Aiming at crane tasks’ characteristics of sequentiality, confliction and uncertainties in steel plants, a crane scheduling method based on Bayesian theory is proposed. Firstly, the rule of generating crane tasks in steel plants is analyzed, and the Bayesian network for crane scheduling process is constructed. Then, the corresponding time-space probability distribution of crane tasks is calculated. Finally, the number of available cranes are assigned to crane tasks, and the crane scheduling plan is generated dynamically. Based on the crane scheduling simulation system, 10900 pieces of actual crane task data are applied. The testing result indicates that the crane scheduling method based on Bayesian theory can effectively complete all crane tasks within the scheduled time. Compared with manual scheduling methods, the total time to complete the task, the frequency of crane avoidance and the additional transportation time caused by crane avoidance are all dramatically reduced. While improving the operation efficiency of cranes, the security risk from crane collision is reduced.