APPLICATION OF EXTREME LEARNING MACHINE AND NEAR INFRARED SPECTROSCOPY TO RAPID ANALYSIS OF TOTAL PHOSPHORUS AND NITRATE NITROGEN IN DENITRIFYING PHOSPHORUS REMOVAL PROCESS.

Total phosphorus and nitrate nitrogen can be analyzed simultaneously and rapidly in denitrifying phosphorus removal with near infrared spectroscopy technology and extreme learning machine. In modeling, the second derivative, multiple scattering correction, wavelet denoising and principal component analysis were applied to preprocess the raw spectral data. The results show that wavelet denoising combined with principal component analysis is the optimal preprocessing method. The model for rapid analysis of total phosphorus showed that the corrected correlation coefficient (rc) and predictive correlation coefficient (rp) were respectively 0.9232 and 0.9024, and that the corrected root mean square error (RMSECV) and predicted root mean square error (RMSEP) were 0.0203 and 1.4013, respectively. The model for rapid analysis of nitrate nitrogen showed that the rc and rp were respectively 0.9028 and 0.8958, and that the RMSECV and RMSEP were 0.0222 and 0.0299, respectively. The study shows that the extreme learning machine, combined with near infrared spectroscopy technology, is capable of simultaneous rapid analysis of total phosphorus and nitrate nitrogen in denitrifying phosphorus removal system, which provides a theoretical basis for real-time monitoring of water quality indicators in wastewater treatment. [ABSTRACT FROM AUTHOR]