本文讨论自校正预报器与知识库系统配合使用时,对高炉铁水含硅量的在线预报问题.自校正预报器按Box和Jenkins的原理构成,预报模型参数用递推近似极大似然法进行在线估计.当炉况稳定时,自校正预报器的精度是满意的,而炉况不稳定时,由知识库系统的输出对自校正预报进行检验和修正.实验表明,综合预报系统的预报精度超过熟练工长.因此,上述系统可作为工长的操作指导. This paper discusses the problem of on-line prediction of the silicon content of hot metal in blast furnace when used together with a knowledge base system.The self-correcting predictor is constructed according to the principle of Box and Jenkins, and the parameters of the prediction model are recursively approximated by maximum likelihood On-line estimation.When the furnace condition is stable, the accuracy of the self-tuning predictor is satisfied, and when the furnace condition is unstable, the self-tuning forecasting is verified and corrected by the output of the knowledge base system.Experiments show that the accuracy of the forecasting system More than skilled workers. Therefore, the above system can be used as a superintendent’s guidance.