基于现场监测和人工神经网络的填埋场二氧化碳释放浓度预测分析  

Analysis of Carbon Dioxide Concentration in Landfill Based on Field Monitoring and Artificial Neural Network

作　　者：谢海建[1,2] 彭滢霏 石阳辉 王巧 严华祥 左欣茹 陈赟[2,3] Xie Haijian;Peng Yingfei;Shi Yanghui;Wang Qiao;Yan Huaxiang;Zuo Xinru;Chen Yun(College of Civil Engineering and Architecture,Zhejiang University,Hangzhou 310058,China;Center for Balance Architecture,Zhejiang University,Hangzhou 310028,China;The Architectural Design&Research Institute of Zhejiang University Co.,Ltd.,Hangzhou 310028,China;College of Resources and Environmental Engineering,Hefei University of Technology,Hefei 230009,China)

机构地区：[1]浙江大学建筑工程学院,杭州310058 [2]浙江大学平衡建筑研究中心,杭州310028 [3]浙江大学建筑设计研究院有限公司,杭州310028 [4]合肥工业大学资源与环境工程学院,合肥230009

出　　处：《科技通报》2024年第2期101-107,共7页Bulletin of Science and Technology

基　　金：国家自然科学基金(52278375,41977223,41931289,41672288)。

摘　　要：垃圾填埋场会产生大量的填埋气,其中40%~60%为CO_(2),CO_(2)会导致温室效应并降低填埋气的利用效率,因此掌握及预测填埋气中CO_(2)的释放规律对于控制其扩散及填埋场温室气体排放评估具有重要意义。本文分析了杭州某填埋场2018年全年CO_(2)浓度,发现CO_(2)与H2S的浓度在春秋季表现出强线性相关性,秋季的皮尔逊相关性系数绝对值为0.77。同时,建立多层感知器(multi layerperceptron,MLP)的人工神经网络模型预测CO_(2)浓度,并选取箱型图法对数据进行前处理以剔除监测设备故障等非自然因素导致的异常值,选取PM_(2.5)、风速、风向、气温、空气湿度作为输入指标,结果表明模型预测结果与现场实测结果约为R=0.7,说明模型效果较好。基于该模型对缺失的9月份数据进行填补,其结果与CO_(2)全年释放规律吻合良好。本文基于现场监测数据建立的神经网络模型可用于填埋场CO_(2)释放的预测评估,对CO_(2)等填埋气释放的控制和填埋场现场管理具有指导意义。Landfills produce a large amount of landfill gas,40%-60%of which is CO_(2).CO_(2) causes greenhouse effect and reduces the efficiency of landfill gas utilization.Understanding the emission mechanisms of CO_(2) from the landfill is key to control its dispersion and release.This study inverstigated the CO_(2) concentration in 2018 of a landfill in Hang-zhou.The concentration of CO_(2) and H2S showed a strong linear correlation in spring and autumn,with the absolute val-ue of the Pearson correlation coefficient of 0.77 in autumn.Meanwhile,an artificial neural network model based on the multi layer perceptron(MLP)was developed to predict and analyze the concentration of CO_(2).In order to eliminate the abnormal values,box chart method was used to screen the concentration data for the whole year of 2018.Five in-put indicators including PM_(2.5),wind speed,wind direction,air temperature and air humidity were finally chosen as the input variables through combination and comparison.The results indicate that the predicted results of the model are in agreement with the field measurement results with about R=0.7.In addition,the missing data in September was ob-tained and the results were consistent with the annual CO_(2) release law.The MLPNN model established based on the on-site monitoring data can be used to predict and evaluate the carbon dioxide release,which is of great significance for the control of CO_(2) and other landfill gas emission.

关 键 词：MLP神经网络 填埋气 温室效应 CO_(2) 气象因素 

分 类 号：X51[环境科学与工程—环境工程]