高炉生产过程的智能预测建模

Intelligent predictive modeling of blast furnace system

针对高炉生产过程的复杂性、非线性以及强耦合、多变量、难测量等特点,提出铁水硅含量的智能复合多变量预测模型对高炉生产过程进行建模。整个系统分为两部分:首先离线建立不同工况下铁水硅含量的多变量预测模型;然后运用模糊逻辑推理建立各模型输出、实际输出与模型权重之间的对应关系,进行多模型智能融合,生成复合模型,并对其进行在线调整以优化预测过程。研究结果表明:采用此方法计算周期短,对被控对象的变化有较强的鲁棒性;该系统预测误差小,能够快速适应工况的变化,实用性好。

In order to predict the production process of complicated nonlinear blast furnace（BF）,the intelligent mixed multi-variable predictive models are proposed.The prediction strategy including two parts is used： the models reflecting different working conditions are established by adopting the support vector machine（SVM） in offline part.Then fuzzy reasoning is employed to train and derive the weight of every model,multiplying those weights by the corresponding models gives the actual prediction output.And the online correction is called to adjust those weights to optimize the prediction system.The intelligent mixed prediction model has been successfully applied and validated in several real-life iron-making scenarios and clearly demonstrates its effectiveness for silicon content prediction in the BF with the less time consumption and the robustness to the change of working conditions.