Based on ultraviolet-visible spectroscopy technology, a system that can be used for real-time online detection of water quality parameters dissolved oxygen (COD) was designed, and the measurement principle and overall structure (light path and circuit) design of the system were introduced. By detecting the COD of the water body, the spectral data of the COD solution was measured, the partial least squares regression (PLS) algorithm prediction model and the voltage solution model were established using the spectral data, and the prediction accuracy of the two models was compared. At the same time, considering the environmental interference factors in the water body, the impact of turbidity and temperature on the COD measurement results was analyzed, and the constructed model was compensated and corrected, which improved the detection accuracy of the system. The experimental results show that, compared with the traditional water quality COD detection method, the system achieves the purpose of fast, high-precision and real-time online detection.