基于支持向量回归(SVR)的NO_x预测排放模型  

An NO_x Emission Prediction Model Based On Support Vector Regression

作　　者：吴也正[1] 顾钧[1] 

机构地区：[1]苏州市环境监测中心站,江苏苏州215004

出　　处：《污染防治技术》2013年第5期24-27,40,共5页Pollution Control Technology

摘　　要：探索高效而准确的PEMS计算模型,并得出较准确的NOx排放浓度预测结果。利用初始参数C,γ和训练组数据建立支持向量回归SVR模型,搜寻最佳参数C,γ,将模型预测结果与测试组数据中的CEMS实测结果进行对比。文中SVR模型的最佳参数C=84.45,γ=1。SVR训练结果与训练组数据目标值十分吻合,模型预测结果与CEMS实测数据变化趋势一致。结论 SVR模型能对固定污染源NOx排放浓度作出较高效准确的预测,可成为PEMS建模过程中较有发展前景的新方法。An effective and precise calculation model for PEMS was explored in order to predict NOx emission concentration more precisely. SVR model by initial parameter pair （C,γ） and the data of training set, were set up, the best parameter pair （C, γ） was searched, and a comparison between the predictive results obtained by SVR model and the CEMS actual monitoring results in the data of testing set was made. In the case presented in this paper, the best parameter pair was （ C = 84.45 ,γ = 1 ）. The SVR training result and the target in the data of training set were quite march, and the predictive results from the model had a same trend with the CEMS actual monitoring results. SVR model could predict the NOx emission concentrations of stationary pollution sources both effectively and prcisely. SVR model was a promising approach to build PEMS.

关 键 词：烟气预测排放监测系统 烟气连续排放监测系统 支持向量回归 NOx排放浓度预测 

分 类 号：X703[环境科学与工程—环境工程]