多种神经网络在华北西部区域城市空气质量预测中的应用  

Forecasting model of air pollution index based on multi- artificial neural network in western region of Northern China

作　　者：谢超[1] 马民涛[1] 于肖肖 

机构地区：[1]北京工业大学环境与能源工程学院,北京100124

出　　处：《环境工程学报》2015年第12期6005-6009,共5页Chinese Journal of Environmental Engineering

摘　　要：据华北西部区域4个主要城市2003—2012年API日报数据和相应时段的地面气象要素数据,利用4种(BP、Elman、T-S模糊、小波)神经网络构建预测模型并预测相应城市大气环境质量。研究结果显示,4种模型在可靠性、预测精度方面均可满足应用要求可用于实际预测;具有动态反馈能力的Elman神经网络的预测精度以及泛用性要优于具静态馈能力的其他3种网络模型,说明动态神经网络更适用于城市大气环境质量预测。4种神经网络的决策权重大小及其排序虽各不相同,但体现出相似规律性,日最低气温、日均气压、前日API对输出数据的影响较大,说明逆温现象引发的持续性、区域性污染是该地区主要环境问题。The daily air pollution index（ API） and corresponding ground meteorological elements data in2003—2012 of the four major cities in the western region of Northern China were collected for the research. The BP neural network,Elman neural network,T-S fuzzy neural network and wavelet neural network were used to build the API forecasting models. The conclusions were as follows： the four ANN-based models with reliability,high prediction accuracy for API forecasting are successful established. The Elman neural network with dynamic feedback capability is better than the other three static models in high prediction accuracy and good generalization. Thus,in this sense,the dynamic neural networks are optimal urban air quality forecasting models. Although the decision weights and their rank of four ANN-based models are different,it still shows a similar regularity. The daily minimum temperature,average daily pressure and the yesterday API were closely related to the output date.The results demonstrated that the persistent regional pollution triggered by the temperature inversion is the major environmental problem in the region.

关 键 词：华北西部 空气污染指数(API) 气象要素 BP神经网络 ELMAN神经网络 T-S模糊神经网络 小波神经网络 

分 类 号：X823[环境科学与工程—环境工程]