Based on the inversion model of wind measurement for meteorological rockets, the method for evaluating the uncertainty of wind inversion results is studied by the error theory and curve fitting least square principle. According to the motion law of the rocket sonde in the air, a mathematical model is established and the inversion function of wind speed and wind direction is derived. According to the error theory, the general expressions of the uncertainty of wind speed and wind direction are given. Based on the principle of polynomial least square fitting, the formulas of the systematic error and random error of wind speed and direction are derived, and the coefficients in the formulas are obtained. Taking the measured data as an example, the above formulas are used to analyze and calculate the atmospheric wind and the uncertainty. The results show that the uncertainty of wind speed decreases with the decrease of height. The uncertainty is 3.5 to 2.8 m/s in the 60 to 50 km, and less than 1 m/s below 50 km. When the wind direction is around the north (0°), the inversion uncertainty of wind direction will increase abnormally, and the inversion uncertainty is basically within 10° in the other cases.