Based on the grey-forecasting model with unfixed parameter, the primary model was ameliorated through optimized analysis on the non-equidistant sequence to improve its fining and forecasting accuracy especially when there are only few experimental data.