Predictive Modeling and Control Analysis of Fuel Ratio in Blast Furnace Ironmaking Process Based on Machine Learning

The fuel ratio (FR) is an important parameter to characterize the carbon emission and energy consumption in the blast furnace (BF) ironmaking process. However, the BF ironmaking process is complex and there are many factors affecting the FR, which makes it difficult to predict and to quantify the impact of each parameter on the FR. In this study, different machine-learning methods were used to build predictive models of FR and to analyze the effects on it of different types of parameters. The dataset was collected over 7 years in a steelmaking factory, with 31 features. Through data cleaning, the data quality and the performance of machine-learning models were improved. It was found that Gaussian process regression (GPR) and extreme gradient boosting (XGBoost) displayed greater flexibility and accuracy in predicting the FR. The permutation feature importance (PFI) method was used to gain insight into how each parameter correlates with the target. The results indicate that random Forest regression gas utilization rate, blast volume, and oxygen enrichment are the main factors affecting the FR, and their degrees of influence are 0.73, 0.076, and 0.075, respectively.