internet of things context:optimization,comparsion,and application[J]. IEEE Access, 2019, 7:150226-150236. doi: 10.1109/ACCESS.2019.2946662
[21]	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition:IEEE Conference on Computer Vision and Pattern Recognition[C]. 2016.

训练次数/次	样本数量/组
训练次数/次	2 000	4 000	6 000	8 000
1	80.60	81.12	90.67	91.17
2	90.64	91.93	90.67	80.67
3	80.50	80.62	90.67	91.83
4	78.60	91.31	80.67	91.83
5	91.81	80.62	80.67	91.83
平均值	84.43	85.12	86.67	89.47

训练次数/次	样本数量/组
训练次数/次	2 000	4 000	6 000	8 000
1	91.81	90.81	91.83	92.33
2	91.81	91.81	91.83	91.33
3	91.14	91.31	91.83	92.33
4	91.49	92.31	92.33	91.83
5	91.30	92.31	92.33	92.33
平均值	91.51	91.71	92.03	92.03