Abstract: GM(1,1) is a common method of cost forecasting,but this traditional grey model cannot reflect the random volatility of the time sequence.The improved GM(1,1) can be obtained through accelerating translation transform and geometric mean transform.This improved model can decrease the random of the original sequence.In this paper this improved GM(1,1) is used to predict the coal cost of a coal mine.The error test shows that this method can be used in the coal cost prediction and provides a satisfactory tool for coal mine cost prediction.