基于Light-Resnet卷积神经网络的电力设备监测数值识别算法

doi:10.11930/j.issn.1004-9649.202310020

摘要/Abstract

摘要：

在智能电网中，精确监测输电、配电及供电关键设备的运行状态对在线运维至关重要。面对人工抄录和巡检的低效，以及监测装置数字化升级的复杂安装、高成本和长周期等挑战，结合图像采集装置与图像处理技术，根据计算资源合理分配任务，开发了一种基于Light-Resnet数值识别算法，该算法通过D-Add损失函数优化网络训练过程，实现电力设备监测数据的远程读取。实验表明：Light-Resnet以6090的参数量在MNIST数据集获得了98.8%的严格准确率，结合边端协同机制，终端侧能耗降低了20.73%。这一算法不仅证明了自身在资源受限环境下的适应性和高效性，同时D-Add损失函数的设计也显著提升了网络的准确度。

关键词: light-resnet, D-add, 边端协同机制, 数值识别, 智能电网

Abstract:

In the smart grid, precise monitoring of the operational status of critical equipment for transmission, distribution and power supply is essential for effective online maintenance. Faced with the inefficiencies of manual recording and inspection, as well as the challenges associated with the complex installation, high cost and lengthy periods required for digital upgrades of monitoring devices, a novel approach that integrates image capture devices with image processing technology has been developed. This approach, leveraging the allocation of computational resources for task distribution, introduces a Light-Resnet-based numerical recognition algorithm, which enhances network training through the optimization of the D-Add loss function, enabling remote reading of electrical equipment monitoring data. Experiments have demonstrated that Light-Resnet achieves a rigorous accuracy rate of 98.8% on the MNIST dataset with only 6090 parameters. When combined with edge computing collaboration mechanisms, it resulted in a 20.73% reduction in power consumption on the terminal side. The proposed algorithm not only proves its adaptability and efficiency in resource-constrained environments but also significantly improves the network's accuracy with design of the D-Add loss function.

Key words: light-resnet, D-add, edge-end collaboration mechanism, numerical recognition, smart grid

孔志恒, 谭冲, 唐培耀, 胡成博, 郑敏. 基于Light-Resnet卷积神经网络的电力设备监测数值识别算法[J]. 中国电力, 2024, 57(8): 206-213.

Zhiheng KONG, Chong TAN, Peiyao TANG, Chengbo HU, Min ZHENG. Numerical Recognition Algorithm for Power Equipment Monitoring Based on Light-Resnet Convolutional Neural Network[J]. Electric Power, 2024, 57(8): 206-213.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202310020

https://www.electricpower.com.cn/CN/Y2024/V57/I8/206

图/表 9

图 1 系统架构示意

Fig.1 System architecture illustration

图 2 Light-Resnet网络结构

Fig.2 Light-Resnet network architecture

图 3 投影捷径及恒等捷径

Fig.3 Projection shortcut & identity shortcut

图 4 损失函数D-Add的计算过程

Fig.4 The calculation procedure of the loss function D-Add

表 1 不同网络结构性能对比

Table 1 Performance comparison of different network architectures

网络架构	$\overline {{N_{{\text{ACC}}}}} $/%	$\overline {{N_{{\text{SACC}}}}} $/%	权重大小
CapsNet	99.99	99.23	6804384
Light-Resnet	100.00	98.79	6090
MOCNN	99.94	96.38	6538

图 5 不同网络结构严格准确率对比

Fig.5 Strict accuracy comparison of different network architectures

表 2 多种损失函数性能对比

Table 2 Performance comparison of multiple loss functions

损失函数	$\overline {{N_{{\text{ACC}}}}} $/%	$\overline {{N_{{\text{SACC}}}}} $/%
CEL	99.97	98.45
HRC	99.99	97.82
D-ADD	100.00	98.79
AVG	99.99	97.95

图 6 多种损失函数严格准确率对比

Fig.6 Strict accuracy comparison of multiple loss functions

表 3 端侧能耗对比

Table 3 Terminal-side power consumption comparison 单位：J

位置	E_dr	E_dd	E_send	E_sensor	E_other	E
边	0	0	49.624	1.462	6.819	57.905
端	2.218	2.878	2.709	1.517	6.427	15.749
边端协同	2.211	0	2.721	1.514	6.599	13.045

