电站锅炉热效率与NOx排放响应特性建模方法  

Response Characteristics Modeling of Efficiency and NO_x Emission for Power Station Boiler

作　　者：赵欢[1] 王培红[1] 陆璐[2] 

机构地区：[1]东南大学能源与环境学院,江苏省南京市210096 [2]徐州空军学院航空四站系,江苏省徐州市221000

出　　处：《中国电机工程学报》2008年第32期96-100,共5页Proceedings of the CSEE

基　　金：国家自然科学基金项目(50370611);高等等学校博士学科点专项科研基金(20060286033)~~

摘　　要：为了解决电站锅炉高效低污染的优化决策问题,建立了基于核主元分析支持向量回归机(kernel principle component analysis ε-support vector regression,KPCA-ε-SVR)与机理模型混合的锅炉热效率和NOx排放特性响应模型。在建模的过程中,针对模型输入变量之间存在非线性、强耦合等特点,采用核主元分析提取输入变量的主元,去除变量之间的相关性;同时采用5-fold交叉验证方法,循环搜索寻优模型的各个参数,确定输入主元个数。该模型与BP神经网络(back propagation neural-networks,BPNN)和支持向量机模型相比较具有良好的泛化能力。In order to improve the efficiency and to reduce NOx emission in combustion, a mixed model of the coal-fired boiler was studied by using kernel principle component analysis （KPCA）, e-support vector regression （ε-SVR） and function-type model. In the process of modeling, principal components of the original inputs were extracted using KPCA for eliminating the nonlinear relationship and strong-coupling among original inputs, and then the 5-fold cross validation method was used to search the optimal free parameters of the model and select the optimal number of principle components. Finally, the mixed model was compared with back propagation neural-networks（BPNN） and ε-SVR model respectively; it is shown that the mixed response characteristics model exceeds the other two models and has powerful prediction precision and generalization.

关 键 词：高效低污染 燃烧优化 核主元分析 支持向量回归机 5-fold交叉验证 

分 类 号：TK31[动力工程及工程热物理—热能工程]