欢迎来到天天文库
浏览记录
ID:57924014
大小:564.60 KB
页数:5页
时间:2020-04-14
《基于AGNES算法优化BP神经网络和GIS系统的大气污染物浓度预测-论文.pdf》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、第31卷第3期中国环境监测V01.31NO.32015年6月EnvironmentalMonitoringinChinaJun.2015基于AGNES算法优化BP神经网络和GIS系统的大气污染物浓度预测姚宁h,马青兰h,张晶,文印,1.太原理工大学:a.环境科学与工程学院,b.现代科技学院,山西太原0300242.广西大学林学院,广西南宁5300043.亚热带农业生物资源保护与利用国家重点实验室,广西南宁530004摘要:建立了大气污染物浓度与影响因子之间的BP神经网络,对城市中各监测点位的次日大气污染物浓度进行预测,采用GIS的插值分析进行污染物空间分布预测,其中BP神经网络的输入向量采用A
2、GNES算法进行处理。以太原市区SO、PM。。浓度预测为例,选择气温、湿度、降水量、大气压强、风速和前5天的污染物浓度等10个参数训练BP神经网络,结果表明,BP神经网络的训练效果较好,预测结果与实际浓度显著相关,分别为0.988、0.976;结合太原市8个监测点位的污染物浓度预测值,运用GIS空间差值法绘出SO:、PM,。的浓度分布预测图,该图与实际情况大体符合,并且与国控大气污染企业的分布显著相关,Pearson相关系数分别为0.969、0.949。关键词:大气污染物;AGNES算法;BP神经网络;GIS中图分类号:X830.3文献标志码:A文章编号:1002—6002(2015)03—0
3、113—05ForecastofAirPollutantC0ncentrati0nsonAGNESAlgorithm.BPNeuralNetworkandGISYAONing,MAQing—lan,ZHANGJing,WENYint1.TaiyuanUniversityofTechnology:a.CollegeofEnvironmentalScienceandEngineering,b.PolytechnicInstitute,Taiyuan030024,China2.ForestryCollegeofGuangxiUniversity,Nanning530004,China3.StateK
4、eyLaboratoryforConservationandUtilizationofSubtropicalAgro—bioresources,Nanning530004,ChinaAbstract:ABPneuralnetworkbetweenairpollutantconcentrationsandimpactfactorsisestablishedinthisresearch.Morrowairpollutantconcentrationsofcityairmonitoringsitescanbeforecastedbytheneuralnetwork.Distributionofcit
5、yairpollutantcanalsobeforecastedbyinterpolationinGIS.InputvectorsofBPneuralnetworkaremanipulatedbyAGNESalgorithm.ThepapertakesconcentrationforecastsofSO2andPM10inTaiyuanurbandistrictasexamples.Temperature,humidity,precipitation,atmosphericpressure,windspeedandprevious5days’concentrationswereinputted
6、totraintheBPneuralnetwork.Resultsindicatethatthetraineffectsweregoodandforecastoutputshavesignificantcorrelationswiththerealdata.Rofthemare0.988and0.976.Combiningwiththeforecastresultsof8airmonitoringsites,interpolationinGISisutilizedtodrawdistributionmaps.Resultsindicatethatthemapswellfittherealcir
7、cumstances.andhavesignificantcorelationswithdistributionofstatecontrolenterprisesinairpollution.Pearsoncoeficientsare0.969and0.949.Keywords:airpollutant;AGNESalgorithm;BPneuralnetwork;GIS近年来,随着国家对大气污染
此文档下载收益归作者所有