参考文献 26

1	李博, 高志远. 人工智能技术在智能电网中的应用分析和展望[J]. 中国电力, 2017, 50 (12): 136- 140.
	LI Bo, GAO Zhiyuan. Analysis and prospect on the application of artificial intelligence technologies in smart grid[J]. Electric Power, 2017, 50 (12): 136- 140.
2	白钒, 胡杰, 何鹏, 等. 基于物联网技术的智能电网基础设施建设数字化管理平台研究[J]. 机械与电子, 2022, 40 (10): 77- 80.
	BAI Fan, HU Jie, HE Peng, et al. Research on digital management platform for smart grid infrastructure construction based on Internet of Things technology[J]. Machinery & Electronics, 2022, 40 (10): 77- 80.
3	ZHAO S T, LI B S, YUAN J S, et al. Research on remote meter automatic reading based on computer vision[C]//2005 IEEE/PES Transmission & Distribution Conference & Exposition: Asia and Pacific. Dalian, China. IEEE, 2005: 1–4.
4	EDWARD V C P. Support vector machine based automatic electric meter reading system[C]//2013 IEEE International Conference on Computational Intelligence and Computing Research. Enathi, India. IEEE, 2013: 1–5.
5	RODRİGUEZ M, BERDUGO G, JABBA D, et al. HD_MR: a new algorithm for number recognition in electrical meters[J]. Turkish Journal of Electrical Engineering & Computer Sciences, 2014, 22, 87- 96.
6	陈英, 李磊, 汪文源, 等. 家用水表字符的识别算法研究[J]. 现代电子技术, 2018, 41 (18): 99- 103.
	CHEN Ying, LI Lei, WANG Wenyuan, et al. Research on character recognition algorithm for domestic water meter[J]. Modern Electronics Technique, 2018, 41 (18): 99- 103.
7	GÓMEZ L, RUSIÑOL M, KARATZAS D. Cutting sayre’s knot: reading scene text without segmentation. application to utility meters[C]//2018 13th IAPR International Workshop on Document Analysis Systems (DAS). Vienna, Austria. IEEE, 2018: 97-102.
8	YANG F, JIN L W, LAI S X, et al. Fully convolutional sequence recognition network for water meter number reading[J]. IEEE Access, 2019, 7, 11679- 11687. DOI
9	DA SILVA MARQUES R C, COSTA SERRA A, FERREIRA FRANÇA J V, et al. Image-based electric consumption recognition via multi-task learning[C]//2019 8th Brazilian Conference on Intelligent Systems (BRACIS). Salvador, Brazil. IEEE, 2019: 419–424.
10	龚安, 张洋, 唐永红. 基于YOLOv3网络的电能表示数识别方法[J]. 计算机系统应用, 2020, 29 (1): 196- 202.
	GONG An, ZHANG Yang, TANG Yonghong. Automatic reading method of electric energy meter based on YOLOv3[J]. Computer Systems & Applications, 2020, 29 (1): 196- 202.
11	顾允迪, 徐望明, 何钦. 字轮式仪表智能图像抄表系统的设计[J]. 液晶与显示, 2023, 38 (7): 985- 996.
	GU Yundi, XU Wangming, HE Qin. Design of image-based intelligent meter reading system for wheel meters[J]. Chinese Journal of Liquid Crystals and Displays, 2023, 38 (7): 985- 996.
12	吉训生, 谭凯凯. 基于卷积神经网络的水表读数识别方法[J]. 传感器与微系统, 2020, 39 (12): 130- 133.
	JI Xunsheng, TAN Kaikai. Water meter character recognition method based on CNN[J]. Transducer and Microsystem Technologies, 2020, 39 (12): 130- 133.
13	HAN D, KIM H. A number recognition system with memory optimized convolutional neural network for smart metering devices[C]//2018 International Conference on Electronics, Information, and Communication (ICEIC). Honolulu, HI, USA. IEEE, 2018: 1–4.
14	LI C S, SU Y K, YUAN R, et al. Light-weight spliced convolution network-based automatic water meter reading in smart city[J]. IEEE Access, 2019, 7, 174359- 174367. DOI
15	JUNAGADE S, JAIN P, SARANGI S, et al. Digital display recognition towards connected sensing systems for precision agriculture[C]//2021 IEEE Global Humanitarian Technology Conference (GHTC). Seattle, WA, USA. IEEE, 2021: 155–162.
16	XIU H, HE J, ZHANG X T, et al. HRC-mCNNs: a hybrid regression and classification multibranch CNNs for automatic meter reading with smart shell[J]. IEEE Internet of Things Journal, 2022, 9 (24): 25752- 25766. DOI
17	FAN Z Z, SHI L R, XI C, et al. Real time power equipment meter recognition based on deep learning[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71, 5017015.
18	DONG Z P, GAO Y, YAN Y H, et al. Vector detection network: an application study on robots reading analog meters in the wild[J]. IEEE Transactions on Artificial Intelligence, 2021, 2 (5): 394- 403. DOI
19	CIPOLLA R, GAL Y, KENDALL A. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA. IEEE, 2018: 7482–7491.
20	EVGENIOU T, PONTIL M. Regularized multi: task learning[C]//Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining. Seattle, WA, USA. ACM, 2004: 109–117.
21	CARUANA R. Multitask learning[M]//Learning to Learn. Boston, MA: Springer US, 1998: 95–133.
22	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA. IEEE, 2016: 770–778.
23	HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Identity mappings in deep residual networks [C]//Computer Vision – ECCV 2016. Cham: Springer International Publishing, 2016, 630: 45.
24	IOFFE S, SZEGEDY C. Batch normalization: accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on Machine Learning - Volume 37. Lille, France. ACM, 2015: 448–456.
25	GLOROT X, BORDES A, BENGIO Y. Deep sparse rectifier neural networks[J]. Journal of Machine Learning Research, 2011, 15, 315- 323.
26	SABOUR S, FROSST N, HINTON G E. Dynamic routing between capsules[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach, California, USA. ACM, 2017: 3859–3869.

