摘要
基于四川雅安城市空气质量预报和大气污染防控的需求以及冬季以颗粒物(PM2.5和PM10)污染为主、夏季以臭氧(O3)污染为主的特点,本文利用雅安市2015-2018年空气污染监测数据以及同期气象观测资料,重点分析雅安市空气污染物PM2.5、PM10和O3浓度变化特征的基础上,利用灰色关联度方法对上述污染物浓度与气象要素的相关关系进行了细致分析;通过BP神经网络进行两者的数学建模,构建了雅安市空气质量短期预报模型,并进行了试预报检验。研究表明:雅安市2015-2017年期间污染物O3、PM2.5、PM10浓度呈上升的趋势,空气质量达标率自2015年的92.7%降低到2017年的82.2%,2018年达标率略有上升为88%,但仍出现了9天中度污染和1天重污染。污染物浓度与气象要素变化相关密切,其中,降雨量和气压与PM2.5和PM10污染关联最大,表明雅安作为四川盆地的"雨城",其降水对颗粒物的湿清除效应是很显著的;而气温和风速与O3污染关联最大,恰好反映了高温和由高温所隐含的强辐射对O3生成的促进作用。由BP神经网络所建立的雅安O3预报模型,其准确度较稳定,各季7天平均相对误差都<19%,并且预报效果排序为夏季>冬季>秋季>春季;由BP神经网络所建立的雅安PM2.5预报模型,其在春季和夏季预测准确度较好,两季7日平均相对误差都<16%,秋季相对误差略高一点,其四季预报准确度排序为夏季>春季>秋季>冬季。此研究结果可为当地空气质量预报业务的开展提供技术支持。
Based on the need of air quality forecast and air pollution prevention in Ya’an,Sichuan,air pollution monitoring data of Ya’an City from 2015 to 2018 and meteorological data at the same period were used to analyze the correlations between the air pollutants concentrations and meteorological factors in detail by Gray Correlation Method.The short-term forecast models of air quality in Ya’an city were constructed with BP Neural Network method,and forecast results were checked too.The results showed that the pollutants concentrations of O3,PM2.5 and PM10 in Ya’an City showed upward trends from 2015 to 2017,with the air quality passing rate dropping from 92.7%to 82.2%,but the passing rate rising slightly to 88%in 2018,however there were still 9 days with moderate pollution and 1 day with heavy pollution.The pollutants concentrations were closely related to meteorological factors,of which rainfall and air pressure were most associated with PM2.5 and PM10 pollutions,indicating that Ya’an,as the"rain city"of Sichuan,had a significant wet removal effect on particulate matter.While temperature and wind speed were most correlated with O3 pollution,which just reflected the promotion of O3 generation by high temperature and strong radiation implied by high temperature.Using the BP neural network,the forecast model of O3 had a stable accuracy,with an average relative error less than 19%for each season in 7 days,and the forecast results were sorted from summer,winter,autumn to spring.The forecast models of PM2.5 had better prediction accuracy in spring and summer,with an average relative error less than 16%in 7 days,a slightly higher relative error in autumn.This results could provide technical support for the development of local air quality forecasting operations in Ya’an.
作者
吴亚平
张琦
王炳赟
王式功
邵平
WU Yaping;ZHANG Qi;WANG Bingyun;WANG Shigong;SHAO Ping(Plateau Atmosphere and Environment Key Laboratory of Sichuan Province,College of Atmospheric Sciences,Chengdu University of Information technology,Chengdu 610225,Sichuan,China;Bureau of Meteorology in Ya'an,Ya’an 625000,Sichuan,China;Academician work center in Zunyi,Zunyi 563000,Guizhou,China)
出处
《高原气象》
CSCD
北大核心
2020年第4期889-898,共10页
Plateau Meteorology
基金
国家自然科学基金项目(91644226,41775147)
四川省雅安市科技局2019年度科技计划项目(2019yyjskf02)。
关键词
大气污染物
气象要素
灰色关联度
BP神经网络
预测模型
Atmospheric pollutants
meteorological factors
gray correlation
BP neural network
forecast model
