【作者】 李彦瑞； 杨春节；

【Author】 Li Yanrui;Yang Chunjie;State Key Laboratory of Industrial Control Technology, College of Control Science and Engineering, Zhejiang University;

【机构】 浙江大学控制科学与工程学院工业控制技术国家重点实验室；

【摘要】 钢铁制造业是我国国民经济重要支柱产业,其中炼铁过程是耗能最大、排放最多、成本最高的环节,是实现"碳达峰"、"碳中和"战略目标的主战场。本文针对炼铁过程中的关键问题——高炉铁水硅含量预测问题,考虑其过程动态性、非线性与时变性的特点,设计了基于可视图的特征提取方法对其进行预测。该方法首先构建动态的可视图,进而提取图的聚集系数、子图模式等特征。提取得到的特征可以有效反应过程的动态特点,提高梯度提升树模型的建模精度。该模型在实际高炉生产数据上进行了验证,结果优于传统方法和循环神经网络。更多还原

【Abstract】 Iron and steel manufacturing industry is an important to China’s national economy. Iron making process has the largest energy consumption, the largest emissions and the highest cost. It is the main battlefield to achieve the strategic goals of "carbon peak" and "carbon neutralization". Considering the dynamic, nonlinear and time-varying characteristics of hot metal silicon content in blast furnace, a method based on viewable feature extraction is designed to predict the silicon content in hot metal. This method first constructs a dynamic visibility graph, and then extracts the aggregation coefficient, subgraph pattern and other features of the graph. The extracted features can effectively reflect the dynamic characteristics of the process and improve the modeling accuracy of the gradient lifting tree model. The model is verified on a practical blast furnace production data, and the results are better than the traditional method and the recurrent neural network.更多还原