基于动态相关向量回归的燃煤锅炉烟气NO_x浓度预测模型

NOx Concentration Prediction in Flue Gas of Coal-fired Boiler Based on Dynamic Relevance Vector Regression

针对NO_x的生成过程的高度非线性、强相关性以及动态特征,提出一种基于互信息和动态相关向量回归的烟气NO_x浓度预测模型。依托某660 MW燃煤锅炉的历史运行数据,建立动态相关向量回归模型。通过和相关向量回归、人工神经网络、极限学习机模型以及动态支持向量回归模型对比分析,提出的NO_x浓度预测模型动态跟踪性能好,预测准确性高,为锅炉燃烧参数调整和SCR系统动态优化提供了基础。

Aiming at the high nonlinearity,strong correlation and dynamic characteristics of the NOx generation process,a prediction model of flue gas NOx concentration based on mutual information(MI)and dynamic relevance vector regression(DRVR)is proposed in this paper.Relying on the historical operating data of a 660MW coal-fired boiler,an iterative relevance vector regression model is established.Through comparison and analysis with RVR model,ANN model,ELM model and DSVR model,the proposed NOx concentration prediction model has good dynamic tracking performance and high prediction accuracy,which provides a basis for boiler combustion parameter adjustment and SCR system optimization.