In this paper, based on the non‐equidistance GM(1,1)Model's mechanism, we obtain a conclusionthat the background value is very important to the accuracy of the gray theoretical modeling. Based onnumerical integration principle, Simpson formula with changeable steps was deduced to optimizing thebackground value of GM(1,1)model. The example shows that the prediction accuracy of GM(1,1)model issignification improved by the using the optimized background formula.