基于Adaboost混合模型的乙烯裂解结焦量软测量  

Soft measurement of ethylene cracking coke content based on Adaboost model

作　　者：刘丽颖[1] 李悦[1] 方鲁杰 胡浩威[1] 张蒙蒙[1] 王立冬[1] 王娟[1] 

机构地区：[1]辽宁石油化工大学,辽宁抚顺113001

出　　处：《自动化与仪器仪表》2015年第6期50-53,共4页Automation & Instrumentation

摘　　要：针对乙烯裂解生产过程中结焦机理模型参数难以测量的问题,提出一种混合模型测量结焦量的方法。该方法基于Adaboost算法集成多输出支持向量机(M-SVM)作为智能模型来校正机理模型难以获得的参数。分析影响机理模型的因素,以机理模型中难以获得的参数作为智能模型的输出变量,机理模型根据得到的参数进行计算。利用Adaboost算法在学习过程中重点训练错分样本点的特点,来提高多输出支持向量机中权值分配精度,从而提高模型的泛化能力和精度。智能模型部分采用M-SVM。仿真结果表明:该混合模型可以克服传统结焦机理模型难以准确测量的缺点,且避免"黑箱"模型过分依赖数据的不足。A hybrid model measuring the amount of coking method is put forward to solve the coking mechanism model parameters in ethylene production process which is difficult to measure. The method is based on Adaboost algorithm integrated multi-output support vector machine(M-SVM) to correct the mechanism is difficult to obtain the parameters of the model. Analysis of factors influencing mechanism model, with the parameters are difficult to obtain in the model as a variable output mechanism of intelligent model, then the mechanism model is calculated according to the obtained parameters. Based on the characteristics of Adaboost algorithm training the fault sample points during the learning process, to improve the M-SVM in the weights allocation precision, and to improve the precision of the model. Intelligent model part adopts M-SVM, and the simulation results show that this model not only overcomes the deficiency of the traditional mechanism model which is difficult accurately implement, but also avoids the 'black box' model to rely too much on the defects of the data.

关 键 词：结焦机理模型 支持向量机 多输出 混合模型 ADABOOST 

分 类 号：TQ221.211[化学工程—有机化工]