基于遗传神经网络模型的空气能见度预测  

Air visibility forecast based on genetic neural network model

作　　者：马楚焱 祖建[1] 付清盼 罗凌霄[1] 

机构地区：[1]西安交通大学数学与统计学院,西安710049

出　　处：《环境工程学报》2015年第4期1905-1910,共6页Chinese Journal of Environmental Engineering

基　　金：国家基础科学人才培养基金项目(J1210059);国家自然科学基金资助项目(11201368);西安交通大学本科生科研训练和实践创新基金项目(2014X07001)

摘　　要：由于雾霾导致的空气能见度降低,给人们的出行带来很多不便。针对这一现象,构建基于遗传神经网络算法的空气能见度预测模型。将与空气能见度相关的7种气象因子和6种污染物浓度因子经过主成份分析后作为输入数据,输出8:00能见度和14:00能见度。该模型能够克服BP神经网络易陷入平坦区域和局部最优解的问题。以西安市2013-1-1—8-16的数据训练遗传神经网络,通过使用灰色模型获得预测时间段8-17—23的输入数据,可以得到这段时间能见度的预测值。通过与BP神经网络模型的比较,发现遗传神经网络预测模型在预测结果的相关性和绝对误差方面均优于BP神经网络模型,因此,可以更准确地预测空气能见度。Due to the poor air visibility caused by smog,it brings a lot of inconvenience to people's commute. Based on this phenomenon,an air visibility forecasting model was proposed,which is based on the genetic neural network algorithm. In the air visibility forecasting model,the data handled by the principal components analysis( PCA) over seven meteorological factors and six pollutant concentrations factors,which were related with the air visibility,were used as the input,and the air visibility at time 8: 00 and 14: 00 were used as the output data. The model can overcome the problems of flat area and local optimal solutions in BP neural network model. In this paper,the genetic neural network model was trained with the data from Jan. 1 to Aug. 16,2013 in the city of Xi'an. The visibility of Aug. 17 to Aug. 23 could be estimated by the trained model with the input data got from grey model. Compared with the BP neural network model,the results show that the genetic neural network based air visibility predicting model performs better than the BP neural network model in terms of correlation coefficient and absolute error,thus,it can provide much more accurate forecast for the air visibility.

关 键 词：能见度 遗传算法 BP神经网络 

分 类 号：P427.2[天文地球—大气科学及气象学] X51[环境科学与工程—环境工程]