Abstract: Alloy yield dynamic libraries were established based on self-learning of historical data and computer technology,through which accurate real-time alloy yield was obtained. Alloy charging optimization model was established for the First Steel-Making Plant of Tianjin Pipe Co. Ltd using the two-stage simplex method. As the operation of steelmaking is online,the alloy yields of different steel grades are obtained,and the accuracy of dosage of the alloy is improved. By optimizing alloy charging,average cost of alloy charging for each steel grade has been reduced. The maximum of the cost reduced is 54.96 yuan·t^-1,and the minimum is 8.57 yuan·t^-1.The cost of alloy charging per ton steel was reduced by 6.76%-11.40%,and the average cost was reduced by 9.74%.