摘要
为了建立高精度的燃煤锅炉NOx排放量预测模型,提出一种考虑时延特征的基于深度学习的燃煤锅炉NOx排放量建模算法。首先,结合机理分析和套索算法(least absolute shrinkage and selection operator,LASSO)算法分析特征变量重要性,选取与NOx排放量最相关的变量,并进一步分析所选取变量与NOx排放量之间的时延相关性,确定模型输入变量NO_x采用经验模态分解方法对输入变量时间序列进行分解,提取其中的频域信息与时域信息,构造建模数据库;最后,设计深度神经网络结构并优化网络参数,建立NOx排放量预测模型。基于火电厂实际运行数据的实验结果表明,在多种工况下,所提出的算法预测误差均小于2%,能够满足实际生产对预测精度的要求。
In order to establish a high precision NO_x emission prediction model for coal-fired boilers, a deep learning-based NO_x emission modeling algorithm considering the time-delay characteristics was proposed. First, the importance of characteristic variables was analyzed by combining mechanism analysis and lasso algorithm, and the variables most related to NO_x emissions were selected. Furthermore, the time-delay correlation between the selected variables and NO_x information and time domain information, and construct the modeling database. Finally, the deep neural network structure and parameters were designed to build a NO_x emission prediction model. The experimental results based on the actual operation data of the thermal power plant show that the prediction error of the proposed algorithm is less than 2% under various working conditions, which can meet the requirements of actual production on the prediction accuracy.
作者
唐振浩
柴向颖
曹生现
牟中华
庞晓娅
TANG Zhenhao;CHAI Xiangying;CAO Shengxian;MU Zhonghua;PANG Xiaoya(School of Automation Engineering,Northeast Electric Power University,Jilin 132012,Jilin Province,China;Electric Power Research Institute,State Grid Gansu Electric Power Corporation,Lanzhou 730070,Gansu Province,China)
出处
《中国电机工程学报》
EI
CSCD
北大核心
2020年第20期6633-6643,共11页
Proceedings of the CSEE
基金
国家自然科学基金项目(61503072)
吉林省科技厅自然科学基金(20190201095JC,20200401085GX)。
关键词
燃煤锅炉
NOX排放
深度学习
时延
经验模态分解
coal-fired boiler
NOx emission
deep learning
time delay
empirical mode decomposition
