基于EMD-SVR的火电厂SCR脱硝系统出口NOx浓度预测研究

Prediction of NOx Concentration at the outlet of SCR Denitrification System in Thermal Power Plant Based on EMD-SVR

摘要:

针对火电厂选择性催化还原（Selective Catalytic Reduction，SCR)脱硝系统出口NOx浓度预测准确率低的问题，提出一种基于经验模态分解（Empirical Mode Decomposition，EMD）和支持向量机回归（Support Vector Machine For Regression，SVR）的火电厂脱硝系统出口NOx浓度预测模型。首先，利用EMD算法将出口NOx浓度数据序列进行分解，得到不同时间尺度下有限个本征模函数(Intrinsic Mode Function，IMF);然后引入SVR算法对NOx浓度分解数据进行建模预测;最后，将所有IMF的预测结果求和作为出口NOx浓度的最终预测值。通过对提出的EMD-SVR与标准的SVR、BP、ELM、EMD-BP和EMD-ELM模型进行对比验证，结果表明，基于EMD-SVR模型的预测精度较高，预测结果与真实值相比较，方向变化统计量（Directional Statistics,Dstat）、平均绝对百分比误差（Mean Absolute Percentage Error，MAPE）和均方根误差（Root Mean Square Error，RMSE）为0.914、1.51%和0.346mg/Nm3。

Abstract:

Aiming at the problem of low prediction accuracy of NOx concentration at the outlet of selective Catalytic Reduction (SCR) denitrification system in thermal power plants, a prediction model of NOx concentration at the outlet of denitrification system in thermal power plant based on Empirical Mode Decomposition (EMD) and Support Vector Machine for Regression (SVR) was proposed.Firstly, the empirical mode decomposition (EMD) algorithm was used to decompose the data series of NOx concentration at the outlet of denitrification system in thermal power plants, and a finite number of intrinsic mode functions (IMF) were obtained at different time scales. Then, the SVR algorithm was used to model and predict the decomposition data of NOx concentration. Finally, the prediction results of all IMFs were added up. The sum is used as the final predictor of the concentration of NOx at the outlet.By comparing the proposed EMD-SVR with standard SVR, BP, ELM, EMD-BP and EMD-ELM models, the results showed that the prediction accuracy based on EMD-SVR model was higher. Compared with the real values, the directional statistics (Dstat), mean absolute percentage error (MAPE) and root mean square error (RMSE) were 0.914, 1.51% and 0.346 mg/Nm3,respectively.