Abstract: In order to achieve efficient, rapid and pollution-free detection of ammonia nitrogen, this study constructed a quantitative prediction model for ammonia nitrogen in biogas slurry based on the fusion of near-infrared（NIR）and ultraviolet-visible（UV-Vis）spectral features. The partial least squares（PLS）regression method was used to screen the optimal spectral data preprocessing methods, and the competitive adaptive reweighted sampling algorithm（CARS）was employed to extract characteristic wavelengths. The characteristic wavelengths were fused for PLS modeling. The results showed that the prediction performance（RPD=5.924 3）of the NIR characteristic wavelength model was better than that of the full spectrum, while the prediction model performance（RPD=4.322 2）of the UV-Vis characteristic wavelength model was slightly lower than that of the full spectrum. Through the fusion of NIR-UV-Vis characteristic wavelengths, the RPD value of the prediction model reached 6.511 2, which greatly improved the accuracy and reliability of ammonia nitrogen detection in biogas slurry.

Key words: biogas slurry, ammonia nitrogen, near-infrared spectroscopy, ultraviolet-visible spectroscopy, spectral fusion