基于Matlab的BP神经网络在大气污染物浓度预测中的应用  

The Application of Concentration Forecasting of Air Pollutant Based on BP Neural Network in Matlab

作　　者：欧阳钧[1] 王爱枝[2] 

机构地区：[1]上海市长宁区环境监测站,上海200052 [2]中国气象科学研究院上海办事处,上海200011

出　　处：《环境科学与管理》2009年第11期176-180,共5页Environmental Science and Management

摘　　要：为了寻求有效控制和改善环境质量的相应措施,选用了英国伦敦Bloomsbury监测站的PM10小时平均浓度监测资料,采用“提前终止法”泛化改进的BP神经网络模型,预测PM1024 h内的小时平均浓度。结果表明：采用BP神经网络法对大气污染物浓度进行预测,预测相对误差在2%-48%之间,且绝大部分在2%-17%之间,预测精度较高,泛化能力较好,为大气污染物浓度预测提供了一种全新的思路和方法。In order to find the measures of controlling and improving environmental quality effectively, The BP neural network model improved by the way of advance termination is used for forecasting the average concentration（lhour） of PM10 in 24 hours, through using the materials of average concentration（lhour） of PM10 monitored by the Bloomsbury Site in London. The results show： The prediction relative errors are between 2 percent and 48 percent, and the most numbers less than 17 percent. The forecasting precision is high and the adaptability is better. It provides a effective method for Concentration Forecasting of Air Pollutant.

关 键 词：MATLAB BP神经网络 大气污染物 预测 

分 类 号：X823[环境科学与工程—环境工程]