Abstract: Crop growth models can be used to predict crop growth and development at different scales, particularly from the station to the regional level. Daily weather data is often required to drive crop growth. The stochastic weather simulator (SWG) can fully meet this demand at present. The research team has constructed a random weather generator, according to the characteristics of China's climate (weather generator using dry and wet spells, WGDWS), in order to replicate the widely-used WGEN series weather generator (daily weather stochastic simulator, daily weather stochastic simulator DWSS). This study aims to evaluate the meteorological data generated by WGDWS and DWSS as the inputs for the wheat growth models. The 57a of daily meteorological data was measured at 8 stations in the northern winter wheat region of China. Then, 300a of daily meteorological data was generated using WGDWS and DWSS generators. The quality of the meteorological element statistical data was also evaluated after generation. Except for the Taiyuan station in Shanxi, the monthly average highest and lowest temperatures were achieved in excellent consistency with the measured values. The monthly average of the total solar radiation showed better consistency, with only the relatively small deviations from the measured values. There was no significant difference between the monthly mean values of the meteorological elements and the measured ones. The overestimation and underestimation of the monthly average rainfall were also found during this time. A crop model wheat intelligent decision-making system was used to evaluate the impact of the measured and generated data on the wheat growth simulation under two treatments (potential production and experience water and fertilizer). The results showed that there was a slight overestimation in the mean wheat yield and biomass that was simulated by WGDWS and DWSS. The better performance of the simulation was achieved, with the determination coefficients of 0.94 and 0.99, respectively, compared with the measured values. In terms of the standard deviation to simulate the yield and biomass, the WGDWS was 13% and 44% lower than the measured weather by more than 20%. In DWSS, these values were 6% and 25%, respectively. Once the standard deviation was more than 10% lower than the actual measurement, WGDWS accounted for 50% and 63%, respectively, while DWSS accounted for 31% and 56%, respectively. Overall, the better simulation was observed on the biomass than on the yield. The interannual variation of the simulated biomass was also smaller than that of the simulated yield. There was a highly consistent simulation of the WGDWS and DWSS for the accumulated temperature, evapotranspiration, and dry matter accumulation. The simulation differences for the accumulated evapotranspiration were slightly greater than those for the accumulated temperature and dry matter. There was no significant difference in the yield and biomass from the measured values. Therefore, the weather data was used instead of the measured values for the wheat growth simulation. The average relative error (MRE) and root mean square error (RMSE) of the physiological indicators simulated by WGDWS were smaller than those of the DWSS, except for the dry matter weight. Compared with the DWSS, the differences in yield and biomass simulated by WGDWS are almost below 10%, with the errors of 10%, accounting for 94% and 100%, respectively. The WGDWS presented a 6% and 25% higher simulated yield and biomass with an error of less than 5%. The better simulation was achieved in the yield, biomass, and phenological period using WGDWS and DWSS. Results showed that the absolute error and standard deviation of the WGDWS were 63 and 57 sets, respectively, with the WGDWS superior to DWSS. There were 14 indicators with significant differences between the two types of generators. Among them, only 1 indicator was DWSS better than that of the WGDWS, indicating that the better simulation of the WGDWS than the DWSS. Therefore, the meteorological data generated by WGDWS was fully used as the input for the crop growth models. There were relatively low simulation errors of the indicators, such as the yield and biomass.