This paper introduces the relevant information about data center. The concept of data center is increasingly popular. This paper summarizes the domestic and foreign scholars research about the current situation of optimization and energy-saving about the air-conditioning system in data center and most scholars still come to stay in the study of the optimization of the equipment and control some equipment in order to achieve the purpose of energy optimization, but there is almost no research on predictive control. This topic applied predictive control to the entire air conditioning system. And it introduced the basic principles of predictive control, the policy-related knowledge of forecast model and data collection strategy. This paper introduces emphatically the BP neural network algorithm principle and application. By determining the amount of inputs and outputs and using BP neural network established the forecast model of the air conditioning system in data center. Eventually the model achieved the expected forecast results with MATLAB simulation and test.