The end-point control technology of basic oxygen furnace (BOF) steelmaking is a very important production process in the later stage of smelting, and it has become a research hotspot in the smelting industry. In this paper, the carbon content and temperature prediction models were established based on a twin support vector regression (TSVR). In order to ensure the stability of the model, the input variables of the model were determined by using grey relational analysis (GRA) and partial correlation analysis (PCA). The results of the prediction models demonstrate that the hit rate of carbon content error bound within 0.005 wt% is 90%, and the hit rate of temperature error bound within 10 ° C is 94%, respectively. Moreover, a double hit rate reached 84%. By comparing with the back propagation (BP) neural network model, the hit rates of the proposed prediction model are higher than those of the BP model. Based on prediction models, a whale optimization control model was established for the calculations of auxiliary materials and oxygen blowing volume. The simulation results illustrate that the proposed control model has the mean absolute error of 9.3114 tons for scrap weight, mean absolute error of 2.5791 tons for lime weight, 0.7919 tons for dolomite weight and 537.8215 Nm 3 for oxygen blowing volume, which are better than the other two control models in the aspects of the calculation accuracy.