本报告依托首钢京唐炼钢厂，通过将机理模型和人工智能算法模型结合分别建立了铁水预处理工序、转炉工序、RH精炼工序、LF精炼工序、连铸工序、连铸工序和工序界面调度的智能管控模型，其中，铁水预处理工序中脱硫剂加入量预测误差在±0.3t范围内的命中率为95.59%，终点硫含量预测误差在±10℃范围内的命中率为99%；转炉工序中的终点钢水温度预测误差在±10℃的范围内的命中率为75.71%；RH精炼工序中的终点钢水温度预测误差在±10℃的范围内达到78.29%；LF精炼工序中终点钢水温度预测模型在±10℃的范围内达到94%；钢包调度模型能够缩短钢包周转时间664min。

The intelligent control modesl of hot metal pretreatment process, converter process, RH refining process, LF refining process, continuous casting process, continuous casting process and process interface scheduling were established by combining mechanism model and artificial intelligence algorithm model in this paper which relies on Shougang Jingtang Steelmaking Plant. Among then, the hit rate of the desulfurizer addition amount prediction error in the hot metal pretreatment process is 95.59% in the range of ±0.3t, and the hit rate of prediction error of the end sulfur content is 99% in the range of ±10℃, and the hit rate of prediction error of the end-point molten steel temperature in the converter process is 78.29% in the range of ±10℃, the hit rate of prediction error of the end-point molten steel temperature in the LF refining is 94% in the range of ±10℃ and the ladle scheduling model can reduce the turnround time of the ladle by 664min.