基于深度-广度联合剪枝的电力设备局部放电轻量化诊断方法

doi:10.19595/j.cnki.1000-6753.tces.221585

Abstract
Figure/Table
References
Related Citation (15)

Download: PDF (2620 KB) HTML (1 KB)
Export: BibTeX | EndNote (RIS)

Abstract Partial discharge (PD) is an early indicator on insulation deterioration that will cause catastrophic failure on the power system, so PD diagnosis is a significant approach to monitor the operating status of the electrical equipment. Recently, deep learning (DL) has gradually reached the mainstream in the field of PD diagnosis and the increasing intelligent terminals near power equipment maybe serve as the carrier for such DL models. However, the existing DL-based PD models tend to occupy higher computing resources, while the current power intelligent terminals usually have small memory space and limited calculation capacity. To address it, a lightweight PD diagnosis method based on depth-width joint pruning is proposed in this paper, which can effectively compress the computational resource consumption of DL model while ensuring the accuracy of PD diagnosis.
Firstly, a set of 1D PRPD matrix is constructed based on the discharge amplitude, phase and pulse number, including four types of PD defect such as point discharge, surface discharge, air-gap discharge and suspended discharge. Then, this method selects MobileNetV2 as the basic model, and an iterable importance factor α is inserted in the training process to assess the importance of each convolution module. According to α, several modules with low importance factors (close to 0) are pruned to simplify the basic model in depth direction. Finally, to further compress this model, it adopts a filter-level pruning approach called filter pruning via geometric median (FPGM) to remove redundant convolution filters, in which the pruning ratio of filters in each layer is adaptively calculated by an enhanced simulated annealing search (ESA). Through cyclical search, a highly compressed model can be generated with almost no loss of accuracy, while greatly reduces the computational cost and time.
The experimental results show that, with the premise of remaining the diagnosis accuracy, the proposed method can automatically design an efficient PD diagnosis model with lightweight architecture and less diagnosis time, achieving 98.23% accuracy, 9.9 times of parameter compression and 2.3 times of inference acceleration. The comparison with different pruning methods such as pre-training pruning, Slimming, FPGM and AutoML for model compression (AMC) shows that, the proposed method has 0.31%~4.53% better diagnosis accuracy than the others’, and its pruning speed is 34 faster than AMC’s. Therefore, it is more appropriate to apply in the compression of PD diagnosis model. Furthermore, compared with other DL-based diagnosis models such as VGG16, ResNet18, ShuffleNetV2 and GhostNet, the proposed method shows similar diagnosis accuracy under different noise levels, and more significantly, the storage and memory consumption reduce to only 0.90 MB and 12.89 MB respectively, and the maximum speedup of diagnosis speed reaches up to 9.3 times.
The following conclusions can be drawn from the above analysis: (1) Driven by the PRPD data,the proposed depth-direction pruning can automatically learn the importance of each convolution module for PD diagnosis task, and then pruning the low importance parts have less impact on the accuracy of PD diagnosis. (2) In the breadth-direction pruning, the pruning ratio of filters in each layer is adaptively determined by ESA search, which achieves the comprehensive optimum among accuracy, parameters number and PD diagnosis time. Therefore, it benefits for avoiding over- or under-compression of DL model. (3) By jointing both depth- and width-directions pruning, the proposed method realizes the automatic design of the architecture for PD diagnosis model, which greatly reduces the storage space, memory consumption and diagnosis time. Compared with traditional deep learning models, it is more suitable for the scenarios of resource-constrained power intelligent terminals.

Key words： Partial discharge deep learning automatic pruning lightweight diagnosis architecture design

Received: 16 August 2022

PACS:

TM85

	Service

	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	Zhang Yi
	Zhu Yongli

Cite this article:

Zhang Yi,Zhu Yongli. A Lightweight Partial Discharge Diagnosis Method of Power Equipment Based on Depth-Width Joint Pruning[J]. Transactions of China Electrotechnical Society, 2023, 38(7): 1935-1945.

URL:

https://dgjsxb.ces-transaction.com/EN/10.19595/j.cnki.1000-6753.tces.221585 OR https://dgjsxb.ces-transaction.com/EN/Y2023/V38/I7/1935

