Kerosene, diesel and gasoline are taken as research objects without considering each oil′s components. By fluorescence spectroscopy analysis, a dimension reduction method that turns three-dimensional spectrum into two-dimensional spectrum is proposed by linking the spectral datum matrixes end to end , drawing its envelope and extracting the apparent statistic (the standard deviation, coefficient of kurtosis and coefficient of skewness are chosen). The method is applical to the systems of certain degree of spectral overlap but low spectral shape similarity. The recognition rate of the oils is over 98% with clustering analysis. According to the Lambert-Beer law, fitting curve of fluorescence intensity and concentration is used for quantitative analysis and standard addition recovery rate is over 95%. Distinguish efficiency is increased by 43%, compared with the parallel factor method used commonly.