摘要
针对火电厂烟气NO_x质量浓度检测分析仪存在延时较长的问题,本文首先采用主元分析及多变量过程监测选出影响脱硝反应器入口烟气NO_x质量浓度的主导因素,再根据分析仪的大致延时时间Ts,选取Ts后反应器入口烟气NO_x质量浓度数据为训练信号,进而利用最小二乘支持向量机(LS-SVM)建立反应器入口烟气NO_x质量浓度的预测模型。采用某电厂历史运行数据对该预测模型进行训练及测试,结果表明:利用主元分析及多变量过程监测方法从多个辅助变量中选出主导因素,简化模型结构的同时也使预测模型具有较好的学习及泛化能力。该方法是反应器入口烟气NO_x质量浓度准确、实时预测的一种可行方法。
According to serious decay of NO_xemission analyzer used in thermal power plants,the principal component analysis and multivariable process monitoring method were employed to select the dominant factors influencing the NO_xconcentration at entrance of the denitration reactor.Then,on the basis of the analyzer＇s decay time,the NO_xcontent data of the flue gas at inlet of the reactor was selected as the pedagogic signal to establish the prediction model for NO_x concentration in flue gas at inlet of the reactor,by using the least squares support vector machine（LS-SVM）.Moreover,combing with the historical operation data of a power plant,the above model was trained and tested.The results indicate that,using the principal component analysis and multivariable process monitoring method to select the dominant factors from multiple auxiliary variables can simplify the model and make the model have a good training ability and generalization ability.This provides a feasible method for accurate and timely measurement of the NO_x content in flue gas at the reactor inlet.
出处
《热力发电》
CAS
北大核心
2016年第7期98-103,共6页
Thermal Power Generation
基金
河北省自然科学基金项目(F2014502059)
