{"title":"Drone image recognition and intelligent power distribution network equipment fault detection based on the transformer model and transfer learning","authors":"Jiayong Zhong, Yongtao Chen, Jin Gao, Xiaohong Lv","doi":"10.3389/fenrg.2024.1364445","DOIUrl":null,"url":null,"abstract":"In today’s era of rapid technological advancement, the emergence of drone technology and intelligent power systems has brought tremendous convenience to society. However, the challenges associated with drone image recognition and intelligent grid device fault detection are becoming increasingly significant. In practical applications, the rapid and accurate identification of drone images and the timely detection of faults in intelligent grid devices are crucial for ensuring aviation safety and the stable operation of power systems. This article aims to integrate Transformer models, transfer learning, and generative adversarial networks to enhance the accuracy and efficiency of drone image recognition and intelligent grid device fault detection.In the methodology section, we first employ the Transformer model, a deep learning model based on self-attention mechanisms that has demonstrated excellent performance in handling image sequences, capturing complex spatial relationships in images. To address limited data issues, we introduce transfer learning, accelerating the learning process in the target domain by training the model on a source domain. To further enhance the model’s robustness and generalization capability, we incorporate generative adversarial networks to generate more representative training samples.In the experimental section, we validate our model using a large dataset of real drone images and intelligent grid device fault data. Our model shows significant improvements in metrics such as specificity, accuracy, recall, and F1-score. Specifically, in the experimental data, we observe a notable advantage of our model over traditional methods in both drone image recognition and intelligent grid device fault detection. Particularly in the detection of intelligent grid device faults, our model successfully captures subtle fault features, achieving an accuracy of over 90%, an improvement of more than 17% compared to traditional methods, and an outstanding F1-score of around 91%.In summary, this article achieves a significant improvement in the fields of drone image recognition and intelligent grid device fault detection by cleverly integrating Transformer models, transfer learning, and generative adversarial networks. Our approach not only holds broad theoretical application prospects but also receives robust support from experimental data, providing strong support for research and applications in related fields.","PeriodicalId":12428,"journal":{"name":"Frontiers in Energy Research","volume":null,"pages":null},"PeriodicalIF":2.6000,"publicationDate":"2024-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Frontiers in Energy Research","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.3389/fenrg.2024.1364445","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENERGY & FUELS","Score":null,"Total":0}
引用次数: 0
Abstract
In today’s era of rapid technological advancement, the emergence of drone technology and intelligent power systems has brought tremendous convenience to society. However, the challenges associated with drone image recognition and intelligent grid device fault detection are becoming increasingly significant. In practical applications, the rapid and accurate identification of drone images and the timely detection of faults in intelligent grid devices are crucial for ensuring aviation safety and the stable operation of power systems. This article aims to integrate Transformer models, transfer learning, and generative adversarial networks to enhance the accuracy and efficiency of drone image recognition and intelligent grid device fault detection.In the methodology section, we first employ the Transformer model, a deep learning model based on self-attention mechanisms that has demonstrated excellent performance in handling image sequences, capturing complex spatial relationships in images. To address limited data issues, we introduce transfer learning, accelerating the learning process in the target domain by training the model on a source domain. To further enhance the model’s robustness and generalization capability, we incorporate generative adversarial networks to generate more representative training samples.In the experimental section, we validate our model using a large dataset of real drone images and intelligent grid device fault data. Our model shows significant improvements in metrics such as specificity, accuracy, recall, and F1-score. Specifically, in the experimental data, we observe a notable advantage of our model over traditional methods in both drone image recognition and intelligent grid device fault detection. Particularly in the detection of intelligent grid device faults, our model successfully captures subtle fault features, achieving an accuracy of over 90%, an improvement of more than 17% compared to traditional methods, and an outstanding F1-score of around 91%.In summary, this article achieves a significant improvement in the fields of drone image recognition and intelligent grid device fault detection by cleverly integrating Transformer models, transfer learning, and generative adversarial networks. Our approach not only holds broad theoretical application prospects but also receives robust support from experimental data, providing strong support for research and applications in related fields.
期刊介绍:
Frontiers in Energy Research makes use of the unique Frontiers platform for open-access publishing and research networking for scientists, which provides an equal opportunity to seek, share and create knowledge. The mission of Frontiers is to place publishing back in the hands of working scientists and to promote an interactive, fair, and efficient review process. Articles are peer-reviewed according to the Frontiers review guidelines, which evaluate manuscripts on objective editorial criteria