In order to evaluate the wind speed forecast ability of Shanghai Meteorological Service-WRF ADAS Rapid Refresh System (SMS-WARR), the 10 m wind was divided into three levels to analyze the error characteristics and corresponding causes of formation， especially during the period of typhoon influence. The results show that the SMS-WARR model overestimates the wind weaker than Grade 6 and underestimates the wind stronger than or equal to Grade 6 in each lead time. The wind speed forecast score decreases for the wind weaker than Grade 6 and increases for the wind stronger than or equal to Grade 6 with the increase of lead time. The forecast and observed wind speeds are obviously linearly correlated between 5%-90% quartiles. During the period of typhoon influence, the area of strong wind (Grade 6) is much larger than the area where the strong wind was actually measured. The area of large deviation is mainly distributed in the area of forecast strong wind. These error characteristics may be related to the scheme setting of the cold and hot starting of the model and the model’s weak predictability to the wind field during the rapid strengthening phase of typhoons.