According to the fact that the accuracy of the model has a close relationship with the background value, this paper integral the both sides of whitenization differential equation at the same time that based on the 1-AGO form of approximate non-homogeneous exponential sequence. After that the grey differential equation is obtained so that the background value has been constructed more reasonable and the grey differential equation, parameter α can be solved. Then let the prediction function as an exponential form, we can find the parsing expressions of parameter α and β through the target function as the minimum of the sum of the square of the relative error. Combine with the above two steps, a new optimization model is offered. Finally, experiments indicate the effect of optimization in this method; they also illustrate the correctness of the conclusion.