摘要
为了研究天津地区PM2.5与5种日常监测的污染物——PM10、SO2、CO、O3和NO2的定量关系,通过天津环保物联网获取了11个环境监测点2013年8月—2014年5月的日监测数据,分别尝试了多元一次回归模型、二次回归模型和基于主成分分析回归模型进行建模与量化分析。相关性上,天津地区各站点PM2.5与5种污染物显著相关,其中与PM10、SO2和NO2相关系数分别达到0.89±0.03、0.63±0.09和0.69±0.06。模型评价结果表明,利用5种污染物监测指标预估PM2.5浓度基本可行,基于主成分分析模型效果最好,拟合度能够达到0.85以上,平均误差在20%左右。
In this study,the daily monitoring data from August 2013 to May 2014 acquired through the Internet of Things for Environmental Protection of Tianjin were analyzed to identify the relationship between PM2. 5and the common pollutants（ PM10,SO2,CO,O3,NO2）. Linear multiple regression,quadratic multiple regression and multiple regression based on principal component analysis are tried to be used in the model,which describes the relationship in form of math. The correlation coefficients show the significant correlation between PM2. 5and the five kinds of common pollutants. The coefficients of PM10,SO2,and NO2 are 0. 89 ± 0. 03,0. 63 ± 0. 09,0. 69± 0. 06,respectively. The concentration of PM2. 5can be estimated by the five kinds of common pollutants according to the evaluation of the model. Statistical result based on principal component analysis shows the best perfermance among three models with 0. 85 goodness-of-fit level and 20% average forecasting error.
出处
《环境工程学报》
CAS
CSCD
北大核心
2015年第10期5017-5023,共7页
Chinese Journal of Environmental Engineering
基金
国家重大科学仪器设备开发专项(2012YQ060165)
天津市科技支撑计划项目(14ZCZDSF00005)
关键词
环境监测
PM2.5
多元回归模型
主成分分析模型
线性相关性
environmental monitoring
PM2.5
multiple regression model
regression model based on princi-pal component analysis
linear correlation
