Ten-Meter Wind Speed Forecast Correction in Southwest China Based on U-Net Neural Network

Accurate forecasting of wind speed holds significant importance for the economic and social development humanity. However, existing numerical weather predictions have certain inaccuracies due to various reasons. Therefore, it is highly necessary perform statistical post-processing on forecasted results. traditional linear methods possess inherent limitations. Hence, in this study, we employed two deep learning methods, namely convolutional neural network (CNN) U-Net network, calibrate forecast Global Ensemble Forecast System (GEFS) predicting 10-m surface Southwest China with a lead time one seven days. Two decaying average method (DAM) unary regression (ULR), are conducted parallel comparison. Results show that original GEFS forecasts yield poorer performance western eastern Sichuan provinces, Yunnan province, within Guizhou province. All four provided correction effects study area, demonstrating best performance. After using U-Net, 1-day time, proportion U-component errors less than 0.5 m/s has increased by 46% compared GEFS. Similarly, V-component wind, 50% Furthermore, mean square error-based error decomposition further diagnose sources different prediction models reveal their calibration capabilities sources. The results indicate DAM ULR correcting Bias2, while all were variable distribution time. demonstrated sequence.