Study on the Performance of the STMAS-WRF Model in Short-term Forecasting of Wind Field in Hangzhou Bay, Zhejiang Province

To evaluate the forecasting performance of the wind speed prediction product based on the STMAS-WRF model along the coastal region of Hangzhou Bay, this study utilizes wind speed observation data at $\mathbf{1 0 0}$-meter turbine heights from the Jiaxing and Shengsi wind farms between February and November 2024. The results indicate that: (1) The STMAS-WRF wind speed forecast product performs well for the Jiaxing and Shengsi wind farms, with forecast results closely aligned with wind speed trends. The correlation coefficients for Days 1 to 3 are $0.59,0.55$, and 0.52, respectively, while the root mean square errors (RMSEs) are $2.94 \mathrm{~m} / \mathrm{s}, 3.13 \mathrm{~m} / \mathrm{s}$, and $3.21 \mathrm{~m} / \mathrm{s}$. As the forecast lead time increases, the correlation decreases, and the RMSE increases. (2) Wind speed forecast errors exhibit a diurnal variation, with the smallest errors occurring in the early morning and the largest at midday. Overall, the forecasted wind speed tends to be higher than observed values. (3) Regarding wind speed ranges, the forecasts for Days $\mathbf{0}-\mathbf{3}$ effectively capture wind speed variations in the $3-12 \mathrm{~m} / \mathrm{s}$ range, but the accuracy for predicting high winds (above $12 \mathrm{~m} / \mathrm{s}$) is relatively lower. In general, the STMAS-WRF model demonstrates good forecasting performance, with error patterns showing regularity. This provides a basis for further calibration of the product to reduce error levels.