基于LSSVM-ARMA的电站锅炉NO_x排放量动态软测量的研究  

Research on Dynamic Soft Sensing of NO_x Emission for Utility Boiler Based on LSSVM-ARMA

作　　者：袁洪 宋选锋 赵征[1] 

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

出　　处：《电力科学与工程》2017年第12期50-54,共5页Electric Power Science and Engineering

基　　金：新能源电力系统国家重点实验室开放课题资助(LAPS16008);中央高校基本科研业务费专项资金资助项目(2017MS133)

摘　　要：建立准确的NO_x排放量模型是锅炉优化降低NO_x的基础。为提高NO_x排放量的预测精度,提出基于最小二乘支持向量机和自回归滑动平均模型的锅炉NO_x排放量动态软测量的方法。基于某电厂330 MW机组的一段历史运行数据,首先,建立最小二乘支持向量机的NO_x排放量静态软测量模型。其次,利用自回归—滑动平均方法实现对静态模型的动态校正。最后,针对2组不同样本验证LSSVM-ARMA模型和LS-SVM模型,得到2个模型的平均误差和均方根误差。结果表明:与LS-SVM模型相比,LS-SVM与ARMA相结合的模型具有更高的预测精度,对于电站锅炉NO_x排放量的预测具有一定的有效性。An accurate NO_x emission model is the cornerstone of the boiler optimization and NO_x reduction. To improve the forecast accuracy of NO_x emissions,a dynamic soft measurement of NO_x emissions is proposed,which is based on least-square support vector machine（ LS-SVM） and auto regress moving average（ ARMA）. According to the historical operation data of a 330 MW unit in some power plant,firstly,a static soft sensor model of NO_x emission based on least-square support vector machine is established. Secondly,the dynamic correction of the static model is realized by the auto regressive moving average method. Finally,the lssvm-arma model and LS-SVM model are validated through two different sets of samples,and hence the average error and root mean square error of the two models are obtained. The results show that compared with the LS-SVM,the combination of LS-SVM and ARMA has higher prediction accuracy,and it is effective for the prediction of NO_x emissions of the power plant boilers.

关 键 词：动态软测量 最小二乘支持向量机 自回归滑动平均 氮氧化物排放量 

分 类 号：TK321[动力工程及工程热物理—热能工程]