Transmission line fault identification method based on weighted federated learning and multimodal residual network
Although data-driven deep learning networks can be used for transmission line fault identification, the generalization performance of the algorithm will be seriously reduced in the face of “data islands” and class imbalance among fault samples in various regions. Therefore, this paper proposed a tra...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | International Journal of Electrical Power & Energy Systems |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S0142061525004089 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Although data-driven deep learning networks can be used for transmission line fault identification, the generalization performance of the algorithm will be seriously reduced in the face of “data islands” and class imbalance among fault samples in various regions. Therefore, this paper proposed a transmission line fault identification method based on weighted federated learning and multi-modal residual network. First, through the statistical analysis of historical fault samples of transmission lines, significant differences in fault types, causes, waveforms, and weather conditions across different regions are revealed. Second, transient waveform images and one-hot encoded weather conditions at the time of fault occurrence are used as inputs for the fault identification network based on multimodal residual classification model. Then, leveraging the federated learning mechanism, aggregating and distributing local model parameters to eliminate the impact of regional differences in fault sample features on the performance of classification model. Finally, the results of case study for real world fault samples show that under the federated learning mechanism, compared to training with data from a single region, the proposed method improves the fault type and cause identification accuracy to 99.63% and 96.17%, respectively, effectively enhancing the identification accuracy and generalization ability. |
|---|---|
| ISSN: | 0142-0615 |