基于DF-AdaboostSVM模型的脱硝入口氮氧化物浓度预测研究  

Research on Prediction of Nitrogen Oxide Concentration at the Denitration Inlet Based on DF-AdaboostSVM Model

作　　者：马立增 张玲 谷宇 吴俣 唐媛媛 汤光华 MA Lizeng;ZHANG Ling;GU Yu;WU Yu;TANG Yuanyuan;TANG Guanghua(CHN Energy BengBu Power Generation Co.,Ltd.,Bengbu 23341l,China;Nanjing Guodian Environmental Protection Technology Co.,Ltd.,Nanjing 210061,China)

机构地区：[1]国能蚌埠发电有限公司,安徽蚌埠233411 [2]南京国电环保科技有限公司,江苏南京210061

出　　处：《锅炉技术》2025年第2期31-37,共7页Boiler Technology

摘　　要：传统煤电机组脱硝系统喷氨不精准导致过量喷氨、氮氧化物排放超标以及喷氨无法投自动等现象,解决上述问题需同时实现喷氨总量精确控制和脱硝反应器氨氮空间分布的均衡配比。针对脱硝系统反应器入口氮氧化物浓度检测滞后性导致喷氨总量控制不精确问题,提出一种基于主导因素(DF)和自适应增强算法(Adaboost)集成支持向量机(SVM)的氮氧化物浓度预测模型。通过DF分析某660MW煤电机组历史运行数据,选择对脱硝入口氮氧化物浓度影响较大的辅助特征参数并确定所选参数相对于氮氧化物浓度的迟滞时间。依据迟滞时间重构数据集,构建DF-AdaboostSVM氮氧化物浓度预测模型。研究结果表明:与限定单一迟滞时间180s、240s和300s建模以及单一SVM模型相比,使用DF迟滞时间重构数据集搭建集成模型有更优秀的预测精度,其平均绝对百分比误差为4.03%,均方根误差为16.74,决定系数为0.91,均优于上述对比模型。由此可见提出的算法和模型更适合脱硝人口氮氧化物浓度预测。The traditional coal-fired power unit denitrification system lacks precision in ammonia injection,resulting in excessivee ammonia injection,excessive nitrogen oxide emissions,and inability to automatically inject ammonia.To solve these problems,it is necessary to achieve precise control of the total amount of ammonia injection and balanced proportion of ammonia nitrogen in the denitrification reactor space.This paper proposes a nitrogen oxide concentration prediction model based on Dominant Factor(DF)and Adaboost integrated Support Vector Machine(SVM)to address the issue of inaccurate control of total ammonia injection due to the lag of nitrogen oxide concentration detection at the inlet of the denitrification system reactor.Firstly,through DF analysis of the historical operating data of a 660MW coal-fired power unit,select the auxiliary characteristic parameters that have a significant impact on the nitrogen oxide concentration at the denitrification inlet and determine the lag time of the selected parameters relative to the nitrogen oxide concentration.Then,based on the lag time reconstruction dataset,a DF-AdaboostSVM nitrogen oxide concentration prediction model is constructed.The research results indicate that compared to modeling with a limited single lag time of 180 s,240 s,and 300 s,as well as a single SVM model,building an integrated model using DF lag time reconstruction dataset has better prediction accuracy,with an average percentage error of 4.03%,root mean square error of 16.74,and R^(2)of 0.91,all of which are superior to the other models mentioned above.From this,it can be seen that the proposed algorithm and model are more suitable for predicting the nitrogen oxide concentration at the denitrification inlet.

关 键 词：主导因素 Adaboost集成 迟滞时间 氮氧化物浓度 预测模型 

分 类 号：X511[环境科学与工程—环境工程]