基于SSA优化的XGBoost-BP融合模型的高炉压差预测

Prediction of blast furnace pressure difference based on a combined model of XGBoost and BP optimized by SSA

高炉压差对高炉的正常运行和冶炼效率有着重要的影响。压差过高会导致悬料,破坏高炉顺行;压差过低会引发炉内气流过吹甚至管道气流现象。因此,将高炉压差维持在一定范围内是保证高炉炉况稳定的前提,对高炉压差提前预测可以帮助操作人员了解高炉的工作状态,及时调整操作参数,进而保证炉况稳定。为了提高高炉压差预测精度,首先,以国内某钢铁企业监测的高炉冶炼数据为基础,建立了基于麻雀搜索算法(Sparrow Search Algorithm,SSA)优化的梯度提升决策树(e Xtreme Gradient Boosting,XGBoost)和BP神经网络(Back Propagation Neural Network,BPNN)融合预测模型。该模型以SSA算法对单一模型的优化为核心,并通过误差倒数法加权预测值来降低模型的预测误差。结果表明,基于SSA优化的XGBoost-BP融合模型预测效果明显高于其他模型,该模型拟合优度达到0.842,具有较高的拟合能力,对比结果显示,融合模型较SSA-BP、SSA-XGBoost模型预测误差更小,收敛速度更快,并且在±0.025×10^(5)Pa的误差范围下,其预测准确率达到了96.13%。最后,基于所提的融合模型建立高炉压差预测系统,该系统不仅起到了指导高炉生产的作用,而且对炼铁工业的转型升级具有一定的现实意义。

The pressure difference of blast furnace has an important impact on the normal operation and smelting efficiency of blast furnace.Excessive pressure difference can cause material suspension and disrupt the smooth operation of the blast furnace;a low pressure difference can cause air flow through the furnace or even pipeline.Therefore,it is the premise of ensuring the stability of the blast furnace on maintaining the blast furnace pressure difference within a certain range.It can help operators understand the working status,adjust the operating parameters in time,and ensure the stability of the blast furnace by predicting the blast furnace pressure difference in advance.In order to improve the prediction accuracy of blast furnace pressure difference prediction,based on the monitoring of blast furnace smelting data in a domestic iron and steel enterprise,a fusion prediction model of eXtreme Gradient Boosting(eXtreme Gradient Boosting,XGBoost)and Back Propagation Neural Network(Back Propagation Neural Network,BPNN)optimized by Sparrow Search Algorithm(Sparrow Search Algorithm,SSA)was established.The core of this model lies in the optimization of the individual models by the SSA algorithm,and reduced the prediction error by weighting the prediction value with error reciprocal method.The results show that the prediction effect of the XGBoost-BP fusion model based on SSA optimization is significantly higher than that of other models.The goodness of fit of the model reaches 0.842,which has high fitting ability.Compaed with SSA-BP and SSA-XGBoost models,the fusion model has smaller prediction error and faster convergence speed.Under the error range of±0.025×10^(5)Pa,the prediction accuracy reaches 96.13%.Finally,based on the fusion model proposed,a blast furnace pressure difference prediction system was established,which not only plays a role in guiding blast furnace production,but also has a certain practical significance for the transformation and upgrading of ironmaking industry.