The most popular way to account for the uncertainty of ether forecasting is through the use of an ensemble prediction system (EPS). Ensemble predictions generally require some form of statistical post-processing to remove biases and provide reliable probabilistic forecasts. In this study, the EMOS technique is used to calibrate the raw outputs of an ensemble for wind speed at 12 meteorological stations over Tehran. A log normal distribution is fitted that the mean of the Gaussian predictive PDF is a bias-corrected weighted average of the ensemble member forecasts; and the variance is a linear function of the ensemble spread. The coefficients of the first two moments are estimated by minimizing the continuous ranked probability score (CRPS) value averaged over a training period. In this research, for minimization of the CRPS function, we use an iterative algorithm named as Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm. When compared to the mean of deterministic ensemble members, the root mean squared error and mean absolute error of the bias-corrected ensemble is decreased considerably.