A prediction method of silicon content in hot metal of blast furnace

Abstract

In blast furnace smelting, the silicon content in hot metal can indirectly reflect the blast furnace temperature and measure the quality of hot metal. For more accurate prediction, according to the reduction reaction, the input parameters affecting the silicon content are selected to form a data set. The Weighted Random Forest (WRF) prediction model and the Scaling Coefficient Particle Swarm Optimization (SCPSO) algorithm are proposed. The prediction method based on SCPSO-WRF has higher prediction hit rate and lower mean error than those traditional methods. The results show that the prediction hit rate and mean error of SCPSO-WRF are 89,1 % and 0,0291 respectively. The prediction method has theoretical research and practical application value.

Keywords

• blast furnace
• hot metal
• silicon
• particle swarm optimization
• random forest