Study on denitration optimization control model based on CNN-LSTM algorithm

doi:10.3969/j.issn.2097-0706.2023.06.004

Abstract

Abstract:

To achieve the "double carbon" target，most thermal power plants have been optimizing their units to meet the new requirements of deep peak shaving.A 2×330 MW power plant adopted selective catalyst reduction flue gas denitration process，but this emission control model can no longer satisfy the requirements of denitration under new operating conditions in view of the aging of equipment，transformation of the system，shortening of catalyst service life，volatile coal quality and requirement of deep peak shaving.A production data optimization and analysis platform for denitration is established based on Hadoop technology，and the production data middle platform is built based on NO_x data from combustion and flue gas treatment.Then，the feature extraction is carried out on the data sequence affecting the NO_x mass concentration through CNN algorithm，and the data mining is executed on the characteristic parameters affecting the denitration control.Making prediction on NO_x emission by LSTM algorithm，and training and updating the CNN-LSTM neural network model on demands can realize closed-loop optimal control on denitration process.

Key words: flue gas denitration, Hadoop, CNN, LSTM, data governance, grey relational analysis, dual-carbon target, deep peak regulation

CLC Number:

TK09

WANG Yonglin, BAI Yongfeng, KONG Xiangshan, HAO Zheng, YANG Pengfei, KONG Dewei. Study on denitration optimization control model based on CNN-LSTM algorithm[J]. Integrated Intelligent Energy, 2023, 45(6): 25-33.

