摘要
在研究北京市能见度变化特征的基础上,利用北京市环境监测站2015-2017年的空气污染物检测数据及同期美国国家环境预报中心的全球预报系统数值资料,筛选出主要的预报因子,分别用神经网络和多元逐步回归法建立预报模型,并进行试预报检验.结果表明,神经网络预报效果优于多元逐步回归,平均预报准确率达到75%(多元逐步回归为66%).神经网络在0~10 km能见度预报方面能够取得更好的效果,预测数据与观测值更为接近.
Based on the study of visibility variations in Beijing,appropriate predictors were selected from air pollutant detection data and global forecast system numerical data of the national centers for environmental prediction(NCEP)from 2015 to 2017.The neural network and the multiple stepwise regression were used to establish a prediction model to conduct the forecasting test.The results indicated that the neural network prediction model had a higher accuracy than the multiple stepwise regression.The average forecast accuracy rate was 75%for neural network and 66%for the multiple stepwise regression.Neural network could achieve a better effect in 0-10 km visibility prediction and the predicted data was closer to the observed value.
作者
周开鹏
黄萌
樊旭
马晓玲
杨子凡
尚可政
Zhou Kai-peng;Huang Meng;Fan Xu;Ma Xiao-ling;Yang Zi-fan;Shang Ke-zheng(College of Atmospheric Science,Lanzhou University,Lanzhou 730000,China;Jiuquan Satellite Launching Center,Kuerle 841001,Xinjiang,China)
出处
《兰州大学学报（自然科学版）》
CAS
CSCD
北大核心
2020年第4期522-526,共5页
Journal of Lanzhou University(Natural Sciences)
基金
国家自然科学基金项目(91644226,41575138)
国家基础科技条件平台建设项目(NCMI-SBS17-201707,NCMI-SJS15-201707)
国家公益性行业(气象)科研专项项目(GYHY201306047)
关键词
能见度预报
神经网络
多元逐步回归
visibility forecast
neural network
the multiple stepwise regression
