Based on the ECMWF reanalysis data and AWS data from 2000 to 2016, the intensity of the weather process and the corresponding physical field, and quantitative transformation of thunderstorm cases with different weather phenomena such as gale and precipitation in Yinchuan Hedong International Airport, the normality tests were conducted on the quantitatively transformed data by using the coefficients of Kurtosis and Skewness, Chi-square and Q-Q plot. The results showed that the samples classified by weather phenomena passed normality, and the unclassified samples basically obeyed normality distribution. Thunderstorms phenomenon and intensity prediction models were developed in five ways of stepwise regression, multiple regression, nonlinear regression, BP artificial neural network and Support Vector Machine(SVM). The forecast results showed that the BP artificial neural network and SVM had strong forecasting ability for weather phenomena, and the classified stepwise, multivariate and nonlinear regression models showed better effects on the intensity of weak weather, heavy rain, simultaneous occurrence with gale and rainfall forecasting. Based on the analysis and research above, an operational platform system combining qualitative and quantitative prediction of different types of thunderstorms in Yinchuan Hedong International Airport was established.