Figures/Tables 18

Fig.1

Fig.2

Fig.3

Fig.4

Table 1

Table 2

Fig.5

Fig.6

Fig.7

Table 3

Table 4

Fig.8

Table 5

Fig.9

Fig.10

Table 6

Fig.11

Fig.12

References 26

[1]	胡小夫, 王云, 郝正, 等. 燃煤电站SCR催化剂寿命在线预测研究[J]. 华电技术, 2021, 43(3):65-69.
	HU Xiaofu, WANG Yun, HAO Zheng, et al. Research on SCR catalyst service life on-line prediction in coal-fired power plants[J]. Huadian Technology, 2021, 43(3):65-69.
[2]	曹喜果, 张永涛, 李雅恬. 基于IQGA-GRNN模型的SCR脱硝出口NO_x质量浓度预测方法[J]. 华电技术, 2021, 43(5):9-14.
	CAO Xiguo, ZHANG Yongtao, LI Yatian. Prediction method for NO_x discharged from SCR denitrification systems based on IQGA-GRNN model[J]. Huadian Technology, 2021, 43(5):9-14.
[3]	赵大周, 王明祥, 阮慧锋, 等. 分布式能源站天然气内燃机Urea-SCR系统模拟优化研究[J]. 华电技术, 2021, 43(5):45-52.
	ZHAO Dazhou, WANG Mingxiang, RUAN Huifeng, et al. Simulation and optimization for Urea-SCR system of the natural gas internal combustion engine in a distributed energy station[J]. Huadian Technology, 2021, 43(5):45-52.
[4]	孟范伟, 徐博, 吕晓永, 等. 神经网络预测控制在SCR烟气脱硝系统中应用[J]. 东北大学学报(自然科学版), 2017, 38(6):778-782. doi: 10.12068/j.issn.1005-3026.2017.06.004
	MENG Fanwei, XU Bo, LYU Xiaoyong, et al. Application of neural network predictive control in SCR flue gas denitration system[J]. Journal of Northeastern University(Natural Science Edition), 2017, 38(6):778-782.
[5]	黄镇柯, 金志远, 朱恒毅, 等. 基于多层感知器神经网络的燃煤电站SCR脱硝系统建模[J]. 湖北电力, 2022, 46(3):100-105.
	HUANG Zhenke, JIN Zhiyuan, ZHU Hengyi, et al. Modeling of SCR denitrification system in coal-fired power station based on multilayer perceptron neural network[J]. Hubei Electric Power, 2022, 46(3):100-105.
[6]	王林, 高林, 郭亦文, 等. 离散粒子群寻优径向基神经网络模型在脱硝控制中应用[J]. 热力发电, 2019, 48(10):22-27,37.
	WANG Lin, GAO Lin, GUO Yiwen, et al. Application of RBF neural network based on discrete particle swarm optimization in denitrification control[J]. Thermal Power Generation, 2019, 48(10):22-27,37.
[7]	王天堃. 基于神经网络模型及预测控制DMC的火电机组脱硝控制策略[J]. 中国电力, 2019, 52(12):140-145.
	WANG Tiankun. The denitration strategy of predictive dynamic matrix control (DMC) combined with BP neural network in fossil fuel power plant[J]. Electric Power, 2019, 52(12):140-145.
[8]	陈静, 洪乔, 姜媛媛. 基于改进动态矩阵算法的SCR烟气脱硝系统[J]. 哈尔滨商业大学学报(自然科学版), 2022, 38(2):171-176.
	CHEN Jing, HONG Qiao, JIANG Yuanyuan. Study on SCR denitrification system based on improved dynamic matrix algorithm[J]. Journal of Harbin University of Commerce (Natural Sciences Edition), 2022, 38(2):171-176.
[9]	陈皓炜, 贾新春, 孙小明, 等. SCR脱硝系统的强化学习复合串级控制[J]. 动力工程学报, 2022, 42(5):421-428. doi: 10.19805/j.cnki.jcspe.2022.05.005
	CHEN Haowei, JIA Xinchun, SUN Xiaoming, et al. Composite cascade control based on reinforcement learning for a SCR denitration system[J]. Journal of Chinese Society of Power Engineering, 2022, 42(5):421-428.
[10]	江顺, 陈荣宇, 林伟君, 等. 基于大数据的智能农业云服务平台设计与实现[J]. 安徽农业科学, 2022, 50(16):190-197.
	JIANG Shun, CHEN Rongyu, LIN Weijun, et al. Design and implementation of intelligent agricultural cloud service platform based on big data technology[J]. Journal of Anhui Agricultural Sciences, 2022, 50(16):190-197.
[11]	王新柱, 赵杰. 基于Hadoop生态系统的电网量测数据共享技术研究[J]. 电气传动自动化, 2022, 44(1):40-43.
	WANG Xinzhu, ZHAO Jie. Research on power grid measurement data sharing technology based on Hadoop ecosystem[J]. Electric Drive Automation, 2022, 44(1):40-43.
[12]	陈行滨, 王周, 郑飘飘, 等. 基于分布式文件系统电力大数据存储实现[J]. 粘接, 2022, 49(6):172-175.
	CHEN Xingbin, WANG Zhou, ZHENG Piaopiao, et al. Realization of power big data storage based on distributed file system[J]. Adhesion, 2022, 49(6):172-175.
[13]	李娟. 基于Hadoop云平台的空间属性数据挖掘技术研究[J]. 南京理工大学学报(自然科学版), 2022, 46(4):419-426.
	LI Juan. Research on spatial attribute data mining technology based on hadoop cloud platform[J]. Journal of Nanjing University of Science and Technology(Natural Sciences Edition), 2022, 46(4):419-426.
[14]	唐建海. 基于Hadoop平台的网络安全趋势大数据挖掘算法[J]. 工业加热, 2022, 51(7):67-70.
	TANG Jianhai. Big data mining algorithm of network security trend based on Hadoop platform[J]. Industrial Heating, 2022, 51(7):67-70.
