Transmission line fault detection and classification based on SA-MobileNetV3
Accurate fault detection and classification help to analyze fault causes and quickly restore faulty phases. Deep learning can automatically extract fault features and identify fault types from the original three-phase voltage and current signals. However, this still imposes challenges such as recogn...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2023-12-01
|
| Series: | Energy Reports |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2352484722026439 |
| _version_ | 1827902741053177856 |
|---|---|
| author | Yanhui Xi Weijie Zhang Feng Zhou Xin Tang Zewen Li Xiangjun Zeng Pinghua Zhang |
| author_facet | Yanhui Xi Weijie Zhang Feng Zhou Xin Tang Zewen Li Xiangjun Zeng Pinghua Zhang |
| author_sort | Yanhui Xi |
| collection | DOAJ |
| description | Accurate fault detection and classification help to analyze fault causes and quickly restore faulty phases. Deep learning can automatically extract fault features and identify fault types from the original three-phase voltage and current signals. However, this still imposes challenges such as recognition accuracy and computational complexity. More importantly, high level fault features cannot be extracted in the one-dimensional time series. This paper presents a robust fault classification method based on SA-MobileNetV3 for transmission systems. Considering that the SE (Squeeze-and-Excitation) attention module cannot aggregate the spatial dimension information on the channel, SA (shuffle attention) module is introduced into MobileNetV3, which can effectively fuse the importance of pixels in different channels and in different locations at the same channel. Also, transforming the time series three-phase voltage and current signals into two-dimensional images based on CWT (continuous wavelet transform) makes the proposed method be similar to image recognition, which can mine high level fault features and classify the faults visually. To verify the effectiveness of the method, a 735kV transmission line model is built for data generation through Simulink. Various kinds of fault conditions and factors are considered to verify the adaptability and generalizability. Simulation results show that the method can quickly and accurately identify 11 types of faults, and the accuracy rate is as high as 99.90%. A comparison between the proposed method and other existing techniques shows the superiority of the proposed SA- MobileNetV3, and better anti-noise performance makes it more suitable for real fault signals taken on-site. |
| first_indexed | 2024-03-13T00:03:28Z |
| format | Article |
| id | doaj.art-0061128cd7934fa58b02364cdcc27f68 |
| institution | Directory Open Access Journal |
| issn | 2352-4847 |
| language | English |
| last_indexed | 2024-03-13T00:03:28Z |
| publishDate | 2023-12-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Energy Reports |
| spelling | doaj.art-0061128cd7934fa58b02364cdcc27f682023-07-13T05:28:49ZengElsevierEnergy Reports2352-48472023-12-019955968Transmission line fault detection and classification based on SA-MobileNetV3Yanhui Xi0Weijie Zhang1Feng Zhou2Xin Tang3Zewen Li4Xiangjun Zeng5Pinghua Zhang6Hunan Province Higher Education Key Laboratory of Power System Safety Operation and Control, Changsha University of Science and Technology, Changsha 410114, Hunan Province, ChinaHunan Province Higher Education Key Laboratory of Power System Safety Operation and Control, Changsha University of Science and Technology, Changsha 410114, Hunan Province, China; Corresponding authors.School of Electronic Information and Electrical Engineering, Changsha University, Changsha 410022, Hunan Province, China; Corresponding authors.Hunan Province Higher Education Key Laboratory of Power System Safety Operation and Control, Changsha University of Science and Technology, Changsha 410114, Hunan Province, China; Corresponding authors.Hunan Province Higher Education Key Laboratory of Power System Safety Operation and Control, Changsha University of Science and Technology, Changsha 410114, Hunan Province, ChinaHunan Province Higher Education Key Laboratory of Power System Safety Operation and Control, Changsha University of Science and Technology, Changsha 410114, Hunan Province, ChinaHunan College of Information, Changsha 410200, Hunan Province, ChinaAccurate fault detection and classification help to analyze fault causes and quickly restore faulty phases. Deep learning can automatically extract fault features and identify fault types from the original three-phase voltage and current signals. However, this still imposes challenges such as recognition accuracy and computational complexity. More importantly, high level fault features cannot be extracted in the one-dimensional time series. This paper presents a robust fault classification method based on SA-MobileNetV3 for transmission systems. Considering that the SE (Squeeze-and-Excitation) attention module cannot aggregate the spatial dimension information on the channel, SA (shuffle attention) module is introduced into MobileNetV3, which can effectively fuse the importance of pixels in different channels and in different locations at the same channel. Also, transforming the time series three-phase voltage and current signals into two-dimensional images based on CWT (continuous wavelet transform) makes the proposed method be similar to image recognition, which can mine high level fault features and classify the faults visually. To verify the effectiveness of the method, a 735kV transmission line model is built for data generation through Simulink. Various kinds of fault conditions and factors are considered to verify the adaptability and generalizability. Simulation results show that the method can quickly and accurately identify 11 types of faults, and the accuracy rate is as high as 99.90%. A comparison between the proposed method and other existing techniques shows the superiority of the proposed SA- MobileNetV3, and better anti-noise performance makes it more suitable for real fault signals taken on-site.http://www.sciencedirect.com/science/article/pii/S2352484722026439MobileNetV3Shuffle attentionContinuous wavelet transformFault classification |
| spellingShingle | Yanhui Xi Weijie Zhang Feng Zhou Xin Tang Zewen Li Xiangjun Zeng Pinghua Zhang Transmission line fault detection and classification based on SA-MobileNetV3 Energy Reports MobileNetV3 Shuffle attention Continuous wavelet transform Fault classification |
| title | Transmission line fault detection and classification based on SA-MobileNetV3 |
| title_full | Transmission line fault detection and classification based on SA-MobileNetV3 |
| title_fullStr | Transmission line fault detection and classification based on SA-MobileNetV3 |
| title_full_unstemmed | Transmission line fault detection and classification based on SA-MobileNetV3 |
| title_short | Transmission line fault detection and classification based on SA-MobileNetV3 |
| title_sort | transmission line fault detection and classification based on sa mobilenetv3 |
| topic | MobileNetV3 Shuffle attention Continuous wavelet transform Fault classification |
| url | http://www.sciencedirect.com/science/article/pii/S2352484722026439 |
| work_keys_str_mv | AT yanhuixi transmissionlinefaultdetectionandclassificationbasedonsamobilenetv3 AT weijiezhang transmissionlinefaultdetectionandclassificationbasedonsamobilenetv3 AT fengzhou transmissionlinefaultdetectionandclassificationbasedonsamobilenetv3 AT xintang transmissionlinefaultdetectionandclassificationbasedonsamobilenetv3 AT zewenli transmissionlinefaultdetectionandclassificationbasedonsamobilenetv3 AT xiangjunzeng transmissionlinefaultdetectionandclassificationbasedonsamobilenetv3 AT pinghuazhang transmissionlinefaultdetectionandclassificationbasedonsamobilenetv3 |