An accurate prediction of grain yield is an important part of national grain security,which is of much important significance for the government to establish a series of related grain policies.First,based on grey relative relational analysis,this paper calculates factor endowments which are corresponded to Hunan,i.e.grain planting area,village employees,output per hectare and effective irrigation area.Then,according to these factors,this paper carries out GM(1,1) residual modification forecasting;taking the predicted data as?relevant factors data and grain yield as characteristic variable data,it constructs a GM(1,N) prediction model on grain yield.According to the related data about grain production from 1995 to 2010,and prediction of grain output data in Hunan Province in 2015,through this model it calculates that the mean absolute percentage error is 0.77% by comparing the predicted data and the actual grain production data from 2008 to 2010.Therefore,this model has high accuracy.Finally,the paper offers policy suggestions which can ensure a rich harvest of Hunan grain production.