Operational-pattern Optimization In Blast Furnace PCI Based On Prediction Model Of Neural Network

Blast Furnace PCI is the main technique to Blast furnace ironmaking. It is difficult tobuild model and optimize the process of Blast Furnace PCI because the mechanism forBlast Furnace PCI process is complicated and many technological parameters can not bemeasured which is also strong coupling. Blast Furnace PCI process has accumulated lots ofproduction data, the condition and operating parameters of the PCI smelting process isextracted from the data and described in the form of operational pattern. the superioroperational patterns library based on the standards of blast furnace expert is also huge, it isdifficult to search the superior operational pattern appropriate to the current working status.To solve this problem, a pattern matching and evolution method is researched in this paper,and operation optimization for the Blast Furnace PCI process is achieved.The silicon content in hot metal tapped from blast furnace is a very important factorwhich should be controlled in iron making process. It is not only a very important criteriain evaluating the quality of hot metal, But also reflect the energy utilization ratio of blastfurnace, Adirect comparison of parameters is also the operator in the actual productionprocess on furnace condition judgement in the same time. So, it is essential to keep itsstabilization for the blast furnace manufacturing. However, In the actual productionprocess，the silicon can not be real-time monitored，it is very difficult for the process ofoptimization operation mode.In this paper, the influence factors on the silicon content inmolten iron of blast furnace and control knowledge based onthe collection, analysis,Combined with the silicon content of molten iron control process ofmany uncertain factors,the neural network of BP prediction model is established.Through the simulation of actual production data show，The model can be effectivefor blast furnace smelting process to predict the silicon content in iron water.Considering the low speed of pattern matching process and low matching accuracybecause of the huge superior operational patterns library, a similarity measure criterioncombined the Euclidean distance is proposed to improve the accuracy of the pattern matching process; First, analyzed two clustering of fine operation mode of Library, and getfive good operation mode sub class; Then compare the similarity of the working status andthe center of sub-class to determine the sub-class that the working status belongs to;Finally in the sub operation modes have been identified in the class,search the optimaloperational patterns in the sub-class based on the similarity measure criteria.Because of the smelting blast furnace coal injection process for integrated processrequirements is very harsh, it is possible that there is no operational pattern appropriate tothe current working status, it is difficult to optimize.To solve this problem,a operationalpattern evolution strategy is proposed, first the relation model between operational patternsand the comprehensive working status evaluation standards is built, then the optimaloperational pattern and the optimal operational parameters appropriate to the currentworking status can be acquired using the Particle Swarm Optimization and the optimaloperational guides the production, the superior operational pattern appropriate to thecurrent working status is sent to the superior operational patterns library and renews theoperational patterns library at the same time.Combined with the Blast Furnace PCI process of a smelter, a operational patternoptimize system using the operation optimization method based on the operational patternmatching and evolution strategy for the Blast Furnace PCI process is exploited and isverified to be effective and feasible by operating data,it provide an good advice for blastfurnace operator.