LSTM과 GRU 딥러닝 IoT 파워미터 기반의 단기 전력사용량 예측
이선민, 선영규, 이지영, 이동구, 조은일, 박대현, 김용범, 심이삭, Jin‐Young Kim
Abstract
이선민, 선영규, 이지영, 이동구, 조은일, 박대현, 김용범, 심이삭, Jin‐Young Kim
Abstract
본 연구에서는 Long Short Term Memory (LSTM) 신경망과 Gated Recurrent Unit(GRU) 신경망을 Internet of Things (IoT) 파워미터에 적용하여 단기 전력사용량 예측방법을 제안하고, 실제 가정의 전력사용량 데이터 를 토대로 예측 성능을 분석한다. 성능평가 지표로써 Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Percentage Error (MPE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE) 를 이용한다. 실험 결과는 GRU 기반의 모델이 LSTM 기반의 모델에 비해 MAPE 기준으로 4.52%, MPE 기준으로 5.59%만큼의 성능개선을 보였다.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
본 연구에서는 Long Short Term Memory (LSTM) 신경망과 Gated Recurrent Unit(GRU) 신경망을 Internet of Things (IoT) 파워미터에 적용하여 단기 전력사용량 예측방법을 제안하고, 실제 가정의 전력사용량 데이터 를 토대로 예측 성능을 분석한다. 성능평가 지표로써 Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Percentage Error (MPE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE) 를 이용한다. 실험 결과는 GRU 기반의 모델이 LSTM 기반의 모델에 비해 MAPE 기준으로 4.52%, MPE 기준으로 5.59%만큼의 성능개선을 보였다.
Key concepts: Mean absolute percentage error, Mean squared error, Mean absolute error, Statistics, Mathematics, Approximation error