Information flow processing in blast furnace and coke ratio prediction based on particle swarm optimisation with BP neural network
The coke ratio of a blast furnace has direct effect on its output and operational efficiency.However,the prediction accuracy of this ratio is typically hindered by the multitude of covariance phenomena inherent in the blast furnace smelting process.To address this challenge,the blast furnace data was initially preprocessed and categorized for separating outlier analysis and treatment.Subsequently,a comprehensive analysis leveraging Pearson's correla-tion,Spearman's rank correlation,Maximum Information Coefficient(MIC),and stepwise regression was conduc-ted to identify the parameter variables most strongly correlated with the coke ratio.This process reduced the initial pool of 32 parameters to 16 key feature variables,which were then sorted and analyzed based on their relevance.Fi-nally,a blast furnace coke ratio prediction model employing an error backpropagation network(BPN)was estab-lished.Furthermore,a particle swarm optimization-based BP neural network(PSO-BP)model was developed to re-fine the prediction capabilities.Evaluation of both models reveals that the PSO-BP model exhibits superior perform-ance,achieving higher accuracy and precision in predicting the coke ratio of the blast furnace.
NETL
NSTL
万方数据
