이 연구는 위성사진을 활용하여 건설지점의 기대풍속을 예측하기 위한 인공신경망 방법론을 제안한다. 제안된 방법은 기존 의 엔지니어의 판단을 대체하여, Auto-Encoder를 사용해 지형적 특성을 정량화하고, 이를 바탕으로 대상지점과 유사한 지역의 관측소 풍속 데이터를 선형 조합해 기대 풍속을 예측한다. 또한, 머신러닝과 인공신경망을 활용한 종단간 풍속 예측 모델을 제안하고, 성능을 비교 분석하였다. 그 결과 관측소의 풍속 데이터의 선형 조합보다는 종단간 모델을 구성하는 방법이 더 높은 정확도를 보였으며, 특히 Graph Neural Network (GNN)이 Multi-Layer Perceptron (MLP)에 비해 상당히 우수한 예측 성능을 나타내었다.

큰 에디 모의과정을 포함한 WRF 모델 (WRF-LES)을 이용하여 수치모델의 수평공간 규모에 따른 대기경계층 모수화 실험과 LES 모의 결과를 지표층 근처의 풍속 예측에 대하여 비교하였다. 수치실험은 복잡한 산악지형과 해안지역을 포함하는 강원도 지역에서 수평해상도 1 km와 333 m 실험을 수행하였다. 수평해상도 1 km 실험은 대기경계층 모수화 방안을 채택하였으며, 333 m 실험에서는 LES를 이용하였다. 복잡한 산악지역에서의 풍속 예측의 정확성은 수평해상도 1 km 실험 보다 333 m 실험에서 향상되었으며 해안지역에서는 1 km 실험에서 관측과 더 일치하였다. 지표층 근처의 큰 난류를 직접 계산하는 LES 실험은 산악지역의 풍속예측 개선에 기여하였다.

This study used a quantile regression model and a non-homogeneous regression model to calibrate probabilistic forecasts of wind speed. These techniques were applied to the forecasts of wind speed over Pyeongchang area using 51-member European Centre for Medium-Range Weather Forecast (ECMWF). Reliability analysis was carried out by using rank histogram to identify the statistical consistency of ensemble forecasts and corresponding observations. The performances were evaluated by rank histogram, mean absolute error, root mean square error and continuous ranked probability score. The results showed that the forecasts of quantile regression and non-homogeneous regression models performed better than the raw ensemble forecasts. However, the differences of prediction skills between quantile regression and nonhomogeneous regression models were insignificant.

This paper applied the ensemble model output statistics (EMOS) with truncated normal distribution, which are easy to implement postprocessing techniques, to calibrate probabilistic forecasts of wind speed that take the form of probability density functions. We also considered the alternative implementations of EMOS, which were EMOS exchangeable model and reduced EMOS model. These techniques were applied to the forecasts of wind speed over Pyeongchang area using 51 members of the Ensemble Prediction System for Global (EPSG). The performances were evaluated by rank histogram, mean absolute error, root mean square error and continuous ranked probability score. The results showed that EMOS models with truncated normal distribution performed better than the raw ensemble and ensemble mean. Especially, the reduced EMOS model exhibited better prediction skill than EMOS exchangeable model in most stations of study area.

This paper used the Bayesian model averaging (BMA) with gamma distribution that takes the form of probability density functions to calibrate probabilistic forecasts of wind speed. We considered the alternative implementation of BMA, which was BMA gamma exchangeable model. This method was applied for forecasting of wind speed over Pyeongchang area using 51 members of the Ensemble Prediction System for Global (EPSG). The performances were evaluated by rank histogram, means absolute error, root mean square error, continuous ranked probability score and skill score. The results showed that BMA gamma exchangeable models performed better in forecasting wind speed, compared to the raw ensemble and ensemble mean.

This study estimated extreme wind speed by means of national wind map, provided by Korea Institute of Energy Research, to sustain the wind towers’ structural stability that is required for the production of mass wind power. The number of research object regions is three in the East, West and South seas. Meteorological resource data were calculated by dividing the data into the various reference periods. In addition, Gumbel distribution method and Extreme Wind Speed Model (EWM) indicated in IEC 61400-3 were adopted to measure the extreme wind speed. In conclusion, it is discovered that the more suitable Gumbel distribution method is to secure the stability of wind towers.