Using spectroscopy sensing technology to measure water COD is the trend of development of modern environmental monitoring. Compared to the traditional chemical analysis is has the benefits of online continuous detection of environmental water samples for real-time monitoring of water COD. This paper collected real water samples, using laser-induced breakdown spectroscopy (LIBS) to obtain water samples of spectral data. Establish water sample COD quantitative prediction model combining Partial Least Squares regression (PLS) by different spectral pretreatment method, then quantitative prediction of LIBS spectrum measurement method of water COD and the relevant model parameters were analyzed. Found that the baseline correction superimposed S-Golay derivative partial least-squares model had better prediction results. The determination coefficient of calibration samples were 0.995 8, while the determination coefficient of prediction were 0.975 3, with RMSEC of 4.438 7 and RMSEP of 9.733 9. The experimental results showed that spectrum sensing technology can be used in the actual environment of water COD quantitative predictive analysis, laid the theoretical foundation for the development of portable water testing equipment.