燃煤机组SCR脱硝入口NO<sub>x</sub>质量浓度的多模型动态预测及协同部署

卢得勇; 韦智友; 刘宇博; 李海强; 丁得龙; 沈波; 李来

doi:10.19666/j.rlfd.202507085

欢迎来到能源电力期刊网！

English Version 期刊登录

您当前的位置：

首页 >

文章列表页 >

燃煤机组SCR脱硝入口NO_x质量浓度的多模型动态预测及协同部署

发电技术论坛 | 更新时间：2026-06-12

- 燃煤机组SCR脱硝入口NO_x质量浓度的多模型动态预测及协同部署
- 燃煤机组SCR脱硝入口NO_x质量浓度的多模型动态预测及协同部署
- 热力发电 2026年55卷第4期页码：156-165
- 作者机构：
  
  1. 浙江浙能兰溪发电有限责任公司
  2. 浙江浙能科技环保集团股份有限公司
- 作者简介：
- 基金信息：
  
  浙江省能源集团有限公司科学技术项目（ZNKJ-2025-034）~~
- DOI：10.19666/j.rlfd.202507085
  中图分类号： TM621;X773
- 纸质出版：2026
- 稿件说明：
移动端阅览
[1]卢得勇,韦智友,刘宇博,等.燃煤机组SCR脱硝入口NO_x质量浓度的多模型动态预测及协同部署[J].热力发电,2026,55(04):156-165.
[1]卢得勇,韦智友,刘宇博,等.燃煤机组SCR脱硝入口NO_x质量浓度的多模型动态预测及协同部署[J].热力发电,2026,55(04):156-165. DOI： 10.19666/j.rlfd.202507085.

DOI：10.19666/j.rlfd.202507085.

摘要

为满足燃煤机组选择性催化还原(selective catalytic reduction

SCR)脱硝系统对氮氧化物(NO

)质量浓度预测的高需求，设计并提出一种兼具高精度和低延迟的软测量及部署方案。基于某660 MW燃煤机组11万余组高维运行数据，在统一平台下系统对比了深度学习(deep learning

DL)与XGBoost(extreme gradient boosting)2种模型，采用时序交叉验证结合网格搜索优化模型超参数，并从预测精度、计算效率及可解释性三方面进行综合评估。在此基础上，提出“边缘–嵌入式”协同部署策略，将DL模型部署于边缘服务器以获取高精度预测结果，XGBoost模型嵌入分散控制系统(distributed control system

DCS)保障实时性。结果表明：DL模型在动态响应与预测精度上表现更优，其测试集均方根误差较XGBoost模型降低约10%，并在工况剧烈波动时DL模型能保持稳定性；变量重要性分析显示，烟气含氧量、燃烧器壁温和总风量是影响NO

生成的主要因素。所提协同部署架构理论上可实现毫秒级推理，并具备离线容错能力。

Abstract

To meet the high requirements of selective catalytic reduction(SCR) systems in coal-fired power plants for accurate and low-latency prediction of nitrogen oxides(NO

) mass concentrations

this study designs and proposes a soft sensing and deployment framework that balances high accuracy and real-time performance. Using more than 110 000 sets of high-dimensional operational data from a 660 MW coal-fired unit

a systematic comparison of deep learning(DL) and XGBoost models was conducted on a unified platform. Time series crossvalidation combined with grid search was employed to optimize hyperparameters

and model performance was comprehensively evaluate

d in terms of predictive accuracy

computational efficiency

and interpretability via local interpretable model-agnostic explanations. On this basis

an “edge-embedded” collaborative deployment strategy was proposed

in which the DL model is deployed on edge servers to deliver high-accuracy predictions

while the XGBoost model is embedded into the distributed control system(DCS) to ensure real-time responsiveness. The results show that the DL model outperforms XGBoost in dynamic response and predictive accuracy

achieving root mean square errors approximately 10% lower than that of the XGBoost

and maintaining stability under highly fluctuating conditions. Variable importance analysis highlights flue gas oxygen content

burner wall temperature

and total air volume as the dominant factors affecting NO

formation. The proposed collaborative architecture can theoretically achieve millisecond-level inference and provide offline fault tolerance

offering a practical pathway for intelligent ammonia injection control and combustion optimization.

