Abstract: The 15th National Games were staged as a joint endeavor by Guangdong, Hong Kong, and Macao. As one of the megacities in the Greater Bay Area, Shenzhen held some important games and the closing ceremony. To provide better meteorological forecast services, some numerical postprocessing techniques were developed to form Shenzhen Fast-updated and Extended Elements Digital (SZ—FEED) forecast system, including nowcasting forecasts of minute precipitation and temperature, three-dimensional wind, and short-term forecast of probabilistic quantitative precipitation. During the nowcasting period, minute-scale precipitation was developed by optimal merging of the traditional and multiple artificial intelligence (AI) algorithms. The optimal merged quantitative precipitation is evaluated better than single method by 2.7%-8.5%. The rapid updated temperature forecasts were improved by blending the numerical forecast variation and the objective observational grid to alleviate the underscore of daily high temperature and reduce the topographic effects. Based on the South China regional model and boundary layer wind profile model, the three-dimensional wind forecasts are constructed at some critical locations by calibrated using the 356-m-high Shenzhen Meteorological Gradient Tower. In the short range, the probabilistic precipitation forecast was improved by taking the advantages of the traditional numerical models and AIbased Pangu-Weather according to the optimized Bayesian model averaging (BMA). The above-mentioned technologies gave good supports for the meteorological service of the 15th National Games of China.