基于RPM映射和GAN-CNN-RF的输电线路故障诊断-电气工程与自动化-CSCIED科技核心评价数据库-手机版

基于RPM映射和GAN-CNN-RF的输电线路故障诊断

Transmission Line Fault Diagnosis Based on RPM Mapping and GAN-CNN-RF

ES评分 0 浏览量：42 下载量：0

DOI	10.12208/j.jeea.20250073
刊名	电气工程与自动化 Journal of Electrical Engineering and Automation
年，卷(期)	2025, 4(3)
作者	马国煜, 陶锴, 吴祎, 余子煜,
作者单位	南京邮电大学江苏南京
摘要	输电线路是电能传输的核心，对电力系统稳定性有重要影响。因此，当输电线路发生故障时，快速准确地识别出故障类型十分关键。在传统的故障诊断方法中，特征的选取过于依赖于主观判断，可能会遗漏重要特征信息。针对此问题，本研究提出一种基于RPM映射和GAN-CNN-RF的输电线路故障诊断方法，首先，采用RPM（Relative Position Matrix）将原始故障信号映射为二维图像。然后，使用CNN（Convolutional Neural Network）自动提取二维图像的典型特征并输入到RF（Random Forest）中实现故障诊断。针对真实故障数据难以获得的问题，本研究采用GAN（Generative Adversarial Network）对真实故障数据的训练集进行扩充，有效地提升了训练效果。采用Simulink模拟数据集和真实数据集两种数据集对模型进行了训练和测试。与EfficientNet，AlexNet，LeNet等模型相比，所提出的CNN-RF在两种数据集上的各项指标均为最优，验证了所提方法在输电线路故障诊断中的有效性。
Abstract	Transmission lines are the core component of electrical energy transmission. Consequently, rapid and accurate identification of fault types is considered crucial when faults occur in transmission lines. For traditional fault diagnosis methods, feature selection is excessively dependent on subjective judgment, which could miss the critical feature information. A transmission line fault diagnosis method based on RPM mapping and GAN-CNN-RF was proposed in this research. First, original signals were transformed into two-dimensional images by RPM (Relative Position Matrix) mapping. Subsequently, typical features of the two-dimensional images were automatically extracted by a CNN (Convolutional Neural Network) and fed into a RF (Random Forest) for fault diagnosis. Considering the difficulty in obtaining the real fault data, the training set of real fault data was expanded by a GAN (Generative Adversarial Network), which effectively enhances the training effect. The model was validated on both Simulink-simulated datasets and real-world datasets. Compared with the models including EfficientNet, AlexNet, and LeNet, the proposed CNN-RF performed well across all evaluation metrics on both datasets. The effectiveness of this method in transmission line fault diagnosis was experimentally confirmed.
关键词	输电线路，故障诊断，RPM，二维特征提取
KeyWord	transmission lines, fault diagnosis, RPM, two-dimensional feature extraction
基金项目
页码	20-29

参考文献
相关文献

[1] 王建,吴昊,张博,等. 不平衡样本下基于迁移学习-AlexNet的输电线路故障辨识方法
[J]. 电力系统自动化, 2022, 46 (22): 182-191.

[2] Ren G, Zha X, Jiang B, et al. Location of Multiple Types of Faults in Active Distribution Networks Considering Synchronization of Power Supply Area Data. Applied Sciences. 2022 Oct 6;12(19):10024.

[3] 张宇博,郝治国,林泽暄,等. 基于深度字典学习的输电线路故障分类方法
[J]. 电力自动化设备, 2022, 42 (11): 159-166.

[4] Chen X, Ji N, Qin X, et al. Transformer Fault Diagnosis Based on the Improved Sparrow Search Algorithm and Random Forest Feature Selection
[C]. 2024 3rd International Conference on Energy and Electrical Power Systems (ICEEPS), Guangzhou, China, 2024, 1086-1091.

[5] Ni Q, Luo H, Liu J, et al. A Feature Vector Learning-Based Method for Diagnosing Main Circuit Ground Faults in Electrical Traction Drive Systems. IEEE Transactions on Power Electronics. 2024;39(2):2537-45.

[6] Wang J, Zhang B, Yin D, et al. Distribution network fault comprehensive identification method based on voltage-ampere curves and deep ensemble learning. International Journal of Electrical Power & Energy Systems. 2025;164.

[7] 余波,高学军,王灿,等. 基于行波特征与KOA-CNN-BiGRU-AM的柔直输电线路故障诊断
[J]. 电力工程技术, 2025, 44 (02): 185-196.

[8] 王浩,杨东升,周博文,等. 基于并联卷积神经网络的多端直流输电线路故障诊断
[J]. 电力系统自动化, 2020, 44 (12): 84-92.

[9] Goodfellow IJ, Pouget-Abadie J, Mirza M, et al. Generative Adversarial Nets.
[C]. 28th Conference on Neural Information Processing Systems (NIPS); 2014.

[10] Bhandari V, Londhe ND, Kshirsagar GB. TL-P3GAN: An Efficient Temporal-Learning-Based Generative Adversarial Network for Precise P300 Signal Generation for P300 Spellers. Ieee Transactions on Cognitive and Developmental Systems. 2024;16(2):692-705.

[11] Chen W, Shi K. A deep learning framework for time series classification using Relative Position Matrix and Convolutional Neural Network. Neurocomputing. 2019;359:384-94.

[12] Tao K, Wang Q, Yue D. Data compression and damage evaluation of underground pipeline with musicalized sonar GMM. IEEE Transactions on Industrial Electronics. 2023 May 1;71(3):3093-102.

[13] Tao K, Chen G, Wang Q, et al. Ultrasonic Curved Coordinate Transform-RAPID With Bayesian Method for the Damage Localization of Pipeline. IEEE Transactions on Industrial Electronics. 2024 Feb 27.

[14] Zheng W, Tao K, Jiang, W. Automatic pattern identification of rock moisture based on the Staff-RF model. Journal of Geophysics and Engineering, 2018;15(2), 438-448.

[15] Zheng W, Jiang J, Tao K. A method based on musical-staff-inspired signal processing model for measuring rock moisture content. Measurement. 2018 Sep 1;125:577-85.

[16] Zhang H, Zhao M, Liu L, Zhong H, Liang Z, Yang Y, et al. Deep Multimodel Cascade Method Based on CNN and Random Forest for Pharmaceutical Particle Detection. Ieee Transactions on Instrumentation and Measurement. 2020;69(9):7028-42.

[17] 徐耀松,张原. 基于改进VMD-MSE方法的输电线路故障特征分析及分类
[J]. 电子测量与仪器学报, 2019, 33 (06): 89-95.

引用本文

马国煜, 陶锴, 吴祎, 余子煜. 基于RPM映射和GAN-CNN-RF的输电线路故障诊断 [J]. 电气工程与自动化. 2025; 4; (3). 20 - 29.

文献评论

相关学者

相关机构