烟道优化协同机器学习的SCR精准喷氨试验研究  

Experimental Study on Precise Ammonia Injection in SCR Based on Collaborative Machine Learning for Flue Optimization

作　　者：张国兴 陈磊 金立梅 张健 周静[3] 谭厚章[3] ZHANG Guoxing;CHEN Lei;JIN Limei;ZHANG Jian;ZHOU Jing;TAN Houzhang(Ningxia Yuanyang Lake Second Power Generation Co.,Ltd.,Lingwu 751400,China;Xi'an Gerui Power Technology Co.,Ltd,Xi'an 710000,China;MOE Key Laboratory of Thermo-Fluid Science and Engineering,Xi'an Jiaotong University,Xi'an 710000)

机构地区：[1]国能宁夏鸳鸯湖第二发电有限公司,宁夏灵武751400 [2]西安格瑞电力科技有限公司,陕西西安710000 [3]西安交通大学热流科学与工程教育部重点实验室,陕西西安710000

出　　处：《工业炉》2025年第2期1-7,13,共8页Industrial Furnace

摘　　要：通过流体力学计算(CFD)和深度神经网络(DNN)模型实现燃煤机组瞬态过程选择性催化还原NO_(x)系统(SCR)入口的NO_(x)浓度预测和喷氨量精准控制。采用CFD对SCR烟道进行仿真优化,采用互信息算法结合DNN模型预测SCR入口的NO_(x)浓度。结果表明:根据CFD结果进行改造后,SCR入口烟道的流速均匀性和NO_(x)浓度分布得到了显著改善;经过DNN模型预测和喷氨量控制,SCR出口AB两侧的NO_(x)浓度分别降至40.45 mg/m~3@6%O_(2)和36.04 mg/m^(3)@6%O_(2),并有效减少NH_(3)逃逸;最终,改造后每度电尿素消耗量降低了16.40%。Through computational fluid dynamics(CFD)and deep neural network(DNN)models,the prediction of NO_(x) concentration and the precise control of ammonia injection at the inlet of selective catalytic reduction system(SCR)for NO_(x)in the transient process of coal-fired units are achieved.CFD is used to simulate and optimize the SCR flue,and the mutual information algorithm combined with DNN model is used to predict the NO_(x) concentration at the SCR inlet.The results show that after modification based on the CFD results,the uniformity of flow velocity and the distribution of NO_(x) concentration within the SCR inlet flue is significantly improved;After the implementation of DNN-based NO,concentration prediction and ammonia injection control,the NO_(x)concentration on both AB sides of SCR outlet decreases to 40.45 mg/m@6%0_(2) and 36.04 mg/m@6%0_(2),respectively,and NH;slip is also effectively controlled.Ultimately,after the renovation,the urea consumption per kilowatt-hour decreases by 16.40%.

关 键 词：烟道优化 机器学习 喷氨控制 NO_(x)排放 

分 类 号：TK16[动力工程及工程热物理—热能工程]