基于改进Adaboost算法的SCR脱硝系统多模型集成建模  

Multi-model Integrated Modeling for SCR Denitration System byImproved Adaboost Algorithm

作　　者：乔源 邢波涛 赵文杰[1] 孙志英 QIAO Yuan;XING Botao;ZHAO Wenjie;SUN Zhiying(School of Control and Computer Engineering,North China Electric Power University,Baoding 071003,China;School of Electrical and Electronic Engineering,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003 [2]华北电力大学电气与电子工程学院,河北保定071003

出　　处：《华北电力大学学报（自然科学版）》2021年第1期90-97,106,共9页Journal of North China Electric Power University：Natural Science Edition

基　　金：国家重点研发计划项目(2016YFB0600701)。

摘　　要：燃煤机组选择性催化还原法(SCR)脱硝系统是复杂的多变量非线性时变系统,且运行工况多变,单一模型难以准确描述系统的多工况运行特性,因此,基于改进自适应提升算法(Adaptive Boosting,Adaboost),提出一种多模型集成建模方法。首先对Adaboost集成算法的损失函数进行优化,引入正则化因子和先验知识参数以提高模型的精确度;然后基于SCR脱硝系统的多工况运行数据,采用改进的Adaboost算法自动调整运行数据的权重,形成多个样本子集;在每个样本子集空间上,通过支持向量机(Support Vector Machine,SVM)建立SCR脱硝系统局部模型,并采用智能算法对局部模型的参数进行优化;最后基于改进的Adaboost算法集成得到SCR脱硝系统集成模型。仿真结果表明,与传统的Adaboost算法相比,改进的Adaboost算法集成的预测模型有效提高了模型预测精度,同时避免了单一模型由于样本分布的不均匀性而导致的模型工况适应性差等问题,提高了模型的复杂工况适应性能。As a complex multi-variable nonlinear time-varying system,the selective catalytic reduction(SCR)denitration system features variable operating conditions,which makes it difficult to accurately describe the system operating characteristics under a single model.Therefore,this paper proposes a multi-model integrated modeling method based on an improved Adaptive Boosting(Adaboost)algorithm.Firstly,we optimize the Adaboost integrated algorithm in terms of loss function and introduce regularization factors and a priori knowledge parameters to improve the model accuracy.Then,based on the multi-operating data of the SCR denitration system,the modified Adaboost algorithm is used to automatically adjust operation weights of the data and to form multiple sample subsets.On each sample subset space,a local model of the SCR denitration system is established by supporting vector machines(SVM),and intelligent algorithms are applied to optimize the parameters of the local model.Finally,based on the improved Adaboost algorithm,we obtain the model of SCR denitration system.The simulation results show that the improved Adaboost algorithm integrated prediction model is able to deal with complex operating conditions in that it effectively improves the model prediction accuracy,and that it avoids poor model operating conditions caused by the unevenness of the sample distribution in a single model.

关 键 词：NOX排放量 多模型集成 ADABOOST算法 支持向量机 

分 类 号：TN081[电子电信—物理电子学]