小样本下基于CNN-DCGAN的电缆局部放电模式识别方法

doi:10.16180/j.cnki.issn1007-7820.2022.07.002

摘要/Abstract

摘要：

在电缆局部放电模式识别过程中,传统人工特征提取依赖特定领域的知识及经验,特征选择和优化工作量较大。针对该问题并为了避免非均衡小样本数据下分类器的过拟合,文中提出了一种在小样本的情况下基于CNN-DCGAN的电缆局部放电模式的识别方法。利用滑动时间窗将局部放电时域信号转化为二维图像信息,构建深度卷积生成对抗网络,在原始数据集的基础上进行样本增强,将原始样本和增强样本作为系统输入,构造卷积神经网络,利用其非线性编码器自动提取局部放电特征,并通过Softmax层训练特征分类模型。实验结果表明,相较于人工特征,基于自动特征提取的CNN分类器识别准确率提高了4.18%。相较于原有数据集,基于样本增强数据集的系统识别准确率提高了3.175%。

关键词: 局部放电, 特征提取, 样本增强, 卷积神经网络, 生成对抗网络, 模式识别, 绝缘缺陷, 时域信号

Abstract:

In the process of cable partial discharge pattern recognition, the traditional manual feature extraction relies on the knowledge and experience of specific fields, and the workload of feature selection and optimization is heavy. In view of this problem and to avoid the overfitting problem of the classifier under the unbalanced small sample data of the model, this study presents a partial discharge pattern recognition method based on CNN-DCGAN in the case of small samples. Partial discharge time domain signals are transformed into two-dimensional image information by sliding time window. The DCGANs are constructed, and the data enhancement is carried out on the basis of the original data set. The original data and the enhanced data are taken as the system input. CNN is constructed, and its nonlinear encoder is used to automatically extract partial discharge features, and the feature classification model is trained by Softmax layer. Experimental results show that compared with artificial features, the recognition accuracy of CNN classifier based on automatic feature extraction is improved by 4.18%. Compared with the original data set, the system recognition accuracy based on the sample enhanced data set is improved by 3.175%.

Key words: partial discharge, feature extraction, data augmentation, convolution neural networks, generative adversarial networks, pattern recognition, insulation defect, time domain signal

中图分类号:

TP181

孙抗,轩旭阳,刘鹏辉,赵来军,龙洁. 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.

SUN Kang,XUAN Xuyang,LIU Penghui,ZHAO Laijun,LONG Jie. Partial Discharge Pattern Recognition of Cable Based on CNN-DCGAN under Small Data[J]. Electronic Science and Technology, 2022, 35(7): 7-13.

