面向城市固废焚烧过程的缺失数据填充及应用  

Flling Method of Missing Data for Municipal Solid W aste Incineration Processes With Its Application

作　　者：汤健 徐雯 夏恒 乔俊飞 TANG Jian;XU Wen;XIA Heng;QIAO Junfei(Faculty of Information Technology,Bejjing Universily of Technology,Beijing 100124,China;Bejjing Laboratory of Smart Environmental Protection,Bejing 100124,China)

机构地区：[1]北京工业大学信息学部,北京100124 [2]智慧环保北京实验室,北京100124

出　　处：《北京工业大学学报》2023年第4期435-448,共14页Journal of Beijing University of Technology

基　　金：北京市自然科学基金资助项目(4212032);国家自然科学基金资助项目(62073006)。

摘　　要：针对城市固废焚烧(municipal solid waste incineration, MSWI)过程中存在的随机和连续数据缺失问题,提出了一种基于专家经验和约简特征集成模型的填充方法.首先,将过程数据缺失情况识别为随机分布、时间维度和特征维度缺失3种类型.接着,基于专家经验对前2种类型进行缺失填充后,面向第3种类型基于分布相似性和互信息相关性为缺失特征预测模型选择建模数据集和约简特征,建立具有互补特性的随机森林、梯度提升决策树和反向传播神经网络子模型对缺失值进行初步预测,利用贝叶斯线性回归(Bayesian linear regression, BLR)构建集成模型以获得最终填充值.最后,利用填充后的MSWI数据建立基于跨层全连接深度森林回归的二噁英排放浓度软测量模型.实验结果表明所提方法提高了MSWI过程数据的质量.In view of the problem of random and continuous data missing in municipal solid waste incineration(MSWI)process,a illing method of missing data based on expert experiences and ensemble model of reduced features was proposed.First,according to the lack conditions of process data,the missing data were divided into three types,i.e.,missing with random distribution,time dimension and feature dimension.Then,the former two ones were filled based on expert experience,and the distribution similarity and mutual information(MI)correlation were used to select the modeling data and reduce the input feature for the third type one.Sub-models with complementary characteristics based on random forest(RF),gradient boosting decision tree(GBDT)and back propagation neural network(BPNN)were used to predict the preliminary missing values.Furthermore,the fusion model based on Bayesian linear regression(BLR)was used to obtain the final data flling values.Finally,a soft-sensor model of dioxin(DXN)emission concentration based on deep forest regression with cross-layer full connection(DFR-clfc)was established to verify the flling effect.Results show that the proposed method improves the data quality of MSWI process.

关 键 词：城市固废焚烧(municipal solid waste incineration MSWI) 数据填充 专家经验 约简特征 集成模型 贝叶斯线性回归(Bayesian linear regression BLR) 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]