关键词

Keywords

references

BONINGARI T, SMIRNIOTIS P G. Impact of nitrogen oxides on the environment and human health:Mn-based materials for the NOx abatement[J]. Current Opinion in Chemical Engineering, 2016, 13:133-141.

MURRAY L T. Lightning NOx and impacts on air quality[J]. Current Pollution Reports, 2016, 2:115-133.

WILSON R. Air pollution, the automobile, and public health[J]. Environment:Science and Policy for Sustainable Development, 1989, 31(4):25-27.

YANG S, LI M, GUO C, et al. Associations of long-term exposure to nitrogen oxides with all-cause and causespecific mortality[J]. Nature Communications, 2025,16(1):1730.

YAMASHITA R, NISHIO M, DO R K G, et al.Convolutional neural networks:an overview and application in radiology[J]. Insights Into Imaging, 2018,9:611-629.

MANISALIDIS I, STAVROPOULOU E, STAVROPOULOS A, et al. Environmental and health impacts of air pollution:a review[J]. Frontiers in Public Health, 2020,8:14.

CHEN T M, KUSCHNER W G, GOKHALE J, et al.Outdoor air pollution:nitrogen dioxide, sulfur dioxide,and carbon monoxide health effects[J]. The American Journal of the Medical Sciences, 2007, 333(4):249-256.

HE K, DING H. Prediction of NOx emissions in thermal power plants using a dynamic soft sensor based on random forest and just-in-time learning methods[J]. Sensors,2024, 24(14):4442.

LV Y, LIU J, YANG T, et al. A novel least squares support vector machine ensemble model for NOx emission prediction of a coal-fired boiler[J]. Energy, 2013, 55:319-329.

SONG M, XUE J, GAO S, et al. Prediction of NOx concentration at SCR inlet based on BMIFS-LSTM[J].Atmosphere, 2022, 13(5):686.

WU Z, ZHANG Y, DONG Z. NOx concentration prediction based on multi-channel fused spectral temporal graph neural network in coal-fired power plants[J].Energy, 2024, 305:132222.

TANG Z, WANG S, CHAI X, et al. Auto-encoderextreme learning machine model for boiler NOx emission concentration prediction[J]. Energy, 2022, 256:124552.

XIE P, GAO M, ZHANG H, et al. Dynamic modeling for NOx emission sequence prediction of SCR system outlet based on sequence to sequence long short-term memory network[J]. Energy, 2020, 190:116482.

LI N, LV Y, HU Y. Prediction of NOx emissions from a coal-fired boiler based on convolutional neural networks with a channel attention mechanism[J]. Energies, 2023,16(1):76.

GOODFELLOW I, BENGIO Y, COURVILLE A, et al.Deep learning[M]. Cambridge:MIT Press, 2016:1.

WEN X, LI K, WANG J. NOx emission predicting for coal-fired boilers based on ensemble learning methods and optimized base learners[J]. Energy, 2023, 264:126171.

YUAN Z, MENG L, GU X, et al. Prediction of NOx emissions for coal-fired power plants with stackedgeneralization ensemble method[J]. Fuel, 2021, 289:119748.

王印松,陈瑞杰.基于Elastic Net特征变量选择的SCR入口NOx软测量模型[J].中国测试, 2021, 47(12):79-86.WANG Yinsong, CHEN Ruijie. Soft sensor model of SCR inlet NOx based on Elastic Net feature variable selection[J]. China Measurement&Test, 2021, 47(12):79-86.

于静,金秀章,刘岳.基于结构改进RBF神经网络的NOx预测模型比较[J].控制工程, 2023, 30(9):1616-1623.YU Jing, JIN Xiuzhang, LIU Yue. Comparison of NOx prediction models based on structurally improved RBF neural network[J]. Control Engineering, 2023, 30(9):1616-1623.

金秀章,于静,刘岳.基于人工鱼群-径向基神经网络的NOx预测模型[J].动力工程学报, 2021, 41(7):551-557.JIN Xiuzhang, YU Jing, LIU Yue. NOx prediction model based on artificial fish swarm-RBF neural network[J].Journal of Power Engineering, 2021, 41(7):551-557.

