The determination of emergency planning zone requires calculating dose levels according to atmospheric dispersion models, Gaussian puff model is suitable for the calculation of emergency planning zone under the light wind and the calm wind conditions. Different weather conditions lead to different atmospheric dispersion conditions, under which the calculation results of emergency planning zone are different. To be conservative, annual meteorological conditions should be considered to calculate the size of emergency planning zone, which takes much time. This paper puts forward an optimization scheme based on the weather stratified sampling method for calculation of emergency planning zone, which extracts part of weather sequences from the annual weather sequences so as to reduce the calculation time and achieve the purpose of optimization. This paper qualitatively analyzes the effect of the stratified sampling method applied on the weather sequences needed by puff models and the result shows that the stratified sampling method in this paper can classify the annual weather sequences into different classes well, which satisfies the purpose of stratified sampling. Then, this paper selects RC01(Release category with intact containment), which represents the most frequent release category of source terms, and compares the size of emergency planning zone and weighted dose consequences in 16 directions at different distances obtained by the chosen weather sequences and annual weather sequences. Results show that the size of emergency planning zone obtained by the chosen weather sequences is consistent with the annual results and the relative deviation of dose consequences between the sampling results and the annual results is less than 7%. The weather stratified sampling method in this paper provides optimization ideas for the emergency planning zone determination.