基于多模型聚类集成的锅炉烟气NO_x排放量预测模型

Prediction model of NO_x emission from coal-fired boiler based on multi-model clustering ensemble

电站锅炉烟气NO_x排放量的预测控制对电站的经济效益和环境污染治理有重要影响。为了提高NO_x排放量预测模型的精度,本文提出了一种基于多模型聚类集成的锅炉烟气NO_x排放量建模方法。首先根据输出NO_x排放量的高低划分数据空间,通过基于相关性分析的变量权重和基于信息熵的分层聚类确定参与聚类的变量,然后利用提出的多模型聚类集成(VMSC)算法聚类得到各子空间的隶属度矩阵,最后采用融合隶属度的最小二乘法对各子空间的最小二乘支持向量机(LS-SVM)模型进行集成。仿真结果表明,通过集成模糊C均值聚类(FCM)和有监督的遗传算法-软模糊聚类(GA-SFCM)的VMSC算法提高了建模的精度,比单一模型的仿真性能更好。

Predictive control o fNOX emission from flue gas o f utility boilers has im portant influence on power plants' economic benefits and environmental pollution control. To improve the accuracy o f theNOX emission prediction model, a m odeling method o f boilerNOX emission based on m ulti-m odel clustering ensemble was proposed. In this method, the data space is firstly divided according to the level o f NOX emission, and the variables that participate in clustering are determ ined by using the variable weight based on relevant analysis and hierarchical clustering utilized information entropy. Then, the proposed algorithm VM SC is used to obtain the new m embership degree m atrix o f each subspace. Finally, the m ultiple least squares support vector machine (LS-SVM ) model o f each subspace is integrated by the least-squares m ethod fused membership degree. The simulation results show that, the VM SC algorithm integrating the soft fuzzy C-means clustering (SFCM ) w ith the genetic algorithm-soft fuzzy clustering (GA-SFCM ) improves the accuracy o f the clustering, and the simulation perform ance is better than the single model.