The Operational Parameters Optimization Based On Data-driven For Blast Furnace Ironmaking Process

As the main way of ironmaking today, blast furnace ironmaking process is complex thatfaced with many difficulties in its optimization control. Firstly, with the complex ore sourceand changeable working conditions, the process involves lots of conversion and transferbetween substance and energy, it is hard to establish the accurate mathematical model.Secondly, The nonlinear relationship between parameters and production indexes in blastfurnace process are difficult to describe and estimate, and parameters has strong couplingcharacteristic from each other, thus lead to some difficulties in optimization. Therefore, thedata mining technology is used to find out the rules of ironmaking process from mass datathat accumulated in production and determine the suitable operational parameters (Coalinjection amount, Blast volume, Oxygen enrichment percentage) quickly and effectively.These are significant for energy-saving, consumption-reducing, economic indicators-increasing of ironmaking process, so doesthe competitiveness-enhancing ofenterprises.In this paper, optimization method of operational parameters is presented based on datamining when combined the mechanisms of blast furnace ironmaking, the characteristics ofoperational parameters and their research status. This method includes the prediction modelof silicon content of hot metal, coke rate, coal ratio and the operational parametersoptimization algorithm to seek the best parameters. In this paper, the main results are asfollows.(1) The framework of operational parameters optimization in blast furnace ironmakingprocess based on data driven is proposed in this paper. This framework includes thepreparation process of pattern matching, pattern matching process and pattern evolutionprocess. First, it is important to establish the optimal library based on comprehensive statusindex according to the framework, then some methods are used to obtain the optimaloperationalparametersthat suitethe statusofblast furnace.(2) The strategy of multi-hierarchy pattern matching strategy based on euclidean distance is proposed. On the basis that state parameters, operational parameters andindicators in ironmaking process are determined, considering the low speed of patternmatching process caused by huge optimal operational patterns library, this strategy is used tocalculate the similarity of patterns and convert it to similarity value, at the same time, theoptimal library are clustered by adopting fuzzy c-means clustering method. Thus can reducethetime ofpattern matching greatly.(3) The strategy of operational pattern evolution based on particle swarmoptimization(PSO) is proposed. Since the blast furnace ironmakingprocess is verycomplicated, the operational pattern which matches completely to work status may not exitin optimal library, it is necessary to use similar operational patterns and PSO algorithmwithglobal searching ability. Then judging the operational parameters whether them meetsstopping criteria according to prediction model of indexes. It will stop iterative update untilmeets the requirements so as to obtain optimal operational parameters suitable for workingstate. Meantime, the optimalpatternshould be put into optimalpattern library.Using above-mentioned methods in this paper, combined with the actual operationprocess of blast furnace in some steel plant, and mass data have been used to verify theeffectiveness of this method. The results show that,it canprovide decision-makingguidance for operators and has a positive effect on the overall optimization of blast furnaceironmaking process.