Abstract：This paper has introduced the technology of artificial network into the modeling of underground water content forecast. Hetao plain and Huabei plain were chosen as the testing ground with a number of known local agro-wells as the verification sites. Induced Polarization (IP) and resistivity sounding and other surface geophysical methods were used to construct the artificial neural network (ANN) model based on such parameters as apparent resistivity, polarization rate, half-life, decay rate and rate of deviation in the relevance. Then, the comprehensive parameter was added to improve the inputting neurons. Finally, the quantitative prediction model of the water content was establiehed. The results of mean-variance test show that this technique has a good effect in the plain area. The study has provided a new concept and a new idea for the forecasting work in hydrogeological exploration.