In order to reduce the error of single-satellite localization from frequency measurements, an algorithm of single-satellite localization from frequency measurements, which combines the initial value calibration and the second order approximation, was proposed. The algorithm obtains distance information from the satellite to the localization target at the time point when the Doppler frequency shift of the satellite receiving signal is zero, and then corrects the geometric relationship between the orbit plane and the localization target by using the distance information, thus it provides a calibration method for the selection of the initial value. By substituting the calibrated initial value into Taylor′s expansion of the localization equation containing the second order numbers, the real position of the localization target was obtained through fewer number of iteration, thus reducing the complexity of localization algorithm and improving the localization accuracy. The simulation results indicate that compared with the Doppler single-satellite localization algorithm, the iteration times and the localization error of the proposed algorithm are reduced substantially. The algorithm is easy to implement, low in computation and small in error, and has high theoretical and practical value in the field of single-satellite localization research.