실 계통 고장 데이터에 적용 가능한 장단기메모리 기반 송전선로 고장 유형 분류모델 (LSTM-based Fault Classification Model in Transmission Lines for Real Fault Data)

서지정보

· 발행기관 : 대한전기학회
· 수록지 정보 : 전기학회논문지 / 73권 / 3호 / 585 ~ 592페이지
· 저자명 : 김태근, 임세헌, 송경민, 윤성국

초록

Most studies on the line fault classification in transmission lines are conducted based on the simulation data because of the lack of fault data in the real world. However, the real fault signals differ from simulation data due to the presence of noise and uncertainty of fault parameters. As a result, the performance of previous fault classification studies based on the simulation data may not work accurately when they are used in real fault data. In this research, we compared the classification performance of some previous fault classification works based on simulation data with real fault data. The result showed that the classification performance was severely reduced with real fault data. To overcome this mismatch between simulation and real fault data, we propose a fault classification model for real fault data. The proposed classification model combined the long short-term memory (LSTM) layer and fully connected (FC) layer with different input features to train time series and non-time series features from the data. The result shows that the proposed model classifies fault type with high accuracy for both simulation data and real fault data.

영어초록

Most studies on the line fault classification in transmission lines are conducted based on the simulation data because of the lack of fault data in the real world. However, the real fault signals differ from simulation data due to the presence of noise and uncertainty of fault parameters. As a result, the performance of previous fault classification studies based on the simulation data may not work accurately when they are used in real fault data. In this research, we compared the classification performance of some previous fault classification works based on simulation data with real fault data. The result showed that the classification performance was severely reduced with real fault data. To overcome this mismatch between simulation and real fault data, we propose a fault classification model for real fault data. The proposed classification model combined the long short-term memory (LSTM) layer and fully connected (FC) layer with different input features to train time series and non-time series features from the data. The result shows that the proposed model classifies fault type with high accuracy for both simulation data and real fault data.

참고자료

· 없음