[15]	刘岳, 于静, 金秀章. 基于特征优化和改进长短期记忆神经网络的NO_x质量浓度预测[J]. 热力发电, 2021, 50(7):162-169.
	LIU Yue, YU Jing, JIN Xiuzhang. NO_x mass concentration prediction based on Feature optimization and improved LSTM network[J]. Thermal Power Generation, 2021, 50(7):162-169.
[16]	方鹏, 高亚栋, 潘国兵, 等. 基于LSTM神经网络的中长期光伏电站发电量预测方法研究[J]. 可再生能源, 2022, 40(1):48-54.
	FANG Peng, GAO Yangdong, PAN Guobing, et al. Research on forecasting method of mid- and long-term photovoltaic power generation based on LSTM neural network[J]. Renewable Energy Resources, 2022, 40(1):48-54.
[17]	郑必君, 金宏伟, 孟瑜炜, 等. 改进型LSTM实现燃煤电厂脱硝预警及优化[J]. 能源工程, 2021(2):53-60.
	ZHENG Bijun, JIN Hongwei, MENG Yuwei, et al. Realization of early warning and optimization of denitrification in coal-fired power plants by improved LSTM realizes[J]. Energy Engineering, 2021(2):53-60.
[18]	姚宁, 金秀章, 李阳峰. 基于改进鲸鱼算法优化Bi-LSTM的脱硝系统NO_x建模[J]. 华北电力大学学报(自然科学版), 2022, 49(6):76-83.
	YAO Ning, JIN Xiuzhang, LI Yangfeng. NO_x Modeling of denitrification system based on Bi-LSTM optimized by improved whale algorithm[J]. Journal of North China Electric Power University(Natural Science Edition), 2022, 49(6):76-83.
[19]	温鑫, 钱玉良, 彭道刚, 等. 基于深度双向LSTM的SCR系统NO_x排放预测模型研究[J]. 热能动力工程, 2020, 35(10):57-64.
	WEN Xin, QIAN Yuliang, PENG Daogang, et al. NO_x emission prediction model of SCR system based on deep bidirectional LSTM[J]. Journal of Engineering for Thermal Energy and Power, 2020, 35(10):57-64.
[20]	简定辉, 李萍, 黄宇航, 等. 基于CNN-LSTM网络的短期电力负荷预测[J]. 电工电气, 2022(9):1-6.
	JIAN Dinghui, LI Ping, HUANG Yuhang, et al. Short-term power load forecasting based on CNN-LSTM[J]. Electrotechnics Electric, 2022(9):1-6.
[21]	黄治军. 基于EEMD和卷积神经网络的电站NO_x排放预测[J]. 热能动力工程, 2022, 37(10):96-103.
	HUANG Zhijun. Prediction of NO_x emissions from power plants based on EEMD and convolutional neural network[J]. Journal of Engineering for Thermal Energy and Power, 2022, 37(10):96-103.
[22]	KARPATHY A, TODERICI G, SHETTY S, et al. Large-scale video classification with convolutional neural networks[C]// Proceedings of the IEEE conference on Computer Vision and Pattern Recognition. 2014:1725-1732.
[23]	TAKAHASHI R, MATSUBARA T, UEHARA K. A novel weight-shared multi-stage CNN for scale robustness[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2018, 29(4):1090-1101. doi: 10.1109/TCSVT.76
[24]	GREFF K, SRIVASTAVA R K, KOUTNIK J, et al. LSTM:A search space odyssey[J]. IEEE Transactions on Neural Networks and Learning Systems, 2017, 28(10):2222-2232. doi: 10.1109/TNNLS.2016.2582924
[25]	韩金磊, 熊萍萍, 孙继红. 基于LSTM和灰色模型的股价时间序列预测研究[J/OL]. 南京信息工程大学学报(自然科学版):1-22(2023-01-06)[2023-03-05]. http://kns.cnki.net/kcms/detail/32.1801.N.20230105.1635.003.html.
	HAN Jinlei, XIONG Pingping, SUN Jihong. Stock price time series prediction based on LSTM and grey model[J/OL]. Journal of Nanjing University of Information Science & Technology(Natural Science Edition):1-22(2023-01-06)[2023-03-05]. http://kns.cnki.net/kcms/detail/32.1801.N.20230105.1635.003.html.
[26]	SRIVASTAVA N, HINTON G, KRIZHEVSKY A, et al. Dropout:A simple way to prevent neural networks from overfitting[J]. The Journal of Machine Learning Research, 2014, 15(1):1929-1958.

定义	长度	描述
wReserved1	word	保留字
wReserved2	word	保留字
wDstNode	word	目的节点号，总为0xffff
wSrcNode	word	源节点号，表示发出信息的节点号
wLen	word	数据包长度，即byData
wType	word	包类型，见下面具体定义
wReserved3	word	保留字
wReserved4	word	保留字
byData[]	byte	数据部分，最大为496

定义	长度	描述
wGID	word	模拟量点的全局标识
wStatus	word	模拟量点的状态
fValue	float	实时值

Kernel	MSE	RMSE	MAE	R2_Score
3	2 182.346	46.716	35.922	0.771
5	2 568.838	50.684	41.732	0.655
7	2 576.705	50.761	41.887	0.653

项目	配置
卷积核(Kernel)	3
卷积层(Convolution Layer)	4×64
最大池化层(Max_Pooling Layer)	1×5
双向LSTM层(Bi-LSTM Layer)	64
Dropout Layer	0.1
全连接层(Liner Layer)	Input:(64×2)，Output:1

模型	MSE	RMSE	MAE
CNN-LSTM	0.057 91	2.273 19	1.578 69
Bi-LSTM	0.063 06	2.300 22	1.600 48
LSTM	0.176 78	3.583 63	2.650 75