摘要
为解决燃煤电厂对选择性催化还原烟气脱硝系统(Selective Catalytic Reduction,SCR)入口NO_(x)质量浓度的测量延迟较大且准确度低的问题,针对某在役600 MW超临界机组,提出基于互信息系数的动态延迟时间分析法,分析输入特征对NO_(x)质量浓度的延迟影响;在此基础上,提出基于时序卷积网络与双向门控循环单元融合注意力机制(TCN-BiGRU-Attention)的燃煤电厂SCR入口NO_(x)质量浓度的预测方法,并引入残差修正提高模型预测效果。实验结果表明:动态延迟分析法能够捕捉输入特征的最优时序映射关系,提升模型的预测效果;与传统LSTM和TCN-GRU等模型对比,预测方法在预测精度和鲁棒性方面具有显著优势;考虑输入特征对预测结果的延迟影响并提出TCN-BiGRU-Attention与残差修正的预测方法,能够实现对SCR入口NO_(x)质量浓度的准确预测,为脱硝提供有效技术指导。
To solve the problem of large delay and low accuracy in measurement of NO_(x)mass concentration at the inlet of selective catalytic reduction(SCR)flue gas denitrification system in coal-fired power plants,based on an in-service 600 MW supercritical unit,a dynamic delay time analysis method based on mutual information coefficients was proposed to analyze the delay effect of input features on NO_(x)mass concentration;then,a prediction method for NO_(x)mass concentration at the inlet of SCR in coal-fired power plants was proposed based on the fusion of time-sequence convolutional network and bi-directionally gated recurrent unit attention mechanism(TCN-BiGRU-Attention),and the residual correction was introduced to improve the model prediction effect.The experimental results show that the proposed dynamic delay analysis method is able to capture the optimal temporal mapping relationship of the input features and improve the prediction effect of the model.The proposed prediction method has significant advantages in terms of prediction accuracy and robustness compared with traditional models such as LSTM and TCNGRU.The proposed prediction method of TCN-BiGRU-Attention with residual correction with considering the delayed effect of input features on the prediction results can realize accurate prediction of NO_(x)mass concentration at the SCR inlet and provide effective technology guidance for denitrification.
作者
鲁润旭
茅大钧
LU Runxu;MAO Dajun(College of Automation Engineering,Shanghai University of Electric Power,Shanghai,China,200090)
出处
《热能动力工程》
北大核心
2025年第4期132-142,170,共12页
Journal of Engineering for Thermal Energy and Power
基金
中国华能集团有限公司2022年度科技项目(HNKJ22-HF22)。
关键词
SCR脱硝系统
动态延迟时间
TCN
双向门控循环单元
注意力机制
残差修正
SCR denitrification system
dynamic delay time
TCN
bi-directionally gated recurrent unit(BiGRU)
attention mechanism
residual correction
作者简介
鲁润旭(2001-),男,上海电力大学硕士研究生;通信作者:茅大钧(1966-),男,上海电力大学教授.
