基于混合数据驱动算法的SCR氮氧化物排放量动态预测模型  

Dynamic Prediction Model for NO_(x) Emission of SCR System Based on Hybrid Data-driven Algorithms

作　　者：唐振浩[1] 王世魁 曹生现[1] 李扬[2] 沈涛[3] TANG Zhenhao;WANG Shikui;CAO Shengxian;LI Yang;SHEN Tao(School of Automation Engineering,Northeast Electric Power University,Jilin 132012,Jilin Province,China;School of Electrical Engineering,Northeast Electric Power University,Jilin 132012,Jilin Province,China;Harbin Boiler Company Limited,Harbin 150040,Heilongjiang Province,China)

机构地区：[1]东北电力大学自动化工程学院,吉林省吉林市132012 [2]东北电力大学电气工程学院,吉林省吉林市132012 [3]哈尔滨锅炉厂有限责任公司,黑龙江省哈尔滨市150040

出　　处：《中国电机工程学报》2022年第9期3295-3306,共12页Proceedings of the CSEE

基　　金：国家自然科学基金项目(61503072);吉林省科技发展计划项目(20190201095JC,20200401085GX)。

摘　　要：针对火电厂选择性催化还原(selective catalytic reduction,SCR)系统建模中存在的时延难确定、模型精度不高等问题,提出一种基于多数据驱动算法混合的动态建模方案。首先,对原始生产数据进行数据预处理,并采用最大信息系数(maximal information coefficient,MIC)估算各变量的延迟时间,对数据重构;然后,采用组合特征选择方法确定输入变量,并对输入时间序列进行变分模态分解;最后,结合极限学习机(extreme learning machine,ELM)和误差修正(error correction,EC)模型等数据驱动算法设计SCR出口NO_(x)混合动态预测模型。基于实际历史运行数据的实验结果表明,所建立模型预测结果的平均百分比误差(mean absolute percentage error,MAPE)为2.61%。模型敏感性分析表明,除喷氨量外,入口氧气浓度及烟气温度对NO_(x)排放量存在显著影响,在SCR过程优化控制中应重点考虑。A dynamic modeling scheme based on the hybrid data-driven algorithm was proposed to determine the delay time and improve prediction accuracy of SCR model.First,the original production data were preprocessed.The delay time of each variable was estimated with the maximal information coefficient algorithm.The data were reconstructed based on the delay time.Then,based on the input variables selected by the combined feature selection method,the input time series was decomposed by VMD.Finally,the hybrid dynamic prediction model for NO_(x)emissions of SCR system was built combining ELM and the error correction model.Experimental results based on actual production data show that the MAPE of predicted results is 2.61%.Model sensitivity analysis shows that besides the amount of ammonia injection,the inlet oxygen concentration and the flue gas temperature have a significant impact on NO_(x) emission,which should be considered in SCR process control and optimization.

关 键 词：选择性催化还原 最大信息系数 变分模态分解 数据驱动 误差修正 

分 类 号：TM61[电气工程—电力系统及自动化]