Abstract:For Penman Monteith formula application limitations, the use of public weather forecast measurable factors and historical meteorological data calculate ET0 as a benchmark for Guangzhou station 2017-01-01—2019-03-31 wind condition after a quantitative forecast of meteorological information in 2017, and 2018 meteorological forecast information as input factors, ET0 as output factors, respectively based on regression support vector machines(SVR)forecasting model and BP neural network prediction model, select has a better performance forecast model to forecast of 2019 years of ET0 and compared with calculated value analysis. Results show that the regression support vector machines reference crop evapotranspiration forecasting model test set mean square error is 0.206, deterministic coefficient is 0.896, the amount of BP neural network reference crop evapotranspiration forecasting model test set mean square error is 0.305, deterministic coefficient is 0.851, the SVR amount of refe-rence crop evapotranspiration forecasting model is obviously better than the mean square error and the correlation factor to the BP neural network; The correlation coefficient between the predicted value based on the SVR model and the calculated value based on the PM formula is 0.761, showing no significant difference, but showing significant correlation and overall coincidence, which can provide relatively accurate ET0 forecast data for irrigation forecasting and decision-making.