基于支持向量回归的PM_(2.5)浓度实时预报  

Real time prediction of PM_(2.5)concentration based on support vector regression algorithms

作　　者：朱亚杰[1] 李琦[1] 侯俊雄 冯逍[1] 范竣翔 

机构地区：[1]北京大学地球与空间科学学院遥感与地理信息系统研究所,北京100871

出　　处：《测绘科学》2016年第1期12-17,22,共7页Science of Surveying and Mapping

基　　金：国家科技支撑计划项目(2012BAC20B06)

摘　　要：为了研究适合于我国当前重污染天气的实时空气质量预报模型,论文利用支持向量回归方法对北京市地面空气质量监测数据和气象数据进行分析,构建了基于支持向量回归的PM2.5浓度实时预报模型。实验表明,该方法能够对未来6日内的日均PM2.5浓度以及未来0~72h内的小时级PM2.5浓度进行预报,且模型训练过程和预报过程都耗时很短,适用于建立PM2.5浓度实时预报系统。In order to develop a real time air quality prediction model suitable for severe pollution condition of China,the paper used the support vector regression algorithm to analyze the surface air quality monitoring data and meteorological data of Beijing city,and then presented an real time prediction model of PM2.5concentration.Experimental result showed that the model could forecast daily PM2.5Concentration within six days and hourly PM2.5 Concentration within 72 hours with high speeds of both model training and model predicting,which makes it a favorable method to build a real time forecasting system of PM2.5Concentration.

关 键 词：支持向量回归 空气质量 PM2.5浓度预报 

分 类 号：P208[天文地球—地图制图学与地理信息工程]