基于BP神经网络的河北中南部空气质量预报研究  

Study on Prediction of Air Quality in Central and Southern Hebei Based on BP Neural Network

作　　者：张珺[1,2] 王式功[1] 杜亮亮[2] 王娜[3] ZHANG Jun;WANG Shi-gong;DU Liang-liang;WANG Na(Department of Atmospheric Science,Lanzhou University,Lanzhou 730000,China;Handan Municipal Meteorological Bureau of Hebei Province,Handan 056001,China;Baoding Municipal Meteorological Bureau of Hebei Province,Baoding 071000,China)

机构地区：[1]兰州大学大气科学系,甘肃兰州730000 [2]河北省邯郸市气象局,河北邯郸056001 [3]河北省保定市气象局,河北保定071000

出　　处：《江西农业学报》2019年第5期96-102,共7页Acta Agriculturae Jiangxi

基　　金：河北省科技计划项目“邯郸市重污染天气预报预警研究”;河北省气象局气象科研项目“河北中南部重污染天气气象条件及其预报研究”

摘　　要：详细阐述了基于相关性较好的初始样本的BP神经网络空气质量预报模型的建立过程。以石家庄和邢台为例,将相邻两日污染物浓度差值作为预报量,利用前一日污染物浓度和当日气象要素日均值为预报因子,两者结合起来进行空气质量预报,并以冬季空气质量模型为例,对空气质量等级预报准确率进行检验。结果表明,石家庄和邢台SO2和O3的等级预报准确率为90%以上,PM2.5、PM10的等级预报准确率均为80%以上,首要污染物预报准确率均为80%以上。总体上,石家庄的空气质量等级预报准确率好于邢台。This paper expounded the process of establishing BP neural network air quality prediction model based on the initial samples with better correlation.Taking Shijiazhuang and Xingtai as examples,this article took the difference value of the concentration of pollutants between adjacent two days as the forecast object,used the concentration of pollutants on previous day and the daily average value of meteorological factors on this day as the forecast factors,and predicted the air quality of Hebei province.Taking the winter air quality prediction model as an example,this study tested the accuracy rate of air quality grade prediction.The results showed that:the prediction accuracy of SO2 and O3 grade in Shijiazhuang and Xingtai was all more than 90%,that of PM2.5 and PM10 grade was all over 80%,and that of the primary pollutants was all over 80%.In general,the accuracy rate of air quality prediction in Shijiazhuang was better than that in Xingtai.

关 键 词：河北中南部 空气污染 BP神经网络 空气质量预报 

分 类 号：X823[环境科学与工程—环境工程] TP183[自动化与计算机技术—控制理论与控制工程]