Abstract: The mining subsidence dynamic prediction describes the relationship between the surface movement deformation and the mining time, and has important guiding significance in controlling the surface deformation and failure of the gob. The time function is the theoretical core of dynamic prediction. The dynamic prediction model of mining subsidence based on it provides real-time and accurate theoretical data basis for the protection of surface buildings. As a new type of time function, the normal distribution time function is complete in time and space, but it has theoretical defects in dynamic prediction. By comparing the difference between the characteristics of the normal distribution time function and the ideal time function, the defect was analyzed by the decrease of the effective integral domain loss of the density function of the time function with the decrease of the time parameter c. In order to expand its application range and improve its prediction accuracy, the whole deviation correction method was used to correct the function to the theoretical position, eliminating the theoretical deviation, and then using the growth function model to optimize the normal distribution time function. The results show that the optimized normal distribution time function corrects the theoretical error of the function value at the dynamic prediction key node, broadens the selection range of the expected parameters, which can be applied to different geological mining conditions and solve the expected error of the original time function end point. The problem of increasing the parameter is increased. The comparison with the measured data of the two working faces in Datong mining area shows that the optimized normal distribution time function is better than the original normal distribution time function in the surface prediction dynamic accuracy, which is more in line with the actual situation and can be predicted with high precision. The movement of the surface of the mining area and the protection of important buildings provide a more reliable data basis.