大气颗粒物源解析的BP网络权重分析模型

A Model of Sources Apportionment of Atmospheric Particulates Based on Weight Analysis of Networks

应用BP网络进行大气颗粒物源解析,将大气采集样本中的元素含量及其污染源成分谱构成训练样本集,用BP网络进行训练.由训练好的网络的权值分布可以计算出网络各输入节点的权重贡献率.根据输入节点的权重贡献率可以确定与其相联系的污染排放源的重要性,从而实现大气颗粒物的源解析.将BP网络权重分析法与其它源解析法应用于实例进行分析比较,结果表明:BP网络权重分析用于大气颗粒物源解析意义明确,方法简便可行.

The approach is based on the weight analysis of BP network. With the data of the element contents in atmospheric samples and the source profiles of atmospheric particulates as training samples, the training was performed using BP network. The weight contribution rates of input points of BP network were calculated with the weight distribution of trained BP network, and the importances of pollutant soures of atmospheric particulates were derived based on the weight contributions rates. Compared with results from other methods of sources apportionment by a specific case, it shows that the application of weight analysis of BP network to sources apportionment of atmospheric particulates has distinct meaning, simplicity and practicability.