Wind energy is originated from the movement of the atmosphere with great randomness and intermittency. Wind speed forecasting is the basis of wind power forecasting in wind power plant, and its accuracy is of great significance. For complex terrain conditions, wind speed forecasting has been the difficulty and focus of research in all countries. In order to improve the accuracy of shortterm wind speed forecasting of wind power plant, this paper uses different boundary layer parameterization schemes to forecast wind speed. The wind speed of each single boundary layer parameterization scheme prediction and the corresponding measured wind data are used to establish the forecast model by random forest algorithm to study the shortterm wind speed in wind power plant. The experimental results show that with the integrated forecasting method of wind speed, the wind speed prediction error is significantly smaller compared to the single boundary layer parameterization scheme. Also, it has a good simulation effect for wind speed and wind direction and other meteorological elements in the study area and can effectively improve the accuracy of wind speed forecasting.