Carbon Content Prediction of Converter Steelmaking End-Point Based on Hybrid Strategy

The endpoint prediction model for converter steelmaking is particularly important for achieving the carbon content at the endpoint. In order to accurately predict the end carbon content of steel, the maximum relevance and Minimum redundancy (mRMR) algorithm, which is simple and easy to process high-dimensional data, is introduced for feature selection, and key factors affecting the end carbon content are screened out as input to the model. An improved Pelican optimization algorithm (MPOA) is proposed to optimize the hyperparameters of hybrid core Extreme Learning Machine (HKELM). Then, the prediction model of mRMR-MPOA-HKELM is constructed and compared with the prediction results of HKELM and MPOA-HKELM. The mean absolute error of mRMR-MPOA-HKELM model is 0.012447, the root mean square error is 0.017671, and the coefficient of determination is 0.92007. The accuracy rate of the prediction error of the end carbon content reaches 93% within ±0.02%, which is more accurate than the other two models.