Semi-variant function as an important mathematical model of Kriging spatial analysis can effectively describe the features of the variants(such as ore grade,thickness of ore body) in some districts of ore deposit.Based on practical variograms,the parameters of variant functions can be estimated more precisely and more automatically by genetic algorithms than by traditional linear regression method.In this paper how to use genetic algorithms to simulate the variant functions for Kriging spatial analysis is introduced,and also for exemplification,this method is applied to analyze the thickness distribution of coal beds in some regions.