Abstract:

Blast furnace is a highly complex production process with the characteristics of nonlinearity, large delay, big noise, parameters distribution, and so on. According to the problem that coke ratio is used as energy consumption evaluation index but can't provide real-time guidance in the blast furnace field, carbon monoxide utilization ratio is studied as the energy consumption index and a prediction method for it based on improved support vector regression(SVR) is proposed. Firstly, relevant operation factors are selected by analyzing the mechanism of blast furnace combined with mutual information method. Next, wavelet transform is used to remove the noise including in the production data. Additionally, relative gray correlation analysis is applied for temporal registration to avoid the time delay of blast furnace operation, and then the prediction model of carbon monoxide utilization ratio is established. Support vector regression method and adaptive particle swarm optimization algorithm(APSO) are united to overcome the randomness of the selection of parameters and improve the accuracy of the prediction model. The simulation results demonstrate that APSO-SVR provides an effective way to predict the carbon monoxide utilization, which serves as a scientific decision support for the following optimization of blast furnace operation as well as energy saving and emission reduction.

Key words: blast furnace, carbon monoxide, model, support vector regression, prediction, adaptive particle swarm optimization

摘要：

高炉冶炼是一个具有非线性、大时滞、大噪声、分布参数等特征的高度复杂生产过程。针对目前高炉现场以焦比为能耗评价指标却无法提供实时指导的问题, 研究以一氧化碳利用率为能耗评价指标, 提出一种基于改进支持向量机的高炉一氧化碳利用率预测方法。首先分析高炉炼铁过程机理, 结合互信息法得出影响一氧化碳利用率的相关操作因素。然后鉴于生产数据含噪高的特点, 采用小波去噪方法去除数据噪声干扰, 并且利用灰色相对关联度分析方法对操作参数进行时序配准, 消除时滞影响, 建立高炉一氧化碳利用率预测模型。在建模过程中, 将自适应粒子群与支持向量机回归方法相结合, 以克服模型参数选择的随机性, 提高了模型预测精度。现场实际数据的预测结果表明所提出方法的有效性, 能够实时精确地预测高炉一氧化碳利用率, 为后续高炉的优化操作和节能减排提供了及时有效的决策支持。

关键词: 高炉, 一氧化碳, 建模, 支持向量机, 预测, 自适应粒子群