Abstract: Fluorescence intensity of twenty two crude oils and twelve marine fuels were detected by a constant wavelength synchronous fluorescence spectrometry at three feature wavelengths 280±2 nm、300±2 nm and 332±2 nm, A logistic regression model with fluorescence and wavelength as variables was established, which was used for rapid quantitative identification of crude oil and marine fuel spilled on the sea. Hosmer-Lemeshow goodness of fit test of 2.43 less than 15.51 and P of 0.97 greater than 0.05, these indicated that the extraction and the utilization of modeling data were sufficient. Receiver operating characteristic curves of oil samples before and after weathering showed that the model had high accuracy for oil species distinguishing, AUC was greater than 0.9, the sensitivity and the specificity were greater than 0.85. Fifteen oil samples except modeling were verified by the model, the calculation results matched with the actual truth very well, the identification accuracy reached 93%. In addition, The model was also applicable to the identification of oil species after weathering.