Growing pressure on the world’s water resources is having major impacts on us. In this paper, we discuss on water resources carrying capacity. We have a case study of Shandong Province which is one of the most arid regions in China. Considering the dynamics of water supply and demand, we combine the Falkenmark indicator and the binary dynamics model to establish an evaluation model of regional water resources carrying capacity. According to the result of our model, Shandong Province is heavily exploited. The per capita water resources in Shandong province were less than 300 m³ in the past ten years. The increasing destruction and increasing waste make the situation even worse. Then ARIMA model and BP neural network is combined to propose a prediction model. We use it to predict the supply and demand of water resources in Shandong Province in the next 15 years