基于IGWO-BP的SCR脱硝效率软测量模型  

Soft Sensor Model of SCR Denitration Efficiency Based on IGWO-BP

作　　者：章文涛 张东平 ZHANG Wentao;ZHANG Dongping(School of Electrical Engineering,Nanjing Institute of Technology,Nanjing 211167,China;School of Environmental Engineering,Nanjing Institute of Technology,Nanjing 211167,China)

机构地区：[1]南京工程学院电力工程学院,南京211167 [2]南京工程学院环境工程学院,南京211167

出　　处：《计算机测量与控制》2021年第10期66-70,76,共6页Computer Measurement &Control

基　　金：江苏省自然科学基金(BK20181023);企业重大科研攻关项目(科18-168)。

摘　　要：针对电厂SCR脱硝装置运行参数多且相互高度耦合,脱硝效率定量描述困难,以及传统BP网络存在的问题,提出一种基于IGWO-BP的脱硝效率软测量模型;该方法将基于主成分分析后的降维数据作为输入变量,采用改进灰狼算法对BP网络初始权值、阈值进行优化,利用优化后的网络对脱硝效率进行预测;该模型已成功应用于大唐洛河发电厂6号机组脱硝装置,结果表明:实际脱硝效率平均绝对百分比误差为2.31%,较传统BP算法与IGWO-BP算法分别降低48.92%和21.69%,具有更高的预测精度。The SCR denitrification device in power plant has many operation parameters and is highly coupled with each other,so it is difficult to describe the denitrification efficiency,and the traditional BP network has some problems.A denitration efficiency prediction model based on IGWO-BP is proposed.Firstly,the dimension reduction data based on principal component analysis is used as input variables,and the improved gray wolf algorithm is used to optimize the initial weights and thresholds of BP network,and the optimized network is used to predict the denitration efficiency.The model has been successfully applied to the denitrification device of No.6 unit in Datang Luohe Power Plant.The results show that the average absolute percentage error of actual denitrification efficiency is 2.31%,which is 48.92% and 21.69% lower than that of traditional BP algorithm and IGWO-BP algorithm,respectively,with higher prediction accuracy.

关 键 词：脱硝效率 神经网络 灰狼算法 主成分分析 IGWO 

分 类 号：TM62[电气工程—电力系统及自动化] TP274[自动化与计算机技术—检测技术与自动化装置]