基于OSC-CKPLS方法的SCR出口NO_x排放预测  

Prediction of SCR outlet NO_x Emission Based on OSC-CKPLS Method

作　　者：闫来清 董泽[2] 孟磊 YAN Laiqing;DONG Ze;MENG Lei(School of Control and Computer Engineering,North China Electric Power University,Beijing 102206,China;Hebei Technology Innovation Center of Simulation&Optimized Control for Power Generation,North China Electric Power University,Baoding 071003,China;Datang Environment Industry Group Co.,Ltd.,Beijing 100192,China)

机构地区：[1]华北电力大学控制与计算机工程学院,北京102206 [2]河北省发电过程仿真与优化控制技术创新中心(华北电力大学),河北保定071003 [3]大唐环境产业集团股份有限公司,北京100192

出　　处：《锅炉技术》2020年第2期7-13,共7页Boiler Technology

基　　金：河北省自然科学基金资助项目(E2018502111);中央高校基本科研基金(2018QN097)。

摘　　要：针对选择性催化还原(selective catalytic reduction,SCR)脱硝系统反应过程复杂,在工况变化时存在非线性、大惯性和强干扰性的特点,难以建立准确的出口NO_x排放浓度模型。利用核偏最小二乘法具有解决变量众多且存在严重相关的非线性工业过程建模的优点,首先引入正交信号校正(orthogonal signal correction,OSC)对相空间重构后的样本进行预处理,剔除与建模无关的信息;然后利用组合核偏最小二乘法(combination kernel partial least squares,CKPLS)具有较好的学习能力和泛化能力的特点,提出OSC-CKPLS方法提高模型性能;最后采用滑动窗口更新,并反馈补偿修正模型。对2个标准数据集进行仿真,分别验证CKPLS、OSC和OSC-CKPLS能够提高模型性能;并对SCR脱硝系统现场数据验证了本文方法的有效性。For the selective catalytic reduction(SCR) denitrification system, the reaction process is complex, and there are nonlinear, great inertia and strong distribution characteristics when the working conditions change, so it is difficult to establish an accurate model of SCR outlet NO_x emission concentration. In this paper, owing to kernel partial least squares which has the advantage of solving the modeling problem of nonlinear industrial processes with many variables and serious correlations, firstly, the orthogonal signal correction(OSC) is introduced, and the sample data based on phase space reconstruction is preprocessed to eliminate the information unrelated to modeling. Then, by using combination kernel partial least squares(CKPLS) which has better learning ability and generalization ability, the proposed OSC-CKPLS method can improve model performance. Finally, the sliding window is used for updating and feedback compensation to modify the model. By simulating two standard datasets, it is verified that the OSC, CKPLS and OSC-CKPLS method can improve the model performance, and the validity of the proposed method is checked by the field data of SCR denitrification system.

关 键 词：选择性催化还原 脱硝系统 正交信号校正 组合核偏最小二乘法 出口NO_x排放预测 

分 类 号：TK39[动力工程及工程热物理—热能工程]