FRIEDMAN J H. Greedy function approximation:a gradient boosting machine[J]. Annals of Statistics, 2001:1189-1232.

GARREAU D, LUXBURG U. Explaining the explainer:a first theoretical analysis of LIME[C]. International Conference on Artificial Intelligence and Statistics.PMLR, 2020:1287-1296.

RIBEIRO M T, SINGH S, GUESTRIN C.“Why should I trust you?” Explaining the predictions of any classifier[C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.2016:1135-1144.

LUNDBERG S M, ERION G, CHEN H, et al. From local explanations to global understanding with explainable AI for trees[J]. Nature Machine Intelligence, 2020, 2(1):56-67.

CARTER A, IMTIAZ S, NATERER G F. Review of interpretable machine learning for process industries[J].Process Safety and Environmental Protection, 2023, 170:647-659.

LIANG Y C, MAIMURY Y, CHEN A H L, et al. Machine learning-based prediction of air quality[J]. Applied Sciences, 2020, 10(24):9151.

MÉNDEZ M, MERAYO M G, NÚÑEZ M. Machine learning algorithms to forecast air quality:a survey[J].Artificial Intelligence Review, 2023, 56(9):10031-10066.

固定污染源烟气(SO2、NO、颗粒物)排放连续监测技术规范:HJ 75—2017[S].北京:中国环境科学出版社,2018:1.Technical specification for continuous emissions monitoring of flue gas from stationary pollution sources(SO2, NOx, particulate matter):HJ 75—2017[S]. Beijing:China Environmental Science Press, 2018:1.

TAN P, HE B, ZHANG C, et al. Dynamic modeling of NOx emission in a 660 MW coal-fired boiler with long short-term memory[J]. Energy, 2019, 176:429-436.

WANG Z, PENG X, CAO S, et al. NOx emission prediction using a lightweight convolutional neural network for cleaner production in a down-fired boiler[J].Journal of Cleaner Production, 2023, 389:136060.

PAPAGIANNAKIS R G, RAKOPOULOS D C,RAKOPOULOS C D. Evaluation of the air oxygen enrichment effects on combustion and emissions of natural gas/diesel dual-fuel engines at various loads and pilot fuel quantities[J]. Energies, 2018, 11(11):3028.

IAVARONE S, PARENTE A. NOx formation in MILD combustion:potential and limitations of existing approaches in CFD[J]. Frontiers in Mechanical Engineering, 2020, 6:13.

陈东东,伍林玲,苑梦兰,等.水泥窑低氮分级燃烧下脱硝工程技术的研究综述[J].化学工程与技术,2023,13(4):252.CHEN Dongdong, WU Linling, YUAN Menglan, et al.Review of denitrification engineering technology under low-nitrogen staged combustion in cement kilns[J].Journal of Chemical Engineering and Technology, 2023,13(4):252.

ALQAHTANI D K, CHEEMA M A, TOOSI A N.Benchmarking deep learning models for object detection on edge computing devices[C]//Proceedings of the International Conference on Service-Oriented Computing.Singapore:Springer Nature Singapore, 2024:142-150.

LU L, CAO Z, CHEN X, et al. Hybrid precision gradient accumulation for CNN-LSTM in sports venue buildings analytics:energy-efficient spatiotemporal modeling[J].Buildings, 2025, 15(16):2926.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

深度学习算法在光伏电站无人机智能运维中应用

机器学习在电站设备状态分析中的应用

基于模型与数据混合驱动的燃气轮机气路故障诊断方法

基于深度学习及混沌优化的燃机电站机组热电负荷优化分配

燃气轮机性能监测与故障预警系统开发及工程应用研究

燃煤机组SCR脱硝入口NOx质量浓度的多模型动态预测及协同部署

燃煤机组SCR脱硝入口NOx质量浓度的多模型动态预测及协同部署

DOI：10.19666/j.rlfd.202507085

摘要

Abstract

关键词

Keywords

references

燃煤机组SCR脱硝入口NO_x质量浓度的多模型动态预测及协同部署

燃煤机组SCR脱硝入口NO_x质量浓度的多模型动态预测及协同部署