基于双向深度学习的电站锅炉SCR脱硝系统入口NOx浓度预测  

Pridiction of NOx Concentration at the Inlet of SCR Denitration System of Utility Boiler Based on Bidirectional Deep Learning

作　　者：王祖林 韩硕 康俊杰 李云飞[1] 藕泉江 WANG Zu-lin;HAN Shuo;KANG Jun-jie;LI Yun-fei;OU Quan-jiang(Hebei Guohua Dingzhou Power Generation Co.,Ltd.,Dingzhou 073000,China;School of Control and Computer,North China Electric Power University,Beijing 102206,China;Beijing Dahua Jieneng Engineering Technology Co.,Ltd.,Beijing 100029,China)

机构地区：[1]河北国华定州发电有限责任公司,定州073000 [2]华北电力大学控制与计算机学院,北京102206 [3]北京达华洁能工程技术有限公司,北京100029

出　　处：《自动化与仪表》2021年第1期82-87,共6页Automation & Instrumentation

摘　　要：该文基于某660 MW电站锅炉的现场运行数据,在进行数据预处理的基础上,利用随机森林算法对输入变量进行特征提取以降低变量维数和消除变量间的相关性,并与双向长短时记忆神经网络(bi-directional long short-term memory,Bi-LSTM)相结合,建立了SCR脱硝系统入口NOx浓度的模型。将上述模型与其他建模方法进行比较,并将该模型实际应用于某电厂,作为精准喷氨控制的基础,结果表明通过Bi-LSTM建立的预测模型具有良好的预测能力,为进一步实施精准喷氨控制提供了模型基础。This article based on the field operation data of a 660 MW utility boiler,on the basis of data preprocessing,the random forest algorithm is used to extract the features of input variables to reduce the dimension of variables and eliminate the correlation between variables.Combined with bi-directional long short-term memory(Bi-LSTM),the model of NOx concentration at the inlet of SCR denitration system is established.The above model is compared with other modeling methods and is applied to a power plant as the basis of precise ammonia injection control.The results show that the prediction model established by Bi-LSTM has good prediction ability,which provides a model basis for further implementation of accurate ammonia injection control.

关 键 词：数据预处理 随机森林 双向长短时记忆神经网络 入口NOx排放 电站锅炉 

分 类 号：TM621.2[电气工程—电力系统及自动化]