In order to monitor the oil pollution of water real time and accurately for the environmental protection, an intelligent online detection system for the mineral oil in water is put forward in the present paper, based on the technology of ultraviolet fluorescence and internet of things (IOT). For this system, the resolution can be improved by using the higher precision asymmetric Czemy-Turner monochromator;the impact of light fluctuations on the results of exploration can be corrected by a bunch reference light;the optical system deviation caused by the instrument vibration can be reduced by optical fiber transmission;the coupling efficiency of fiber and output signal can be increased by a special fiber beam;the real-time measurement, data processing and remote control can be achieved by the control module and wireless communication module. This system has characteristics of high integration, high precision and good stability etc. The concentration of the unknown sample can be accurately calculated by the methods of parallel algorithms of chemometric metrology and the calculation errors caused by different components can be reduced by the theory of chemical correction factor analysis. The fluorescence spectra of three kinds of sample solution, diesel, engine and crude oil in preparative concentration of 10, 25, 50 and 100 mg·L-1 were measured by this system respectively. The absorption wavelengths of the above-mentioned three oils were measured to be 256, 365 and 397 nm by a grating spectrometer;their absorbances were measured to be 0.028, 0.036 and 0.041 by fluorescence spectrophotometer, respectively. Their fluorescence emission wavelengths are 355, 419 and 457 nm respectively. Finally the concentration detection limits of the mineral oil in water of diesel, engine and crude oil were obtained, i.e. 0.03, 0.04 and 0.06 mg·L-1 respectively. Their relative errors are 2.1%, 1.0% and 2.8% respectively.