Abstract: The chemical oxygen demand (COD) of seawater directly determines the degree of pollution of seawater water quality. The fluorescence interference in the traditional ultraviolet(UV)-visible light detection is relatively large, and the infrared absorption peak of water molecules is seriously affected when the chemical oxygen demand of seawater is detected by the near infrared light. A COD detection method for seawater based on Raman spectroscopy was proposed. The simulated seawater samples of different concentrations were used as the subjects to determine the characteristic Raman displacement of 981.6 cm^-1. After pre-processing of Raman spectroscopy, by partial least square method, a regression model was established for the relative strength of Raman spectrum and the seawater COD values measured by potassium permanganate method. The experimental results show that the correlation coefficient between the training set and the prediction set reaches 0.99, the determination coefficient of the verification set reaches 0.990 9, and the predicted mean square root error is 0.79 mg/L.