Abstract: A method to simplify the model for forecasting the soil moisture of tobacco field by aiming at the same growing stage and area is proposed and a model to forecast the soil moisture of tobacco field based on BP neural network is established, in which the temperature, sunlight, rainfall, evaporation, and initial moisture of soil at a stage beginning are taken as the input and the soil moisture of tobacco field as the output. The research result shows that the prediction model of soil moisture has a good accuracy and has a wide range of adaptability.