基于KPCA和Logistic-SSA-BP的烧结矿FeO含量预测
Prediction of FeO content in sinter based on KPCA and logistics-SSA-BP
惠佳豪 1邢相栋 1郑兆颖 1王宇星 1吕明1
作者信息
- 1. 西安建筑科技大学冶金工程学院,陕西西安 710055
- 折叠
摘要
烧结矿作为高炉炼铁过程中的重要原料之一,其FeO含量的控制对于炼铁工艺、铁水质量和能源消耗等方面都具有重要影响.针对目前研究过程中存在特征选择偏离实际和预测模型泛化能力差等问题,提出了一种基于核主成分分析(KPCA)和Logistic-SSA-BP的烧结矿FeO含量预测模型.使用皮尔逊(Pearson)和KPCA对特征参数进行筛选和降维,并结合Logistic-SSA-BP优化算法,对采集到的数据进行训练、学习和验证.实验结果表明,预测的绝对误差稳定在[0,0.21]范围内,预测值与实际值相差在士0.2以内命中率达到98.75％,预测模型的性能表现较好,评价指标MSE、MAE和RMSE达到0.013、0.101、0.115.该预测模型能够准确预测烧结矿FeO含量,为高炉操作人员在建立配料方案和执行工艺操作时提供指导方向.
Abstract
Sinter is one of the important raw materials in the blast furnace ironmaking process.The control of FeO content has an important impact on ironmaking process,iron quality and energy consumption.Due to the problems of feature selection deviating from reality and poor generalization ability of prediction model in the current research process,the prediction model of FeO content of sinter based on kernel principal component analysis(KPCA)and Logistics-SSA-BP was proposed.The feature parameters were screened and dimensionality reduced by Pearson and KPCA.Combined with Logistic-SSA-BP optimization algorithm,the collected data are trained,learned and verified.The experimental results show that the absolute error is stable within the range of[0,0.21],and the hit rate reaches 98.75％within+0.2 between the predicted value and the actual value.The performance of the prediction model is better,and the evaluation indexes of MSE,MAE,and RMSE reach 0.013、0.101,0.115.The prediction model could accurately predict the FeO content of sinter,which could provide direction to blast furnace operators when establishing batching programs and executing process operations.
关键词
烧结矿/FeO含量/Pearson/KPCA/预测模型Key words
sinter/FeO content/Pearson/KPCA/prediction model引用本文复制引用
基金项目
国家自然科学基金面上项目(52174325)
陕西省重点研发计划(2019TSLGY05-05)
陕西省创新能力支撑计划(2023-CX-TD-53)
出版年
2024
NETL
NSTL
万方数据