Abstract: Combined with the characteristics of the roadway monitoring point analysis, that Gauss white noise interference of wind speed data is not stationary process conditions, we proposed Kalman filtering data preprocessing. Combined with the characteristics of wind speed and wind speed sensor sequence range of measurement errors, we described Kalman filter algorithm to select the initial value of the filtering process and the parameters. Wind speed data through 1 000 consecutive inspection in a mine of Gansu Province were verified by Kalman filtering of wind speed data preprocessing methods. The results show that wind from the data pre-processing is better.