转炉终点钢水磷含量的预测对控制钢材质量和降低生产成本具有重要意义.为提高转炉终点磷含量的控制精准度和减少原料消耗,提出一种在Stacking集成学习框架下融合多个改进型极限学习机(Improved Ex-treme Learning Machine,IELM)的转炉终点磷含量预测模型.首先,根据理论基础和相关性分析确定模型的输入特征;其次,针对ELM(Extreme Learning Machine,ELM)的参数选取问题,提出一种改进的猎人-猎物优化(Improved Hunter-Prey Optimization,IHPO)算法对其进行参数寻优得到IELM;最后,在Stacking集成算法框架下,融合多个IELM初级学习器,以高斯过程回归(Gaussian Process Regression,GPR)作为次级学习器,建立Stacking-IELM-GPR集成学习模型.与7种单一模型和2种Bagging同质集成算法进行对比,结果表明,所提出模型在预测精度和误差性能指标方面表现最优,且预测误差在±0.003％间的命中率为92.86％.
Research on prediction of phosphorus content at endpoint of converter integrated with IHPO-ELM
The prediction of phosphorus content in molten steel at the endpoint of the converter is of great significance for controlling steel quality and reducing production costs.To improve the accuracy of controlling the phosphorus content at the endpoint of the converter and reduce raw material consumption,a prediction model for the phosphorus content at the endpoint of the converter was proposed,which integrated multiple improved Extreme Learning Machines(IELM)under the Stacking ensemble learning framework.Firstly,the input features of the model were determined according to the theoretical basis and correlation analysis;secondly,for the parameter selection problem of ELM(Extreme Learning Machine,ELM),an improved Hunter-Prey Optimization(IHPO)algorithm was proposed to optimize its parameters to obtain IELM;finally,under the framework of Stacking integrated algorithm,multiple IELM primary learners were fused,and Gaussian Process Regression(GPR)was used as the secondary learner to establish the Stacking IELM-GPR Ensemble learning model.Compared with 7 single models and 2 Bagging homogeneous ensemble algorithms,the results show that the proposed model performs best in terms of prediction accuracy and error performance indicators,and the hit rate of prediction error within±0.003％is 92.86％.
NETL
NSTL
万方数据