[1] 宋思蒙, 钱勇, 王辉, 等. 基于方向梯度直方图属性空间的局部放电模式识别改进算法[J]. 电工技术学报, 2021, 36(10): 2153-2160.
Song Simeng, Qian Yong, Wang Hui, et al.Improved algorithm for partial discharge pattern recognition based on histogram of oriented gradient attribute space[J]. Transactions of China Electrotechnical Society, 2021, 36(10): 2153-2160.
[2] 李泽, 王辉, 钱勇, 等. 基于加速鲁棒特征的含噪局部放电模式识别[J]. 电工技术学报, 2022, 37(3): 775-785.
Li Ze, Wang Hui, Qian Yong, et al.Pattern recognition of partial discharge in the presence of noise based on speeded up robust features[J]. Transactions of China Electrotechnical Society, 2022, 37(3): 775-785.
[3] Gao Kai, Tan Kexiong, Li Fuqi, et al.PD pattern recognition for stator bar models with six kinds of characteristic vectors using BP network[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2002, 9(3): 381-389.
[4] Ibrahim K, Sharkawy R M, Salama M M A, et al. Realization of partial discharge signals in transformer oils utilizing advanced computational techniques[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2012, 19(6): 1971-1981.
[5] 周利军, 刘聪, 权圣威, 等. 基于点对称变换的乙丙橡胶电缆终端缺陷诊断[J]. 电工技术学报, 2022, 37(9): 2388-2398.
Zhou Lijun, Liu Cong, Quan Shengwei, et al.Defect diagnosis of EPR cable terminal based on symmetrized dot pattern[J]. Transactions of China Electrotechnical Society, 2022, 37(9): 2388-2398.
[6] Barrios S, Buldain D, Comech M P, et al.Partial discharge identification in MV switchgear using scalogram representations and convolutional AutoEncoder[J]. IEEE Transactions on Power Delivery, 2021, 36(6): 3448-3455.
[7] 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.
[8] 万晓琪, 宋辉, 罗林根, 等. 卷积神经网络在局部放电图像模式识别中的应用[J]. 电网技术, 2019, 43(6): 2219-2226.
Wan Xiaoqi, Song Hui, Luo Lingen, et al.Application of convolutional neural networks in pattern recognition of partial discharge image[J]. Power System Technology, 2019, 43(6): 2219-2226.
[9] 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.
Zhu Yufeng, Xu Yongpeng, Chen Xiaoxin, et al.Pattern recognition of partial discharges in DC XLPE cables based on convolutional neural network[J]. Transactions of China Electrotechnical Society, 2020, 35(3): 659-668.
[10] Borghei M, Ghassemi M.A deep learning approach for discrimination of single- and multi-source corona discharges[J]. IEEE Transactions on Plasma Science, 2021, 49(9): 2936-2945.
[11] 张聪聪, 王刚, 高栋, 等. 基于卷积网络的GIS局部放电缺陷诊断方法与应用[J]. 电工电能新技术, 2021, 40(3): 72-80.
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, 2021, 40(3): 72-80.
[12] 陈伟根, 张知先, 李剑, 等. 电气设备状态参量智能传感技术[J]. 中国电机工程学报, 2020, 40(增刊1): 323-342.
Chen Weigen, Zhang Zhixian, Li Jian, et al.Intelligent sensing technology for power equipment state parameters[J]. Proceedings of the CSEE, 2020, 40(S1): 323-342.
[13] 马富齐, 王波, 董旭柱, 等. 电力工业安全影像解译:基本概念与技术框架[J]. 中国电机工程学报, 2022, 42(2): 458-475.
Ma Fuqi, Wang Bo, Dong Xuzhu, et al.Safety image interpretation of power industry: basic concepts and technical framework[J]. Proceedings of the CSEE, 2022, 42(2): 458-475.
[14] 黄彦钦, 余浩, 尹钧毅, 等. 电力物联网数据传输方案:现状与基于5 G技术的展望[J]. 电工技术学报, 2021, 36(17): 3581-3593.
Huang Yanqin, Yu Hao, Yin Junyi, et al.Data transmission schemes of power internet of things: present and outlook based on 5 G technology[J]. Transactions of China Electrotechnical Society, 2021, 36(17): 3581-3593.
[15] 赵仕策, 赵洪山, 寿佩瑶. 智能电力设备关键技术及运维探讨[J]. 电力系统自动化, 2020, 44(20): 1-10.
Zhao Shice, Zhao Hongshan, Shou Peiyao.Discussion on key technology and operation & maintenance of intelligent power equipment[J]. Automation of Electric Power Systems, 2020, 44(20): 1-10.
[16] Sandler M, Howard A, Zhu Menglong, et al.MobileNetV2: inverted residuals and linear bottlenecks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018: 4510-4520.
[17] Wang Yanxin, Yan Jing, Sun Qifeng, et al.A MobileNets convolutional neural network for GIS partial discharge pattern recognition in the ubiquitous power internet of things context: optimization, comparison, and application[J]. IEEE Access, 2019, 7: 150226-150236.
[18] Luo Jianhao, Zhang Hao, Zhou Hongyu, et al.ThiNet: pruning CNN filters for a thinner net[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41(10): 2525-2538.
[19] Liu Zhuang, Li Jianguo, Shen Zhiqiang, et al.Learning efficient convolutional networks through network slimming[C]//2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 2017: 2755-2763.
[20] He Yang, Liu Ping, Wang Ziwei, et al.Filter pruning via geometric median for deep convolutional neural networks acceleration[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, 2020: 4335-4344.
[21] He Yihui, Lin Ji, Liu Zhijian, et al.AMC: AutoML for model compression and acceleration on mobile devices[C]//Computer Vision - ECCV 2018, Munich, Germany, 2018: 815-832.
[22] 王旭红, 李浩, 樊绍胜, 等. 基于改进SSD的电力设备红外图像异常自动检测方法[J]. 电工技术学报, 2020, 35(增刊1): 302-310.
Wang Xuhong, Li Hao, Fan Shaosheng, et al.Infrared image anomaly automatic detection method for power equipment based on improved single shot multi box detection[J]. Transactions of China Electrotechnical Society, 2020, 35(S1): 302-310.
[23] Zhai Xiaodong, Qiao Fei, Ma Yumin, et al.A novel fault diagnosis method under dynamic working conditions based on a CNN with an adaptive learning rate[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-12.
[24] Meng Fanxu, Cheng Hao, Zhuang Jiaxin, et al.RMNet: equivalently removing residual connection from networks[EB/OL]. https://arxiv.org/abs/2111.00687.
[25] Ma Ningning, Zhang Xiangyu, Zheng Haitao, et al.ShuffleNet V2: practical guidelines for efficient CNN architecture design[C]//Computer Vision-ECCV 2018, Munich, Germany, 2018: 122-138.
[26] Han Kai, Wang Yunhe, Tian Qi, et al.GhostNet: more features from cheap operations[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 2020: 1577-1586.