作者：马宁1,2, 董泽1
作者单位：1. 华北电力大学河北省发电过程仿真与优化控制工程技术研究中心, 河北 保定 071003;
2. 华北电力大学控制与计算机工程学院, 北京 102206
关键词：NOx排放;燃煤锅炉;偏最小二乘;支持向量回归
摘要：
针对影响电站燃煤锅炉NOx排放的热工变量数量较多，且存在较强的耦合相关性的问题，提出一种基于偏最小二乘（PLS）和支持向量回归（SVR）相结合的PLS-SVR建模方法用以建立燃煤锅炉NOx排放模型。PLS-SVR模型首先利用PLS提取输入变量的特征信息消除变量之间的耦合特性并降低输入变量维度，然后将提取的特征信息作为SVR模型的输入。以某超超临界1 000 MW机组锅炉为对象，选取23个相关变量作为模型的输入建立电站锅炉NOx排放模型（PLS-SVR模型），该模型对测试样本的预测误差指标MRSE为4.072。将PLS-SVR模型的预测结果和建模时间与ANN模型和未经特征提取的SVR模型对比，结果表明PLS-SCR模型具有更好高的预测精度和更快的建模速度。
Modeling of NOx emission characteristics of coal-fired boiler based on PLS-SVR
MA Ning1,2, DONG Ze1
1. Hebei Engineering Research Center of Simulation & Optimized Control for Power Generation, North China Electric Power University, Baoding 071003, China;
2. School of Control and Computer Engineering, North China Electric Power University, Beijing 102206, China
Abstract: In order to solve the problem that the number of thermal variables affecting the NOx emission of coal-fired boilers in power plants is large and there is strong coupling correlation, a PLS-SVR modeling method based on partial least squares (PLS) and support vector regression (SVR) is proposed. PLS-SVR model firstly uses PLS method to extract the feature information of input variables, eliminating the coupling between variables and reducing the dimension of input variables, and then the extracted the feature information is used as the input of SVR model. Taking a 1 000 MW ultra-supercritical unit boiler as an object, a PLS-SVR model of NOx emission from utility boiler is established by selecting 23 relevant variables as input of the model. The prediction error index MRSE of the model is 4.072. The prediction results and modeling time of PLS-SVR model are compared with ANN model and SVR model without feature extraction. The results show that PLS-SCR model has better prediction accuracy and faster modeling speed.
Keywords: NOx emission;coal-fired boiler;partial least squares;support vector regression
2019, 45(11):138-142 收稿日期: 2018-09-18;收到修改稿日期: 2018-11-29
基金项目: 国家自然科学基金（71471060）
作者简介: 马宁(1992-),男,河北迁安市人,博士研究生,研究方向为热工系统建模
参考文献
[1] 王培红, 李磊磊, 陈强, 等. 人工智能技术在电站锅炉燃烧优化中的应用研究[J]. 中国电机工程学报, 2004, 24(4):184-188
[2] 刘定平, 陈敏生, 陆继东, 等. 电站锅炉高效低污染燃烧优化控制系统设计[J]. 电力自动化设备, 2006, 26(5):46-49
[3] 牛培峰, 赵振, 马云鹏, 等. 基于风驱动算法的锅炉NO x排放模型优化[J]. 动力工程学报, 2016, 36(9):732-738
[4] 刘芳, 张德珍, 赵文杰, 等. 电站锅炉燃烧系统的神经网络建模[J]. 电力科学与工程, 2010, 6(6):33-37
[5] 周昊, 丁芳, 黄燕, 等. 核心向量机的电站锅炉NO x排放特性大数据建模[J]. 中国电机工程学报, 2016, 36(3):717-722
[6] SI F, ROMERO C E, YAO Z, et al. Optimization of coal-fired boiler SCRs based on modified support vector machine models and genetic algorithms[J]. Fuel, 2009, 88(5):806-816
[7] 王桂增, 叶昊. 主元分析与偏最小二乘法[M]. 北京:清华大学出版社, 2012.
[8] BAFFI G, MARTIN E B, MORRIS A J. Non-linear projection to latent structures revisited:the quadratic PLS algorithm[J]. Computers and Chemical Engineering, 2009, 23:395-411
[9] 刘靖旭. 支持向量回归的模型选择及应用研究[D]. 长沙:国防科学技术大学, 2006.
[10] SMREKAR, J, POTOCNIK P, SENGACNIK A. Multi-step-ahead prediction of NO x emissions for a coal-based boiler[J]. Applied Energy, 2013, 106:89-99
