基于小波去噪和PCA-ANFIS的SCR脱硝系统建模  

Modeling of SCR denitrification system based on wavelet denoising and PCA-ANFIS

作　　者：张晓雯 向文国[1] 陈时熠[1] 刘全军 徐龙飞 ZHANG Xiaowen;XIANG Wenguo;CHEN Shiyi;LIU Quanjun;XU Longfei(Key Laboratory of Energy Thermal Conversion and Control of Ministry of Education,Southeast University,Nanjing 210096,China;Jiangsu Xinhai Power Generation Co.,Ltd.,Lianyungang 222023,China)

机构地区：[1]东南大学能源热转换及过程测控教育部重点实验室,江苏南京210096 [2]江苏新海发电有限公司,江苏连云港222023

出　　处：《热力发电》2021年第6期114-120,共7页Thermal Power Generation

基　　金：国家科技重大专项(2017-I-0002-0002)。

摘　　要：针对选择性催化还原(SCR)脱硝系统非线性、时变和大滞后的特点,本文提出基于小波去噪和主成分分析、自适应神经模糊推理系统(PCA-ANFIS)而建立的SCR脱硝系统预测模型。通过分析不同阈值选取原则及不同小波基和分解层数的去噪效果,选取最适合系统数据去噪的rigrsure原则、软阈值函数、Sym10小波3层分解方式,对数据进行去噪处理,并利用主成分分析法进行数据降维。然后基于减法聚类构建ANFIS模型的初始网络结构,采用混合算法优化模型参数。最后利用某燃煤机组实际运行数据对模型进行验证,并与BP神经网络模型预测结果进行对比。结果表明,基于小波去噪和PCA-ANFIS的SCR脱硝系统模型具有较好的拟合精度和泛化能力。Against the non-linear,time-varying and time-delayed characteristics of selective catalytic reduction(SCR)denitrification system,this paper established a predictive model of the SCR denitrification system based on wavelet denoising,principal component analysis and adaptive neuro-fuzzy inference system(ANFIS).By analyzing the results of different threshold selection principles,wavelet bases and decomposition levels,the most suitable denoising method for the system data signals was selected,and the principal component analysis method was adopted for data dimensionality reduction.Then,the initial network structure of the ANFIS model was constructed based on subtractive clustering,and the model parameters were optimized by a hybrid algorithm.Finally,the model was verified by the operational data of a coal-fired power plant,and the results were compared with the BP neural network model.The results showed that the ANFIS denitrification system model based on wavelet denoising and principal component analysis has higher fitting accuracy and better generalization ability.

关 键 词：SCR脱硝 小波去噪 主成分分析 ANFIS 减法聚类 预测控制 

分 类 号：X773[环境科学与工程—环境工程]