기계학습을 활용한 공항 활주로 지상 바람의 예측

• Published in: 한국항공운항학회지 Vol. 32; no. 3; pp. 15 - 28
• Main Authors: 이승민, 이승재, 강하림, 함숙정, 송재익, 김기남, Seung-Min Lee, Seung-Jae Lee, Harim Kang, Sook Jung Ham, Jae Ik Song, Ki Nam Kim
• Format: Journal Article
• Language: Korean
• Published: 한국항공운항학회 01.09.2024
• Subjects: 항공우주공학
• ISSN: 1225-9705; 2466-1791

Wind forecast is one of the key meteorological factors required for safe aircraft takeoff and landing. In this study, we developed an artificial intelligence-based wind compensation method by learning the Korea Air Force Weather Research and Forecast (KAF-WRF) forecast data and the Airfield Meteorological Observation System (AMOS) data at five airports using Support Vector Machine (SVM). The SVM wind prediction models were composed of three types according to the learning period (30 days, 40 days, and 60 days) using seven KAF-WRF variables as training data, and the wind prediction performance at the five airports was evaluated using Root Mean Squared Errors (RMSE). According to the results, the SVM wind prediction model trained using U (east-west) and V (north-south) components performed approximately 18% better than the model trained using wind speed and wind direction. The wind correction of KAF-WRF with AMOS observations via SVM outperformed the conventional KAF-WRF wind predictions in eight out of ten cases, capturing abrupt changes in wind direction and speed with a 25% reduction in RMSE.