图/表 15

图1

图2

图3

图4

图5

图6

图7

图8

图9

表1

表2

表3

图10

表4

图11

参考文献 21

[1]	杨帆, 王干军, 彭小圣, 等. 基于卷积神经网络的高压电缆局部放电模式识别[J]. 电力自动化设备, 2018, 38(5):123-128.
	Yang Fan, Wang Ganjun, Peng Xiaosheng, et al. Partial discharge pattern recognition of high voltage cable based on convolutional neural network[J]. Electric Power Automation Equipment, 2018, 38(5):123-128.
[2]	Yang F, Xu Y, Qian Y, et al. Application of correlation analysis techniques in feature extraction and selection for DC partial discharge signals of XLPE cables[J]. Power System Technology, 2018, 42(5):1653-1660.
[3]	Peng X, Li J, Wang G, et al. Random forest based optimal feature selection for partial discharge pattern recognition in HV cables[J]. IEEE Transactions on Power Delivery, 2019, 34(4):1715-1724. doi: 10.1109/TPWRD.2019.2918316
[4]	李程, 李强, 张启超, 等. 基于支持向量机递归特征消除的电缆局部放电特征寻优[J]. 电气技术, 2020, 21(1):67-71.
	Li Cheng, Li Qiang, Zhang Qichao, et al. Partial discharge characteristic optimization of cables based on support vector machine-recursive feature elimination[J]. Electrical Engineering, 2020, 21(1):67 -71.
[5]	Li L, Tang J, Liu Y. Partial discharge recognition in gas insulated switchgear based on multi-information fusion[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2012, 22(2):1080-1087. doi: 10.1109/TDEI.2015.7076809
[6]	毛振宇, 李方利, 叶玉明, 等. 基于BP神经网络算法的电缆局部放电类型模式识别[J]. 机电信息, 2019, 27(10):20-22.
	Mao Zhenyu, Li Fangli, Ye Yuming, et al. Partial discharge pattern recognition of cable based on BP neural network algorithm[J]. Mechanical and Electrical Information, 2019, 27(10):20-22.
[7]	刘兵, 郑剑. 基于卷积神经网络的变压器局部放电模式识别[J]. 高压电器, 2017, 53(5):70-74.
	Liu Bing, Zheng Jian. Partial discharge pattern recognition in power transformers based on convolutional neural network[J]. High Voltage Apparatus, 2017, 53(5):70-74.
[8]	Song H, Dai J, Sheng G,et.al. GIS partial discharge pattern recognition via deep convolutional neural network under complex data source[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2018, 25(2):678-685. doi: 10.1109/TDEI.2018.006930
[9]	Li G, Wang X, Li X, et al. Partial discharge recognition with a multi-resolution convolutional neural network[J]. Sensors, 2018, 18(10):1-10. doi: 10.3390/s18010001
[10]	Liu F, Shen C, Lin G, et al. Learning depth from single monocular images using deep convolutional neural fields[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(10):2024-2039. doi: 10.1109/TPAMI.2015.2505283
[11]	葛靖, 刘子龙. 基于CNN和LSTM的睡眠呼吸暂停检测算法[J]. 电子科技, 2021, 34(2):21-26.
	Ge Jing, Liu Zilong. The algorithm based on CNN and LSTM for sleep apnea syndrome detection[J]. Electronic Science and Technology, 2021, 34(2):21-26.
[12]	马波, 蔡伟东, 赵大力. 基于GAN样本生成技术的智能诊断方法[J]. 振动与冲击, 2020, 39(18):153-160.
	Ma Bo, Cai Weidong,Zhao Dali.Intelligent diagnosis method based on GAN sample generation technology[J]. Journal of Vibration and Shock, 2020, 39(18):153-160.
[13]	孙秋野, 胡旌伟, 杨凌霄, 等. 基于GAN技术的自能源混合建模与参数辨识方法[J]. 自动化学报, 2018, 44(5):901-914.
	Sun Qiuye, Hu Jingwei, Yang Lingxiao, et al. We-energy hybrid modeling and parameter identification with GAN technology[J]. Acta Automatica Sinica, 2018, 44(5):901-914.
[14]	谈元鹏, 刘伟, 赵紫璇, 等. 面向配电异构数据的生成对抗式数据增殖技术研究[J]. 供用电, 2019, 36(10):36-40.
	Tan Yuanpeng, Liu Wei, Zhao Zixuan, et al. Generative adversarial learning based date generation technology for distribution heterogeneous data[J]. Distribution & Utilization, 2019, 36(10):36-40.
[15]	Zhang A, He J, Lin Y, et al. Recognition of partial discharge of cable accessories based on convolutional neural network with small data set[J]. COMPEL-the International Journal for Computation and Mathematics in Electrical and Electronic Engineering, 2020, 39(2):431-446. doi: 10.1108/COMPEL-08-2019-0317
[16]	Bae S H, Yoon K J. Confidence-based data association and discriminative deep appearance learning for robust online multi-object tracking[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(3):595-610. doi: 10.1109/TPAMI.2017.2691769
[17]	Dong C, Loy C C, He K. Image super-resolution using deep convolutional networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(2):295-307.
[18]	Chong U P. Signal model-based fault detection and diagnosis for induction motors using features of vibration signal in Two-Dimension domain[J]. Strojniski Vestnik, 2011, 57(9):655-666.
[19]	曹思灿. 基于生成对抗网络的轴承故障诊断方法研究[D]. 武汉: 华中科技大学, 2019.
	Cao Sican. Generative adversarial network based methods for rolling bearing fault diagnosis[D]. Wuhan: Huazhong University of Science and Technology, 2019.
[20]	曾钰廷. 基于深度学习的物体检测与跟踪方法的研究[D]. 南昌: 东华理工大学, 2018.
	Zeng Yuting. Object detection and tracking based on the deep learning[D]. Nanchang: East China University of Technology, 2018.
[21]	Peng X, Yang F, Wang G, et al. A convolutional neural network-based deep learning methodology for recognition of partial discharge patterns from high-voltage cables[J]. IEEE Transactions on Power Delivery, 2019, 34(4):1460-1469. doi: 10.1109/TPWRD.2019.2906086

缺陷类型	击穿电压/kV	实验电压/kV
外导电层爬电	7.7	5.6
绝缘内部气隙	25.0	5.3
绝缘表面划伤	17.0	5.6
绝缘表面金属污秽	51.5	20.0

网络层	大小
Input	64×64二维图像
Conv1	3×3×16
Pool1	3×3×16
Conv2	3×3×32
Pool2	3×3×32
Conv3	3×3×64
Pool3	3×3×64
FC	128

训练数据集	测试数据集	识别精度
D₁	D₂	93.425%
D₁+D₃	D₂	96.600%