Grey prediction theory is an important part of grey system theory. As one of the important models of grey prediction theory, GM (1,1) model has been widely used in economy, management, energy and other fields. In order to improve the prediction accuracy of the classical GM (1,1) model, this paper proposes a combined optimization method, that is, the difference equation is used to replace the static equation in the classical model, and the variable weight is used to construct the background value to reduce the system error caused by human intervention. Taking the domestic soybean annual consumption data as an example, the validity of the combined model is verified. The results show that the prediction accuracy of the combined optimization model is significantly better than that of the classical GM (1,1) model.