[1]	于宗超, 文明, 李湘华, 谢欣涛, 杨洪明. 融合群体智慧的分布式智能电网高效发展管理策略[J]. 中国电力, 2024, 57(10): 57-68.
[2]	吴岩, 王光政. 基于CiteSpace的配电网韧性评估与提升研究综述与展望[J]. 中国电力, 2023, 56(12): 100-112, 137.
[3]	徐杰, 汪石农. 面向智能电网电压崩溃路径的优化粒子群算法[J]. 中国电力, 2023, 56(1): 142-149.
[4]	李峰, 彭慧敏, 袁虎玲, 万芳茹, 黄天罡, 李威. UPFC在线运行优化辅助决策应用功能设计[J]. 中国电力, 2021, 54(7): 76-82.
[5]	佘蕊, 张宁池, 王艳茹, 郭丹丹, 马文洁, 刘卉, 张洁. 面向电力物联网的5G通信认知无线电NOMA系统研究[J]. 中国电力, 2021, 54(5): 35-45.
[6]	曾永浩, 范心明, 彭元泉, 李新, 张子麒, 宋关羽. 三端柔性多状态开关运行控制与示范应用[J]. 中国电力, 2021, 54(11): 97-103.
[7]	柴彬, 刘飞, 江平开, 江秀臣, 黄兴溢. 面向电力设备检测与诊断的压电材料及器件[J]. 中国电力, 2021, 54(10): 105-116.
[8]	麻建中, 胡凯波, 於立峰. 工控运维中基于语义分析的智能电网控制攻击防御[J]. 中国电力, 2020, 53(9): 214-220.
[9]	李博, 方彤. 北斗卫星导航系统（BDS）在智能电网的应用与展望[J]. 中国电力, 2020, 53(8): 107-116.
[10]	贾惠彬, 盖永贺, 李保罡, 郑宏达. 基于强化学习的电力通信网故障恢复方法[J]. 中国电力, 2020, 53(6): 34-40.
[11]	曾懿辉, 何通, 郭圣, 熊勇良, 崔颖铷, 左剑, 罗昊. 基于差分定位的输电线路多旋翼无人机智能巡检[J]. 中国电力, 2019, 52(7): 24-30.
[12]	钱斌, 蔡梓文, 肖勇, 杨劲锋, 董歆滢, 谷科. 基于边缘计算的电表计量系统数据协同检测方案[J]. 中国电力, 2019, 52(11): 145-152.
[13]	周静, 孙媛媛, 胡紫巍, 卢利锋, 刘国军. 智能电网信息通信架构演进探讨[J]. 中国电力, 2018, 51(3): 131-135.
[14]	王凯军, 郑伟民, 孙可, 靳康萌, 张沛. 智能电网增值业务的调研分析及个性化服务制定[J]. 中国电力, 2018, 51(2): 67-74.
[15]	张俊, 王飞跃, 林洁瑜. 平行智能与智慧能源：全面融合人因的社会能源技术[J]. 中国电力, 2018, 51(10): 26-31.