Abstract: In some Yellow River irrigation areas, most of irrigation water returns to its source as return water and the study of return water has a significant meaning to reuse it. The paper determines the major factors on the return water using correlation analysis method and discusses the quantitative analysis method of return water with the neural network model based on Levenberg-Marquardt algorithm. Through studying examples, the model can accurately simulate and predict the return water volume and has the application value in the study of return water volume of irrigation area.