基于互信息和PSO-LSSVM的脱硝反应器入口NO_x浓度预测  

Prediction of NO_x concentration at the entrance of denitration reactor based on mutual information and PSO-LSSVM

作　　者：刘延泉[1] 杨堃 王如蓓[1,2] LIU Yanquan;YANG Kun;WANG Rubei(School of Control and Computer Engineering,North China Electric Power University,Baoding 071003,China;Hebei Power Generation Process Simulation and Optimization Control Engineering TechnologyResearch Center,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003 [2]河北省发电过程仿真与优化控制工程技术研究中心(华北电力大学),河北保定071003

出　　处：《电力科学与工程》2018年第9期67-72,共6页Electric Power Science and Engineering

摘　　要：针对火电机组烟气排放连续监测系统对NO_x气体测量滞后的问题,将互信息变量选择和通过粒子群算法优化参数的最小二乘支持向量机模型相结合,建立了脱硝反应器入口NO_x浓度的预测模型。影响NO_x生成的数据类型繁多,变量之间存在很强的耦合性,而使用互信息对变量进行筛选,在考虑到主导变量与辅助变量间最大的相关性的同时,降低了选取的辅助变量间的冗余性。将互信息法筛选出来的辅助变量作为PSO-LSSVM模型的输入,经Matlab仿真模拟,并将此模型与未经互信息进行变量筛选的PSO-LSSVM模型进行比较,表明该模型可以较好地实现对火电机组脱硝反应器入口NO_x浓度的预估,同时为得到SCR脱硝反应器的最佳喷氨量打下了良好的基础。Aiming at the problem of NOx gas measurement lag in the continuous monitoring system of flue gas emission in thermal power unit,this paper combines the mutual information variable selection and the least squares support vector machine model which is optimized by particle swarm optimization,and establishes a prediction model for the concentration of NOx in the denitrification reactor entrance.There are many types of data that affect the NOx generation,and there exists the strong coupling between variables. In view of this,the application of mutual information to filter variables reduces the redundancy of the selected auxiliary variables,which takes into account the maximum correlation between the dominant variable and the auxiliary variable. The auxiliary variables selected by the mutual information method are used as the input of the PSO-LSSVM model. The model is simulated in MATLAB,and the model is compared with the PSO-LSSVM model without mutual information for variable selection. It shows that the model can well realize the prediction of NOx concentration at the inlet of the denitrification reactor of the thermal power unit,and it lays a solid foundation for the acquisition of the optimum ammonia injection amount of the SCR denitrification reactor.

关 键 词：相似度函数 互信息 最小二乘支持向量机 粒子群优化 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]