融合高炉煤气流机理分析和数据深度挖掘的炉况诊断和预测研究

Accurate diagnosis and prediction of the blast furnace conditions is the premise to realize stability and effective operation for blast furnace. It is also the basic guarantee of realizing blast furnace energy saving, emission reduction and improving enterprise benefit efficiency. However, the characteristics of sealing and large time delay of the smelting process of the nonlinear blast furnace have brought great challenges to the accurate diagnosis and prediction on the furnace conditions. This project intends to make full use of the information of the gas flow which has the advantages of containing rich information and good timeliness. Combine the mechanism analysis method and the data depth mining technology to realize the accurate diagnosis and prediction on blast furnace conditions in closed environment with large time delays. Firstly, establish the relationships between the gas flow indicators and 7 kinds of furnace condition by analyzing the mechanism property and deeply mining the association rules of the gas flow. Secondly, choose the indicators influencing the gas flow based on the mechanism, compute the time delays and determine the characteristic variables for prediction mode with different gas flows. Further, the small sample problems of gas flow under abnormal blast furnace condition are studied, and the construction of the prediction model of the gas flow indicators fusing the time varying characteristics of the blast furnace system is also addressed. Finally, the blast furnace condition analysis diagnosis and forecast system which can be applied to No. 2 blast furnace in Liugang for verification is developed. The completion of the project will greatly enrich and develop the complex industrial processes modeling theory. Moreover, the research results will provide theory and technical support for the accurate diagnosis and prediction of blast furnace conditions.

高炉炉况的准确诊断和预测是实现高炉稳定顺行的前提，也是实现高炉节能减排，提高企业效益的基本保障。而高炉冶炼过程的封闭性和大时滞非线性特点，给炉况的准确诊断和预测带来巨大挑战。项目拟充分利用携带信息量丰富和时效性好的煤气流信息，融合机理分析方法和数据深度挖掘技术，实现对封闭环境下大时滞非线性高炉系统的炉况诊断和预测。首先，分析煤气流的机理特性，深度挖掘煤气流数据所蕴含的关联规则，确定煤气流指标与7种炉况的关系。其次，基于机理选择影响煤气流各指标的因素，计算其滞后时间，进而确定不同煤气流指标预测模型的特征变量。进一步，研究异常炉况下煤气流的小样本问题，建立融合高炉系统时变特点的煤气流指标数据驱动预测模型。最后，开发炉况诊断和预测系统，并应用于柳钢2号高炉进行测试验证。项目的完成将丰富和发展复杂工业过程的建模理论，研究成果有望为高炉炉况的准确诊断和预测提供理论支撑和技术支持。

高炉炉况的准确诊断和预测是实现高炉稳定顺行的前提，也是实现高炉节能减排，提高企业效益的基本保障。而高炉冶炼过程的封闭性和大时滞非线性特点，给炉况的准确诊断和预测带来巨大挑战。本项目充分利用具有携带信息量丰富和时效性好的煤气流信息，融合机理分析方法和数据深度挖掘技术，针对封闭环境下大时滞非线性高炉系统的炉况诊断和预测这一关键科学问题进行深入研究。首先，全面分析煤气流的机理特性，深度挖掘煤气流数据所蕴含的关联规则，进而以机理为准则确定煤气流指标与炉况的关系。其次，从机理角度选择影响煤气流各指标的因素，从数据统计角度计算各指标影响因素的滞后时间，根据煤气流机理的耦合特点构建合适的特征选择方法，进而确定不同煤气流指标预测模型的特征变量。进一步，研究异常炉况下煤气流的小样本问题，以及融合高炉系统时变特点的煤气流指标预测建模问题。最后，开发炉况诊断和预测系统，并应用于柳钢2号高炉进行测试验证。我们从机理角度定性地分析煤气流指标与炉况的关系，并通过数据深度挖掘方法获得煤气流指标与炉况的关联规则，在此基础上基于机理分析结果对数据深度挖掘的关联规则进行评价筛选，从而确定煤气流指标与炉况的确切关系。我们从机理角度选择影响煤气流各指标的影响因素，并利用最大信息系数法确定各个影响因素的滞后时间，在此基础上利用群正则化法确定不同煤气流指标预测模型的特征变量，从而为煤气流指标建模确定最优输入变量。我们利用特征加噪的方法对异常炉况下煤气流小样本进行扩展，构建融合高炉时变特点的Direchlet过程无限混合模型，并引入隐含变量利用Bayes方法推理求解。最后，我们开发了炉况诊断和预测系统，并应用于柳钢2号高炉进行了验证和改进。本项目的完成将丰富和发展复杂工业过程的建模理论，研究成果有望为高炉炉况的准确诊断和预测提供理论支撑和技术支持。这对缓解我国能源资源紧张与减排形势严峻的局面具有重要的意义。