摘要
建立了大气污染物浓度与影响因子之间的BP神经网络,对城市中各监测点位的次日大气污染物浓度进行预测,采用GIS的插值分析进行污染物空间分布预测,其中BP神经网络的输入向量采用AGNES算法进行处理。以太原市区SO2、PM10浓度预测为例,选择气温、湿度、降水量、大气压强、风速和前5天的污染物浓度等10个参数训练BP神经网络,结果表明,BP神经网络的训练效果较好,预测结果与实际浓度显著相关,R2分别为0.988、0.976;结合太原市8个监测点位的污染物浓度预测值,运用GIS空间差值法绘出SO2、PM10的浓度分布预测图,该图与实际情况大体符合,并且与国控大气污染企业的分布显著相关,Pearson相关系数分别为0.969、0.949。
A BP neural network between air pollutant concentrations and impact factors is established in this research. Morrow air pollutant concentrations of city air monitoring sites can be forecasted by the neural network. Distribution of city air pollutant can also be forecasted by interpolation in GIS. Input vectors of BP neural network are manipulated by AGNES algorithm. The paper takes concentration forecasts of SO2 and PM10 in Taiyuan urban district as examples. Temperature, humidity, precipitation, atmospheric pressure, wind speed and previous 5 days＇ concentrations were inputted to train the BP neural network. Results indicate that the train effects were good and forecast outputs have significant correlations with the real data. R2 of them are 0. 988 and 0. 976. Combining with the forecast results of 8 air monitoring sites,interpolation in GIS is utilized to draw distribution maps. Results indicate that the maps well fit the real circumstances, and have significant correlations with distribution of state control enterprises in air pollution. Pearson coefficients are 0. 969 and 0. 949.
出处
《中国环境监测》
CAS
CSCD
北大核心
2015年第3期113-117,共5页
Environmental Monitoring in China
