The wind forecasting results of the mesoscale meteorological model are closely related to the adaptability of the parameterization schemes. Two pilot wind farms with typical underlying surface are selected as simulation regions, which are the undulating hilly terrain of Inner Mongolia and the flat land sea border terrain of Jiangsu respectively. Six kinds of physical process parameterizations of WRFV32 (Weather Research and Forecast Model) are adopted in forecasting the wind speed and direction of the two wind farms in January 2010 and July 2010, and the impacts of parameterizations difference on wind forecasting are analyzed. The results show that: (1) In the experimental area of Inner Mongolia,the MRF boundary Layer parameterization scheme is more reasonable than MYJ does in describing the structure of the boundary layer; the microphysical scheme WSM3 forecasts the wind speed better than WSM6 in summer and vice versa in winter; the land surface process scheme increases the influence on the boundary layer structure of complicated terrain, especially after summer rainfall, and the Noah scheme is better than that without any land surface process. (2) In the experimental area of Jiangsu, the MRF boundary Layer parameterization schemes describes the structure of the boundary layer more reasonable than MYJ; the land surface process RUC is superior to the heat diffusion and Noah schemes in January; the heat diffusion scheme is superior to the RUC and Noah in July. (3) The forecasting wind direction statistics of the six parameterization combination schemes and the actual recorded wind direction statistics are in good agreement, and the wind direction probability distribution is similar, and the prevailing wind direction is consistent and stable.