Prediction Modeling Of Blast Furnace Molten Iron Quality Parameters Based On Ensemble Learning

The steel industry is an important pillar of the national economy,and blast furnace ironmaking is an important part of steel production.Blast furnace ironmaking is a complex production process that transforms solid iron ore into liquid molten iron through complex and changeable physical and chemical reactions in a closed environment with high temperature,high pressure,and multiple phases and multiple fields.The quality of the output molten iron directly affects the quality,condition and price of the subsequent steel.In order to achieve the goals of high yield,low consumption,long life,and stable operation of blast furnace ironmaking,the quality parameters of molten iron—Molten iron temperature(MIT),silicon content([Si]),phosphorus content([P]),and sulfur—are required.([S])).However,the physical and chemical reactions inside the blast furnace are complicated,and they are in a state of coupling of high temperature,high pressure,and multiple phases and multiple fields.At the same time,due to the limitations of the existing detection technology,it is difficult to directly detect the quality parameters of blast furnace molten iron online,and offline testing is required.1 to 2 hours,there is a long lag.Therefore,it is very important to establish an accurate,stable and reliable parameter model of molten iron quality.At present,the modeling methods of blast furnace molten iron quality include mechanism model,reasoning model and data-driven model.The mechanism model requires strict assumptions.The application of the inference model is limited when the model is complex,while the data-driven model does not require an in-depth understanding of the complex and variable machinery inside the blast furnace.Only the data can be used to build a blast furnace hot metal quality model.,Become a hot spot in the research and application of blast furnace hot metal quality.In order to establish an accurate,stable,and reliable model of molten iron quality,this paper relies on the National Natural Science Foundation of China’s major project"Control method and implementation technology of high-performance operation of large blast furnaces"(project number:61290323).The object,based on data-driven intelligent algorithms,carries out research on the prediction modeling of blast furnace molten iron quality parameters based on ensemble learning.The specific work is as follows:(1)Aiming at how to effectively improve the quality of blast furnace modeling data,this paper uses operations such as uniform time granularity,removal of outliers,and normalization of data to achieve the sorting and cleaning of blast furnace data;and how to choose from many blast furnace body parameters For the most effective modeling input variable problem,this paper uses the grey correlation analysis method to screen out the key variables with the strongest correlation with the multivariate molten iron quality index as the input variables for modeling,thereby improving the efficiency of blast furnace molten iron quality modeling.(2)Aiming at the problem of how to effectively improve the modeling accuracy of blast furnace multivariate molten iron quality,a rms error probability weighted ensemble learning modeling algorithm was proposed for predictive modeling of blast funace multivariate molten iron quality.The proposed algorithm uses random vector functional-link networks(RVFLNs)with a very fast modeling speed as the sub-model,and uses the kernel density estimation method to estimate the root mean square error probability density function curve of the sub-model,and The root-mean-square error probability of each sub-model is obtained and used as its own weight,followed by weighted summation to obtain the final rms error-weighted integrated RVFLNs model.Finally,industrial experiments conducted show that the proposed algorithm can quickly and accurately predict the quality of multiple molten irons based on the changes in the input data of the blast furnace ironmaking process in real time,and has a high prediction accuracy for the daily operation and production of blast furnace ironmaking Provide valuable information.(3)Aiming at the problems of unstable furnace conditions and large fluctuations in blast furnace data,the prediction results of the multiple molten iron quality have large deviations and low credibility,which cannot provide guidance for daily operation and adjustment of the blast furnace.Blast furnace molten iron quality interval prediction modeling algorithm is used to build a molten iron quality interval prediction model.First,in order to improve the accuracy of modeling,a stacking-based molten iron quality model was established.Then,in order to characterize the credibility of the prediction results,an interval prediction method is introduced for simultaneous prediction of multiple molten iron quality values and prediction intervals.Finally,industrial experiments have shown that the algorithm proposed in this paper can accurately predict multiple molten iron quality prediction values and prediction intervals at the same time,and the prediction intervals have high credibility,providing valuable information for field operators,it has great reference value.