基于BP神经网络的循环流化床锅炉生成NO_x质量浓度在线软测量  

Online soft measurement of NO_x mass concentration for circulating fluidized bed boiler based on BP neural network

作　　者：白建云[1] 朱竹军 张培华 BAI Jianyun ZHU Zhujun ZHANG Peihua(Department of Automation, Shanxi University, Taiyuan 030013, China Shanxi Pingshuo Gangue Power Generation Co. , Ltd. , Shuozhou 036800, China)

机构地区：[1]山西大学自动化系,山西太原030013 [2]山西平朔煤矸石发电有限公司,山西朔州036800

出　　处：《热力发电》2016年第12期78-83,共6页Thermal Power Generation

基　　金：山西省科技攻关项目(20140313-1);山西省煤基重大项目(MD2014-03-06-03)

摘　　要：采用选择性非催化还原(SNCR)技术的循环流化床(CFB)锅炉产生的NO_x质量浓度无法直接在线测量,阻碍了NO_x排放的自动控制及经济运行。对此,本文以某300MW CFB锅炉为例,采用软测量技术,分析影响烟气中生成NO_x的主要因素为机组负荷和氧体积分数,对这2个因素分别采集现场数据,建立BP神经网络模型,并将该模型应用于新华DCS系统,实现了锅炉生成NO_x质量浓度的在线预测。预测结果表明,软测量所得NO_x质量浓度比传感器实测数据提前120s,有利于改善烟气脱硝控制系统的调节效果。In circulating fluidized bed （CFB） boilers using selective non catalytic reduction technology, the produced NOx mass concentration can not be directly measured online,which prevents automatic control of the NOx emission and the unit economic operation. Taking a 300 MW CFB boiler using soft measurement technology as the example,the main factors influencing the NO. mass concentration are unit load and oxy- gen volume fraction, field data of each factor were collected respectively,and was BP neural network model built. Applying the model to the Xinhua DCS system, thus the on-line boiler NOx mass concentration was realized,and by which the NOx mass concentration was obtained 120 s earlier than by sensor measure. So the control effect of flue gas desulfurization system is improved.

关 键 词：软测量 锅炉 循环流化床 SNCR 生成NOx质量浓度 BP神经网络 新华DCS 

分 类 号：TK39[动力工程及工程热物理—热能工程]