Realization of BP neural network modeling based on NOXof CFB boiler in DCS

International Conference on Energy Engineering and Environmental Protection 2017

能源学;生态环境科学

Bai, Jianyun^1 ; Zhu, Zhujun^1 ; Wang, Qi^1 ; Ying, Jiang^1

Department of Automation, Shanxi University, Taiyuan

030013, China^1

关键词: Accurate modeling;    Ammonia injection;    Average errors;    BP neural network model;    Mass concentration;    Online prediction;    Reducing reactions;    Step disturbances;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/121/5/052025/pdf
DOI  :  10.1088/1755-1315/121/5/052025

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

In the CFB boiler installed with SNCR denitrification system, the mass concentration of NOXis difficult to be predicted by the conventional mathematical model, and the step response mathematical model, obtained by using the step disturbance test of ammonia injection,is inaccurate. this paper presents two kinds of BP neural network model, according to the relationship between the generated mass concentration of NOXand the load, the ratio of air to coal without using the SNCR system, as well as the relationship between the tested mass concentration of NOXand the load, the ratio of air to coal and the amount of ammonia using the SNCR system. then itrealized the on-line prediction of the mass concentration of NOXand the remaining mass concentration of NOXafter reductionreaction in DCS system. the practical results show that the average error per hour between generation and the prediction of the amount of NOXmass concentration is within 10 mg/Nm3,the reducing reaction of measured and predicted hourly average error is within 2 mg/Nm3, all in error range, which provides a more accurate model for solvingthe problem on NOXautomatic control of SNCR system.