Aiming at the problem of flood volume forecast in semi-arid area, we selected 15 flood data of the Caoping hydrological station in Chabagou watershed from 1980 to 2010. According to the measured single storm, flood control data in Chabagou watershed, considering the warp land dam area, storm rainfall, rainfall center, antecedent rainfall and others factors, combined with SPSS and MATLAB software, multivariate linear regression model and BP neural network model for the prediction of flood volume were built. The prediction results of the two models show that multivariate linear regression model and BP neural network model can be better applied to flood volume prediction, and the BP neural network model is superior to multivariate linear regression model. Research result can provide decision-making basis for dam safety in the flood season.