Comparison analysis of downscaling forecast methods on wind speed in wind power field

The small-scale CALMET model and bilinear interpolation (BLI) method were used to downscale the output of mesoscale numerical weather prediction from the WRF model using both 9 km and 3 km resolutions. The simulated and observed wind speeds were compared. The results show that forecast accuracy of wind speed after downscaling with the CALMET model is higher with 9 km resolution than that with 3 km resolution. At the same time, the forecast effect of wind speed is also good using the BLI method because of the spatial resolution of mesoscale numerical weather prediction model is high. Wind speed is divided into three levels, namely, 0 m·s^-1≤wind speed<5 m·s^-1, 5 m·s^-1≤wind speed<10 m·s^-1 and wind speed≥10 m·s^-1, and the percentage of prediction errors of three levels are tested. It shows that forecast results for wind speed ≥10 m·s^-1 are poor no matter whether the CALMET model or the BLI method. It is therefore important to correct strong wind forecast for improvement of wind speed forecast accuracy.