基于XGBoost和反向自适应粒子群的烧结配料智能优化方法  

Optimization of sinter batching based on intelligent method of XGBoost and reverse adaptive population particle swarm optimization

作　　者：陈健 杨春节[2] 胡兵[3] 钱卫东[3] 苏志祁 CHEN Jian;YANG Chunjie;HU Bing;QIAN Weidong;SU Zhiqi(Polytechnic Institute,Zhejiang University,Hangzhou 310015,China;College of Control Science and Engineering,Zhejiang University,Hangzhou 310027,China;Shanghai Baosight Software Co.,Ltd.,Shanghai 201203,China;Guangxi Liuzhou Steel Dongxin Technology Co.,Ltd.,Liuzhou 545002,China)

机构地区：[1]浙江大学工程师学院,浙江杭州310015 [2]浙江大学控制科学与工程学院,浙江杭州310027 [3]上海宝信软件股份有限公司,上海201203 [4]广西柳钢东信科技有限公司,广西柳州545002

出　　处：《冶金自动化》2023年第3期71-80,共10页Metallurgical Industry Automation

基　　金：国家自然科学基金重点项目(61933015);中央高校基本科研业务费专项资金资助项目(K20200002)(浙江大学NGICS大平台)。

摘　　要：原料配制是烧结工艺的关键环节,不仅影响烧结成本,而且还关系到烧结矿质量。针对现有烧结配料方案优化目标相对单一、模型更新不太及时的问题,提出了基于XGBoost和反向自适应粒子群优化(reverse adaptive population particle swarm optimization,RAPPSO)算法的烧结配料智能优化方法,以优化烧结矿的TFe含量和成本。首先,依托数字孪生数据库提供的实时数据,建立基于XGBoost的烧结矿TFe含量在线预测模型。然后,提出了一种RAPPSO算法,该算法可以根据迭代次数自动调整反向种群大小,经过标准测试函数验证,该算法具有良好的收敛性和搜索能力。最后,综合考虑烧结矿TFe含量和配料成本来构建优化模型的目标函数,建立烧结配料智能优化模型并使用RAPPSO算法进行求解。经烧结数字孪生系统验证,该方法使配料成本降低8.27元/t,烧结矿TFe质量分数提升0.57%,这表明所提配料优化方法具有良好的性能。Raw fuel formulation is a key part of the sintering process,which not only affects the sintering cost,but also relates to the quality of sintered ore.Aiming at the problems of relatively single optimization objective and less timely model update of existing sintering batching scheme,an intelligent optimization method of sintering batching based on XGBoost and reverse adaptive population particle swarm optimization(RAPPSO)algorithm was proposed to optimize the TFe content of sinter ore and cost.Firstly,an online prediction model based on XGBoost for the TFe content of sintered ore was established by relying on the real-time data provided by the digital twin database.Then,a RAPPSO algorithm was proposed,which could automatically adjust the inverse population size according to the number of iterations,and was proved to have good convergence and search ability by standard test function tests.Finally,the objective function of the optimization model was constructed by integrating the sintered ore TFe content and the batching cost,and the intelligent optimization model of sintered batching was established and solved by the RAPPSO algorithm.As verified by the sintering digital twin system,the method reduces the batching cost by 8.27 RMB/t and increases the sintered ore TFe content by 0.57%,indicating that the proposed batching optimization method has good performance.

关 键 词：烧结配料 TFe含量 配料成本 智能优化 XGBoost RAPPSO 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TF046.4[自动化与计算机技术—控制科学与工程]