In order to explore a simple and convenient method for obtaining specific water capacity，the relevant basic physical and chemical parameters of the soil，the soil water characteristic curve and specific water capacity are tested in the soil of the Loess Plateau in Shanxi Province． The parameters of the specific water capacity model are fitted and measured． Based on an analysis of the relationship between the basic physical and chemical parameters of each soil and the parameters of the water capacity model，a BP neural network prediction model is established on the soil texture，soil bulk density，soil organic matter content，and soil inorganic salt content． The results show that the BP neural network prediction model based on soil texture，soil bulk density，soil organic matter content and soil inorganic salt content are feasible，and the average relative error between the measured and predicted values of the water capacity model parameters is lower than 10%， prediction and accuracy are better． The research results provide theoretical and technical supports for obtaining specific water capacity，and promoting the development of soil transfer function theory．