摘要
通过基于过程机理和经验的变量初筛、拉依达法则的显著误差处理、MIC最大信息系数法的变量相关性分析,提出了一种实时数据处理规则,可提高建模数据的质量。基于催化重整装置苯产品中甲苯和非芳烃含量的生产管控需求,采用BP神经网络算法,建立了苯产品中非芳烃和甲苯含量预测模型,所建模型对两个产品预测的均方根误差分别为0.0124和0.0463,平均相对误差分别为1.036%和3.312%。采用遗传算法NSGA-Ⅱ求解,可使苯中非芳烃和甲苯质量分数分别降低24.38%和82.58%,所建模型可为装置生产优化方案分析提供支持,所提出的建模方法可用于相关装置的智慧化平台建设。
A real-time data processing rule was proposed based on the process mechanism and experience of variable screening,significant error processing of Raida rule,and variable correlation analysis of MIC maximum information coefficient method,and the quality of modeling data was improved.Based on the production control demand of toluene and non-aromatic hydrocarbon content in benzene products of catalytic reforming unit,a prediction model of non-aromatic hydrocarbon and toluene content in benzene products was established by using BP neural network algorithm.The root mean square error of the two products predicted by the model was 0.0124 and 0.0463 respectively,and the average relative error was 1.036%and 3.312%respectively.Using genetic algorithm NSGA-Ⅱ,the content of non-aromatic hydrocarbon and toluene in benzene could be reduced by 24.38%and 82.58%respectively.The model could be used to support the analysis of optimization scheme of the plant production,and the proposed modelling method could be used for the intelligent platform construction of the related devices.
作者
潘艳秋
张超阳
李鹏飞
俞路
Pan Yanqiu;Zhang Chaoyang;Li Pengfei;Yu Lu(School of Chemical Engineering,Dalian University of Technology,Dalian,Liaoning 116024)
出处
《石油炼制与化工》
CAS
CSCD
北大核心
2023年第9期131-136,共6页
Petroleum Processing and Petrochemicals
