According to the jumping relation between the basic form and the whitenization differential equation in NGM(1,1,k), the model is optimized. Firstly, the whitenization differential equation is integrated in the interval $[k-1,k]$ to obtain the new parameter estimation basic form of NGM(1,1,k). The front and before background value are defined and the precise calculation formulas are derived. The causes of error are also analyzed in the geometric relationship. Then the desirability function is built based on the sum of the error square, to determine the optimal constant value in the time response function. The optimized modeling steps of NGM(1,1,k) are summarized, and it is proved that the model has a non-homogeneous white exponential superposition. Finally, the optimization model and the classical model are contrasted by two examples, the good results obtained show the effectiveness and practicality of the proposed optimization model.