基于CIFE-MLP模型的脱硝入口氮氧化物浓度预测研究  

Prediction of NOx concentration at the entrance of denitration system based on CIFE-MLP

作　　者：骆宏飞 周传明 火鸿宾 吴俣 程恩路 LUO Hongfei;ZHOU Chuanming;HUO Hongbin;WU Yu;CHENG Enlu(China Energy BengBu Power Generation Co.,Ltd,Bengbu 233411,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：《电力科技与环保》2023年第1期43-48,共6页Electric Power Technology and Environmental Protection

基　　金：国能蚌埠发电有限公司科技项目(GN蚌发采[2022]121号)。

摘　　要：为解决燃煤机组深度减排和调峰约束下,由于脱硝系统氮氧化物测量大滞后、快时变所引起的测量系统无法及时反馈氮氧化物浓度变化的缺陷,优化解决过量喷氨产生过度氨逃逸、喷氨不足引起的氮氧化物排放超标、喷氨设施无法自动投运等问题,提出了一种基于互信息特征选择-多层感知机神经网络模型(conditional information feature extraction-multi layer perceptron,CIFE-MLP)的脱硝系统入口处氮氧化物浓度预测模型,利用某600 MW火电机组历史运行数据进行模拟实验验证。结果表明,将特征变量相较于输出变量(脱硝系统入口处氮氧化物浓度)的时间前移180 s后,使用CIFE-MLP建立的预测模型准确率为95.41%,决定系数R2为0.82,能够准确预测脱硝系统入口处氮氧化物浓度的变化趋势,解决了氮氧化物浓度测量延迟的问题。对燃煤机组深度减排和调峰运行操作具有一定的参考意义。To solve the defect that the measurement system can not timely feedback the concentration change of generated nitrogen oxides due to the characteristics of large lag and fast time variation of nitrogen oxides measurement of denitration system under the constraints of thorough emission reduction and peak shaving of coal-fired units,and optimize and solve the problems of excessive ammonia escape caused by excessive ammonia injection,excessive nitrogen oxides emission caused by insufficient ammonia injection,failing to realize the automatic operation of ammonia injection facilities,and so on,a prediction model of NOx concentration at the inlet of denitration system based on the CIFE-MLP model is proposed.The simulation experiment is carried out using the historical operation data of a 600MW thermal power unit.The time of the characteristic variable compared with the output variable(inlet NOx concentration)is moved forward by 180s.The characteristic variable with a large mutual information value is screened using the mutual information feature selection algorithm.Then,the prediction model is established using the multilayer perceptron neural network.The results show that the accuracy of the model is 95.41%and the determination coefficient R~2 of the model is 0.82,which can better predict the changing trend of nitrogen oxide concentration at the entrance of the denitration system and solve the problem of delay in measuring nitrogen oxide concentration.This research has certain reference significance for thorough emission reduction and peak shaving operation of coal-fired units.

关 键 词：氮氧化物浓度 多层感知机 互信息特征选择 

分 类 号：X511[环境科学与工程—环境工程]