Abstract: There is a relationship between the variations of PM_2.5 concentration and the happen of disaster weather as the haze,the accurate prediction of the changed trend of PM_2.5 concentration plays an important guiding role in the effective prevention and control of the wide range of disaster weather as the haze.Considering the gray characteristics of PM_2.5 concentration is disturbed by the various factors,the article adopts the way of the same dimension gray recurrence and supplies the new gray information in time,establishes the gray relational computing model of the main factors of PM_2.5.At the same time,the GM(1,N) is improved by introducing the gray control parameters,which can meet accurate prediction on the condition of multi-effects factors.The above approaches are combined,the adaptation related-combinatorial mode GCM(1,N) is established and can be applied for different metadata N.By the analysis of the measured data obtained in Beijing,the precision of the calculation and prediction model GCM(1,N) has reached to 89.75%~96.44%,the relative error of the prediction of PM_2.5 concentration is between 7.32% and 15.21%,the results are satisfactory.