LSTM-based Deep Learning Method for Excessive Neutral-to-ground Voltage (NTGV) Localization in Secondary Distribution Systems
The excessive neutral-to-ground voltage (NTGV) on the secondary distribution system (SDS) may lead to unnecessary losses and safety hazards. Methods for troubleshooting and monitoring the root cause of these issues are limited. This paper introduces a deep learning (DL) method that utilizes real-wor...
| Published in: | 2024 IEEE 4th International Conference in Power Engineering Applications: Powering the Future: Innovations for Sustainable Development, ICPEA 2024 |
|---|---|
| Main Author: | |
| Format: | Conference paper |
| Language: | English |
| Published: |
Institute of Electrical and Electronics Engineers Inc.
2024
|
| Online Access: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85191746184&doi=10.1109%2fICPEA60617.2024.10498326&partnerID=40&md5=043bd1bbcad40f7c33345408c777e187 |
| Summary: | The excessive neutral-to-ground voltage (NTGV) on the secondary distribution system (SDS) may lead to unnecessary losses and safety hazards. Methods for troubleshooting and monitoring the root cause of these issues are limited. This paper introduces a deep learning (DL) method that utilizes real-world data to pinpoint the source of the problem due to ground fault events, whether upstream or downstream of the measurement point. The method employs a specialized recurrent neural network (RNN), specifically long short-term memory (LSTM), adept at processing time-series signals. Using 921 two-cycle NTGV time series data from various SDS locations, the study shows that the proposed method effectively locates the source of NTGV with 99.2% accuracy on the test dataset. The developed architecture is novel, representing the first implementation for feature learning and classification of NTGV source problems. © 2024 IEEE. |
|---|---|
| ISSN: | |
| DOI: | 10.1109/ICPEA60617.2024.10498326 |