Blast Furnase Semlting Process Decision-making Model Research Of Pulverized Coal Injectioin Based On Data-driven

Blast furnace pulverized coal technology is achieving one of the important means of energy saving and emission reduction.On the economic perspective,pulverized coal injection smelting process insteads of coke for expensive,meanwhile coal injection technology also is one of the means on controling in the process of blast furnace smelting furnace temperature from the bottom of the blast furnace.Due to the amount of the pulverized coal injection in blast furnace smelting decision is depended on the experience of blast furnace long observation the detection index of the parameters in the process of blast furnace operation to determine the size of the coal injection volume of the current time.As a result,the blast furnace coal injection quantity decision of subjectivity.This topic is based on the real-time pulverized coal injection in blast furnace smelting process of decisions are taken as the research object,liuzhou iron and steel no.2 blast furnace as the research background,based on the analysis of mechanism of pulverized coal for blast furnace smelting effects of each part and decision influence factors of blast furnace coal injection quantity,based on the data analysis,screening of different variables as the input of the model characteristics,really for the blast furnace smelting condition was improved,the model makes it more close to the actual blast furnace running condition.Specific research work are as follows:(1)Considering industrial site sensor and the process mechanism of blast furnace smelting process and collect data which is in view of the blast furnace and in the smelting process due to dirty environment outlier data,this paper using boxplot method to eliminate abnormal data,using correlation analysis,determine the characteristics of sequence under different time delay affects the quantity of coal injection decision,set different correlation coefficient threshold value,depending on the coefficient of threshold selection variables as the input into the subsequent data modeling features.(2)Considering the characteristics of different correlation coefficient threshold selection variables of the modeling data,the influence of different variables under the threshold selection to support vector regression model,to verify the best threshold value of correlation coefficient modeling effect.At the same time,considering the characteristics of the data distribution of sensitive data modeling model features,this paper introduced the encoding neural network to improve the support vector regression model,the characteristics of the data before into the regression model to reconstruct the feature space,ensure that the data noise smoothing,the regression model for a large number of the existence of noise blast furnace data has certain robust performance.And in the fourth chapter simulation experiments verify the improved support vector regression model based on the coding the model accuracy due to the classical support vector regression model.(3)The model using clustering algorithm for clustering the data preprocessing to deal with the blast furnace smelting blast furnace in the process of smelting condition of more problems.The clustering algorithm for smelting state is similar to the sample data which is generated as a cluster,then each data of pulverized coal injection subset in the smelting process of modeling,modeling can make more targeted training model training data set,it can effectively improve the accuracy of the models and can get better training effect.The final data simulation proves the validity of this method and the precision of the model.