Hyperparameter Analysis-based Long-Short Term Memory (LSTM) for Power Quality Disturbances Classification
Power quality disturbances are a critical issue in electrical power systems, as they can lead to more severe problems with electrical machines or equipment, resulting in significant losses. While such disturbances are rare, using binary classification to differentiate between normal and abnormal pow...
| 出版年: | 2024 IEEE International Conference on Power and Energy, PECon 2024 |
|---|---|
| 第一著者: | |
| フォーマット: | Conference paper |
| 言語: | English |
| 出版事項: |
Institute of Electrical and Electronics Engineers Inc.
2024
|
| オンライン･アクセス: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85217379198&doi=10.1109%2fPECON62060.2024.10827412&partnerID=40&md5=ca1bb516f1ea2e09cc96fd117950055d |
| 要約: | Power quality disturbances are a critical issue in electrical power systems, as they can lead to more severe problems with electrical machines or equipment, resulting in significant losses. While such disturbances are rare, using binary classification to differentiate between normal and abnormal power quality signals has proven effective in detecting disruptions in power systems. This paper proposes a binary classification approach for identifying power quality disturbances, employing the Long Short-Term Memory (LSTM) algorithm. A hyperparameter analysis was conducted to optimize performance, and the results indicated that the highest accuracy of 91.67% was achieved with a sigmoid activation function, a learning rate of 0.0001, 40 epochs, and a batch size of 128. ©2024 IEEE. |
|---|---|
| ISSN: | |
| DOI: | 10.1109/PECON62060.2024.10827412 |