Large warehouse is the main type of grain granary for state grain depot. In the summer, due to the continuous heat outside, grain pile will reach a higher temperature, moreover, the growth of microorganisms will lead to further internal heat for grain heap, which would harm the security of stored grain. In order to safeguard the quality of the stored grain and control the temperature, it is important to research and apply granary temperature field prediction system. BP neural network forecasting model was studied based on neural network model. The actual monitoring data of grain in warehouse were selected to emulate on MATLAB platform, and construct models. The factors which affected the grain temperature field were analyzed, and the weights of the factors were determined by SPSS statistical software, the results of the principal component analysis were verified by neural network method.