基于模糊聚类的PM_(2.5)拟合组分选择模型的研究

提出了一种新的PM2.5源成分谱拟合组分选择模型,在充分考虑拟合过程的物理意义的基础上,采用聚类正确率作为组分选择的依据.实验验证,该模型能够准确获取较好的拟合主组分,相比与经验选或者手动盲选所得拟合结果,我们提出的模型将成功拟合(误差范围在0~0.05之间)的比例由40%提升到83%.

The fitting component selection model of PM2.5 based on fuzzy clustering

In current research, there is a lack of uniform standards for components selection in PM_2.5 source profile apportionment. Researchers tend to choose the component manually and empirically, leading to a subsequent poor fitting result, or even failures. Concerning on this problem, this paper has proposed an innovative component selection model of PM_2.5 source profiles apportionment. On the basis of the physical representative of each component, the proposed model calculates the accuracy of fuzzy clustering as the standard score for selection. The experiments prove that our model outperforms the traditional empirical models. The successful rate for fitting, measured by the fitting errors in 0 to 0.05, grows to 83% by implementing our model, in contrast to rate of 40% from the traditional selection model.