摘要：Total phosphorus and nitrate nitrogen are the most important indicators in denitrification phosphorus removal. Traditionally, it is time-consuming and complex to measure the concentration of total phosphorus and nitrate nitrogen, which usually lead to much pollution. In addition, the total phosphorus and nitrate nitrogen have to be analyzed separately with different methods, which lead to high cost in time and labor. Near infrared spectroscopy was used for simultaneous analysis of total phosphorus and nitrate nitrogen in denitrifying phosphorus removal. Near infrared spectra of total phosphorus and nitrate nitrogen were modeled through back-propagation neural network (BP neural network). To reduce redundant information, near infrared spectra were pretreated respectively with second derivative, multiple scattering correction and wavelet denoising, combined with principle component analysis. The results showed that wavelet denoising, coupled with principle component analysis, produced the most effective calibration models for total phosphorus and nitrate nitrogen. The calibration models were then tested with samples of the validation set, and the robustness and the validity of these models was attested by the correlation coefficient (rc) and root mean square error of prediction (RMSEP). As to the model for total phosphorus, the rc is 0.9422, and the RMSEP is 2.360. As to the model for nitrate nitrogen, the rc was 0.9006, and the RMSEP was 2.944. Total phosphorus and nitrate nitrogen are the most important indicators of the effect of denitrification phosphorus removal. In summary, this study indicated near infrared spectroscopy as a convenient method to monitor simultaneously total phosphorus and nitrate nitrogen in denitrification phosphorus removal, and to evaluate the effect of denitrification phosphorus removal.
