Calibrations of Ten-Meter Wind Speed Prediction over the Yunnan-Kweichow Plateau Based on the U-Net Neural Network.

In our study, we used a deep learning approach, namely U-Net, to enhance the forecasting accuracy of 10-meter wind speed in the Yunnan-Kweichow Plateau. We conducted the analysis using the GEFS for a forecast lead time ranging from 1 to 7 days. To assess the effectiveness of the U-Net model, we compared it with two conventional linear statistical techniques: the DAM and ULR. The findings revealed that the GEFS forecasts exhibited inadequate accuracy in predicting wind speed in the central region of the plateau. However, they performed well in the western and northern regions. All three methods demonstrated some level of calibration effect on the wind speed predictions in this region, but the U-Net model having the most effective calibration performance.